Re-evaluation Of Human Health Effects Of Atrazine:

Transcript

Re-Evaluation of Human Health Effects of Atrazine: Review of Cancer Epidemiology, Non-cancer Experimental Animal and In vitro Studies and Drinking Water Monitoring Frequency Presented Jointly To The FIFRA Scientific Advisory Panel By: U.S. Environmental Protection Agency Office of Pesticide Programs Health Effects Division and Environmental Fate and Effects Division in collaboration with the Office of Research and Development Presented On: July 26-29, 2011 Page 1 of 184 TABLE OF CONTENTS 1. INTRODUCTION ...............................................................................................................7 2. MODE OF ACTION (MOA) & EXPERIMENTAL TOXICOLOGY ....................................10 2.1 HISTORICAL BACKGROUND ............................................................................................................................................ 10 2.2 2010 SCIENTIFIC RE-EVALUATION .................................................................................................................................. 11 2.3 MODE OF ACTION (MOA) AND ADVERSE HEALTH OUTCOMES............................................................................................. 12 2.3.1 Background .................................................................................................................................................... 12 2.3.2 Reproductive Senescence and Mammary tumors in rats: Well established MOA ......................................... 14 2.3.3 Adverse Health Outcomes – LH Changes as a Sentinel Effect ......................................................................... 16 2.3.3.1 Factors influencing GnRH Modulation of LH following atrazine exposure .................................................................. 17 2.3.3.1.2 GnRH in primates ............................................................................................................................................. 17 2.3.3.2 Females of Reproductive Age: Role of LH, Progesterone and Corticosterone in Atrazine Induced Changes in Ovarian Cyclicity ...................................................................................................................................................................... 18 2.3.3.2.1 Estrogen and progesterone regulation of LH secretion ..................................................................................... 19 2.3.3.2.2 Differential effects of single versus multiple atrazine exposures on the LH surge ........................................... 21 2.3.3.2.3 LH Attenuation in Humans – Pharmaceutical Experience .................................................................................. 24 2.3.3.3 Prepubertal Individuals: Atrazine-induced delays in sexual maturation .................................................................... 25 2.3.3.3.1 Delays in Female Sexual Maturation .................................................................................................................. 26 2.3.3.3.2 Delays in Male Sexual Maturation .................................................................................................................... 27 2.3.3.4 Atrazine-associated decrease in prolactin and prostatitis in young rats ................................................................... 29 2.3.4 Mammary gland whole mounts ..................................................................................................................... 30 2.3.5 Summary ........................................................................................................................................................ 31 3. EPIDEMIOLOGIC EVIDENCE OF THE CARCINOGENIC POTENTIAL OF ATRAZINE ............................................................................................................................................33 3.1 INTRODUCTION ........................................................................................................................................................... 33 3.1.1 Background.................................................................................................................................................... 33 3.1.2 Identification of Epidemiology Studies included in Current Review ............................................................... 34 3.2 EPIDEMIOLOGY LITERATURE OF THE CARCINOGENIC POTENTIAL OF ATRAZINE ......................................................................... 35 3.2.1 Overview of Epidemiologic Database ............................................................................................................ 36 3.2.2 Agricultural Health Study (AHS) Atrazine Cancer Point Estimates ................................................................ 37 3.2.2.1 Study Design and Methods....................................................................................................................................... 37 3.2.2.2 AHS Strengths and Weaknesses ............................................................................................................................... 39 3.2.3 Reproductive and Endocrine System Cancers ................................................................................................ 39 3.2.3.1 3.2.3.2 3.2.3.3 3.2.3.4 Prostate Cancer ........................................................................................................................................................ 40 Breast Cancer ........................................................................................................................................................... 43 Ovarian Cancer ......................................................................................................................................................... 48 Thyroid Cancer ......................................................................................................................................................... 52 3.2.4 Lymphohematopoietic Cancers ..................................................................................................................... 53 3.2.4.1 3.2.4.2 3.2.4.3 3.2.4.4 Population-Based Case-Control Studies: Midwestern U.S........................................................................................ 54 Hospital-Based Case-Control Studies (France) ......................................................................................................... 56 AHS Estimates of Atrazine and Lymphohematopoietic Cancers............................................................................... 58 Conclusions: Lymphohematopoietic Cancers ........................................................................................................... 59 3.2.5 Other Cancer Sites ......................................................................................................................................... 60 3.2.5.1 Brain/Glioma ............................................................................................................................................................ 60 Page 2 of 184 3.2.5.2 Pediatric Cancers ...................................................................................................................................................... 61 3.2.5.3 Colon Cancer ............................................................................................................................................................ 63 3.2.5.4 Other AHS ................................................................................................................................................................. 64 3.3 SUMMARY OF EVIDENCE CONCERNING THE CARCINOGENIC POTENTIAL OF ATRAZINE ............................................................... 65 3.3.1 Synthesis and Integration of Toxicology and Epidemiology Literature.......................................................... 65 3.3.2 EPA Conclusion ................................................................................................................................................. 71 4. PROPOSED UPDATES TO THE DOSE-RESPONSE ASSESSMENT ........................72 4.1 BENCHMARK DOSE ANALYSIS ........................................................................................................................................ 72 4.1.1 Identification of Critical Studies ...................................................................................................................... 73 4.1.2 Methods & Results ......................................................................................................................................... 74 4.1.3 Proposed Point of Departure ......................................................................................................................... 78 4.2 PHARMACOKINETIC ANALYSIS .......................................................................................................................................... 79 4.2.1 Background ...................................................................................................................................................... 79 4.2.2 Updates on the Pharmacokinetics of Atrazine and Its Metabolites in the Rat ............................................... 80 4.2.3 Oral Gavage Dosing of Sprague Dawley Rats with Atrazine .......................................................................... 81 4.2.4 Dietary Exposures of Sprague Dawley Rats to Atrazine ................................................................................. 87 4.2.5 Plasma Clearance Discrepancy Based on Studies with Radiolabeled Atrazine ............................................... 88 4.2.6 Additional Rat Radiolabeled Atrazine Studies Evaluated for Plasma Clearance ............................................ 90 4.2.7 Pharmacokinetic Studies with Radiolabeled Atrazine in Monkeys ................................................................. 93 4.2.8 Pharmacokinetic Studies with Atrazine in Humans ........................................................................................ 94 4.2.8.1 Human Blood Pharmacokinetic Information on Atrazine and its Chloro-s-triazine Metabolites .............................. 95 4.2.8.2 Human Urine Pharmacokinetic Information on Atrazine and its Chloro-s-triazine Metabolites .............................. 95 4.2.9 Updates on the Metabolism of Atrazine ........................................................................................................ 97 4.2.10 Summary of Pharmacokinetic Information .................................................................................................. 98 4.2.11 Progress on Physiologically Based Pharmacokinetic Modeling Efforts ........................................................ 99 4.2.12 Proposed Interim Pharmacokinetic Modeling Approach ............................................................................ 100 4.2.13 Implications for Water Monitoring Frequency ........................................................................................... 103 4.2.13.1 Rat Forward Internal Dosimetry for LH Attenuation ............................................................................................. 104 4.2.13.2 Human Reverse Dosimetry from Rat Internal Dosimetry for LH Attenuation ....................................................... 105 4.2.14 Calculating Average Chlorotriazine Levels in Drinking Water for Human Plasma Triazine AUC Estimates 107 5. SCIENTIFIC CONSIDERATIONS IN POTENTIAL SENSITIVITY OF INFANTS & CHILDREN........................................................................................................................109 5.1 BACKGROUND .......................................................................................................................................................... 109 5.2 EXPERIMENTAL TOXICOLOGY ....................................................................................................................................... 110 5.2.1 Gestational Exposure ................................................................................................................................... 110 5.2.2 Multi-Lifestage Exposure .............................................................................................................................. 111 5.3. SUMMARY............................................................................................................................................................... 112 6.EVALUATING ATRAZINE DRINKING WATER MONITORING DATA FOR USE IN HUMAN HEALTH ASSESSMENTS .................................................................................114 6.1 INTRODUCTION ......................................................................................................................................................... 114 6.2 KEY QUESTIONS TO ADDRESS IN ANALYZING MONITORING DATA........................................................................................ 115 6.3 SHORT-TERM TEMPORAL VARIABILITY: DAILY VARIATIONS WITHIN A GIVEN YEAR .................................................................. 117 6.3.1 General Approach for Assessing Uncertainties in Monitoring Related to Short-term Temporal Variability 118 6.3.2 Comparative Analysis of Monitoring Frequency Performance ..................................................................... 120 6.3.3 Geostatistical Analysis and Stochastic Application ...................................................................................... 125 Page 3 of 184 6.3.4 Mechanistic Approach: PRZM Hybrid Model ................................................................................................ 132 6.3.5 Watershed Atrazine Analysis Model............................................................................................................. 133 6.3.6 Generating Atrazine Time Series from Weekly CWS Monitoring for Use in Dietary Exposure Assessments 134 6.4 YEAR-TO-YEAR TEMPORAL VARIABILITY ......................................................................................................................... 137 6.5 SPATIAL VARIABILITY .................................................................................................................................................. 139 6.6 SUMMARY................................................................................................................................................................ 140 7. CASE STUDY: IMPLICATIONS OF MOA & TOXICITY PROFILE ON WATER MONITORING ...................................................................................................................142 7.1 INTRODUCTION ......................................................................................................................................................... 142 7.2 ATRAZINE TIME SERIES ESTIMATES FOR USE IN DIETARY EXPOSURE ASSESSMENTS ................................................................. 143 7.3 ESTIMATING HUMAN PLASMA AVERAGE DAILY AUC FROM ROLLING AVERAGE ATRAZINE CONCENTRATIONS ................................. 145 8. REFERENCES ............................................................................................................149 LIST OF FIGURES FIGURE 1: LH SUPPRESSION AND ADVERSE OUTCOMES OBSERVED IN RATS ....................................................................................... 14 FIGURE 2: AUGMENTATION BY A SUBCUTANEOUS INJECTION OF PROGESTERONE ON AN LH SURGE IN OVARIECTOMIZED RATS IMPLANTED WITH AN ESTRADIOL CAPSULE AT THE TIME OF SURGERY. PROGESTERONE WAS ADMINISTERED JUST PRIOR TO THE SAMPLING OF SMALL ALIQUOTS OF BLOOD BY TAIL NICK OVER THE AFTERNOON OF THE THIRD DAY AFTER OVARIECTOMY. CONCENTRATIONS ARE INDICATED IN NG/ML ± SEM .............................................................................................................................................................. 21 FIGURE 3: THE LH SURGE (NG/ML + SEM) AND AREAS UNDER THE CURVE IN RESPONSE TO 1 AND 4 DAYS OF ATRAZINE TREATMENT IN OVARIECTOMIZED / ESTRADIOL-PRIMED LONG-EVANS RATS. THE NUMBERS IN PARENTHESES INDICATE THE GROUP SIZES FOR EACH OF THE TREATMENTS. NUMBERS WITHIN THE 100 MG/KG COLUMNS AT THE RIGHT INDICATE THE PERCENT CHANGE FROM THE 0 MG/KG GROUP. *P<0.05. ......................................................................................................................................................... 23 FIGURE 4: EPA BMDS EXPONENTIAL MODEL (CONSTANT VARIANCE; BMR = 1 SD FROM CONTROL MEAN) RESULTS FOR ATTENUATION OF LH SURGE IN FEMALE RATS ADMINISTERED ATRAZINE BY GAVAGE - NHEERL DATA. ................................................................. 76 FIGURE 5: EPA BMDS HILL MODEL (CONSTANT VARIANCE; BMR = 1 SD FROM CONTROL MEAN) RESULTS FOR “AREA UNDER THE TRIAZINE EQUIVALENT PLASMA-CONCENTRATION TIME CURVE” AS THE DOSE METRIC - NHEERL DATA (ATTENUATION OF LH SURGE IN FEMALE RATS ADMINISTERED ATRAZINE BY GAVAGE). ...................................................................................................................... 77 FIGURE 6: EPA BMDS HILL MODEL (CONSTANT VARIANCE; BMR = 1 SD FROM CONTROL MEAN) RESULTS FOR “DAILY STEADY STATE AUC FOR TRIAZINE EQUIVALENTS” AS THE DOSE METRIC - NHEERL DATA (ATTENUATION OF LH SURGE IN FEMALE RATS ADMINISTERED ATRAZINE BY GAVAGE). ................................................................................................................................................... 78 FIGURE 7: METABOLIC SCHEME FOR ATRAZINE ............................................................................................................................. 80 FIGURE 8: PLASMA PROFILE OF ATRAZINE FOR 24 HOURS FOLLOWING ORAL GAVAGE DOSING OF RATS WITH 3, 10, OR 50 MG/KG BW ATRAZINE. EACH TIME POINT REPRESENTS AVERAGE OF SIX MEASUREMENTS ± SD. DATA IS FROM CODER 2011. ............................................ 82 FIGURE 9: THE PARALLEL PLASMA PROFILES OF ATRAZINE VERSUS DEA OR DIA FOLLOWING ORAL GAVAGE DOSING OF RATS WITH 3 MG/KG BW ATRAZINE. DATA IS FROM CODER 2011. ............................................................................................................................. 84 FIGURE 10: 24-H PLASMA PROFILE OF ATRAZINE AND ITS CHLORINATED METABOLITES DEA, DIA, AND DACT FOLLOWING ORAL GAVAGE DOSING OF RATS WITH 3 MG/KG BW ATRAZINE. EACH TIME POINT REPRESENTS AVERAGE OF SIX MEASUREMENTS ± SD. DATA IS FROM CODER 2011. ............................................................................................................................................................... 85 FIGURE 11: PLASMA PROFILE OF ATRAZINE, DEA, DIA, AND DACT FROM REPEATED ONCE A DAY ORAL GAVAGE DOSING OF RATS WITH 3 MG/KG BW ATRAZINE FOR 4 DAYS. VALUES REPRESENT THE MEAN OF SIX REPLICATES ± SD. DATA IS FROM CODER 2011. ............................ 86 Page 4 of 184 FIGURE 12: PLASMA PROFILE OF ATRAZINE, DEA, DIA, AND DACT FROM DIETARY DOSING OF RATS WITH ATRAZINE AT A NOMINAL DOSE OF 3 MG/KG BW/DAY ATRAZINE FOR FOUR DAYS. DATA AT EACH TIME POINT REPRESENTS MEASUREMENTS IN 6 ANIMALS. DATA POINTS ARE NOT CONNECTED TO HELP ELUCIDATE PATTERN. DATA IS FROM CODER 2011. ............................................................................ 87 FIGURE 13: PLASMA PROFILES OF RADIOLABELED TRIAZINE EQUIVALENTS RESULTING FROM REPEATED ONCE DAILY DOSING OF RATS WITH RADIOLABELED ATRAZINE AT 1, 3, 7, 10, 50 AND 100 MG/KG BW ATRAZINE. DATA IS FROM THEDE 1987. .................................... 90 FIGURE 14: PLASMA PROFILE AND ANALYSIS OF ELIMINATION OF A SINGLE ORAL GAVAGE DOSE OF 1 AND 100 MG/KG BW RADIOLABELED ATRAZINE IN RATS. DATA IS FROM PAUL ET AL. 1993. ........................................................................................................... 92 FIGURE 15: PLASMA PROFILE AND ANALYSIS OF ELIMINATION OF REPEATED ORAL GAVAGE DOSING WITH EITHER 0.4 OR 4 MG/KG BW 14CATRAZINE IN RATS FOR SEVEN DAYS. DATA IS FROM SIMONEAUX 1985. .................................................................................... 93 FIGURE 16: UPDATED METABOLIC PROFILE FOR ATRAZINE. NEWLY IDENTIFIED METABOLITES ARE ABOVE THE DASHED LINE FOR COMPARISON TO FIGURE 7. NEW METABOLITES ARE FROM LEBLANC ET AL. 2011 AND JOO ET AL. 2011. .............................................................. 98 FIGURE 17: PROPOSED INTERIM PHARMACOKINETIC MODEL FOR ATRAZINE AND ITS METABOLITES ........................................................ 101 FIGURE 18: PROPOSED LINKAGE BETWEEN ATRAZINE EXPOSURE AND OPTIMAL LH ATTENUATION IN RAT STUDIES..................................... 104 FIGURE 19: PROPOSED LINKAGES BETWEEN ATRAZINE EXPOSURE AND LH ATTENUATION IN ADULT HUMANS. THE ASTERISKS INDICATE THE PARAMETERS THAT WERE ALLOMETRICALLY SCALED FROM ADULT RATS. .................................................................................. 106 FIGURE 20: HYPOTHETICAL WATER CHEMOGRAPH SHOWING THE TYPICAL VARIATION IN CHLOROTRIAZINE LEVELS AS A FUNCTION OF TIME. ... 106 FIGURE 21: THE TRAPEZOID RULE FOR ESTIMATING THE AREA UNDER A WATER CHEMOGRAPH FOR A GIVEN DURATION OF EXPOSURE. .......... 107 FIGURE 22: ANALYSIS STRATEGY FOR MODELING PESTICIDE TIME SERIES CONCENTRATIONS FROM MONITORING DATA. .......................... 126 FIGURE 23: VARIOGRAMS FOR 4-DAY AND 7-DAY SAMPLING FREQUENCY FOR THE MAUMEE AND MO-01 2007 DATA: A.) 4-DAY 1995 MAUMEE; B.) 7-DAY 1995 MAUMEE; C.) 4-DAY GRAB MO-01 2007; AND D.) 7-DAY GRAB MO-01 2007................................ 128 FIGURE 24: REPRESENTATIVE REALIZATIONS OF KRIGED ATRAZINE TIME SERIES FROM (A) 4-DAY, (B) 7-DAY, (C) 14-DAY, AND (D) 28-DAY SAMPLING OF THE NCWQR MAUMEE RIVER 1995 DATA.................................................................................................... 129 FIGURE 25: REPRESENTATIVE REALIZATIONS OF KRIGED ATRAZINE TIME SERIES FROM (A) 4-DAY, (B) 7-DAY, (C) 14-DAY, AND (D) 28-DAY SAMPLING OF THE MO-01 2007 DATA. ........................................................................................................................... 130 FIGURE 26: ESTIMATES OF DAILY AND ROLLING-AVERAGE ATRAZINE CONCENTRATIONS FROM THE ACTUAL TIME SERIES (BLACK LINE) AND 5TH AND 95TH PERCENTILES (DASHED RED LINES) OF REALIZATIONS OF KRIGED TIME SERIES FROM 7-DAY SAMPLING OF THE MAUMEE RIVER 1995 DATA. ............................................................................................................................................................... 135 FIGURE 27: ESTIMATES OF DAILY AND ROLLING-AVERAGE ATRAZINE CONCENTRATIONS FROM FROM THE ACTUAL TIME SERIES (BLACK LINE) AND 5TH AND 95TH PERCENTILES (DASHED RED LINES) OF REALIZATIONS OF KRIGED TIME SERIES FROM 7-DAY SAMPLING OF THE MO-01 2007 DATA. ............................................................................................................................................................... 136 FIGURE 28: LOCATION OF CWS IN THE AMP IN RELATION TO VULNERABLE WATERSHEDS IDENTIFIED BY WARP BASED ON ATRAZINE USE ON CORN AND SORGHUM.................................................................................................................................................... 140 FIGURE 29: ESTIMATES OF DAILY AND ROLLING-AVERAGE ATRAZINE CONCENTRATIONS FROM THE ACTUAL TIME SERIES (BLACK LINE) AND 5TH AND 95TH PERCENTILES (DASHED RED LINES) OF REALIZATIONS OF KRIGED TIME SERIES FROM 7-DAY SAMPLING OF THE MAUMEE RIVER 1995 DATA. ............................................................................................................................................................... 144 FIGURE 30: ESTIMATES OF DAILY AND ROLLING-AVERAGE ATRAZINE CONCENTRATIONS FROM THE ACTUAL TIME SERIES (BLACK LINE) AND 5TH AND 95TH PERCENTILES (DASHED RED LINES) OF REALIZATIONS OF KRIGED TIME SERIES FROM 7-DAY SAMPLING OF THE MO-01 2007 DATA. ........................................................................................................................................................................ 145 FIGURE 31: ESTIMATES OF ROLLING AVERAGE DAILY HUMAN PLASMA AUC VALUES CORRESPONDING TO THE ACTUAL TIME SERIES (BLACK LINE) AND 5TH AND 95TH PERCENTILES (DASHED RED LINES) FOR THE MAUMEE RIVER 1995 DATASET ACCORDING TO FIGURE 29. THE DASHED LINE IS FOR COMPARISON TO THE RAT PLASMA POD AUC WITH A 300X APPLIED UNCERTAINTY FACTOR. ...................................... 147 FIGURE 32: ESTIMATES OF DAILY ROLLING-AVERAGE ATRAZINE CONCENTRATIONS FROM THE ACTUAL TIME SERIES (BLACK LINE) AND 5TH AND 95TH PERCENTILES (DASHED RED LINES) OF REALIZATIONS OF KRIGED TIME SERIES FROM 7-DAY SAMPLING OF THE MO-01 2007 DATA. ................................................................................................................................................................................ 148 Page 5 of 184 LIST OF TABLES TABLE 1: SUMMARY OF ADMINISTERED DOSE BMD ANALYSES........................................................................................................ 75 TABLE 2: BMD MODELING RESULTS WITH AVERAGE STEADY STATE PLASMA TRIAZINES AND AUC ESTIMATES............................................ 76 TABLE 3: PHARMACOKINETIC PARAMETERS ESTIMATED FOR ATRAZINE AND ITS CHLORINATED METABOLITES DEA, DIA, AND DACT FOLLOWING ORAL GAVAGE DOSING OF RATS WITH 3, 10, AND 50 MG/KG BW ATRAZINE. DATA ANALYZED IS FROM CODER 2011. ........................ 83 TABLE 4: PHARMACOKINETIC ELIMINATION PARAMETERS ESTIMATED FOR ATRAZINE AND ITS CHLORINATED METABOLITES DEA, DIA, AND DACT FOLLOWING DOSING OF RATS WITH ATRAZINE THROUGH THE DIET AT NOMINAL DOSES OF 3, 10, AND 50 MG/KG BW. DATA IS FROM CODER 2011........................................................................................................................................................ 88 TABLE 5: FIRST ORDER ELIMINATION RATE CONSTANTS ESTIMATED FROM PLASMA DATA IN THE THEDE 1987 STUDY INVOLVING REPEATED ONCE DAILY DOSING OF RATS WITH RADIOLABELED ATRAZINE (THEDE 1987). ..................................................................................... 90 TABLE 6: ESTIMATED PLASMA PHARMACOKINETIC PARAMETERS IN MONKEYS FOLLOWING ORAL GAVAGE DOSING WITH 1, 10, AND 100 MG 14 [ C]-ATRAZINE. DATA IS FROM HUI ET AL. 2011................................................................................................................. 94 TABLE 7: ESTIMATED PHARMACOKINETIC PARAMETERS FROM HUMAN BLOOD AND URINE .................................................................... 96 TABLE 8: EVALUATION OF THE ONE-COMPARTMENT LINEAR MODEL FOR RELATING AN ATRAZINE DOSE TO STEADY STATE PLASMA TRIAZINES.. 102 TABLE 9: SUMMARY OF AVAILABLE ATRAZINE EXPERIMENTAL TOXICOLOGY STUDIES FROM THE GESTATIONAL PERIOD................................ 111 TABLE 10: DOSING REGIMEN MULTI-LIFESTAGE STUDY................................................................................................................ 112 TABLE 11: SUMMARY OF 10,000 SIMULATIONS OF MONITORING ESTIMATES AT DIFFERENT SAMPLING INTERVALS FOR THE AEEMP MO-01 SITE IN 2007. .............................................................................................................................................................. 122 TABLE 12: SUMMARY OF 10,000 SIMULATIONS OF MONITORING ESTIMATES AT DIFFERENT SAMPLING INTERVALS FOR THE NCWQR MAUMEE RIVER SITE IN 1995. ..................................................................................................................................................... 123 TABLE 13: VARIOGRAM MODELS USING THE NCWQR MAUMEE RIVER 1995 AND AEEMP MO-01 2007 TIME SERIES WITH DIFFERENT SAMPLING FREQUENCIES. .............................................................................................................................................. 127 TABLE 14: COMPARISON OF LOWEST AND HIGHEST YEARLY MAXIMUM DETECTIONS FOR CWS IN THE AMP BASED ON NUMBER OF YEARS SAMPLED. ................................................................................................................................................................... 138 Page 6 of 184 1. INTRODUCTION Atrazine – a chlorotriazine herbicide – is currently one of the most widely used agricultural herbicides in the United States, with approximately 70 million pounds of active ingredient applied domestically per year. The Interim Reregistration Eligibility Document (IRED) for atrazine was finalized in January 2003 with an addendum added in October 2003 (US EPA 2003). The Agency completed the reregistration eligibility document (RED) for atrazine in April 2006. In recent years, numerous experimental toxicology and epidemiologic studies evaluating the toxicity profile and/or mode of action (MOA) of atrazine have become available. To consider the extent to which these new studies may impact the Agency’s human health risk characterization for atrazine, the Office of Pesticide Programs (OPP) in collaboration with the Office of Research and Development (ORD) is re-evaluating atrazine and its chloro-s-triazines metabolites [deethyl-atrazine (DEA), deisopropyl-atrazine (DIA), and diamino-s-chlorotriazine (DACT)]. 1 As part of this re-evaluation, the Agency held three meetings of the FIFRA SAP in 2010 which involved atrazine. The July 2011 meeting includes a variety of challenging scientific topics initially discussed at these three 2010 meetings; the July meeting will cover the most recent analyses on each. Specifically, the July SAP meeting will cover issues related to non-cancer effects, particularly related to effects on Luteinizing Hormone (LH) to several different lifestages and considering the temporal component of these health outcomes and how such temporality impacts the confidence/uncertainty in water monitoring data. In addition, the Agency will discuss cancer epidemiology studies and the integration of such cancer epidemiology studies with animal toxicology studies and mode of action information (MOA). The February, 2010 SAP meeting focused on a number of broad scientific issues related to the use of epidemiology and human incident data in risk assessment. One of the case studies presented at the February meeting involved a collaborative project between the Agency and investigators at the Agricultural Health Study (AHS) (http://aghealth.nci.nih.gov). The goal of this collaborative effort is to compare exposure assessment approaches used for pesticide applicator exposure assessment and is part of a broad goal of OPP to better use epidemiology data in risk assessment. In this broad context, the collaborative project is considered an on-going methods development effort; the Agency will not solicit comments from the Panel on the status of this effort. However, a short summary of the on-going analyses are provided simply as an update to the Panel and public on that effort. Two of the major goals of the atrazine human health re-evaluation are to consider the degree to which the current drinking water monitoring program conducted by Syngenta 1 Henceforth mention of atrazine in this document also infers its chloro-s-triazine metabolites (DEA, DIA, and DACT) unless otherwise specified. Page 7 of 184 (and as required by EPA) is sufficient to protect human health and whether or not a new human health risk assessment needs to be developed. As part of the re-evaluation of the human health effects of atrazine, the Agency has evaluated the hazard potential of atrazine and the drinking water monitoring strategy together. Specifically, the frequency of drinking water monitoring is related to the temporal pattern of the toxicological endpoint of concern used for the risk assessment. Generally, longer durations of toxicological concern (e.g., a long-term chronic effect) require a less frequent drinking water sampling design to approximate longer term exposures. However, as the duration of concern shortens, the frequency and timing of sampling become more important in determining how well the sample data capture short-duration exposures. The current drinking water monitoring frequency involves weekly sampling during the growing season and bi-weekly (i.e., every other week) sampling for the remainder of the year. This frequency is considered adequate for the 90-day rolling average duration used in the previous risk assessment. Relying on atrazine’s established neuroendocrine mode of action (MOA), this draft issue paper will discuss experimental toxicology findings in the context of adverse health outcomes and lifestages/populations that may be affected by atrazine exposure. Attenuation of the LH surge continues to be the most sensitive effect (i.e., occurs at the lowest dose) identified to date in the atrazine database. Perturbations of LH signal a disruption of the hormonal environment in the individual and serve as a sentinel effect used to establish a point of departure (PoD) for the risk assessment that would be health protective for the other effects noted in the database. These other effects occur at higher doses than the LH surge attenuation and include delays in puberty onset, disruption of estrous cycles, and and reduced prolactin from milk early in life and prostatitis in young adult rats; they provide insight into the temporal relationship between atrazine exposure and adverse health outcomes. At the September 2010 SAP, the Agency proposed a range of exposure durations relevant for the toxic effects of 4-28 days. A single value was not proposed due to uncertainties in the duration required to elicit toxic effects, particularly in humans. The SAP commented in the December 2010 report that “the imprecision in the Agency’s proposed sampling frequency seems justified. This may be about as precise an estimate as can be obtained when starting with the experimental animal data and the exposure requirements for LH surge suppression…” At the July 2011 meeting, the Agency will consider internal dosimetry of atrazine and its chlorinated metabolites in the context of potential adverse health outcomes relevant to specific lifestages and subpopulations, will update the benchmark dose analysis and internal dosimetry calculations based on new and additional pharmacokinetic data available for atrazine, and review the current status of efforts to develop a physiologically-based pharmacokinetic (PBPK) model. The Food Quality Protection Act (FQPA) requires the Agency to give special attention to the potential risk to infants and children. Specifically, FQPA instructs EPA, in making its “reasonable certainty of no harm” finding, that in “the case of threshold effects, an additional tenfold margin of safety for the pesticide chemical residue and other sources of exposure shall be applied for infants and children to take into account potential pre- and Page 8 of 184 post-natal toxicity and completeness of data with respect to exposure and toxicity to infants and children.” At the September 2010 SAP, the Agency considered the experimental toxicology and epidemiology studies considering the potential for increased susceptibility of infants and children; this literature review will be updated as part of the July 2011 SAP. Although the new science supports a duration of exposure shorter than 90-days like used in the previous risk assessment, this shortening of the duration of exposure may not necessarily translate to increased water monitoring. At the July 2011, the Agency has evaluated the confidence/uncertainty in existing water monitoring data; this analysis has been informed using advice from the Panel at the April and September 2010 meetings along with preliminary conclusions about the completeness of the data on exposure and toxicity to infants and children. The potential cancer effects of atrazine have been considered by the SAP on multiple occasions beginning in 1988 with an evaluation of rat mammary gland tumor response (FIFRA SAP, 1988). Subsequent to this meeting, substantial research was conducted on atrazine's hormonal or neuroendocrine mode of action (MOA) and the Agency returned to the SAP in 2000 (FIFRA SAP, 2000a) for advice on atrazine’s MOA leading to mammary gland tumors, reproductive and developmental effects in rats, as well as the human relevance of these findings. The SAP agreed with the Agency’s proposal for atrazine’s neuroendocrine mode of action, and they further concluded that it is unlikely that the mechanism by which atrazine induces mammary tumors in female Sprague Dawley rats could be operational in humans. At the April, 2010 SAP, the Agency reviewed new experimental toxicology studies on cancer published since the last atrazine risk assessment in 2003. The Agency concluded, and the SAP concurred, that the new experimental toxicology studies on cancer did not alter or contradict the major key events in the neuroendocrine MOA leading to mammary gland tumors in the rat or the conclusion that MOA leading to mammary gland tumors in the rat is not relevant to humans (USEPA, 2010b; FIFRA SAP, 2010). In 2003, the Agency presented its evaluation on prostate cancer and exposure to atrazine to the FIFRA SAP. At that meeting, the SAP evaluated the available epidemiology studies suggesting an association between atrazine and prostate cancer in workers at an atrazine manufacturing plant in St. Gabriel Louisiana. The SAP concurred with EPA’s conclusion that the data did not suggest an association between atrazine exposure and prostate cancer, but that the increase in incidence of prostate cancer in workers at the plant could likely be explained by the increase in Prostate Serum Antigen (PSA) screening at the plant (FIFRA SAP, 2003a). At the 2003 meeting, the SAP recommended a more thorough and systematic review of the epidemiologic literature on the topic of the potential or observed carcinogenic effects of atrazine exposure. The Agency developed such a comprehensive review of the cancer epidemiologic literature; this analysis will be reviewed as part of the July 2011 SAP. In addition, the Agency has developed an integrative analysis of cancer epidemiology studies and experimental toxicology studies to consider the cancer potential of atrazine. Page 9 of 184 As summarized above, this draft issue discussed a wide array of scientific issues related to atrazine. The document is organized by the following: • • • • • • Section 2 of this paper briefly describes the linkage between atrazine-induced hormonal perturbations and adverse health outcomes in females of reproductive age, prepubescent children, and newborns. Section 3 discusses the Agency’s evaluation of the cancer epidemiology data (the non-cancer epidemiology data was reviewed during the September 2010 SAP meeting) and includes a brief summary of the Agency’s review of the pre-2003 data as well as a review of the epidemiology studies published since the 2003 IRED. Section 4 describes the Agency’s proposed updates to the dose response assessment incorporating recommendations received during the September SAP meeting. Section 5 discusses the scientific considerations in potential sensitivity of infants and children. Section 6 discusses the Agency’s proposed approaches to evaluating water sampling strategies and frequency of monitoring. Section 7 presents a case study integrating all the concepts discussed in the previous Sections and connects them to the water chemographs to get information on lifestage/temporal/outcome specific risks using an internal dose metric. This Section will also illustrate how all these factors intersect to inform the FQPA Safety Factor determination. 2. MODE OF ACTION (MOA) & EXPERIMENTAL TOXICOLOGY 2.1 Historical Background Over the past one and a half years, the Agency has conducted a comprehensive and extensive evaluation of atrazine. This review encompassing experimental toxicity data, pharmacokinetic data, epidemiological data, and water monitoring data was undertaken to determine if new data collected since the 2003 human health risk assessment needed to be updated. The 2003 human health risk assessment for atrazine is complex and incorporated significant data development, regulatory evaluation, and FIFRA SAP review. Prior to the 2010 re-evaluation, the Agency sought advice from the SAP on human health issues during three previous meetings (FIFRA SAP, 1988, 2000, 2003). Atrazine was first taken to the SAP for evaluation of rat mammary gland tumor response seen in adult cancer studies in 1988 (FIFRA SAP, 1988). At that time, the SAP noted that a “hormonal influence” might be an important consideration in the development of these mammary gland tumors in adult rats. The Agency returned to the SAP in 2000 (FIFRA SAP, 2000) for advice on atrazine’s MOA leading to mammary gland tumors, reproductive and developmental effects in rats, as well as the human relevance of these findings. Page 10 of 184 The SAP agreed with the Agency’s proposal for atrazine’s neuroendocrine MOA, and they further concluded that it is unlikely that the mechanism by which atrazine induces mammary tumors in adult female Sprague Dawley rats could be operational in humans. Based on the 2000 SAP guidance, EPA changed its position on atrazine and reclassified it from a “possible carcinogen” to “not likely to be carcinogenic in humans”. Nevertheless, the SAP further concluded that it is not unreasonable to expect that atrazine might cause adverse effects on hypothalamic-pituitary-gonadal (HPG) function in humans if exposures were sufficiently high and that the MOA was relevant for developmental and reproductive effects (FIFRA SAP, 2000). It is these developmental and reproductive effects of atrazine which have been the focus of significant research in the last few years. The Agency concluded, and the 2010 SAP concurred, that new research conducted in the last seven years supports the neuroendocrine MOA identified in the 2003 IRED and the conclusions regarding the human relevance of the mode of action for rat mammary gland tumors. As described in this draft issue paper, the critical and most sensitive adverse effects continue to be related to the reproductive/developmental consequences of perturbations to the HPG axis. In 2003, the Agency presented its evaluation on prostate cancer to the FIFRA SAP. At that meeting, the SAP evaluated the available epidemiology studies suggesting an association between atrazine and prostate cancer in workers at an atrazine manufacturing plant in St. Gabriel, Louisiana. The SAP concurred with EPA’s conclusion that the data did not suggest an association between atrazine exposure and prostate cancer, but that the increase in incidence of prostate cancer in workers at the plant could likely be explained by the increase in Prostate Specific Antigen (PSA) screening at the plant (FIFRA SAP, 2003). The SAP also stated that a more thorough and systematic review of the biologic and epidemiologic literature on the topic of the potential or observed carcinogenic effects of atrazine exposure was required before concluding that atrazine exposure is an unlikely explanation for at least part of the excess of prostate cancer in the St. Gabriel manufacturing plant. Review of the cancer epidemiologic studies with atrazine is now complete and the Agency is seeking the SAP’s comments on its evaluation and conclusions as part of this SAP meeting. 2.2 2010 Scientific Re-Evaluation In preparation for the 2010 SAP meetings, the Agency reviewed more than 100 new experimental studies on a variety of topics including reproductive/developmental toxicity in males and females, neurotoxicity, immunotoxicity, mammary gland development, mixtures effects, andrology, aromatase and insulin resistance and mitochondrial function impairment. Additional studies have been reviewed since the September 2010 SAP. The Agency’s reviews on all these topics are provided in Appendix A. In addition summaries of the new literature are provided on each of the scientific areas in Appendix A. Page 11 of 184 In the 2003 IRED, the Agency identified perturbations of the neuroendocrine system (particularly LH regulation) leading to reproductive toxicity as the most biologically plausible and sensitive effects attributable to atrazine exposure. These adverse outcomes include disruption of estrous cyclicity and delays in puberty onset (males and females) occurring as a consequence of disruptions to the HPG axis in rats. An additional effect – not directly linked to LH disruption – is the finding of decreased suckling-induced prolactin release in milk early in life (perinatally) leads to increased incidence of prostatitis in young adult ratsis an atrazine-induced increase in prostatitis in animals exposed perinatally to atrazine. During the September 2010 SAP meeting, the Agency addressed the hypothesis that atrazine exposure during critical developmental stages may result in delays in mammary gland development (MGD). The SAP evaluated and agreed with the Agency’s review of the publicly available data that the significant limitations of these studies preclude definitive conclusions. The Panel, however, recommended that additional research be conducted including a comparison of the subjective scoring methodology used to assess MGD with an objective approach such as morphometric analysis. Agency scientists in the Office of Research and Development (ORD) have since conducted these studies and the results of this research are discussed in Appendix A. 2.3 Mode of Action (MOA) and Adverse Health Outcomes 2.3.1 Background The Agency defines a MOA as “a sequence of key events and processes, starting with interaction of an agent with a cell, proceeding through operational and anatomical changes, and resulting in the adverse effect. The current evaluation of the post-2003 data supports the neuroendocrine MOA and key events originally identified in the 2003 IRED. In addition, new research has become available that extends our understanding of the neuroendocrine events that occur following atrazine exposure and that are germane to our understanding of the processes responsible for the adverse outcomes identified in different rodent models. Thus, this issue paper will briefly discuss atrazine’s established neuroendocrine MOA and then proceed to discuss how this MOA informs our understanding of the reproductive and developmental effects observed after atrazine exposure. For most pesticides, there is little information on the mode of action and even fewer with epidemiology studies that can be used in the risk assessment process. As such, the Agency makes assumptions about the relevance of animal findings to humans and quantitative animal to human extrapolation. In the case of atrazine, the wealth of data across many scientific disciplines allows for a highly refined assessment for atrazine using MOA understanding, human relevance of animal studies informed qualitatively by epidemiology studies, and refined analysis of critical durations of exposure. These refinements will continue to improve as the PBPK model currently being developed is refined. Page 12 of 184 The 2003 IRED concluded that the most probable explanation for many of the cancer and non-cancer effects observed in laboratory species following atrazine exposure could be a disruption of LH release. In addition to the premature reproductive senescence and development of mammary gland tumors in the Sprague Dawley rat, the MOA also provides an explanation for the broader range of adverse outcomes attributed to atrazine exposure including altered ovarian function in the adult female, delays in the onset of puberty in both the male and female, impaired pregnancy maintenance (i.e., full litter resorptions) and premature reproductive senescence. Components of this MOA also provide an explanation for the decrease in serum testosterone observed at high experimental doses in studies of the male rat. In the sections below, the MOA for atrazine for the development of mammary tumors in the female rats is described followed by an evaluation of the role of LH level disruption in eliciting other adverse outcomes. In the course of the current evaluation, the Agency identified an effect of atrazine on reproductive function in both male and female rats. These effects provide insight into evaluating the vulnerability of specific lifestages such as sexual development, puberty, and the perturbation of adult reproductive performance (including premature aging). These effects can be linked to the atrazine-induced changes in LH secretion. Consequently, the Agency will continue to use changes in LH secretion as the basis of the atrazine risk assessment. As such, any of the identified adverse outcomes would be protected since they occur at doses higher than those eliciting changes in LH (Figure 1). Page 13 of 184 Figure 1: LH Suppression and Adverse Outcomes Observed in Rats Atrazine-induced changes in the hormonal milieu lead to a cascade of effects on reproductive function in male and female rats. The decrease in LH is a precursor event to reproductive effects both on a quantitative (i.e., occurs at lower doses) and temporal basis (occurs after 4 days of exposure). An atrazine related suppression of suckling-induced prolactin release in the lactating dams, is another hormonal change leading to an adverse effect (prostatitis) in the rat animal model. 2.3.2 Reproductive Senescence and Mammary tumors in rats: Well established MOA The key events initially postulated leading to the development of mammary gland tumors were described in the 2003 IRED. Briefly, this pathway begins with a perturbation of the brain’s ability to regulate LH secretion, a disruption of ovarian function and eventually the development of an endocrine environment that is favorable to the initiation of mammary gland tumors. This MOA is described below: Page 14 of 184 • Atrazine alters the secretion of LH through gonadotropin releasing hormone (GnRH) modulation: It is clear that atrazine impairs the secretion of LH from the pituitary. In the female rat, a number of reports describe the dose and time response for the suppression of the ovulatory surge of LH that typically occurs on the afternoon of vaginal proestrus (Cooper et al., 2007). It remains to be determined whether or not the changes in GnRH release are the result of a direct effect of atrazine (or its metabolites) on the GnRH neurons themselves, an indirect action of the chemical in other neurons that regulate GnRH activity, or both. Nevertheless, changes in GnRH neuronal activity represent a change in the key signal emanating from the central nervous system (CNS) and regulating gonadotropin (LH & FSH [follicular stimulating hormone]) secretion. This effect on the LH surge in the female rat is considered to be the primary reason that dietary atrazine leads to a premature reproductive senescence in this species. • A loss of the LH surge is responsible for age-dependent disruption of ovarian cycles in the rat: The disruption of the positive feedback processes regulating the LH surge in the female rat is considered to be the primary cause of reproductive senescence in this species. In the young adult female, the LH surge is a key event in the ovarian cycle leading to ovulation. In the aging female (beginning approximately 1 year of age), the regular 4-5 day estrous cycles typical of the youngadult female gives way to a pattern of persistent or constant estrus. The transition from regular cycles to persistent estrus is the result of a loss of the ability of the female to sustain appropriate LH secretion. The preponderance of evidence indicates that the age-related impairment of LH secretion reflects the loss of the positive feedback mechanisms controlling the generation of the surge. This includes altered neurotransmitter and neuropeptide control of gonadotropin releasing hormone (GnRH) the neuropeptide that represents the final common signal from the brain regulating LH release. Thus, aging in the female rat is the result of an agerelated change in the hypothalamic (GnRH/LH) control of the reproductive cycle. • Atrazine accelerates the rate of reproductive aging in rats: Similar to aging, the atrazine-induced changes in GnRH pulses lead to an attenuation of the proestrous, ovulatory LH surge (lower amplitude and reduced area under the curve). Therefore, it is not surprising that chronic exposure to dietary atrazine in young-adult females results in premature reproductive senescence. Similar to normal aging, these prematurely senescent females do not ovulate but develop persistent ovarian follicles which continue to secrete estradiol (i.e., polyfollicular ovaries). Since there is no ovulation, there are no corpora lutea in the polyfollicular ovary and thus progesterone secretion from the ovary is nil. At the same time there is increased prolactin secretion from the pituitary. It is this increase serum estradiol and prolactin and low serum progesterone concentrations that creates the endocrine milieu conducive to the development of mammary gland tumors. Page 15 of 184 • Reproductive aging (menopause) in humans initiated differently than in rats. Unlike rats, reproductive senescence in humans (menopause) is caused by the depletion of follicles and a concomitant decrease in estrogen instead of changes in the LH surge (which remains normal during menopause). As described above, reproductive senescence in rodents is driven by changes in the regulation of LH at the level of the brain. Thus, the altered hormonal environment observed in the postreproductive female rat is not present in the human. In the absence of elevated estrogen levels in humans, prolactin levels remain constant or even decrease slightly. Consequently, there is no sustained prolactin stimulation of the proliferative processes in the mammary gland and thus no tumor development. Although there are similarities in hypothalamic/pituitary function, there is a fundamental difference between rats and humans, and thus, the key events in the MOA leading to mammary gland tumors in rats are not relevant for breast tumorigenesis in humans. 2.3.3 Adverse Health Outcomes – LH Changes as a Sentinel Effect Although tumor development following chronic atrazine exposure is an adverse outcome mediated by a disruption of the neuroendocrine axis controlling reproduction in the female rat, this MOA for mammary tumor development is considered highly unlikely to occur in humans due to the different pattern of aging between the two species. However, it is plausible that the same initial atrazine-induced endocrine perturbations that induce tumors in rats (i.e., altered GnRH) could occur in humans. In fact, the Agency considered the atrazine-induced disruption of the LH surge, in rats, as the key event of the cascade of changes leading to the adverse reproductive outcomes following atrazine exposure. Relevant to this MOA, a number of recent studies have further characterized the cellular and neuroendocrine changes responsible for how atrazine interferes with the regulation of LH secretion. The preponderance of evidence provides added support for the hypothesis that atrazine modifies the hypothalamic (GnRH) control of pituitary function. The ovulatory surge of LH depends on the activity of GnRH neurons (Kalra and Kalra, 1983) with the afternoon rise in LH the result of modifications in GnRH pulsatile secretion into the hypothalamic portal system (Fox and Smith, 1985; Bergendahl et al., 1996; Veldhuis et al., 2008). Research from two laboratories has shown that atrazine interferes with the pattern and amplitude of GnRH release by slowing the frequency of the pulses and lowering the overall amount of GnRH released into the portal system (Cooper et al ., 2007; Foradori et al ., 2009). It is important to note that the modulation of GnRH/LH during the peripubertal period is not limited to rodents but is seen across several species including primates, in which Terasawa and coworkers have demonstrated that all aspects of GnRH pulsatility (pulse amplitude, frequency, and mean GnRH release) are markedly increased during the peripubertal period (Terasawa et al., 1984). Testing the hypothesis that atrazine-induced changes in the regulation of LH ultimately alter gonadal function in rodents, several studies reported adverse effects on reproductive Page 16 of 184 development and adult function including delayed puberty in both sexes (Stoker et al., 2000; Laws et al ., 2000), disruption of regular ovarian cycles in the adult female (Cooper et al., 1996, 2000), inability to maintain pregnancy (Narotsky et al., 2001) and reduced testicular hormone secretion in the male (Stoker et al., 2000; Trentacoste et al., 2001; Rosenberg et al., 2007) after atrazine exposure. Of these potential adverse outcomes, the two that appear to be the most sensitive (i.e. occurring at the lowest dose levels) and/or occurring after the shorter duration of exposure are the disruption of the ovarian cycles and the delays in puberty onset. 2.3.3.1 Factors influencing GnRH Modulation of LH following atrazine exposure As indicated, the accepted mode of action (MOA) for the atrazine-induced impairment of LH secretion is a disruption in the central (CNS) mechanisms controlling LH release [(i.e., impaired gonadotropin releasing hormone (GnRH) release] (Cooper et al., 2007; Foradori et al., 2009). This MOA was reviewed and endorsed by the 2003 atrazine Scientific Advisory Panel (SAP). Since then, additional research from several laboratories (McMullen et al., 2004; Foradori et al., 2009) has provided further evidence supporting this MOA. Still, little is known about the exact molecular or neuronal targets responsible for an atrazineinduced change in GnRH/LH release. It is assumed that the changes in GnRH occur as a result of either (i) a direct effect of the atrazine and/or its metabolites on the GnRH neurons themselves, (ii) an effect on the neurons known to regulate GnRH (and thus affect GnRH cellular activity indirectly) or (iii) an effect on both the GnRH neurons and those neurons regulating GnRH activity. These three alternatives would imply that the CNS is the primary target for atrazine. Preliminary data suggest that atrazine exposure may lead to: (i) a decrease in c-fos in GnRH neurons (Foradori et al, 2009) or (ii) a disruption of catecholaminergic activity after in vitro exposure of undifferentiated PC12 cells [dopamine and norepinephrine have both been implicated in the generation of the LH surge] (Das et al., 2000). However, detailed studies examining any of the putative CNS mechanism controlling GnRH pulsatility are not available. 2.3.3.1.2 GnRH in primates The integral role of GnRH secretion in the function of the HPG axis is widely accepted in humans and non-human primates. Hypothalamic amenorrhea (abnormal absence of menstruation) is a prime example that impairment of pulsatile GnRH secretion can have an adverse functional impact on menstrual cycles. With our increasing understanding of the brain’s role in reproduction, the neuroendocrine similarities and disparities among mammalian species have become more apparent. Although there are anatomical differences in the distribution of the GnRH neurons, however the functional role of these neurons are similar in rodents and primates. In terms of intercellular connectivity, it appears that the regulatory input to human and nonhuman primate GnRH neurons is at Page 17 of 184 least comparable to those in rodents. With a few noted exceptions, the response of GnRH neuronal activity to the positive and negative feedback to endogenous and/or exogenous estrogen and progesterone is the same. It is noteworthy, that impairments in the activity of one or more of the factors regulating GnRH release (e.g., kisspeptin, galanin, and corticotrophin releasing hormone [CRH]) have been implicated with functional hypothalamic amenorrhea in women (Meczekalski et al., 2008). These observations support the use of the rat as a surrogate species for the evaluation of environmental toxicants on the regulation of the LH surge. 2.3.3.2 Females of Reproductive Age: Role of LH, Progesterone and Corticosterone in Atrazine Induced Changes in Ovarian Cyclicity In the previous sections, the evidence indicating that atrazine has a clear effect on the hypothalamic control of LH through regulation of the GnRH pulse generator was discussed. In this section these observations are discussed within the context of atrazine-induced changes in progesterone and corticosterone that adversely impact the steroid hormone feedback regulation of LH secretion. The significance of these atrazine-mediated changes stems from the observation that they occur both in vitro and in vivo through possibly different events in the toxicity pathway. In vitro work by Sanderson et al. (2000) and Hecker et al. using H295R cells demonstrated an atrazine-mediated increase in estradiol/estrone leading researchers to hypothesize that atrazine exerted its effect on steroidogenesis through alterations in aromatase activity. However, atrazine has also been shown to increase both progesterone (Tinfo et al., 2010) and testosterone (Pogrmic-Majkic et al., 2010) production in rat granulosa cells and primary rat Leydig cell cultures, respectively. In vivo, an immediate increase in the levels of circulating adrenocorticotropic hormone (ACTH) occurs in response to atrazine exposure. This in turn stimulates the production and release of progesterone and corticosterone from the adrenal cortex. Considered together, these data not only suggest that the effects of atrazine may involve several levels of biological organization beyond the HPG axis but that brief atrazine exposures (as brief as 4 days) elicit changes in the adrenal hormonal milieu that may have the potential to modify LH release. Whether these brief changes may result in adverse health outcomes, however, is not clear at this time. Nonetheless, it is important to note that the change in LH surge that serves as the basis for the atrazine risk assessment is considered a precursor event to the reproductive and developmental effects noted after atrazine exposure. A study recently submitted to the Agency was designed to evaluate the impact on fertility or reproductive performance following a brief (4-5 days) atrazine exposure at doses up to 100 mg/kg/day (Coder 2011). In the study, Sprague Dawley rats were exposed either by gavage (≤ 100 mg/kg/day) or diet (≤51 mg/kg/day). The animals were then mated on the day of proestrous. No effects on reproductive performance (mating, fertility, or conception Page 18 of 184 indices) were observed at any dose level regardless of strain or method of administration (dietary or gavage). Moreover, no effects on pre-implantation loss, implantation sites, fetal development and/or offspring survival were reported. Based on these observations, the author concluded that “the magnitude of the atrazine-induced preovulatory surge suppression shown in previous studies was not sufficient to influence ovulation or reproductive performance of SD or LE female rats as evaluated in this study.” However, the interpretation of these data is confounded. LH measurements were not conducted as part of this study; although it is reasonable to assume the LH surge would be attenuated at the dose levels administered. The absence of these data, however, does introduce a confounding factor to the study’s interpretation since it is known that mating rats on the day of proestrous causes an LH surge due to coital stimulation causing the rats to become reflex ovulators and resulting in normal pregnancies. Furthermore, it has been previously shown with other chemicals affecting the LH surge (e.g. pentobarbital) that steroid hormone regulation of mating behavior in rats is not affected by this treatment. In a study intended to replicate the work by Cooper et al. (2007) and to assess the effects of brief atrazine exposure (4-14 days) on the LH surge, ovariectomized Sprague Dawley rats received atrazine via gavage at 0, 6.5, 50, and 100 mg/kg/day for 4, 8, or 14 days (Coder, 2011). A statistically significant decrease in LH was observed after 4 days of atrazine exposure at doses ≥ 50 mg/kg/day. Though a statistically significant decrease in LH was also noted at the 6.5 mg/kg/day dose level, the authors discounted it because there was no effect in the AUC for this measure. However, when a BMD analysis is conducted on these data, the BMDL is 3.4 mg/kg/day. Thus, the Agency considered the decrease in LH at 6.5 mg/kg/day to be treatment-related. A longer atrazine exposure (8 or 14 days) also resulted in a statistically significant decrease in LH. Furthermore, after 14 days of exposure the number of animals exhibiting a lengthened estrous cycle (6-7 days vs 4 days control) increased (16-20%) at doses ≥ 50 mg/kg/day. 2.3.3.2.1 Estrogen and progesterone regulation of LH secretion Estradiol and progesterone have long been known to play coordinate roles in the generation of the LH surge (Mann & Barraclough, 1973). In a female rat exhibiting a normal 4 day estrous cycle, estradiol begins to rise on diestrus day 2, reaching a midday apex on the following day of proestrus. This increase serves to increase the responsiveness of the hypothalamic mechanisms that promote the secretion of GnRH into the portal vessels and the subsequent surge of LH on the afternoon of proestrus. One outcome of such an increase in estrogen is the increase in progesterone receptors within the positive feedback sites of the hypothalamus. Serum progesterone concentrations increase beginning early on the afternoon of vaginal proestrous within this sequence, a rise in estrogen followed by a rise in progesterone which act together to generate a surge (Brown-Grant and Naftolin, 1972). Importantly, it should be noted that in addition to ovarian derived progesterone, the adrenal cortex contributes a significant portion of the serum progesterone at this time (Feder and Ruf, 1968; Shaikh, 1974). As atrazine was found to Page 19 of 184 stimulate adrenal progesterone and corticosterone release, the possibility that these brief changes in the adrenal steroid milieu could be a part of the atrazine MOA for LH suppression was examined in a rat study conducted by EPA’s Office of Research and Development (ORD) . The background and rationale for evaluating the adrenal effects are presented below followed by a summary of the study. Estradiol alone can initiate a surge, as demonstrated in ovariectomized controls implanted with silastic capsules containing this steroid (see blue Figure 2). The time course for this induction is several days. Over this period, estradiol stimulates an increase in hypothalamic and pituitary progesterone receptors (Camp and Barraclough, 1985; Brown et al., 1987; Turgeon and Waring, 2000), which is essential for any subsequent effect of progesterone on the LH surge. In contrast to the effect of estradiol, progesterone alone is unable to induce a surge (e.g., Caligaris et al., 1971). However, when combined with estradiol, progesterone can either enhance or suppress LH secretion depending on the timing and sequence of exposure to these two hormones. The time course for an influence of progesterone under estrogen priming is different. In ovariectomized female rats, primed with estradiol for 72 h, progesterone exposure around 10 AM in the morning of the surge, will act synergistically with the estrogen to cause a large augmentation in LH (see red line Figure 2). This synergistic effect of progesterone is dependent on the estrogen-induced increase in progesterone receptors mentioned previously. This effect is well-known in rodents (e.g., Caligaris et al., 1968; Brown-Grant and Naftolin, 1972; Banks and Freeman, 1978; 1980) and non-human primates (e.g., Terasawa et al., 1984) and has been identified in women. In fact, Odell and Swerdloff (1968) found that a progestin injection in post-menopausal women previously treated with ethinyl estradiol could induce a surge-like elevation in circulating LH in samples taken 24 hours later. In contrast to the augmentation in the LH surge by a progesterone exposure that immediately precedes an estradiol-induced rise in LH, longer exposure is essentially inhibitory to the surge (e.g., Caligaris et al., 1971; Banks & Freeman, 1978; 1980; Prilusky et al., 1984). Such extended treatment will cause a down-regulation by progesterone of its own receptors which appears to be a factor underlying the suppressive effect on LH (Turgeon and Waring, 2000). This influence of progesterone is the basis for the action of many oral contraceptives. Furthermore, increased levels of progesterone during a woman’s pregnancy, secreted first from the ovaries and subsequently from the placenta, prevent the appearance of a surge and ovulation over that time. Finally, it is noteworthy that treatment with progestins also lowers serum gonadotropins and testosterone production in men (e.g., Amory, 2004). Page 20 of 184 OVX- Estradiol Implant / sc Progesterone 40 30 E + P 20 10 E only 0 -6 -4 -2 Peak +2 +4 +6 Hours from Peak Figure 2: Augmentation by a subcutaneous injection of progesterone on an LH surge in ovariectomized rats implanted with an estradiol capsule at the time of surgery. Progesterone was administered just prior to the sampling of small aliquots of blood by tail nick over the afternoon of the third day after ovariectomy. Concentrations are indicated in ng/ml ± SEM 2.3.3.2.2 Differential effects of single versus multiple atrazine exposures on the LH surge During the September 2010 SAP, the Panel raised the issue of single vs multiple atrazine exposure effects on the LH surge. In their report to the Agency the SAP commented: “Data are clear in identifying that a greater-than-one pulse of exposure to atrazine is necessary for attenuation of the LH surge. For example, single high doses (over 100 mg/kg) administered on the morning of proestrus did not alter characteristics of the LH surge occurring later the same day. Additional data clearly demonstrate a once daily dose for 4 days and beginning on estrus can induce significant inhibition of the LH surge peak. In this instance, a dose response is observed. However, what is not clear is if less than 4, but greater than 1 days’ exposure is sufficient to alter the LH surge. Further complicating the matter, it is not clear if a 4-day exposure, beginning on a different day of the cycle, will result in changes in the LH surge similar to those when dosing begins on the morning of proestrus. Understanding of the relationship between duration of exposure and phase of the cycle will be key in translating rodent data to humans for risk assessment purposes.” In response to the Panel’s comments, EPA scientists in the Office of Research and Development have undertaken a series of experiments to try to elucidate the nexus between phase of the cycle and duration of exposure. This research, is in the early stages. However, initial results suggest that, unlike previous assumptions, a single high dose of atrazine (100 mg/kg bw) can affect the LH surge. Notably, the single atrazine exposure leads to an increase in LH instead of the decrease that is seen after 4 days of exposure. Page 21 of 184 As noted above the biphasic effect of progesterone and the suppressive effect of corticosterone on the LH surge suggested that these two hormones may also play a role in the suppression of the LH surge following a 4-day exposure to atrazine. That is, the demonstrated atrazine-induced increase in these two adrenal hormones could modulate the secretion of LH by feeding back onto both the positive and negative control of the GnRH pulse generator. Indeed, if this were the case, it would be expected that a single dose of atrazine administered to the ovariectomized, estrogen-primed rat would first increase the LH surge because of the immediate increase in adrenal progesterone induced by the atrazine treatment. In contrast, repeated administration of atrazine (i.e., 4 days) to the same ovariectomized, estrogen primed female would result in a decrease in the surge because of the repeated increases in adrenal progesterone and corticosterone levels immediately following each exposure to this herbicide. Figure 3 demonstrates that this is what happens in the ovariectomized model. In response to a single exposure to atrazine the amplitude of the LH surge and the area under the curve were both significantly increased. It should be noted that these preliminary data are only available at high doses (100 mg/kg/day) far exceeding the dose levels currently being used for risk assessment purposes. Research to further elucidate the dose-response curve for this acute effect is ongoing. Page 22 of 184 1 Day Dosing Atrz (1300h) Ovariectomized / E2 prime - Atrazine 1 Day sampling 1400-1600-1800-2000h 40 10 * mg/kg * (11) 8 Serum LH (ng/ml) OVX / E2-implant 0 Serum LH (ng/ml) Area Under the Curve 9 100 7 (10) 6 * 5 4 3 30 181% 20 10 2 1 0 0 -6 -4 -2 Peak +2 0 +4 100 A t r azine (m g/kg) Time Relative to Peak Ovariectomized / E2 prime - Atrazine 4 Days 40 10 OVX / E2-implant mg/kg 9 35 (10) 100 7 (10) 6 p= 0.06 5 4 * 3 Serum LH (ng/ml) Area Under the Curve 0 8 Serum LH (ng/ml) 4 Days Dosing Atrz (1300h) sampling 1400-1600-1800-2000h 30 25 20 15 * 10 2 5 1 0 0 -4 -2 Peak +2 +4 54% 0 100 A t r azine (m g/kg) Time Relative to Peak Figure 3: The LH surge (ng/ml + SEM) and areas under the curve in response to 1 and 4 days of atrazine treatment in ovariectomized / estradiol-primed Long-Evans rats. The numbers in parentheses indicate the group sizes for each of the treatments. Numbers within the 100 mg/kg columns at the right indicate the percent change from the 0 mg/kg group. *p<0.05. Also, as predicted from our understanding of the continued exposure to progesterone and corticosterone, four daily doses of atrazine resulted in a significant decrease in LH Page 23 of 184 secretion. Again, this decrease was associated with an atrazine-induced increase in both progesterone and corticosterone. The details of this study are provided Goldman et al., 2011 (Appendix A). In summary, these data demonstrate two important factors pertaining to the temporal response to atrazine exposure. First, a single high dose of atrazine (100 mg/kg bw) affects the HPG axis by increasing the amount of LH secreted during the afternoon LH surge. The fact that this same single dose of atrazine increases adrenal hormone secretion (primarily progesterone) would explain the enhancement of the surge. Second, multiple (4 daily) doses of atrazine resulted in an attenuation of the LH surge. Again, as atrazine was found to induce an immediate and significant increase in progesterone and corticosterone secretion following each dose (Fraites et al., 2009; Goldman, 2011), it is important to consider the implications of a brief LH surge attenuation. 2.3.3.2.3 LH Attenuation in Humans – Pharmaceutical Experience Little is known about the direct effects of atrazine in humans and available epidemiology studies are not sufficiently robust for making strong conclusions. However, one can infer information about possible windows of susceptibility from what is known about human physiology and from the pharmaceutical literature. In considering the relevance and temporal concordance of a toxic effect in a test species to human health, a variety of factors need to be considered. Among them is the potential window of susceptibility to the toxic effect. For atrazine, the experimental data available indicate that, in rats, a 4-day exposure is sufficient to attenuate the LH surge – the most sensitive endpoint in the atrazine risk assessment. Since the estrous cycle in rats is 4 days long, this window of susceptibility is not unexpected. However, in humans, the menstrual cycle lasts – on average – 28 days. Thus, the question arises as to whether a brief exposure (e.g., a few days) in humans could lead to an attenuation of the LH surge. Evidence of chemically-induced decreases in GnRH or LH secretion is sparse in humans and non-human primates relative to rodents. The available evidence in humans comes primarily from the pharmaceutical arena. Nal-Glu, Cetrorelix®, and Ganirelix are three GnRH antagonists used to block the LH surge and ovulation in women prior to in vitro fertilization (IVF) procedures. In a series of experiments, regularly ovulating women received two 5 mg injections of Nal-Glu on days 8 and 11 of the follicular phase of the natural cycle (Frydman et al. 1992). This treatment resulted in a block of the spontaneous LH surge. This work was further corroborated by Olivennes et al. who demonstrated that a single 3 mg administration of the GnRH antagonist Cetrorelix® on day 8 of the follicular phase was sufficient to block the LH surge. Ganirelix exposure during the late follicular phase of the menstrual cycle has also been demonstrated to inhibit the LH surge and ovulation by competing with the endogenous GnRH for receptor binding (Fauser et al., 2002). One must consider these studies with caution with respect to how they relate to potential effects of atrazine since the potency and pharmacokinetics of these Page 24 of 184 pharmaceuticals relative to atrazine is unknown. However, these studies do help qualitatively inform a potential window of vulnerability to chemicals disrupting the HPG axis in women. Specifically, all of these pharmaceutical agents are administered during the late follicular phase of the menstrual cycle (days 8-12 of the follicular phase) 2. Thus, one can infer that the follicular phase (lasting ≈12 days) and possibly the late follicular phase (days 8-12 of the follicular phase) of the menstrual cycle may be a possible window of susceptibility in humans. Given the differences between the rat and human ovarian cycles, as well as the differences in potency and likely MOA of pharmaceutical agents, these data are only considered in the context of helping to identify if there is a time period during the menstrual cycle that is particularly vulnerable to chemicals interfering with the LH surge. It is not intended to draw any correlations between the mode of action, dose levels, or frequency of dosing needed to elicit the LH suppression after exposure to a pharmaceutical agent or atrazine. Although theoretically possible, a 4 day exposure to atrazine leading to LH surge attenuation in humans and an adverse health outcome is unlikely, given the available dose response, pharmacokinetics, and exposure data. As discussed in Section 4 of this paper, the area under the plasma concentration-time curve (AUC) for atrazine and its chlorinated metabolites is the critical dose metric. Based on the available drinking water monitoring data, it is highly unlikely that the plasma AUC associated with the LH surge attenuation would be achieved in such a short period of time (i.e., a high concentration of atrazine and/or its metabolites would have to be present in the drinking water). Although a 4 day atrazine exposure resulting in LH suppression is a conservative assumption, the Agency is proposing to bound the critical window of exposure from 4 to 28 days as potentially relevant to potential human health outcomes. As the September SAP noted in their report, “this may be about as precise an estimate as can be obtained when starting with the experimental animal data and the exposure requirements for LH surge suppression as opposed to using outcomes that are more unequivocally adverse.” 2.3.3.3 Prepubertal Individuals: Atrazine-induced delays in sexual maturation In addition to the disruption in ovarian cyclicity, atrazine exposure has also been implicated in delays in sexual maturation in both males and females following both perinatal and peripubertal exposure. Pubertal development is directly related to the progressive increases in the neurosecretory activity of GnRH neurons. In female rats, sheep, monkeys, and humans (Grumbach, 2002), detailed analyses of peripubertal LH secretory patterns have been conducted to provide surrogate measures of GnRH release throughout pubertal maturation. These studies have revealed that the initial stages of pubertal maturation are mediated by an acceleration of GnRH pulse generator activity (GnRH pulse frequency), an increase in the amplitude of GnRH pulses, or both of these alterations in GnRH 2 In humans, the follicular phase lasts approximately 12 days, assuming a 28-day menstrual cycle Page 25 of 184 neurosecretion. Terasawa and colleagues have directly measured GnRH release patterns in female rhesus macaques and determined that all features of GnRH pulsatility (pulse amplitude, pulse frequency, and mean GnRH release) are significantly increased throughout pubertal maturation in this species. The work of Sisk et al. (2001) in the rat are consistent with the hypothesis that maturation of the female rodent’s reproductive axis is dependent upon a pubertal increase in GnRH pulse generator activity and a progressive increase in the ability of the hypothalamus to generate surge-like releases of GnRH. The cellular and molecular mechanisms mediating the pubertal acceleration of GnRH pulsatility and the stimulation of putative GnRH mini-surges remain to be resolved. 2.3.3.3.1 Delays in Female Sexual Maturation A number of studies have been conducted to evaluate the impact of atrazine and/or its metabolites on pubertal development and estrous cyclicity in female rats (Laws et al., 2000, 2003; Ashby et al., 2002; Davis et al., 2011; Rayner et al., 2004). Collectively, these studies have shown that atrazine delays the onset of puberty, as measured by a delay in the age of vaginal opening (VO) and first estrus (Safranski et al., 1993). These findings are consistent with the postulated neuroendocrine MOA for atrazine. Female sexual maturation is the culmination of a complex cascade of cellular events at the hypothalamuspituitary-gonadal (HPG) levels that ultimately lead to the attainment of reproductive capacity. Activation of the HPG axis, resulting in the pulsatile secretion of GnRH that triggers a precisely regulated hormonal cascade of gonadotropins (LH and FSH) and ovarian steroids, is critical to puberty onset. Disruption of GnRH and LH release can lead to delays in pubertal development. Peripubertal exposure to atrazine and/or DACT for 19-23 days delays pubertal development in female rats at doses ≥ 34 mg/kg/day (Laws et al., 2000, Ashby et al. 2002, Laws et al. 2003). In some instances animals failed to attain VO by the time of their scheduled necropsy on PND 41. Interestingly, these delays in VO were not accompanied by lower body weights. Thus, the delays do not appear to be related to generalized growth delays. Gestational exposure to high doses of atrazine (100 mg/kg/day) during late gestation (GD 14-21) have been shown to delay sexual maturation of female offspring, however, exposures to lower doses (≤ 20 mg/kg/day) does not affect the age of pubertal onset. A recent study by Davis et al. (2011) evaluated the effects of prenatal exposure to atrazine on pubertal and postnatal reproductive indices in the female (Sprague Dawley) rat. Exposures from GD 14-21 at doses ranging from 1-20 mg/kg/day did not elicit a delay in VO or the timing of the first estrus. However, at 100 mg/kg/day atrazine exposure led to a significant decrease in pup weight (seen at birth but resolved by PND21) and most importantly a delay in VO. These results are consistent with the observations by Rayner and coworkers (2004) that atrazine exposure via the milk of dams exposed to 100 mg/kg/day during GD15-19 led to a delay in VO without affecting estrous cyclicity once sexual maturation was reached. Page 26 of 184 While delays in female puberty onset – as determined by the time of VO – occur at doses ≈ 20 times higher than the doses resulting in disruption of the LH surge, it is important to note that the duration of exposure sufficient to cause delays in VO ranges between 5 (prenatal exposure) and 23 days (peripubertal exposure). Thus, using the point of departure for the LH surge attenuation as the basis for the risk assessment is protective of this effect. These data also suggest that the critical duration of exposure that could lead to this effect – although at high doses – can be as brief as 5 days. 2.3.3.3.2 Delays in Male Sexual Maturation Over the last decade a number of studies demonstrated that atrazine delays male puberty in different strains of rats following both peripubertal and perinatal exposure (Stoker et al., 2000; Friedmann, 2002 ;Trentacoste et al., 2001; Rayner et al., 2006 and Rosenberg et al., 2008; Pogrimic et al., 2009). These studies support the hypothesis that impaired reproductive development is the result of an apparent delay in the maturation of the GnRH pulse generating mechanism and lower LH concentrations leading to insufficient stimulation of the gonads during the period that puberty would normally occur. The low testosterone concentrations result in delayed maturation of the androgen dependent sex accessory tissues. A reduction in testosterone levels following atrazine exposure has been reported in a number of studies in mammals, as well as other species revealing a consistency in the effects of atrazine on androgens. It is well known that the development of the size of the penis and cornification of the epithelium of the prepuce and preputial separation in immature rats are regulated by androgens (Marshall, 1966). A decrease in testosterone secretion during the juvenile period can delay preputial separation (Lyons et al., 1942) and reduce the size of the androgen-dependent tissues, such as the ventral prostate and seminal vesicles. Normally, testosterone levels rise gradually from PND 20-40, and abruptly double by PND 50 (Matsumoto et al., 1986; Monosson et al., 1999). In the male rat, atrazine exposure resulted in delays in the onset of puberty, as determined by assessment of preputial separation (PPS) . Stoker et al. (2001) found that peri-pubertal exposure to atrazine delayed puberty for 1.5 to 2.5 days, and decreased the size of seminal vesicles and prostate. At PND 45, intratesticular testosterone was also decreased. In the studies examining the Sprague Dawley (SD) rat (Trentacoste et al., 2001), decreases in ventral prostate and seminal vesicle weights were observed along with 3-4 day delays in puberty onset. In addition, Friedmann (2002) found a decrease in intratesticular and serum testosterone following exposure to the peripubertal male SD rat. Thus, all three studies found decreases in either serum or intratesticular testosterone and Trentacoste et al. also found a decrease in LH. It should be noted that all these effects were observed at doses that were at least one order of magnitude higher than the NOAEL for LH surge attenuation currently used for risk assessment purposes. Page 27 of 184 Ventral prostate, seminal vesicles, lateral prostates and epididymal weights are significantly reduced at doses just above the doses which delay preputial separation (Stoker et al., 2001). These effects provide further evidence of a decrease in androgen stimulation. Decreases in androgen hormone concentrations also correspond with the effects observed on pubertal endpoints. For example, Stoker et al., 2001 demonstrated that intra-testicular testosterone was significantly decreased following atrazine exposure peripubertally. This has corresponded to the work of others showing that serum testosterone is decreased in SD rats when dosed during a similar period of PND 22 to 47 (Trentacoste et al., 2001; Friedmann, 2002). The study by Stoker et al., 2002 also examined the effect of the atrazine metabolites, DACT, DIA and DEA on pubertal development in the rat. In this study, equimolar doses of the metabolite were compared to the effect seen with atrazine in the earlier study (Stoker et al., 2000). PPS was significantly delayed by DEA, DIA, and DACT. When the males were sacrificed on PND 53 treatments with DEA, DIA or DACT at atrazine equivalent dose (AED) ≥50 mg/kg/day, caused a significant reduction in prostate weight. Seminal vesicle epididymal weights and serum testosterone levels were also reduced This study confirmed that the chlor-s-triazine metabolites of atrazine also result in similar effects on pubertal development. It should be mentioned that other factors, in addition to those changes within the HPG, can affect the timing of puberty in the male rat including nutritional status (leptin/body weight) and environmental (pineal gland/secretion of indole amines) influences (Ebling and Foster, 1989). 3 Nutritional status and body weight are known to have effects on puberty, which are suggested to reflect the metabolic signals in the brain that serve as indices of the metabolic state. These metabolic alterations associated with weight loss or decrease in growth rate are inhibitory to the reproductive system and may be related to substances in the body that can alter the release of GnRH such as insulin, amino acids necessary for precursors of neurotransmitter synthesis and essential fatty acids (Ojeda and Urbanski, 1994). The peripubertal exposure male study (Stoker et al., 2000) using Wistar rats showed decreased body weight at the 200 mg/kg dose, but a pair-fed group was included to discern the effects of changes in growth. There were changes in puberty at doses in which no body weight decreases were observed. However, in a study with Sprague Dawley rats Trentacoste et al., found that the food restriction group (receiving approximately 10% less food which was equivalent to the 100 mg/kg atrazine treatment food intake) showed the same effects on LH, testosterone and reproductive tract development. A recent paper examined the effects of food restriction on the endpoints in the pubertal male exposure 3 The direction of change in daylength provides the seasonal time cue for the timing of puberty in many mammalian species. The pattern of melatonin secretion from the pineal gland transduces the environmental light/dark cycle into a signal influencing the neuroendocrine control of sexual maturation. Page 28 of 184 period and found that reductions in body weight up to 10% did not influence pubertal onset or the growth of androgen dependent tissues in the rat (Laws et al., 2007). Perinatally a single 100 mg/kg atrazine dose to the dam resulted in an effect on pup growth that may involve changes in milk quality, as the offspring showed a decreased weight of 7% on PND 4 (Rayner et al., 2006). Work by Rosenberg and co-workers (2008) showed that pre-natal (GD14-parturition) exposure to atrazine at doses ≥10 mg/kg/day led to a significant delay in preputial separation (1-3 days) while anogenital distance on PND21 albeit at significantly higher doses. Measuring testosterone levels in both the serum and intratesticular fluid of PND 60 rats revealed that testosterone levels were decreased at doses ≥ 50 mg/kg/day. It should be noted that all of the effects on PPS and AGD occurred in conjunction with nonsignifcant decreases in pup body weights (Rosenberg et al.,2008). These alterations of early postnatal body weight could play a role in the delay in puberty in these male offspring. In summary, the MOA of altered pubertal progression in the rat and decreased androgen production by the testes should be assumed to be relevant to humans, as GnRH and the regulation of pituitary hormones stimulate the gonadal secretion of steroids and is critical for the initiation of the set of events and time course of the HPG regulation of puberty in the human. All the effects on sexual maturation and altered androgen status reported after ≈ 30 days of exposure occur at dose levels ≥ 5X higher than those leading to the LH surge disruption which serves as the basis of the Agency’s current risk assessment. 2.3.3.4 Atrazine-associated decrease in prolactin and prostatitis in young rats In rodents, non-bacterial prostate inflammation is typically noted in older males (e.g. greater than one year of age) and can be induced with elevated prolactin concentrations (hyperprolactinemia) (Tangbanluekal and Robinette. 1994). In 1999, Stoker et al. reported an increase in prostatitis in the male offspring of mothers exposed orally to atrazine from PND 1 to 4. This effect was not linked to an alteration to LH, but rather the atrazine related suppression of suckling-induced prolactin release in the lactating dams. An increase in the incidence of prostatitis was observed in the 120 day old male offspring of dams treated with atrazine (≥ 12.5 mg/kg/day) from postnatal day 1-4. An increase in the incidence of prostatitis was also reported by Rayner et al. (2007) dams were exposed to 100 mg/kg/day atrazine during GD 15-19. The dose level eliciting the increase in the incidence in prostatitis in the offspring is > 10-fold higher than the dose leading to the LH surge attenuation used as the basis for the Agency’s risk assessment. In order to understand the significance of this observation, it is necessary to understand the development of the tuberoinfundibular dopaminergic (TIDA) neurons located within the hypothalamus and their role in regulating prolactin secretion in the adult. Prolactin plays a crucial role in the neonatal brain for normal TIDA neuron development. In the adult Page 29 of 184 offspring, the impaired TIDA regulation is reflected by elevated prolactin levels (hyperprolactinemia) (Shyr et al., 1986, Stoker et al., 1999; 2000). It is this elevated level of circulating prolactin in the adult males that has been linked to an increased incidence of prostatitis. Thus, an increased incidence of prostatitis in the offspring of dams exposed to atrazine during the critical time for TIDA neurons activation (first postnatal week) may be attributed to elevated blood prolactin concentrations due to impaired TIDA neuronal maturation (Stoker et al., 1999). In summary, the data indicate that atrazine induces prostatitis at doses ≥ 12.5 mg/kg/day and that – in rats – early postnatal exposure is a critical window of susceptibility to this effect. Stanko et al. (2010) evaluated reproductive development in male rats exposed prenatally to either atrazine or an atrazine metabolite mixture (AMM). The authors reported an increase incidence of prostatitis at doses ≥ 0.9 mg/kg AMM and at 100 mg/kg atrazine. This study is the companion study to the Enoch et al. (2007) study. While Enoch et al. assessed the female offspring of dams exposed during GD 15-19, Stanko and coworkers evaluated the male offspring. Therefore, the limitations identified by the Agency – and confirmed by the September 2010 SAP – in the Enoch et al. publication are also applicable to Stanko et al. The Agency concluded that the shortcomings in the study design, data reporting and analysis compromise the interpretation of the studies. A more detailed description of this study is available in Appendix A of this document. In rats, the dorsal and lateral prostates are the most homologous to the human prostate (Price, 1963) and lateral prostate is most sensitive to prolactin, so the rat is a good model to study the effects of disrupted prolactin on the developing prostate. It is unknown when the TIDA neurons develop in the human fetus or whether this development is dependent on the maternal prolactin concentrations (Ben-Jonathan et al., 2007 review). As such, it is unclear what – if any – is the impact of atrazine exposure on developing prostatitis in humans. 2.3.4 Mammary gland whole mounts Rodent mammary glands are widely used as experimental models for the human breast, and the pre- and postnatal developmental windows of the mammary glands lend several advantages for studying disruptions in normal development (Fenton 2006). The effects of prenatal atrazine exposure on mammary gland development in the rat have been investigated by several groups with conflicting results. Enoch et al. (2007) reported that Long Evans rats prenatally exposed to atrazine metabolite mixtures as low as 0.09 mg/kg/d during late gestation (GD 15-19) exhibited a persistent disruption in mammary gland development up until postnatal day (PND) 60. Similarly, Rayner et al. (2005) found that regardless of the length of gestational exposure (the longest being GD 13-19), 100 mg/kg atrazine stunted mammary gland growth up through PND 67. In contrast, Hovey et al. (2011) observed no differences in mammary gland morphology in Long Evans rats exposed to atrazine from GD 13-19. There were, however, differences in the methods of evaluation Page 30 of 184 employed in these studies. Enoch et al. (2007) and Rayner et al. (2005) using a subjective scoring method that relied on a holistic assessment of the mammary tissue, while Hovey et al. (2011) utilized a quantitative approach. A brief review of this study is available in Appendix A of this document. In response to a recommendation during the September 2010 SAP meeting, the Agency has conducted a set of experiments investigating the potential impact of in utero atrazine exposure on mammary gland development in Sprague Dawley rats using both the subjective scoring methodology described by Enoch et al. and a computer-based quantitative methodology (morphometric analysis) (Davis et al. 2011). Mammary gland development was assessed using whole mounts of the 4th abdominal mammary gland of PND 45 old pups. PND 45 pups were selected based on consensus guidelines established following a recent 2009 workshop (http://www.silentspring.org/pdf/our_research/MammaryGlandWholeMountRoundRobinSu mmary.pdf). No differences in the quantitative measures or in the subjective scores were found indicating that gestational atrazine exposure had no demonstrable effect on normal mammary gland development. 2.3.5 Summary The neuroendocrine MOA of atrazine leads to a perturbation of the hormonal milieu in laboratory animals. This perturbation – in turn – leads to a series of adverse outcomes at different lifestages as observed in rats. Quantitatively, the most sensitive effect is the ≈ 33% LH surge attenuation (BMDL = 2.56 mg/kg/day) observed after female rats of reproductive age are exposed to atrazine for 4 days. The Agency is using the 33% LH surge attenuation after a 4 day exposure as a precursor event to protect for other adverse outcomes including estrous cyclicity disruption, and delays in sexual maturation occurring at higher doses in laboratory animals. In the case of atrazine, it has been noted that in addition to dose, duration of exposure is an important parameter that must be considered in evaluating the relationship between dose and attenuation of the LH surge. Currently available data indicate that a 4-day exposure is sufficient to elicit a decrease of the LH surge in rats. Interestingly, this is the length of the estrous cycle in rats and also the exposure duration needed for atrazine to reach pseudosteady state – a topic that will be discussed in greater detail in Section 4 of this draft issue paper. Preliminary data, however, suggest that in rats even shorter atrazine exposures can result in LH changes, albeit at high doses (100 mg/kg/day). Other effects of concern, such as delays in puberty onset and decrease in suckling-induced prolactin release and eventually prostatitis in young rats, identified in the animal toxicity database occur at higher doses but have a different temporal profile compared to the LH surge attenuation. For instance, prostatitis can be seen in rats exposed to 12.5 mg/kg/day of atrazine for 3 days shortly after birth. Similarly, atrazine-induced delays in puberty onset have been reported in both peripubertal male and female rats after exposures to atrazine (≥12.5 mg/kg/day) for Page 31 of 184 approximately 20-30 days. Although drawing a direct temporal correlation between the effects seen in the rat animal model and potential human health outcomes is not feasible at this time, it is prudent to consider the possibility of a critical temporal window of < 90 days (the current rolling average drinking water concentration). This may have an impact on the frequency of drinking water monitoring that may be required. In Section 7 of this issue paper the Agency will present a case study integrating the animal toxicity, pharmacokinetic and water monitoring data to illustrate the relationships between critical windows of exposure and the sampling frequency needed to assess whether atrazine levels exceed the Agency’s level of concern for critical intervals. Page 32 of 184 3. EPIDEMIOLOGIC EVIDENCE OF THE CARCINOGENIC POTENTIAL OF ATRAZINE 3.1 Introduction 3.1.1 Background This section provides EPA’s review of recently published (September 2003 – May 2011) epidemiologic evidence of the carcinogenic potential of the herbicide atrazine. This review updates EPA’s analysis of this observational database presented in two previously convened FIFRA science advisory panels (SAP): June 2000, and July 2003. Briefly for reference, the June 2000 FIFRA SAP considered studies identified by EPA that evaluated atrazine or triazine exposure in association with breast, ovarian, and prostate cancers as well as non-Hodgkin lymphoma (NHL) and other lymphohematopoietic cancer sub-types. The Final SAP report which provided recommendations to EPA concerning interpretation of these data is presented in Appendix A. The SAP recommended EPA consider the evidence in support of an association between atrazine and prostate and breast cancer to be inconclusive. With regards to ovarian cancers, the SAP commented that while one study was suggestive of an association, it required replication and improved study methods to better inform causal inference. In addition, the Panel urged EPA to re-evaluate the studies related to NHL, as the Panel believed there was some supportive evidence observed. Considering the SAP recommendation, EPA concluded that while the data were not of sufficient quality or quantity to conclude there was an association, EPA should continue to monitor the literature, and re-convene future SAPs when additional epidemiologic data were identified ((EPA), 2000). In 2003, the EPA presented its analysis of several epidemiology studies which evaluated the potential association between atrazine and prostate cancer. Mainly, this review consisted of studies which indicated a possible increased risk of prostate cancer among an occupational cohort of triazine manufacturing plant workers (i.e., the St. Gabriel manufacturing plant cohort). The Agency also considered a preliminary nested case control study from the Agricultural Health Study (AHS), and a correlation analysis of this association in this evaluation. The Final SAP report which provided recommendations to EPA concerning use of these data is presented in Appendix B. EPA concluded, in consultation with the FIFRA SAP, that the increased incidence of prostate cancer in this cohort was likely in part or in whole due to intensive prostate cancer screening program, the prostate specific antigen (PSA) test was available on site to employees and many employees participated in the program (EPA), 2003). Therefore, detection bias was a likely explanation for the apparent risk concerns presented in the occupational studies. Concerning the AHS study on prostate cancer, the Panel noted that although results of this initial, hypothesis-generating analysis did not provide evidence of a positive association between atrazine and prostate cancer, differences between agricultural and manufacturing exposure profiles, as well other demographic differences between the two populations, hinder a direct comparison of results. The Panel in its Final report stated the data were Page 33 of 184 insufficient to support a conclusion regarding the carcinogenic potential of atrazine in the human population. The Panel recommended EPA consider additional, follow-up studies from the AHS as well as other peer-review sources, to clarify the nature of the association with atrazine. As a result of this FIFRA SAP (2003) consultation, EPA concluded: “The Agency does not find any results among the available studies that would lead us to conclude that potential cancer risk is likely from exposure to atrazine. EPA plans to revisit this conclusion upon receipt of new studies, especially those from [National Cancer Institute] NCI’s Agricultural Health Study on atrazine and all cancers, prostate cancer, and non-Hodgkin’s lymphoma, all of which are planned for completion in the next 1-2 years [sic].” (Blondell & Dellarco, October, 28, 2003 2003 #1026) EPA believes it is appropriate to present this analysis at this time as several new evaluations on the topic have been published since 2003, including several evaluations by researchers with the Agricultural Health Study (AHS). 3.1.2 Identification of Epidemiology Studies included in Current Review This review of the epidemiologic literature of an association between atrazine exposure and specific anatomical cancer sites includes studies previously reviewed by EPA and presented to the FIFRA SAP panels (June 2000 and July 2003), as well as more recent peer-reviewed publications, including a new evaluation by AHS researchers published in May 2011. For the purpose of the current evaluation, emphasis is placed on epidemiologic studies published since the time of the last EPA review, i.e., September 2003 to May 2011. However, to provide a comprehensive evaluation, EPA presents epidemiology studies of atrazine or triazine exposure in association with any cancer outcome published between 1985 and 2011 in this review. The current review is intended to update the Agency’s cancer epidemiology evaluation to determine whether new information alters the Agency’s previous conclusion regarding evidence of the carcinogenic potential of atrazine in the human population. EPA values and appreciates the counsel provided by previous FIFRA SAPs (February and September 2010) regarding the draft “Framework” for incorporating epidemiology data into the risk assessment process (U. EPA, 2010). Specifically, these previous Panels encouraged EPA to employ a clear and consistent methodology for identifying studies included in the review of epidemiologic literature, including a delineation of excluded studies; consistent application of guidelines for adjudicating the results of individual observational studies, e.g., consistent application of inclusion and exclusion criteria, potential for bias in the studies; and, a simple and understandable method to synthesize the evidence presented in the epidemiology database, as well as integrate with the experimental animal database. While EPA is in the process of refining the proposed Framework, EPA strives to adhere to the principles offered by previous Panels in this respect. However, EPA also acknowledges that the current literature review reflects a necessarily hybrid approach to performing a review of this literature, given the history of the Page 34 of 184 previous atrazine cancer scientific evaluations, and the decision to a priori include all previously reviewed materials as well as all AHS relevant atrazine cancer point estimates in the current EPA review. Upon EPA’s request, previous FIFRA SAP panels (February and September 2010) provided several thoughtful suggestions to EPA regarding an epidemiological literature review methodology. Specifically, the Panel noted the importance of a clear and transparent literature search methodology, articulation of study inclusion and exclusion criteria, and the use of a scoring system with which multiple raters may adjudicate the quality of individual studies, among other suggestions. In the current review, inclusion criteria were purposely broad and reflected only that atrazine or triazine is measured in association with a cancer outcome. Upon title and abstract review, studies were excluded if they did not measure an association between atrazine or triazine exposure and a health outcome. Concerning the use of a scoring system, EPA considered this option but instead decided a broader and open evaluation of the epidemiologic literature was merited at this time while noting the strengths and limitations of each study. However, among the list of studies remaining after title and abstract review, a group of scientists jointly adjudicated the relevance of the study for inclusion, as well as “grouped” studies as generally “strongly” or “weakly” providing evidence to inform the discussion. A full description of the literature review methodology including a delineation of the search string utilized is presented in Appendix C. EPA considered several factors indicative of good quality epidemiologic analysis in performing this review. Among these factors were the ability to accurately measure the health effect (the outcome) as well as atrazine exposure; the identification and accurate measurement of potential confounding factors; the sample size and statistical power to measure the association; the appropriateness of the statistical analysis to the study design; and, the interpretation of the overall study results. Key among these factors is the method of exposure assessment. Specifically, whether authors were able to perform qualitative or quantitative exposure assessment; the ability to measure atrazine exposure during critical time window for the health effect; the range of the exposure captured in the study; and, the relevance of the biomarker (if measured) to actual environmental exposures. In combining evidence from individual studies and across disciplines, i.e., toxicology and epidemiology, EPA considered factors such as presence of an exposure-response relation, temporality of exposure-disease measurement, strength of individual risk point estimates, consistency of findings across studies, knowledge of biologically plausible hypotheses to support the observed findings, as well as professional judgment. 3.2 Epidemiology Literature of the Carcinogenic Potential of Atrazine The present review includes research articles presenting evidence concerning the carcinogenic potential of atrazine in the human population. However, the aspects of this database which have already been part of an external review, i.e., FIFRA SAP or other Page 35 of 184 external review processes, are not presented with the same level of detail in this chapter or in the study reviews. The relevant FIFRA SAP final reports which provide a thorough review of these studies as well as an overall synthesis of the epidemiologic evidence at prior points in time are included as appendices to this chapter for reference (Appendices A and B). Organized by anatomical cancer site, the chapter first briefly recollects studies previously reviewed, followed by a more detailed summary of the recent epidemiologic research. At the conclusion of each section, a synthesis of the epidemiologic data is proffered. EPA’s review of individual investigations are included in Appendix D (studies published prior to 2003), and Appendix E (studies published subsequent to 2003, including several from the AHS). In addition, a summary table of study characteristics is presented in Appendix F for the recently published epidemiology studies, the focus of the current EPA review. Ultimately, a synthesis and integration of the cancer epidemiology studies with the results of experimental toxicological investigations is presented in Section 3.3, using the modified Bradford Hill criteria as an organizing tool, in addition to scientific judgment. Importantly, EPA notes at this time that there was no evidence of tumors noted in any of the experimental toxicology studies performed, for any of the cancer sites evaluated in the epidemiology studies discussed below, with the exception of mammary gland tumors in rats which were later determined to have development through a biological mechanism not operational in humans (See full discussion in Section 3.3 below). 3.2.1 Overview of Epidemiologic Database The atrazine cancer epidemiology database includes 40 studies reflecting cohort, population-based case-control or nested case-control study designs. Among this number, several ecologic studies were identified, but were not weighed heavily in the weight of evidence evaluation, per the advice of previous Panels (U. EPA, 2010). Studies reflected general population exposure, occupational manufacturing as well as agricultural exposure. Various methods of atrazine and/or triazine exposure assessment were employed across these studies. In general, the studies varied in terms of strengths and weaknesses. The observational studies included in EPA’s current review of the atrazine cancer epidemiology database mainly consider the herbicide’s potential carcinogenic effect upon the reproductive and endocrine systems as well as the lymphohematopoietic systems. Approximately one-third of the cancer studies published subsequent to 2003 relate to these types of cancer endpoints. Investigations of atrazine exposure in association with leukemias and lymphoma include Hodgkin’s disease, non-Hodgkin lymphoma (NHL), multiple myeloma, leukemia, among other specific sub-types. Two new evaluations of atrazine exposure in association with lymphomas and NHL specifically have been published since 2003, however, a majority of studies published in this database published prior to 2003, investigate atrazine’s link with leukemias and lymphomas. Other cancer sites considered in the studies reflected in the current review include brain/glioma and colorectal cancers, as well as two evaluations of atrazine exposure in association with pediatric cancers including childhood leukemia. Given the importance of the AHS investigations to Page 36 of 184 the question of atrazine carcinogenicity, as articulated by the FIFRA SAP (2003) (See Appendix B), all other atrazine-cancer point estimates measured in either nested casecontrol or cohort studies are also included for reference. For this reason, a brief review of AHS design and methods, and the strengths and limitations of this cohort study are presented below. Results of specific AHS evaluations are incorporated within respective cancer-specific discussion sections. Following the discussion of the AHS, cancer-site specific studies are presented. Within each sub-section studies published prior to 2003 are briefly recalled, followed by a more expansive discussion of the recent epidemiologic data (2003-2011). EPA’s review of each individual study is presented in Appendix D (pre-2003) and E (post-2003). A synthesis of the observational studies within each cancer site-specific section as well as across all cancer endpoints researched is offered at the conclusion of each section. EPA presents its integrative discussion of both the toxicological and also the epidemiological evidence in Section 3.3. 3.2.2 Agricultural Health Study (AHS) Atrazine Cancer Point Estimates Since 2003 there have been six nested case-control analyses within the AHS which evaluated use of a number of agricultural pesticides, including atrazine, in association with specific anatomical cancer sites. These include investigations of prostate (Alavanja et al., 2003), lung (Alavanja et al., 2004), colorectal (Lee et al., 2007), breast (Engel et al., 2005) (among female spouses and female applicators in the AHS), pancreatic (Andreotti et al., 2009), and melanoma (Dennis, Lynch, Sandler, & Alavanja, 2010). In addition, researchers also published two cohort analyses specifically examining the association between atrazine use and cancer outcomes; multiple cancer sites were evaluated in each study (Beane Freeman et al., 2011; J. A. Rusiecki et al., 2004). This section provides general information concerning the design and methods of the AHS. The results of these investigations are considered within respective cancer-site discussion sections presented in this section. 3.2.2.1 Study Design and Methods The Agricultural Health Study (AHS) is a large, prospective cohort study comprised of private and commercial pesticide applicators in Iowa and North Carolina. The study includes 57,310 licensed pesticide applicators, comprising 52,394 private applicators and 4,916 commercial applicators, as well as 32,346 spouses of pesticide applicators. Researchers successfully recruited 82% of the target population into the AHS cohort which represents a high recruitment rate for an epidemiologic study. While farmer and pesticide applicator participants in the AHS are generally healthier than the overall U.S. population, AHS participants have a higher rate of some cancers including prostate, lip, multiple myeloma and ovarian cancers (Alavanja et al., 2005; Koutros et al., 2010). Lower rates of smoking and alcohol use, and the physically demanding nature of the occupation are health protective factors in the AHS population, and pesticides, viruses (animals), bacteria, Page 37 of 184 sunlight, and other non-pesticide chemicals may be putative risk factors for carcinogenesis among these farmers. Pesticide applicators must obtain a license from the state to purchase and use restricteduse pesticides. Participants in the AHS were enrolled into the study at county pesticide licensing facilities and were asked to complete a 21-page written enrollment questionnaire. Enrollment occurred over a period of 4-years, December 13, 1993 to December 31, 1997 (“Phase I”). The Phase 1 “enrollment” questionnaire solicited information concerning types of agricultural crops and livestock produced, types of pesticides used, pesticide application methods, use of personal protective equipment, and, lifestyle factors such as smoking and alcohol intake. A portion of the cohort also completed a “take-home” questionnaire (~40%) in which additional lifestyle (e.g., diet) and chemical exposure information (i.e., an additional 28 chemicals) was ascertained. To determine case- and vital-status, cohort members were linked to state cancer registries in Iowa and North Carolina and the state death registries and the National Death Index, respectively. Linking was accomplished through use of social security number, first and last name, date of birth, and gender. Study participants who were alive but no longer residing in either study states were identified through the Federal Internal Revenue Service address file, motor vehicle registration records, and the restricted-use pesticide license and certification records of the departments of agriculture in Iowa and North Carolina. At this time, less than 2% of the cohort has been “lost-to-follow-up,” i.e., can no longer be located for purposes of the study. This is a very low percentage lost to follow-up, reducing the chance of selection bias in this prospective cohort study. To measure atrazine exposure, AHS researchers considered three metrics across the studies including an atrazine cancer point estimate: (i) whether participants ever used the insecticide; (ii) the cumulative exposure days (the total number of exposure days per year, and total number of years in lifetime); and, (iii) an intensity weighted cumulative exposure metric. Exposure information was collected via self-report questionnaire responses. The “ever used” metric dichotomized atrazine exposure into ever/never categories where “ever used” was considered as at least used one time in lifetime. The lifetime exposure- days metric was calculated by taking the product of the total number of years atrazine was applied and the number of days per year atrazine was applied. The intensity-weighted lifetime exposure day metric includes: the frequency and duration of application (lifetime exposure days), application methods, mixing and equipment repair status, and use of personal protective equipment. These factors were weighted to reflect the intensity of the exposure based on monitoring data available in the published literature. Methods of pesticide exposure assessment used in the AHS have been reviewed extensively and shown to have high reliability and validity (Blair et al., 2002; Hoppin, Yucel, Dosemeci, & Sandler, 2002). Page 38 of 184 There have been six nested case control studies published within the AHS since 2003: prostate, breast, pancreatic, lung, colorectal, and cutaneous melanoma. In addition, there have been two cohort analyses of the relation between atrazine and several different cancer sites. The results of these specific evaluations are integrated within the cancerspecific discussions presented below. 3.2.2.2 AHS Strengths and Weaknesses The AHS includes both strengths and weaknesses. The large sample size and sufficient statistical power to investigate the potential association between atrazine use and several cancer sites including prostate cancer is a strength of the study. In addition, given the prospective study design the potential for recall bias (differential exposure measurement error) is eliminated, as pesticide exposure information was collected prior to the identification of disease. Exposure measurement error exists in this study (Blair et al., 2011), however it is likely non-differential in nature and would attenuate, and not falsely inflate, measured risk estimates. Selection bias is considered of low potential as very few participants have been lost to follow-up in this cohort (i.e., few participants dropped out of study or were unable to be followed through other means). In this cohort, cancer cases were ascertained through linkage with state cancer registries, and include histopathological confirmation as well as information as to stage and grade of cancer at diagnosis. High quality outcome ascertainment reduces the potential for outcome measurement error. The AHS collects a vast amount of information for use in statistical analysis including prostate cancer risk factors such as age, race and family history of prostate cancer; cancer risk factors such as smoking and alcohol use; and, use of other correlated pesticides. Lastly, the AHS exposure measurement tools have been illustrated to be valid (Hoppin et al., 2002) and reliable (Blair et al., 2002), and provide a quantitative measure of pesticide exposure over the working lifetime. In combination, these studies provide an important database of epidemiologic evidence pertaining to atrazine carcinogenicity. However, it is important to note that most of these studies are hypothesis-generating in nature and did not propose a specific biological mechanism as to the carcinogenic mode of action of atrazine per se. The recent evaluation (Beane Freeman et al., 2011) was able to test hypotheses related to prostate, breast, ovarian and NHL as these sites have been previously suggested to be related to atrazine use. These results are further discussed below. 3.2.3 Reproductive and Endocrine System Cancers Because atrazine has a known neuroendocrine mode of action several researchers have hypothesized a role for atrazine in the etiology of certain reproductive and endocrine system cancers. Investigations of a link between the herbicide and prostate, breast, ovarian, and thyroid cancer are available for consideration. Several studies have identified that farmers have a greater risk of prostate cancer, although overall cancer rates are less Page 39 of 184 than the rates in the general population (Alavanja et al., 2005; Koutros et al., 2010). Similarly, there is some indication for an increased risk of ovarian cancer among women who live near or work in agriculture (Alavanja et al., 2005). However, few analyses to date have hypothesized a role for atrazine specifically, i.e., hypothesis-testing in nature, and even fewer have investigated an association with female reproductive system cancers. In many of these investigations, the number of female cancer cases exposed to atrazine is small. 3.2.3.1 Prostate Cancer In 2003, EPA reviewed several studies that investigated the association between atrazine exposure and prostate cancer, and presented a review of these data to the FIFRA SAP (EPA, 2003b) [See Appendix]. Briefly, an increased risk of prostate cancer was observed among an occupational cohort of triazine manufacturing plant workers (P. A. MacLennan et al., 2002; Sathiakumar & Delzell, 1997; N. Sathiakumar, E. Delzell, & P. Cole, 1996), however further analyses by EPA and others, and supported by external peer review including the FIFRA SAP, concluded that the availability of an on-site prostate cancer screening program, i.e., the PSA test, likely explained some or all of the increased prostate cancer incidence (Blondell & Dellarco, October, 28, 2003 2003 #1026; EPA, 2003b). Authors hypothesized that the elevated cancer rates was due to detection bias, and that the increased prevalence of PSA screening could explain some or all of the increase in prostate cancer incidence observed in this occupational population. This is supported by the observation that the increase in cancer risk was only observed among active workers who had access to the screening, and a lack of exposure response gradient using a jobexposure matrix approach to measure exposure . As a result of several internal as well as external peer review processes, reviewers determined that given the limitations of the study design, small sample size and lack of refined exposure measurement, a possible role for atrazine could not be ruled out, however it did not appear likely. Subsequent nested case-control analyses of the triazine manufacturing plant workers using refined exposure measurements did not alter the conclusion (P. A. Hessel et al., 2004). However, the same limitations as noted above, primarily the small sample size and limited exposure information in this occupational study precluded a determination that an association between triazine exposure and prostate cancer was present or absent. Further, the results of a nested case control study within the AHS population indicated little evidence of an association with prostate cancer (OR 0.94, 95% CI 0.78, 1.14), and there was no observation of an exposure-response trend (p trend>0.10) (Alavanja et al., 2003). In another study, analyses using state cancer registry data (CA), and publicly available pesticide use data indicate some evidence of a statistical correlation, however limitations of the exposure analysis restricts use in causal inference (P. Mills, 1998; P. K. Mills & Yang, 2003). However, recently published AHS cohort analysis results addressed this question in cohort analyses with sufficient statistical power to detect an association if one exists Page 40 of 184 (Beane Freeman et al., 2011; J. A. Rusiecki et al., 2004). These data are presented below, followed by a synthesis of the studies in this database. In 2004, Rusiecki et al. published a cohort analysis among participants in the AHS of atrazine exposure in association with many different cancer endpoints, including prostate cancer. AHS study design and methods are noted in section 3.2.2.1. With reference to prostate cancer, authors reported no evidence of a statistical association between atrazine exposure and prostate cancer using either cumulative exposure or an intensity-weighted cumulative exposure metric. Specifically, across quintiles of atrazine lifetime exposure days, the following odds ratios were estimated in comparison to the lowest exposed group (1-20 days): 21-56 days, OR 0.89 (95% CI, 0.66 to 1.21); 57-178.5 days, OR 0.75 (95% CI, 0.56 to 1.03); and greater than 178.5 days OR 0.88 (95% CI, 0.63 to 1.23) (J. A. Rusiecki et al., 2004). The test for linear trend was not significant, p=0.26. Results were similar using intensity-weighted lifetime exposure days. Authors state that in this study there was sufficient statistical power to detect an association of 1.3 in the highest quartile, assuming a linear trend across quintiles of atrazine exposure (1-β=0.89). A new evaluation within the AHS cohort study with over twice as many prostate cancer cases indicated similar results. Risk estimates in the upper quartiles of either lifetime exposure days or intensity-weighted lifetime days as compared to the lowest quartile of exposure were not significantly different from the null (Beane Freeman et al., 2011). Compared to the lowest exposure group, odds ratio of prostate cancer in the upper quartiles of atrazine exposure were 1.04, 1.06, 1.04, respectively. In addition, there was no visual or statistical evidence of any exposure-response trend (p for linear trend 0.81). The latter analysis reflected use of an updated exposure algorithm which reduced exposure measurement error in comparison to the original algorithm used in the earlier analysis. While reflecting two different study periods as well as two different case series (prostate cancer cases updated through 2007), these two investigations within the AHS are similar in design and methods. The initial investigation by Rusiecki et al. (2004) was hypothesis generating in nature, however previous research, as noted above, provided support for a possible association with prostate cancer. The follow-up investigation examined the same question, within the same cohort, at a different point in time, and in many ways confirms the original finding. Differences between the two studies include an increased number of prostate cancer cases in the latter investigation, and use of the updated exposure algorithm in the latter analysis. Both studies had sufficient statistical power to detect an association with atrazine, if an association exists. Notably, the analysis by Rusiecki (2004) and BeaneFreeman (2011) both adjusted for use of other pesticides including several that are typically highly correlated with atrazine use including alachlor and 2,4-D as well as other triazines including cyanazine and metribuzin. Overall, the results of these two cohort analysis in which atrazine use was compared between cases and non-cases of prostate cancer, among other cancer sites evaluated, did not suggest an association is present among pesticide applicators in the AHS, who demographically were primarily U.S. white men. Page 41 of 184 Conclusions: Prostate Cancer - The results of several internal and external peer review processes concluded that the association between triazine exposure among an occupational cohort of manufacturing plant employees can likely be explained in part or in whole by the availability of intensive PSA screening, i.e., evidence of detection bias (EPA, 2003b). Previously published analyses using California state cancer registry and pesticide use data (P. Mills, 1998; P. K. Mills & Yang, 2003) suggest a possibility of a correlation and/or an association with atrazine or triazine (simazine measure Mills et al. 2003), however other limitations of these studies reduce their use in a weight of evidence analysis, primarily the method of aggregate exposure assessment, and lack of an atrazine risk estimate. The results of two cohort analyses within the AHS with sufficient statistical power to detect an association if an association exists did not observe statistical evidence of a link between atrazine use and prostate cancer. The magnitude and direction of the association was not different from unity (i.e., a null association), and there was no evidence of an exposure response trend using either lifetime exposure days or intensity-weighted lifetime exposure days. The AHS study design assured the atrazine exposure preceded the disease outcome. Sensitivity analyses performed using both the low-exposed and nonexposed groups as referent did not differ. Notably both analyses simultaneously controlled for use of correlated pesticides in the statistical analyses, for which the analyses were adequately powered to investigate, and no associations were observed. As noted by previous SAPs (EPA, 2003b), the exposure profiles of the triazine manufacturing plant workers and the AHS private (farmers) and commercial applicators differ, and lack of association among farmers does not preclude the existence of a positive association with triazine manufacturing plant workers. The higher magnitude of the exposure, the different exposure profile, and the distinct population characteristics (age, race, decade of exposure, other pesticide/chemical exposures) may indeed result in two different measures of association. However, to date, the information is not sufficient to support this supposition, and the AHS population is likely among the most highly exposed occupational groups in the nation. Although the primary risk factors for prostate cancer are age and family history, the specific causes of prostate cancer remain unknown. In general, epidemiological evidence of a role of atrazine in prostate cancer is minimal. In spite of the herbicide’s neuroendocrine MOA, a clear role of atrazine in prostate carcinogenesis is not understood, and evidence from the experimental toxicology literature would argue against a possible mode of action for atrazine in prostate tumor formation (See section 3.3 below). At this time the lack of evidence of a statistical association in the two cohort analyses available from the AHS which reflects a prospective design assuring temporality of exposure-disease measurement, consistency of results between manufacturing (P. Hessel et al., 2004; P. A. MacLennan et al., 2002) and agricultural (Alavanja et al., 2003; Beane Freeman et al., 2011; J. A. Rusiecki et al., 2004) populations, and lack of compelling biological hypothesis Page 42 of 184 as to the mode of prostate carcinogenesis from the experimental studies, argue against a link between atrazine and prostate cancer. 3.2.3.2 Breast Cancer Several authors have evaluated the association between atrazine/triazine exposure and breast cancer. The rationale for these investigations is the known neuroendocrine health effects of atrazine, and earlier experimental studies which reported mammary gland tumors in female rats and studies reporting an effect on mammary gland development. However, more recent toxicological data suggest that the mode of action for atrazine’s effect on mammary gland development in rats is not operational in humans (EPA, 2010; EPA, 2000) (See Section 3.3). Nevertheless, there have been three ecologic studies evaluating this potential association in the human population, two were published prior to 2003 and previously reviewed by EPA (Bassil et al., 2007; Hopenhayn-Rich, Stump, & Browning, 2002; M. A. Kettles, S. R. Browning, T. S. Prince, & S. W. Horstman, 1997), and one was more recent (Muir et al., 2004). The ecologic evaluations are collectively discussed below. In addition, a population-based case control study in Wisconsin (J. A. McElroy et al., 2007) using atrazine well water concentrations to measure exposure, and two recent evaluations from the AHS (Beane Freeman et al., 2011; Engel et al., 2005) provide additional information to examine this question. There are three separate ecologic analyses of the association between groundwater/drinking water exposure to atrazine and breast cancer. Two were previously reviewed by EPA, and one is considered for the first time herein. Kettles et al. (1997) conducted an ecologic study to examine the association between breast cancer incidence and triazine herbicide exposure; in 2002 Hopenhayn-Rich updated the analysis. The Kettles et al. (1997) study calculated county breast cancer incidence rates using Kentucky’s statewide cancer registry for the years 1991-1994. Triazine herbicide exposure was assessed using data on ground and drinking water contamination, corn acreage, and pesticide usage by applicators. Researchers report a small, statistically significant association between breast cancer and water contamination. However, the investigators also indicate that the results were not consistent across study years. While the water contaminant associations were not consistent between time periods, the investigators did report that their composite exposure variable was associated with breast cancer during both 1991-92 and 1993-94 (i.e., Medium and High Exposed County ORs = 1.09 (1.041.14), 1.07 (1.01-1.14) for 1991-92 and 1.14 (1.08-1.19), 1.20 (1.13-1.28) for 1993-94) (M. Kettles, S. Browning, T. Prince, & S. Horstman, 1997). This study is discussed in the final SAP report from the June 2000 meeting (Appendix A) and concluded these data were insufficient to make a determination as to the relation between atrazine and breast cancer. In 2002, Hopenhayn-Rich updated this analysis by refining exposure measurement and including three additional years of cancer registry data, but utilized similar methods in other ways. In this ecologic evaluation, authors reported that their study did not provide evidence Page 43 of 184 that supports an association between atrazine exposure and county level incidence of breast cancer(Hopenhayn-Rich et al., 2002). Muir et al. (2004) performed an ecologic analysis of this association in two counties in England, Lincolnshire and Leichestershire. Researchers included all cases of breast cancer diagnosed among women, ages 45 years and greater in these two counties between the years 1988 to 1991. Cancer cases were ascertained through regional cancer registry (Trent Cancer Registry). Pesticide exposure was measured at the group level. A pesticide usage survey for 1991 was used to determine the total kilogram of active ingredient applied per kilometer by each ward within the two counties of England represented in the study. Investigators included atrazine, aldicarb, lindane and cyanazine in the study as these were identified as potentially estrogenic substances by the study authors and hypothesized to play an etiologic role in breast cancer development. Authors reported no significant correlations (“clusters”) of breast cancer by geographic area. Furthermore, authors observed no association between average pesticide usages by ward and breast cancer incidence rate ratio (IRR) in urban or rural areas of Lincolnshire, or urban areas of Leichershire. Authors did note a statistically significant (p<0.05) positive change in slope of breast cancer incidence rates and 1-unit change in kg of atrazine used per km2 (slope 0.17 (0.03, 0.31)) (Muir et al., 2004). The researchers noted the inherent limitations of the study design, but suggest the results of this work supports excluding the possibility of a strong, positive association between atrazine usage and breast cancer among women diagnosed later in life (otherwise these studies would have detected greater associations). Overall, these three ecologic studies provide limited, inconsistent evidence as to whether an association exists between atrazine exposure and breast cancer. The slight suggestion of an association in Kettles (1997) was not observed by Hopenhayn-Rich (2002) using similar methods and similar study population. The study by Muir et al. (2004) does not present internally consistent results, positive slope in only one of two study areas, and the weak magnitude and statistical significance of the slope (coefficient) does not strongly argue in favor of an association. Given the nature of the study design, these data offer weak, inconsistent evidence of a possible link between atrazine and breast cancer. Mills et al. (2006) performed a hypothesis-testing, cohort study using ecologic exposure assessment methods to evaluate the association between atrazine exposure and breast cancer among Latinas diagnosed with breast cancer. Invasive breast cancer cases were identified from the California state cancer registry (CCR) between the years 1988 to 1999, and compared with county level pesticide exposure, and grouped demographic data. Pesticide use information from the years 1970 to 1988 was obtained from the California state pesticide use registry (PUR). Authors were interested in six pesticides, including atrazine and simazine. Authors calculated standardized incidence ratios (SIR) for 19881993 and then for 1994-1999 to reflect secular changes in breast cancer trends (screening, diagnostic tools, knowledge in population), and to examine latency between exposure and diagnosis. Adjusting for age and other co-variables, authors reported no association Page 44 of 184 between atrazine exposure and breast cancer comparing the highest and lowest quartiles of exposure (1988-1993: OR 0.86 (95% CI 0.70, 1.05); 1994-1999 OR 0.87 (95% CI 0.73, 1.04) Mills et al. (2006) performed a large, cohort study of atrazine exposure and breast cancer in a highly exposed group. Outcome classification was good, based upon high quality cancer registry, and using state pesticide use records is an innovative method to assess exposure. By design, authors assured pesticide exposure preceded disease diagnosis. Established breast cancer risk factors were indeed significantly associated with breast cancer in this study, which is reassuring of the study methods. The major limitation of this study is the ecologic nature of the exposure assessment. Authors assumed PUR reports for the county and group level confounder information were a good approximation of pesticide exposure of women diagnosed with breast cancer and co-variable information, which likely resulted in exposure misclassification. Because of the lack of individual level exposure and confounder information, an association cannot be ruled out based upon this study. However, in general, this study lacks evidence of an association. McElroy et al. (2007) performed a population-based case-control study of atrazine exposure and breast cancer among 22,495 women residing in the state of Wisconsin. However, given the timing of exposure and disease information collection, the study was essentially cross-sectional in nature. Cases included women who were diagnosed with breast cancer during the years 1988 through 2001, and reported to the state tumor registry. Controls were randomly selected from state DMV (20-65 years) and Health Care Finance Administration (65-79 years) records to roughly approximate the age distribution of cases (10-year age categories), and for whom no previous breast cancer diagnosis was made. Investigators were interested in atrazine exposure from untreated well water used for drinking water. Therefore, authors restricted the study population to those who resided in a rural area for whom well-water would likely supply local drinking water defined as a region with no public water system. Upon applications of the eligibility and restriction criteria, 3,275 case and 3,669 controls were selected into this study. The concentration of atrazine and atrazine metabolites in well water which provided the sole residential drinking water source for participants in this study was measured in a state survey of pesticide contamination during three survey years: 1994, 1996, and 2001. To estimate atrazine exposure between sampling years of 1994, 1996, and 2001 to the entire study period of 1988-2001, authors utilized a “national neighborhood interpolation” method using ArcGIS. Authors considered atrazine well water concentration within the entire state and within the highest agricultural area of the state separately, the south central WI region. Overall, authors reported little evidence of an association between atrazine concentration in well water and breast cancer. Comparing those in the moderately high category to the lowest exposure category, minimal increased odds of disease were reported (OR 1.10, 95% CI 0.90, 1.40) (J. McElroy et al., 2007). Among participants in the highest atrazine exposure area, odds of breast cancer was non-statistically significantly elevated to 1.3 (95% CI 0.30, 5.0), however precision was Page 45 of 184 low and the number of cases in the upper category was small (n<5) (J. McElroy et al., 2007). This study presents limited evidence of a role for atrazine drinking water exposure in relation to breast cancer. These study methods are an improvement over previously published population-based studies of atrazine and breast cancer; however there are likely significant information biases (resulting in exposure misclassification), and heterogeneity in outcome classification, in addition to low sample size in the upper-exposure categories that could explain the essentially null findings. There are several strengths to this method of exposure assessment including the fact that this is an improvement over previous ecologic studies, the interpolation across study years is supported by the fact that atrazine is relatively persistent in water, and the restriction to rural residents likely isolates the most highly exposed to atrazine in drinking water. Using these exposure methods, however, there are numerous possibilities for exposure measurement error in this study, several of which are acknowledged by the authors. Concerning residential history, the exposure measure in this study for a woman diagnosed at 60 years likely represents her adult life exposure, however a woman diagnosed at 25 years, the exposure may represents her early life exposure experience. Additionally, the outcome measurement may be heterogeneous: breast cancers diagnosed early in life (3rd decade) as opposed to later in life (7th decade of life) most likely reflect different etiologies, and known differences between estrogen receptor positive tumors in relation to possible influence by an endocrine disrupting pesticide were not evaluated. The effect of both exposure and outcome measurement error may have attenuated these results. However, as the authors state, there is no strong indication of an association, if there is a positive association in nature, it is likely modest. These data provide little evidence of a strong association between breast cancer and atrazine exposure, but this study cannot exclude the possibility of a poorly measured moderate association. Within the AHS cohort, two studies evaluated an association between atrazine use and breast cancer: a nested case-control analyses among female spouses of pesticide applicators (Engel et al. 2005), and a cohort analysis including female pesticide applicators (Beane-Freeman et al. 2011). AHS study design and methods are discussed above in section 3.2.2.1. Engel et al. (2005) compared ever-use of several pesticides, including atrazine, and breast cancer rates among female spouses of enrolled pesticide applicators. Authors were able to assess both direct, ever-use of pesticide in lifetime, and indirect pesticide exposure, reported use by the pesticide applicator husband. However, atrazinespecific exposure information was only available for the ever-use metric, i.e., detailed exposure information including duration and frequency of use was not available. The female spouses who responded to the supplemental questionnaire also provided information concerning other demographic data, reproductive health and known breast cancer risk factors. As noted above, cancer outcome information was ascertained through linkage with state cancer registries and is considered to be more than 95% complete (Alavanja et al., 1996). Adjusting for age, race and state, authors did not observe evidence Page 46 of 184 of an association between ever use of atrazine and incident breast cancer among female spouses of enrolled applicators when measuring direct exposure (OR ever use 0.7, 95% CI 0.40, 1.2), or indirect exposure (husband’s ever use OR 1.1, 95% CI 0.7, 1.6) (Engel et al., 2005). Slight differences in estimated risk likely due to increased stability of indirect measure (greater number indirectly exposed women). Considering use of any pesticides, authors observed no trend by duration of use (direct report of number years, days per year), or proximity to pesticide treated area. A slight trend by number of times per year wife washes clothes worn by husband when applying pesticide was observed, but confidence bounds overlapped across indirect exposure category. While exposure measurement is limited in this study, ever-use directly reported by only 11 atrazine exposed cases, the study lacks evidence of an association between atrazine use and breast cancer. In the recent cohort analysis of atrazine use, authors estimated that association with breast cancer among female applicators enrolled in the AHS (Beane Freeman et al., 2011). In this study, authors used the ever-use metric only; there were too few breast cancer cases among female pesticide applicators to evaluate an exposure-response trend using quartiles of atrazine exposure. Authors elected to evaluate this cancer site due to a priori interest based upon previous epidemiologic research, and atrazine’s neuroendocrine health effects. Authors reported no significant association among nine atrazine exposed cases and 27 controls (OR 1.14, 95% CI 0.47, 2.50) (Beane Freeman et al., 2011). Similar results were observed, according to these authors, when comparing breast cancer rates between women greater than and less than the median number of lifetime exposure days. Authors adjusted for age only in this study, as other breast cancer risk factors were not confounders in this sample and the use of fewer co-variables retains statistical power in a small sample. The strengths of the AHS in terms of design, exposure and outcome measurement, and confounder information have been noted above. In this analysis, the small number of female breast cancer cases and resultant inability to explore an exposure-response relation is a limitation. However, strengths of the study include temporal association, exposure precedes disease. There have been seven evaluations of the association between atrazine or triazine exposure and breast cancer since 1997 identified in this review. Three of these studies utilized an ecologic study design, and offered limited, inconsistent evidence of a link in the human population, and are more supportive of the null hypothesis, no true association. Mills et al. (2006) evaluated the question among Latinas in the state of California using state cancer registry data, and aggregate level exposure analysis. These authors reported no evidence of an association in two time periods evaluated, however the measurement error in the method of exposure assessment, and uncertainties presented may have attenuated these results in this highly exposed population of Latina agricultural workers. AHS investigators evaluated the breast cancer-atrazine link among female spouses of licensed pesticide applicators (direct and indirect exposure) as well as female applicators themselves. Neither investigation observed evidence of an association. Although the AHS is generally a very strong study, estimating risks among women exposed to pesticides is Page 47 of 184 more limited than among men. Neither Engel et al. (2005) nor Beane-Freeman et al. (2011) observed strong statistical associations, nor were they able to explore exposure response relation due to the small number of atrazine exposed females, measuring only ever-use of atrazine during the lifetime. Both studies had wide non-statistically significant confidence intervals indicating lack of power to detect a difference if one truly was present. Research indicates chemical exposure during different points in the life cycle may critically impact cancer risks. None of these investigations were able to explore this aspect of the question. Considering this database in total, studies available reflect different designs; varied target populations including general population in high-use areas, agricultural workers, and occupational groups; evaluated both direct and indirect pesticide exposure; explored exposure through drinking water, or occupational (dermal/inhalation) pathways; ever-use of the herbicide as well as exposure-response, among other sources of variation. Overall, these studies lack strong evidence of an association between breast cancer and atrazine exposure among low general population levels, or presumably higher occupational exposure levels. Limitations across these studies include aggregate exposure assessment, exposure measurement issues in those studies that attempt to estimate exposure at the individual-level, and small number of atrazine exposed females in occupational settings (AHS). These studies are mainly initial, exploratory analyses, with the exception of BeaneFreeman et al. (2011) which estimated the association based upon a priori hypothesis. However, none of the investigations were able to explore critical windows of exposure, or exposure-response associations with sufficient number of exposed females in the higher exposure categories. At this time, the database lacks evidence of a relation between atrazine exposure and breast cancer. 3.2.3.3 Ovarian Cancer Because of the neuroendocrine effects of atrazine exposure, some epidemiology studies have been performed evaluating the link between atrazine and ovarian cancer, however there are only a few. An elevated risk of ovarian cancer was reported comparing the AHS population to the general population of the respective study states (IA/NC) (Alavanja et al., 2005; Waggoner et al., 2011). Within the atrazine cancer epidemiology database, two evaluations included a point estimate for atrazine exposure in relation to ovarian cancer risk; one was previously reviewed by EPA (EPA, 2000) (See Appendix B-1). Briefly, for reference, Donna et al. (1989) performed a population-based case control study in Italy in the mid-1980s. Cases were ascertained through hospital registries, assumed to be complete among all cases in the region, and controls were selected from electoral roles of the same region. Among the 65 cases and 126 controls selected, pesticide exposure was assessed through questionnaire and classified as ‘definite,’ ‘probable,’ or ‘non-exposed’ based upon questionnaire responses which were independently rated by two industrial hygienists. Relevant confounding variables were also collected and included in the analysis, e.g., Page 48 of 184 ovarian cancer risk factors. Authors observed two-fold elevated odds of ovarian cancer among those who were classified as definitely exposed to triazines (OR 2.7, 95% CI 1.0, 6.9) (Donna et al., 1984; Donna et al., 1989). In addition, authors observed evidence of exposure-response by duration of exposure and by probability of exposure (i.e., none, probable, definite). This study supported earlier findings from the same research group, however the earlier publication measured herbicide use only (Donna et al., 1984). A strength of the study was the inclusion of histopathologically confirmed cases which reduces outcome measurement error. In addition, authors were able to obtain relatively detailed exposure measures through interviewer administered questionnaire, however, exposure measurement error remains an uncertainty in this study. Case ascertainment was estimated to be 96%, and controls were similar to the source population indicating selection bias was unlikely. The relatively small sample size and the probabilistic assignment of exposure category are weaknesses. As noted in previous EPA reviews of this study, the inability to adjust for other pesticide exposure is an uncertainty. Factors in favor of a true association with ovarian cancer include the magnitude and statistical significance of the reported association in more than one study, evidence of an exposure-response trend, temporality of exposure-disease relation through study design decisions, and general plausibility of a role for triazine in endocrine cancers. Synthesis of these data with experimental studies is discussed in Section 3.3. In addition to this evaluations, Hopenhayn-Rich (2002) performed an ecologic analysis of triazine exposure, measured in state drinking water monitoring data, and ovarian and breast cancer. Regarding ovarian cancer and triazine/atrazine use authors observed no evidence of an association, and authors in fact reported an inverse association between the mid- and high-triazine exposure groups (ecologic assessment) (Hopenhayn-Rich et al., 2002). A separate ecologic analysis did not observe a correlation between atrazine exposure and ovarian cancer in women in Ontario (J. A. Van Leeuwen, D. Waltner-Toews, T. Abernathy, B. Smit, & M. Shoukri, 1999). Under the conditions of these studies, the results do not lend evidence in support of an association with atrazine exposure. In a hypothesis-testing investigation, Young et al. (2005) performed a population-based case control study of ovarian cancer in association with triazine exposure in women who resided in the Central Valley, CA, an area of high agricultural land use and pesticide exposure. Histopathologically confirmed epithelial ovarian cancer cases were ascertained through two regional registries, and controls were selected through random digit dialing (RDD) of female residents in the study area who had a land-line telephone. Ovarian cancer cases diagnosed in 2000-01 were eligible; pesticide exposure information was obtained through a variety of sources and reflected the time period 1974-1999. Response rates were modest in this study: 40% for cases and 57% for controls. Researchers assessed occupational and residential history as well as other ovarian cancer risk factors through a self-administered questionnaire distributed to study participants. Occupational history included information such as location of agricultural job site, crops worked, job titles, years Page 49 of 184 of employment among other characteristics. However, participants were not directly queried as to specific pesticides used in the workplace. Authors estimated triazine exposure indirectly by linking California state pesticide use reports for the study period to participants’ reported job location, years of employment, and crops worked.. In addition, weighting factors were used to estimate intensity of exposure by job title, e.g., mixing and/loading pesticides versus harvesting. Only 8.8% of participants reported using triazines, and participants used simazine and cyanzine more than atrazine. Adjusting for several endocrine cancer risk factors, authors observed a non-statistically significantly elevated risk of ovarian cancer in relation to total triazines (OR 1.34 (95% confidence interval (95% CI) 0.77,2.33), but not in association with atrazine OR 0.76 (95% CI 0.16, 3.55), estimate based upon 2 and 12 exposed cases and controls (H. Young, P. Mills, D. Riordan, & R. Cress, 2005). Examining quintiles of total triazine exposure, a suggestion of an exposure-response trend upon visual inspection, however the test for linear trend was not significant at the p<0.10 significance level. Using an alternative statistical modeling approach, authors modeled triazine exposure as a continuous variable; no evidence of a significant slope was observed. Sensitivity analyses regarding weights used in the job exposure matrix, and the investigation of effect modification by reproductive cancer risk factors were not additionally informative. Authors observed some evidence of a positive association between occupational and residential triazine exposure and ovarian cancer, however there was no statistical evidence observed for an association with ever-use of atrazine, and no statistical evidence of an exposure response trend for duration of exposure to triazines. The low prevalence of atrazine use among women in this cohort (triazine prevalence 8.8%, atrainze prevalence likely <5%), and the number of exposed cases and controls was small (2 and 12, respectively), likely explains the imprecise risk estimate for atrazine exposure. A study with a larger sample size would enhance the precision of the risk estimate, and clarify the true direction of the association, if any. In addition, the exposure assessment method likely led to non-differential exposure misclassification, attenuating the risk estimates. The low to moderate participation rates may indicate the possibility of non-response bias, i.e., those who refused to participate may have a different exposure profile than those who agreed. The reported comparability between included and excluded cases with respect to age and stage and grade of tumor at diagnosis somewhat reduces the potential for this type of systematic bias, however authors acknowledged this uncertainty in the study. Exposure information was collected to reflect the time period prior to ovarian cancer diagnosis, reducing the chance of temporal bias, and exposure was assessed similarly between cases and controls. Statistical analyses were clear and appropriate, including the selection of confounding variables in the analysis. However, triazine-specific models may have been too small to support adjustment for several of the ovarian cancer risk factors included in models, further reducing statistical power. Other specific strengths and weaknesses of this study are discussed in Appendix B-5. Overall, the authors conclude that this study cannot Page 50 of 184 support or refute the existence of a true association between atrazine/triazine exposure and ovarian cancer. A recent AHS investigation provided an estimate of the risk of ovarian cancer among female pesticide applicators in relation to ever use of atrazine during the working lifetime. There were too few atrazine exposed female applicators to explore an exposure-response relationship. Controlling for age only, as other ovarian cancer risk factors were not confounders in the analysis, authors observed a non-statistically significant elevated odds of ovarian cancer, OR 2.91 (95% CI 0.56, 13.60) in association with ever-use of atrazine (Beane Freeman et al., 2011). It should be noted that this analysis reflected only nine ovarian cancer cases, 4 of whom reported ever using atrazine. All of these cases reported lifetime number of days used below the median for all atrazine exposed cancer cases, i.e., the ovarian cancer cases were among the lower end of the exposure distribution. Authors state that the role of chance, or possibly another unknown source of unmeasured confounding may explain these results; however, there is some plausibility of the result given the neuroendocrine effects of atrazine. Strengths and weaknesses of the AHS are noted in section 3.2.2.1 above, however with reference to this evaluation, strengths of the study include hypothesis-testing nature of the estimation, the ability to investigate this question among a group of occupationally exposed women, and lack of confounding bias in the analysis. However, the lack of ability to explore an exposure-response association, and the imprecise nature of the risk estimate due to a small number of atrazine exposed cancer cases are limitations which must be addressed in further studies. There are four epidemiologic investigations which consider the relation between atrazine or triazine use and ovarian cancer: (1) an ecologic analysis which reported an inverse association; (2) a population-based case-control study in Italy which reported elevated odds of ovarian cancer for the most highly exposed group, but for which the study was limited in its ability to adjust for other pesticide and confounder exposures; (3) a population-based case-control study in California using publicly available data on outcome and exposure to estimate risks and, therefore, is reflective of a large sample size, but only a small number of atrazine exposed cases and controls (2 and 14 respectively); and, (4) the hypothesistesting risk estimate presented in a recent AHS cohort analysis that reported elevated odds of the cancer, but which only utilized an ever-use exposure metric. These data are supplemented by general information that the incidence and mortality of ovarian cancer are elevated among the AHS occupational cohort in comparison to the general population, but this information is clearly not specific to atrazine. Each study through design or methods reflects an appropriate temporal association (exposure precedes disease). However, given the typically late stage at ovarian cancer diagnosis, and long period of latency (Adami, 2002), some uncertainty regarding the temporal association exists. Although elevated associations were observed in both Donna et al. (1989) and Beane-Freeman et al. (2011), estimates were somewhat imprecise most likely due to the small number of exposed and highly exposed ovarian cancer cases, and exposure-response relationships were not fully assessed in these studies. Overall, this database presents some indication of a possible Page 51 of 184 association between atrazine and/or triazine use and ovarian cancer; however, the small sample sizes, lack of ability to control for other pesticide use and possibility of unmeasured confounders in the association limit the ability to infer a causal association at this time. The key events in the neuroendocrine mode of action for atrazine in relation to ovarian cancer is discussed in Section 3.3 below. 3.2.3.4 Thyroid Cancer EPA is aware of only one study of the possible association between atrazine use and thyroid cancer; a recently published atrazine cohort analysis through the AHS explored this association. However, EPA notes that the relation was only assessed in male pesticide applicators; there was only one atrazine exposed female applicator in the cohort. The etiology of thyroid cancer is not well understood, although family history of thyroid cancer and radiation exposure are known risk factors. Also, the etiology and epidemiology of thyroid cancer differs between men and women and differs among the types of thyroid tumors. In the AHS, Beane-Freeman et al. (2011) compared lifetime atrazine exposure days among twenty-nine exposed thyroid cancer cases in comparison with non-cases of thyroid cancer. Author’s ascertained thyroid cancers among the cohort through linkage with state cancer registries, and self-administered questionnaires provided information on use of other pesticides, demographic characteristics and other cancer risk factors. Adjusting for several potential confounding factors, researchers observed a statistically significant, four-fold increased odds of thyroid cancer comparing the most highly exposed group to the referent (the lowest exposed) (OR intensity-weighted upper quartile 4.84 (95% CI 1.31, 17.93)) (Beane Freeman et al., 2011). The association remained when researchers controlled for body mass index, a known risk factor for thyroid cancer (OR 4.30 (95% CI 1.21, 15.28)). Non-significant, but positive 2- to 4-fold associations were consistently observed across quartiles of exposure using both the cumulative exposure metric and the intensity-weighted exposure metric. The test for linear trend was significant at p<0.10, but not at p<0.05, for intensity-weighted lifetime days. However, the lack of a clear exposure-response trend, and the small number of atrazine exposed cases in some categories, e.g., n>5, were limitations to this study and present some uncertainties in the interpretation of these results. The authors conclude that this investigation, given the strengths of the AHS in design and methods (mentioned above), provide some evidence of a positive link between atrazine and thyroid cancer in male pesticide applicators who reported using atrazine during their lifetime. EPA identified no other evaluations of this relation in the literature. There is no known biological mechanism postulated for atrazine at this time (See Section 3.3). Authors noted that another AHS investigation examined pesticide use in association with hypothyroidism, a suspected risk factor for thyroid cancer, and observed no association with atrazine Page 52 of 184 (Goldner et al., 2010). Strengths of this one investigation of the topic include the design and method of the AHS, i.e., prospective, extensive pesticide exposure information that has been validated and reliable, appropriate temporal association to estimate risk, ability to examine an exposure-response relation, and general plausibility of result. However, the lack of an observed trend, paucity of other epidemiologic or experimental data providing any comparative information, low statistical power and ability to measure risk with less uncertainty due to small sample size, and lack of mechanistic information. 3.2.4 Lymphohematopoietic Cancers In this section, cancer epidemiology studies of the association between atrazine and/or triazine use and lymphohematopoietic cancers are reviewed. Over the last twenty years, several studies have hypothesized a role for farming exposures and pesticides specifically in relation to leukemia and lymphomas. Specifically, a series of population-based case control studies located in the mid-western U.S. performed by the National Cancer Institute (NCI) estimated the risk between pesticide use including atrazine and several lymphohematopoietic outcomes (Brown et al., 1990; Brown, Burmeister, Everett, & Blair, 1993; Burmeister, 1990; Cantor et al., 1992; Hoar et al., 1986; Hoar Zahm, Weisenburger, Cantor, Holmes, & Blair, 1993; Weisenburger, 1990). Among these studies, however, only two specifically tested the hypothesis of atrazine exposure in association with one of these cancer subtypes, non-Hodgkin lymphoma (NHL) (Hoar Zahm et al., 1993; Zahm et al., 1993). In addition, cohort evaluations comparing cancer incidence (standardized incidence rates, SIR) and cancer mortality (standardized mortality rates, SMR) rates between an occupationally group exposed to triazines and the general population also estimated risk of NHL and other sub-types (P. MacLennan et al., 2002; P. A. MacLennan, E. Delzell, N. Sathiakumar, & S. L. Myers, 2003; Sathiakumar & Delzell, 1997; N Sathiakumar, E Delzell, & P Cole, 1996). Each of these evaluations has been previously reviewed by EPA (Blondell & Dellcarco, October, 28, 2003 2003 #1026; EPA, 2000, 2003a) [See Appendix B.3]. Since the time of the last EPA review, DeRoos et al. (2003) published a pooled analysis of the association between pesticide use and NHL pooling data from three population-based case control studies (IA/MN, KA, and NE) performed by the NCI in the 1980’s. These authors employed a hierarchical regression technique which allowed for co-adjustment for exposure to 47 different pesticides when measuring the specific putative pesticide-NHL association including atrazine (De Roos et al., 2003). EPA also identified two hospitalbased case-control studies performed in France investigating the possible link between atrazine and other pesticides with leukemias and lymphomas during two separate study periods (Clavel et al., 1996; Orsi et al., 2009). In addition, two cohort analyses in the AHS investigating the use of atrazine in relation to all cancer outcomes including leukemias and lymphomas have been published since EPA last reviewed this database (Beane Freeman et al., 2011; J. A. Rusiecki et al., 2004). Each of these studies is briefly discussed below. Page 53 of 184 3.2.4.1 Population-Based Case-Control Studies: Midwestern U.S. In the early 1990’s, the NCI established three population-based case control study populations in the Midwestern U.S. to examine the association between several different lymphohematopoietic cancers and farm-related exposures including pesticides. Across these studies, cases of lymphohematopoietic (leukemias and lymphomas) were identified through state-based cancer registries (IA, KA), or state-wide cancer surveillance, and controls were identified from the general population. Demographic and pesticide exposure was obtained through in-person (IA, MN) or telephone-based (KS, NE), structured interview questionnaires administered by trained study staff members. Duration and intensity of pesticide exposure was available for a subset of participants, and a small pesticide validation study using sales records was performed (Hoar et al., 1986). Authors solicited information about ever use of specific pesticides, farm activities, residential history, nonfarm occupations, and first and last year of use of specific pesticides. Pesticide exposure information was obtained for 38 herbicides, 24 insecticides, 34 crop insecticides, and 16 fungicides. Atrazine exposure information was collected in most of these studies; Hoar et al. (1986) however only measured triazine exposure. Across these evaluations, researchers were able to control for the influence of other pesticide exposure as well as cancer risk factors in the statistical analysis. Brown et al. (1990) evaluated odds of leukemia in association with pesticide use, and reported no evidence of a statistical association: ever use of either triazine (OR 1.1 (95% CI 0.8, 1.5), and atrazine use among 38 and 108 exposed cases and controls, respectively (OR 1.0 (95% CI 0.6, 1.5). In a separate evaluation, no evidence of an association was observed with multiple myeloma, although the number of atrazine exposure cases and controls were small: 12 and 74, respectively (Brown et al., 1993). Regarding an association with non-Hodgkin lymphoma (NHL), Hoar et al. (1986) reported that NHL risk, but not soft tissue sarcoma or Hodgkin’s disease, increases with duration (number of years) and frequency (days/year) of herbicide use (p for trend<0.05), and the association between triazine use (ever/never) and NHL are elevated 2-fold (OR 2.5 (95% CI 1.2, 5.4) (Hoar et al., 1986). When users of phenoxy or uracil herbicides, other highly prevalent herbicides, are excluded from the ever-triazine exposure group the association remained positively elevated, but uncertainty increased, (OR 2.2 (95% CI 0.4, 9.1)). The latter analysis is based on only 3 and 11 triazine exposed cases and controls. Cantor et al. (1992) reported some evidence of a link with farming exposures and NHL: corn growers (OR 1.4 (95% CI 0.9, 2.4), herbicide users (OR 1.3 (95% CI 1.0, 1.6), and triazine users (OR 1.6 (95% CI 1.0, 2.5) (Cantor et al., 1992). However, the risk of NHL in association with ever-use of atrazine is only weakly suggested 20% increased odds of NHL (95% CI 0.9, 1.8), adjustment for use of other pesticides did not alter estimates. Combining datasets from all three of the NCI sponsored, Hoar-Zahm 1993 reported a significant association between atrazine use and NHL (OR 1.4 (95% CI 1.1, 1.8); however, Page 54 of 184 in contrast to Cantor et al. (1992), adjustment for the use of other pesticides, organophosphates and 2,4-dichlorophenoxyacetic acid (2,4-D) in particular, diminished this estimate considerably (Hoar Zahm et al., 1993).Notably, there was no appreciable difference in the risk of NHL among ever-users of atrazine and those who reported personally handling atrazine in this analysis. When evaluation the same association among women only, authors observed little evidence of an association, although the number of atrazine exposed female NHL cases was limited (Zahm et al., 1993). Overall, these authors suggest their evaluation presented little evidence of an association between atrazine use and NHL. Strengths of the studies include use of low likelihood of selection bias, high quality outcome ascertainment including histopathological confirmation, performance of exposure validation study (using pesticide sales records), and ability to adjust for other pesticides in the atrazine-NHL association. Limitations include small numbers of exposed cases in some instances, the hypothesis-generating nature of these investigations, and probable exposure-measurement error each of which may lead to uncertainties in the risk estimates reported. EPA notes other details are noted in Appendix D. The question as to the role of co-exposure to other pesticides when measuring the association between atrazine, or any other pesticide, and NHL was evaluated in a pooled analysis using these three population-based case control studies (IA/MN, NE, KS). DeRoos et al. (2003) estimated the association between atrazine exposure and NHL, as well as several other pesticides, in an analysis which concurrently controlled for the influence of all (n=47) other pesticide exposure using sophisticated analytic techniques. Hierarchical regression techniques use available information (pesticide class, functional group, carcinogenic classification) to estimate a “prior” distribution of the pesticide-NHL risk relations, and adjust observed associations (results of pooled analysis) toward the “prior” distribution (De Roos et al., 2003). In this way, authors used information external to the studies themselves to inform the shape of the exposure-response curve. Hierarchical regression can add precision and accuracy to risk estimates, however precision and accuracy are only enhanced relative to the quality of the “prior” distribution. Researchers reported statistically-significant 50% increased odds of NHL in association with exposure to atrazine (hierarchical regression adjusted odds ratio (OR) 1.5 (95% confidence interval (95% CI), 1.0 to 2.2)), adjusting for exposure to all (n=47) other pesticides (De Roos et al., 2003). It is notable that this result, which is essentially a weighted average of the three NCI case control studies discussed above (Hoar et al. 1986; Cantor et al. 1992; Zahm et al. 1993), is closer to 1.0 than the adjustments performed in the logistic model presented in the pooled analysis of the same question performed by Zahm et al. (1993). The analytic methods in DeRoos et al. (2003) including the ability to mutually adjust for many other pesticides may explain the different risk estimate. Overall, DeRoos et al. (2003) cautioned that these data only provide limited evidence of any specific pesticidePage 55 of 184 NHL associations as results of many different statistical tests, the role of chance may explain these (positive) findings. EPA also identified epidemiology studies external to the NCI series of case-control studies, and reported its interpretation of these data in the earlier evaluation of this database. Burmeister et al. (1990) studied the relation between triazine exposure and multiple myeloma and reported a non-significant 30% increased odds of exposure among those with MM (Burmeister, 1990). In addition, in a correlation analysis using county level pesticide use data (CA DPR), i.e., ecologic exposure assessment, and state cancer registry statistics, Mills et al. (1998) reported a moderate correlation between leukemia and atrazine use (P. Mills, 1998) (See Appendix D). In addition, an ecologic study comparing all cancers reported to a regional registry and triazine usage, authors reported no evidence of a significant correlation with NHL (J. Van Leeuwen, D. Waltner-Toews, T. Abernathy, B. Smit, & M. Shoukri, 1999). Mills et al. (1998) in the same correlation analysis as noted above, reported no evidence of a correlation between pesticide use and NHL among most subpopulations examined, with the exception of Hispanic females. These authors reported a non-significant positive correlation with NHL among this group. It was also noted in previous EPA reviews that an occupational cohort study reported elevated standardized mortality rates (SMRs) for NHL (n=2 cases) among triazine manufacturing plant workers in comparison to the general population of the area (P. A. MacLennan et al., 2003; P. A. MacLennan et al., 2002). These analyses updated comparisons made between manufacturing workers occupationally exposed to triazines and the general public published in the mid-1990’s and also reported elevated rates of NHL among workers (Sathiakumar & Delzell, 1997; N. Sathiakumar et al., 1996). Overall, these studies offer weak evidence to our consideration of this question, and have been superseded by more recent investigations. 3.2.4.2 Hospital-Based Case-Control Studies (France) Clavel et al. (1996) performed a hospital-based case control study of hairy-cell leukemia (HCL) and pesticide use, including triazines. Authors identified cases through oncology departments of 18 regional hospitals, and control patients through other hospital departments. Participation rates were modest, however participants were blinded as to the study hypothesis, therefore it is unlikely participants declined to participate due to pesticide usage. Pesticide exposure was assessed via self-administered questionnaire distributed via mail, verified by an industrial hygiene expert in pesticide use who also classified exposure intensity based upon information provided in the questionnaire. Authors reported elevated odds of hairy cell leukemia for all farmers (OR 1.5, lower limit 1.0), corn growers (OR 3.2, 95% CI 1.6, 6.2), ever use of herbicides (OR 1.8, 95% CI 1.0, 3.1), and ever use of triazines (OR 2.4, 95% CI 1.2, 3.5) (Clavel et al., 1996). However, when authors examined the association among never users of organophosphates only, the association with triazines was attenuated and becomes statistically non-significant OR 2.0 (95% CI 0.7, 5.6). The correlation between organohphosphate and triazines is high in this study population. Page 56 of 184 In a study with similar design and methods and study population, performed between 200004, Orsi et al. (2009) evaluated the association between farming including specific pesticides, and lymphoid neoplasms. Outcomes examined included non-Hodgkin’s lymphoma (NHL), multiple myeloma (MM), Hodgkin’s lymphoma (HL), and lymphoid proliferative syndrome (LPS). Cancer cases were identified through medical record evaluation in oncology centers at six hospitals in France; and, controls were identified from the patients who attended different departments of the same hospitals (e.g., rheumatology, orthopedics) unrelated to cancer treatment. Pesticide exposure information was collected first by self-administered questionnaire, followed by a structured in-person questionnaire administered by trained staff. Industrial hygienists with expertise in pesticide use evaluated participant responses. Authors modeled both ever-use of pesticides, as well as total number of years employed in agriculture as a measure of duration of exposure. Authors reported associations between employment in farming, and corn farming specifically in addition to ever use of specific pesticides in association with NHL, HL, and MM. For reported triazine use specifically, associations were observed with HCL (OR 5.1, 95% CI 1.4-19.3) (Orsi et al., 2009). In addition, triazine use was associated with NHL (diffuse large cell lymphoma 8/20 OR 2.1 [95% CI 0.8–5.0]) and follicular lymphoma (FL) 4/20, OR 2.3 [95% CI, 0.7–7.7]), but not with chronic lymphocytic leukemia 4/17 OR 0.9 [95% CI 0.3–3.0] (Orsi et al., 2009). Small sample size, particularly low numbers of triazine exposed cases of lymphoma sub-types resulted in imprecise measures of the association. The strong magnitude of the association suggests additional studies employing a larger sample may result in more stable, positive associations; however, the width of the confidence limits which include 1.0 in each instances suggests results could be due to chance. The statistically significant association with hairy-cell leukemia is noted; however, this confidence interval is also quite wide. Authors suggest an association between triazine use and lymphohematopoietic cancers may be present, specifically with HCL, NHL and FL. Important to this effort, EPA notes triazines and not atrazine per se was measured. Strengths of these studies include the performance of several sensitivity analyses to reduce some sources of uncertainty, and use of two levels of pesticide exposure data collection by trained, blinded interviewers. However, the relatively low statistical power and relatively wide confidence bounds suggest lack of precision in the estimates and an inability to exclude chance as an explanation for these findings. In addition, as acknowledged by the authors, systematic bias could not be entirely ruled out with respect to representativeness of the control series of the population from which cases arose. Exposure assessment reflected ever-use over a lifetime and disallowed more refined analysis for risk assessment. Overall, the potential for bias in terms of selection and exposure measurement error, the highly correlated nature of multiple pesticide use and the inability to control for this in the Orsi et al. (2009) study, among other critiques noted in the appendices weaken these results. Page 57 of 184 3.2.4.3 AHS Estimates of Atrazine and Lymphohematopoietic Cancers Within the AHS, Rusiecki et al. (2004) performed a cohort analysis of cancer (all) in association with lifetime use of atrazine. While authors examined the link between atrazine use and several cancer outcomes, the results for NHL and MM are discussed here only. A brief discussion of AHS methods and design as well as strengths and weaknesses is presented above in Section 3.2.3.1. Using a measure of both cumulative exposure (total number of days per working lifetime, lifetime exposure days (LTED)) and also intensity-weighted cumulative exposure (intensityweighted LTED, IWED), exposure modified by use of personal protective equipment (PPE) and other use practices, researchers reported evidence of a positive, non-statistically significant associations with NHL (OR(IWED) 1.75 (95% CI 0.73 to 4.20, n=20), as well as with multiple myeloma (OR 2.17 (95% CI 0.45 to 10.32) (same comparison) comparing the upper-quartile of atrazine exposure (175 exposure days) with the lowest exposed group (120 exposure days) (J. A. Rusiecki et al., 2004). A suggestion of a positive trend across quartiles of atrazine exposure was noted, albeit the test for trend was not significant (p>0.10). No evidence of a statistical association between atrazine exposure and leukemia was observed. In this analysis, authors controlled for age, sex, alcohol consumption, residence on a farm, smoking status, educational level, family history of cancer, state of residence, and use of 10 most highly correlated pesticides with atrazine (J. A. Rusiecki et al., 2004). Recently, Beane-Freeman et al. (2011) updated this analysis within the AHS with approximately twice as many cancer cases. Adjusting for standard confounding variables as well as other pesticide use authors observed no evidence of an association with any lymphohematopoietic cancer (Beane Freeman et al., 2011). Specifically, authors reported no association with NHL or any NHL sub-type, or with leukemia or multiple myeloma. Regarding NHL specifically, the updated analysis included 152 cases while the initial evaluation included 68 cases. The increased statistical stability produced by the larger sample size appears to have clarified the nature of the association between atrazine and NHL in the AHS cohort of private and commercial pesticide applicators. Therefore, while the result of an initial, hypothesis-generating study in the AHS reported suggestive, positive associations with NHL and MM, due to elevated risks in the upperquartile, and the suggestion, albeit not statistically significant, of an exposure-response trend, recent re-evaluation of the data did not replicate this initial finding. The recent evaluation with twice as many cases of non-solid tumors report no evidence of an association with atrazine use within the AHS cohort. The increased sample size may have reduced the role of random variability in the estimates (i.e., wide confidence intervals), and the use of an updated AHS exposure algorithm may have reduced exposure measurement error in such a way as to move the estimate toward the null values. Positive associations initially observed may also be the result of multiple statistical tests; however, earlier studies (discussed above) suggest a possible link with NHL. Page 58 of 184 3.2.4.4 Conclusions: Lymphohematopoietic Cancers Several investigators have evaluated the association between pesticide exposure including atrazine/triazine exposure specifically and lymphohematopoietic cancers since the early 1990’s. Studies published prior to the 2003 EPA review provided little evidence of a link between leukemia or multiple myeloma and atrazine use (Brown et al., 1990; Brown et al., 1993; Burmeister, 1990), however the small sample size and role of random variability in addition to non-differential exposure measurement error may have influenced these results. The recent AHS study with enhanced design, lifetime exposure measurement, outcome ascertainment, and increased sample size, reported risk estimates for leukemia and multiple myeloma that did not differ from the null (Beane Freeman et al., 2011). The latter study reduces uncertainty around this association, and is supportive of the conclusion of no association among occupationally exposed applicators. The results of both individual case-control studies and pooled analyses based upon these studies provided some evidence of an association between atrazine and NHL (Cantor et al., 1992; De Roos et al., 2003; Hoar Zahm et al., 1993; Zahm et al., 1993), however the authors themselves cautioned against considering specific pesticide-NHL associations reported as evidence of a causal link. For example, NHL risk was estimated in association with numerous pesticides, including 33 herbicides, in these studies; no adjustment for multiple statistical tests was performed. However, these studies provided adequate control for systematic biases such as selection bias (comparability of controls to population from which cases arose), recall bias (participants blinded to study hypothesis), and some control for confounding due to use of other pesticides. The hospital-based case control studies performed in France during two separate study periods present some evidence of a positive link, however the potential for systematic error in the conduct of these studies including selection bias due to method of control identification, weakens the strength of this evidence in the overall assessment of the epidemiologic database (Clavel et al., 1996; Orsi et al., 2009). In 2004, the initial AHS evaluation (Rusiecki et al. 2004) suggested a positive, elevated risk of multiple myeloma, but neither the risk estimates nor the test for trend was statistically significant although a positive exposure response trend was suggested in the data (J. A. Rusiecki et al., 2004). However, this finding was not replicated in subsequent AHS cohort analysis (Beane Freeman et al., 2011). These authors observed no evidence of a statistical association between atrazine use and leukemia, multiple myeloma, NHL and NHL subtypes, although there were fewer than 5 exposed cases per exposure category in some of the analyses. Differences in study design, i.e., population-based case-control and cohort, and exposure assessment methods between the earlier studies and the recent AHS study, as well as difference in sample size and the number of exposed cases between the earlier and later AHS cohort analysis, may explain these findings. Overall, the database lacks evidence of an association between atrazine and triazine exposure and these lymphoma and leukemia sub-types, including NHL. Page 59 of 184 3.2.5 Other Cancer Sites In addition to studies of atrazine’s potential influence on lymphohematopoietic cancers or reproductive and endocrine system tumors noted above, EPA has identified additional investigations of a potential link between atrazine with other cancer sites in both adults and children. These studies reflect atrazine’s potential influence upon tumors in the brain (glioma) and digestive tract including colon. Other cancer endpoints evaluated within AHS studies, but not yet discussed in the chapter, are also summarized herein to aid completeness of the data presentation. These studies are briefly mentioned herein. 3.2.5.1 Brain/Glioma EPA identified a hypothesis-generating evaluation of the relation between farm exposures, including atrazine, and brain cancer. Results were reported separately for men (Ruder et al., 2004) and women (Carreon et al., 2005). The study was performed in the upper Midwest U.S. by a National Institute of Occupational Safety and Health (NIOSH) research team. Authors performed a population-based case control investigation of the association between pesticide exposure and gliomas, a type of brain cancers, among persons living in the upper mid-western states of Iowa, Michigan, Minnesota, and Wisconsin. In the study among women, cases were ascertained (1995-97) through interactions with medical offices or neurosurgeons, and were histopathologically confirmed. Proxy respondents were used in instances in which the case was deceased. Controls were non-cases of glioma. Controls may have had other cancers, and were identified through either state DMV (ages 18-65) or state health care files (ages 65-80). Pesticide exposure as well as other demographic information was obtained through self-reported questionnaire; trained industrial hygienists reviewed written responses and compared reported pesticide trade names with database to determine active ingredients used. Data collection included comparability of methods between cases and controls including a standardized interview, intermittent re-training of interviewers within the study period, and blinding of interviewers to the study hypothesis. Among 341 female cases and 527 controls, authors reported little evidence of an association between triazine or atrazine use and gliomas. The results include: ever-use of an herbicide, OR 1.0 (95% CI 0.6, 1.5); triazine OR 1.0 (95% CI 0.6, 1.7); and atrazine OR 1.1 (95% CI 0.7, 1.8) (Carreon et al., 2005). Results for men were similarly unremarkable. Authors observed no association between glioma and triazine use (OR 0.9, 95% CI 0.6, 1.2), and they did not separately estimate glioma odds in relation to atrazine use (Ruder et al., 2004). Although both evaluations were derived from the same study, Carreon et al. (2005) provided more extensive detail concerning study design and methods. For example, details concerning case and control participation rates were not provided in the study by Ruder et al. (2004). However, in both evaluations the risk estimate for atrazine excluding proxy respondents was similar. Page 60 of 184 Authors performed a hypothesis-generating evaluation of pesticide exposure in relation to glioma diagnosis. There are many strengths to this study including obtaining information from those who refused to participate in the study to compare characteristics and assess potential systematic bias; use of a 2-year lag in exposure through design to assure exposure among cases is not over-estimated; moderately high recruitment rate; and, use of several quality assurance and control methods including standardized interview techniques, as well as interviewer training and blinding to study hypothesis to reduce observer bias. However, there are some limitations as well including the method of selection of cases and controls and the possibly under-ascertainment of cases, possible heterogeneity in outcome classification, and quality of exposure measurement. Overall, the uncertainties presented in this study make it difficult to discern whether the lack of an observed association reflects the true population risk of glioma in relation to pesticide exposure, or reflects errors in study design, i.e., selection bias or exposure measurement error as noted by the study authors. Other than results from the NIOSH studies, EPA identified two ecologic evaluations of the possible link between atrazine exposure and brain cancer. Mills et al. (1998) suggested a moderate correlation between this exposure and disease, while Van Leeuwen et al. (1999) reported no significant association (P. Mills, 1998; J. A. Van Leeuwen et al., 1999). Overall, the NIOSH studies provide insufficient evidence to determine whether an association between atrazine or pesticide exposure and glioma exists. EPA notes that the result evaluation within the AHS does not report a significant, positive association with any brain cancer (Beane Freeman et al., 2011). 3.2.5.2 Pediatric Cancers In an ecologic analysis, Thorpe et al. (2005) compared concentration of atrazine and other environmental contaminants in residential drinking water, in Maryland with state cancer registry data on childhood cancers to estimate risk of childhood cancers in association with contaminants. Several pediatric cancer sites were evaluated. Authors reported nonsignificantly elevated odds of bone cancer and leukemia among 689 pediatric cancers reported during the study period. There are few studies of pediatric cancers in association with pesticide use, and fewer in relation to atrazine use (Thorpe & Shirmohammadi, 2005). Therefore this study contributes to our understanding of the potential relation. Moderately elevated risk estimates were observed in this ecologic analysis, albeit non-statistically significant. However, in addition to the standard limitations of ecologic studies for use in causal inference, this study assumed all participants used well water for drinking water, did not address the uncertainty presented by residential mobility during pregnancy which may have also contributed to childhood cancers, and could not provide much information as to whether prenatal or postnatal exposures may be more strongly associated with pediatric cancers. Significant uncertainties are present in this study, which limits its use in determining whether a true link with atrazine exists. Page 61 of 184 Rull et al. (2009) performed a population-based case control study of pediatric acute lymphocytic leukemia (ALL) within the Northern California Childhood Leukemia Study (1995-2002). Maternal residence at birth was compared with state pesticide use records (PUR) to estimate pesticide exposure. Authors considered exposure to 118 active ingredients for which there was some evidence of cancer, developmental or reproductive health effects, or evidence of endocrine disruption (adjudicated by authors); or, frequency of use was high. Using proximity of maternal home as a measure of exposure, authors evaluated pesticide exposure over the child’s lifetime as well as during the first year of life only. Authors reported borderline significant associations between those in the moderately exposed category of triazines (OR 1.9 (95% CI 1.0, 3.7), but not among those in the most highly exposed category (R. P. Rull et al., 2009). Adjusting for exposure to other classes of pesticides, the magnitude of the association in the moderately exposed category increased, but the precision decreased (OR 4.1, (95% CI 1.5, 11.1)) (R. Rull et al., 2009). Associations were also observed for herbicide categories, but results were not statistically significant (moderately exposed: OR 1.2 (95% CI 0.8, 1.9), and highly exposed: OR 0.9 (95% CI 0.5, 1.5)) for lifetime exposure to herbicides. There was no association between pediatric triazine exposure during the first year of life and ALL risk in this study. Triazines to which this study population was exposed included atrazine, cyanazine, prometryn, pymetrozine, and simazine. Authors performed a hypothesis-generating, population based case-control study of pediatric ALL in relation to pesticide exposure. Strengths of the study include method of case ascertainment, within 72 hours of diagnosis; the recruitment rate was high (84%); and the cases and controls were similar on several (matching) factors, reducing the potential for selection bias and residual confounding. The major limitation of this study is the method of exposure assessment, i.e., using maternal proximity to farm-site as a proxy for exposure, and the lack of information on critical windows of exposure for development of a pediatric cancer, and lack of precision of estimation of triazine with acute lymphocytic leukemia. Overall, given both the noted limitations of this study as well as the lack of statistical evidence of an association, the study is not strongly supportive of a link with atrazine. Although not triazine or atrazine specific, within the AHS, authors evaluated childhood risk of cancer among children of parents enrolled in the AHS and reported an overall slight excess risk (OR 1.36, 95% CI 1.03, 1.79) (Flower et al., 2004). When examining specific cancers, risk of lymphoma was overall elevated among children of AHS families. Authors observed few pesticide-specific associations; but statistical power may have limited the ability to detect associations. Authors suggested non-pesticide farm exposures may explain these overall findings, such as Epstein-Barr virus. At this time the database is insufficient to determine whether an association exists between atrazine or triazine exposure and pediatric cancers. Importantly for pediatric cancers, Page 62 of 184 neither of these initial evaluations measured pesticide exposure during critical windows of development, including maternal exposures. The ecologic study is not of sufficient quality to support the hypothesis of a relation between atrazine and pediatric cancers. The lack of an exposure-response trend by proximity to field, and imprecision of risk estimate due to small number of cases in upper exposure category as well as some degree of non-differential exposure misclassification, among other factors, weakens the ability to draw a conclusion based upon these data. 3.2.5.3 Colon Cancer Hoar et al. (1985) published a brief report of a case control analysis of pesticide use and colon cancer. Authors selected cases diagnosed with colon cancer in Kansas during 197682, and population-based controls. As a brief report, limited detail concerning data collection procedures or statistical methods used to evaluate the association between colon cancer and exposure to herbicides was provided. Hoar et al. (1985) reported that employment on a farm was associated with a non-significant increase in colon cancer risk (OR = 1.6, 95% CI: 0.8 – 3.5), which was similar to the risk in farmers who did not report using herbicides, and reported use of triazines was not associated with colon cancer (OR 1.4, 95% CI 0.2, 7.9) (Hoar, Blair, Holmes, Boysen, & Robel, 1985). With regard to selfreported use of herbicides, the investigators reported that the risk of colon cancer among farmers who reported mixing herbicides was similar to the risk in farmers who never reported using herbicides and there was a lack of dose-response relationship between years of herbicide use and colon cancer risk. Based on these reported findings, the investigators suggest that the study results do not support an association between colon cancer and herbicide exposure including triazines. In addition to this estimate, previously mentioned analyses in which cancer rates were compared between a cohort of triazine manufacturing plant workers and the general population did not report an elevated rate of either colon or rectal cancer incidence or mortality (P. MacLennan, E. Delzell, N. Sathiakumar, & S. Myers, 2003; P. MacLennan et al., 2002), and an ecologic analysis did not observe a link between atrazine and colon cancer (J. A. Van Leeuwen et al., 1999). Lee et al. (2007) reported colorectal cancer risk estimates for several pesticides including atrazine among 305 incident cases of colorectal cancer and 56, 813 non-colorectal cancer cases. Overall, authors observed no evidence of an association between atrazine use and colorectal cancer (OR 0.9 (95% CI 0.7, 1.2)), or colon cancer (OR 0.8 ( 95% CI 0.6, 1.1)), or cancer of the rectum (OR 1.2 (95% CI 0.7, 2.0)) (Lee et al., 2007). In this investigation, authors only reported the atrazine ever-use exposure metric, as no evidence of an atrazine exposure-response trend was observed. Recent cohort analyses within the AHS did not observe evidence of an association between lifetime atrazine use and either colon or rectal cancer (Beane Freeman et al., 2011; J. A. Rusiecki et al., 2004). The magnitude and direction of the risk estimates were suggestive of no association, and among the studies with sufficient sample size to evaluate an exposure-response trend, Page 63 of 184 none was observed. While the database is a small one, the more recent evaluations within the AHS have several strengths over previously published studies include the prospective design, larger sample size, and measurement of other potentially confounding variables including use of other pesticides. Among a nested case-control study and two cohort analyses which estimated colorectal cancer risk in association with atrazine, no evidence of an association was observed. Authors did not propose a mechanism through which atrazine could influence colorectal cancer risk; animal studies did not observed tumors in these sites (See Section 3.3). Overall, this database lacks evidence of an association. 3.2.5.4 Other AHS In the interest of providing a complete delineation of the atrazine or triazine and cancer point estimates reported in the AHS, EPA notes the following additional results which have not yet been mentioned in cancer specific sections above. Neither the earlier nor the updated cohort analysis noted an association with all cancers combined, nor with several additionally evaluated cancer sites including: oral, esophagus, pancreas, melanoma, kidney, larnynx, brain or liver. Suggestive findings regarding other anatomical cancer sites noted by Rusiecki et al. (2004) in the earlier cohort analysis were not replicated in the updated study (Beane Freeman et al., 2011). Specifically, the odds of either lung or bladder cancer were not significantly different by atrazine exposure category in the more recent AHS study (Beane Freeman et al., 2011). Other anatomical cancer sites for which some statistical evidence of an association was noted, but for which no exposure-response trend was observed, or lacked an a priori hypotheses, were not considered to provide strong biological evidence of an association (esophageal and oral cancers, and inverse associations with pancreatic and laryngeal cancers). As noted above, AHS researchers also performed several nested case-control analyses since the last EPA evaluation. Among these, neither the study of lung cancer nor cutaneous melanoma included specific risk estimates for atrazine use, due to lack of an observed exposure-response association, however EPA notes that lack of an observed association is important evidence to consider in the totality of the atrazine cancer epidemiology database (Alavanja et al., 2004; Dennis et al., 2010). Andreotti et al. (2009), compared pesticide use, including atrazine, among 93 pancreatic cancer cases and 82,503 cancer-free applicator and applicator spouse controls, and reported no evidence of an association with atrazine use (OR 0.70 (95% CI 0.4-1.2)) (Andreotti et al., 2009). Due to the low numbers of atrazine-exposure pancreatic cancer cases, no exposure-response analysis was attempted. Across these AHS studies, evidence suggesting a statistical association is lacking, results did not significantly differ from unity. Page 64 of 184 3.3 Summary of Evidence Concerning the Carcinogenic Potential of Atrazine 3.3.1 Synthesis and Integration of Toxicology and Epidemiology Literature Over the past several decades, there have been numerous experimental toxicological as well as epidemiologic evaluations of the carcinogenic potential of atrazine. EPA has presented its evaluation of earlier iterations of this combined database in both 2000 and 2003 as noted in section 3.1 (Blondell & Dellarco, October, 28, 2003 2003 #1026; EPA, 2000, 2003a). In April 2010, EPA presented the most current assessment of the atrazine cancer toxicological database (EPA, 2010). In this section, EPA will consider the most current assessment of the atrazine cancer epidemiology database in the context of the available experimental data, combining the evidence from these two streams of information. EPA has extensively considered the experimental studies of the carcinogenic potential of atrazine. Briefly, in the 1986 review, EPA considered atrazine a “Group C” carcinogen, i.e., a possibly human carcinogen, based upon increased incidence of mammary gland tumors in Sprage-Dawley rats, and lack of mechanistic data to evaluate the mode of action and the relevance in the human population. Subsequently, mechanistic data were produced concerning the mode of action of atrazine which indicated that the mode of mammary tumor development in the rat was not operational in the human (EPA, 2000; SAP 2000). There were no other tumor sites noted in the animal bioassays (mice and rats) other than the mammary tumors in female rats. Additional mechanistic experimental data included laboratory-based genotoxicity and mutagenicity studies, as well as studies of atrazine’s tumor promoting potential. Study results indicated that atrazine was neither mutagenic or genotoxic, nor was there evidence of tumor promotion in the thyroid, pituitary, ovarian or breast cancer experimental systems evaluated (EPA, 2010; EPA, 2000). Studies of the tumor promotion potential of atrazine included both whole body and human cancer cell line systems. Furthermore, experimental studies published or submitted to EPA since the previous EPA review (i.e., the 2003 IRED) did not reveal additional evidence of tumor formation or promotion in animals (EPA, 2010). Based on the weight of evidence (including both empirical and mechanistic studies), atrazine exacerbates a condition to which the SD rat is normally predisposed due to an endocrine environment that is the normal cause of reproductive senescence in this strain for which there is no human counterpart. Thus, the experimental evidence does not indicate a role for atrazine in the carcinogenic process in humans. In its review of earlier iterations of the atrazine cancer epidemiology database, studies published prior to the 2003 IRED, EPA stated that the data were insufficient to conclude whether an association between atrazine and cancer was present, and that the database lacked adequate information upon which EPA could perform a cancer risk assessment (EPA, 2000, 2003a, 2003b). In this chapter, EPA has considered the results of atrazine Page 65 of 184 cancer epidemiologic evidence published between the mid-1980’s to the present, inclusive of epidemiologic data previously reviewed as well as newly published studies. Epidemiologic studies have mainly focused upon cancers of the lymphohematopoietic system (leukemia and lymphoma including NHL), the reproductive and endocrine systems (prostate, breast, ovarian and thyroid cancer), and a small number of studies considering other cancer endpoints. While there were earlier suggestions of a possible link between atrazine and lymphohematopoietic cancers in a few studies published in the early 1990’s, overall, the evidence was weak. Even among the studies which reported a suggestive, positive relation between atrazine and NHL, authors cautioned against the ability to link any one pesticide with NHL in these studies due to the number of statistical comparisons made, the modest magnitude and marginal statistical significance of the associations estimated, the attenuation of the association upon mutual adjustment for other pesticide exposure, and lack of an exposure-response trends, among other factors (De Roos et al., 2003; Hoar Zahm et al., 1993; Zahm et al., 1993). Two hospital-based case-control studies in France also offered some additional information to consider the question; however concern regarding possible systematic error in the selection of the control series, role of confounding bias, lack of an exposure-response trend and often small sample size in the evaluation of specific leukemia and lymphoma sub-types, limits use of these data (Clavel et al., 1996; Orsi et al., 2009). More recently, within the AHS, although initial cohort analysis results reported some evidence of a non-statistically significant, positive relation between lifetime atrazine use and NHL and multiple myeloma based upon suggestion of an exposure-response trend (J. A. Rusiecki et al., 2004), a follow-up study with twice the number of cancer cases did not observe an association between atrazine and any leukemia or lymphoma sub-type (Beane Freeman et al., 2011). In addition, within the experimental database, there is no evidence of an increase in lymphoma non-solid tumor formation, including NHL, when test animals are dosed with atrazine, simazine or propazine. In addition, there is no evidence in the cancer epidemiology literature that hormones or other reproductive factors are related to the development of leukemias or lymphomas (Adami, 2002). Therefore, at this time, there is no experimental evidence to suggest a role for the herbicide in the etiology of these tumors, nor do the epidemiology studies performed within agricultural, occupational populations, and demographically consisted primarily of white, males, suggest evidence of an association. Given the known neuroendocrine effects of atrazine, several investigators evaluated a potential link between atrazine exposure and cancers of the reproductive and endocrine systems including prostate, breast, ovarian as well as thyroid cancer. In 2003, EPA concluded, in consultation with the FIFRA SAP, that results of studies of triazine manufacturing plant workers which suggested an association with prostate cancer were likely due to the presence of an on-site prostate cancer screening program, the PSA test, in which most of the male employees participated during the time period of this study (EPA, 2003a, 2003b). Follow-up studies within this occupational cohort reflecting more refined Page 66 of 184 triazine exposure information did not alter this conclusion (P. A. Hessel et al., 2004). Two studies within the California cancer registry suggested a possible correlation and/or association with atrazine (P. Mills, 1998) or simazine (P. K. Mills & Yang, 2003), however the aggregate level exposure measurement and lack of risk estimate for atrazine per se in the latter study, presents uncertainty in the use of these data. An initial nested case-control studies within the AHS (Alavanja et al., 2003) observed no evidence of a relation between ever-use of atrazine and prostate cancer, and, more recently, two cohort evaluations with sufficient statistical power to detect an association between atrazine use and prostate cancer, if an association exists, did not report evidence of a link with prostate cancer, i.e., risk estimates did not differ from the null (Beane Freeman et al., 2011; J. A. Rusiecki et al., 2004). During the 2003 FIFRA SAP evaluation, the Panel suggested that differences in exposure profile, including the periodicity and intensity of exposure, between triazine manufacturing plant workers and corn and soybean pesticide applicators, precluded consideration of these data as part of a whole database (EPA, 2003b). However, EPA believes that the AHS cohort represents one of the most highly exposed occupational cohorts available for study, and lack of evidence of an association in this cohort which has adequate statistical power to detect an association, is strongly suggestive of no evidence of an association among occupationally exposed primarily white, males in the U.S. population. Furthermore, the results of chronic toxicology studies in two species (rat and dog) indicated no evidence of increased incidence of prostate tumors. As noted briefly above, there is no evidence of a role for atrazine in prostate tumor formation in either the animal bioassays, or within in vitro studies of tumor promotion, genotoxicity or mutagenicity with atrazine. There is no known mechanism for atrazine’s possible role in prostate tumor formation in humans. Considering both the epidemiologic and toxicologic studies, the current database lacks evidence of an association between atrazine and prostate cancer. The results of three ecologic epidemiology studies provided inconsistent and generally weak evidence concerning a possible association between atrazine and breast cancer (Hopenhayn-Rich et al., 2002; M. A. Kettles et al., 1997; Muir et al., 2004). One ecologic study indicated a small, but significant 20% increased odds of the disease among the exposed (Kettles et al. 1997), however a follow-up analysis using the same methods in the same study population failed to replicate the finding (Hopenhayn-Rich et al. 2002). Muir et al. (2004) found evidence of a positive association with breast cancer, but in only one of the two study sites, thereby inconsistent evidence. A study among Latinas who reported employment in agricultural work residing in California, using state cancer registry data and reported pesticide use and other self-reported data, lacked evidence in support of an association with breast cancer, however, the ecologic nature of the exposure assessment again precludes making any strong conclusions from this study (P. Mills & Yang, 2006). In a population-based case control study in a high-use area of the upper-midwest (WI), authors reported no evidence of an association with breast cancer, however the upperexposure quartiles included fewer than 5 exposed cases (J. A. McElroy et al., 2007). Within Page 67 of 184 the AHS, considering either female spouses of pesticide applicators (Engel et al., 2005) or female applicators themselves (Beane Freeman et al., 2011), no association between everuse of atrazine and breast cancer was observed. However, these authors were not able to address exposure-response trends given the relatively small number of cases included in the analysis. Strengths of this database include the reflection of several different target populations across the studies including women in the general population, although residing in high-use areas, female agricultural workers, female spouses residing on a farm or residing with a commercial pesticide applicator, and female professional applicators, as well as study designs. In addition, investigations measured both indirect, i.e., exposure through laundering atrazine contaminated clothing, as well as direct use of the herbicide. Limitations include use of aggregate exposure measurement techniques (ecologic) as well as lack of ability to investigate exposure response association, and relatively small sample of atrazine exposed females in many of the studies. The observational data are insufficient to determine whether or not a link between atrazine and breast cancer may exist, and at this time, are suggestive of no true association. Further, the initial observation of mammary tumors in early experimental animal studies (female Sprague-Dawley rats), were determined through follow-up mechanistic studies to be due to a mode of action that is not operative in humans, i.e., reproductive senescence due to high estrogen production resulting in constant estrogen stimulation of the mammary tissue. Therefore, at this time, considering the current toxicological and experimental evidence, the database lacks evidence of an association between atrazine and breast cancer. The epidemiologic database for the relation between atrazine and ovarian cancer is small, and weakly suggestive of a possible association across three studies, however the possible role of random variability, bias or confounding in the risk estimates observed cannot be excluded. A recent ecologic analysis suggested an inverse (protective) association between atrazine exposure and ovarian cancer in an area of agricultural use in the U.S. (Hopenhayn-Rich et al., 2002), and an earlier ecologic study reported no evidence of an association (J. A. Van Leeuwen et al., 1999). However, an earlier case-control study in Italy suggested a potential 1.5- to 2-fold increased odds of ovarian cancer among those “possibly” and “definitely” exposed to triazines (Donna et al., 1989). Earlier evaluation of this study by both EPA and the FIFRA SAP suggested uncontrolled confounding by other pesticide exposure may explain results. Using state cancer registry data and pesticide reporting survey information, Young et al. (2005) reported slight increased risk for total triazine users, but not for atrazine exposed women (H. Young et al., 2005). However, the prevalence of atrazine exposure was less than eight percent in this study, limiting the ability to detect an association. Recently, AHS researchers reported an elevated odds of ovarian cancer among ever-users of atrazine in a population of female pesticide applicators. Investigators were not able to evaluate an exposure-response relation, and only reflected nine ovarian cancer cases, four of whom reported atrazine use. Overall, there are weak, but somewhat suggestive observations of an association between ovarian cancer and atrazine use, however significant uncertainties Page 68 of 184 in the observational database exist. At this time, data are insufficient to inform whether an association may exist. The experimental animal studies do not strongly suggest a role for atrazine in ovarian cancer etiology, however uncertainties in both databases must be considered. Animal bioassays, as noted above, do not show increased incidence of ovarian tumor formation upon administration of atrazine, simazine or propazine (common mechanism group). The mode of action through which atrazine influences neuro-endocrine effects, i.e., suppression of the lutenizing hormone (LH) surge and reduced number of ovulatory cycles, the “incessant ovulation” hypothesis, would likely lead to a decreased, and not an increased risk of this cancer in women. Suppression of the LH surge would result in altered pituitary and steroid hormone concentration in the ovary, and reduce the mitogenic potential of estrogens in the target tissue as well as the trauma of ovulatory cycles suspected to be linked to ovarian cancer etiology. However, epidemiologic evidence against the “incessant ovulation” hypothesis exists, including the observation that both pregnancy and lactation result in increased, and not decreased, levels of estradiol and follicle-stimulating hormone (FSH), respectively, and also appear to be protective against ovarian cancer (Henderson, Ponder, & Ross, 2003). Furthermore, recent in vitro studies performed by EPA indicate that while it does not appear that atrazine is an aromatase inhibitor, atrazine does appear to increase CYP19 (the aromatase gene) messenger RNA (mRNA) synthesis, and may increase concentration of aromatase. The aromatase enzyme converts testosterone to estradiol; increasing aromatase could lead to increased concentrations of estradiol. However, an increase in aromatase has not been detected in vivo. Although there are some indications of an association with ovarian cancer and a biologically plausible role for the herbicide, although no known alternative mode of action at this time, overall considering the strengths, weaknesses and uncertainties of the limited number of studies available, the current database lacks strong evidence of a true association. Among all the epidemiology studies included in this evaluation, only one study offers a risk estimate for the association between thyroid cancer and atrazine exposure: the recent AHS cohort analysis (Beane Freeman et al., 2011). In this evaluation, authors report evidence of a statistically significant four-fold increased odds of thyroid cancer among male pesticide applicators; the estimate remains elevated and significant when adjusting for body mass index, a known risk factor for thyroid cancer. However, authors note the lack of a clear exposure-response trend, and the small number of atrazine exposed cases in some categories, e.g., n<5. These and other limitations to this study present some uncertainties in the interpretation of these results. There is no evidence of an increased incidence of thyroid tumors in experimental animal bioassays, nor does atrazine act as a tumor promoter in experimental systems, or illustrate the ability to alter thyroid hormone levels or thyroid-stimulating hormone (EPA, 2010). In the AHS, atrazine was not associated with thyroid disease (Goldner et al., 2010). Epidemiologic evidence indicates a likely link between reproductive factors including steroid and peptide hormones in the etiology of this cancer; the striking difference in the trends and pattern of thyroid cancer between men and Page 69 of 184 women support this hypothesis (Adami, 2002). However, a clear mode of action is lacking, and no toxicological evidence exists as to a mode of action for atrazine in thyroid tumor development. Thus, at this time, evidence is insufficient to determine the nature of the association between thyroid cancer and atrazine. In addition to these more commonly assessed atrazine-cancer associations, EPA identified a few other studies. Considering pediatric cancers, the evidence is currently insufficient to make a determination as to whether a relation exists. One ecologic study of the association between atrazine concentration in drinking water and all pediatric cancers reported to the Maryland state cancer registry, only provided unadjusted risk estimates as the result of simple contingency table analysis. Other methodological limitations noted by EPA in Appendix E render this study of low quality. In the AHS, parental atrazine use is weakly, non-significantly associated with childhood cancers (combined) (OR 1.27 95% CI 0.70, 2.30), but multiple statistical tests were performed (Flower et al., 2004). An evaluation of acute lymphocytic leukemia (ALL) among children in the Northern California childhood cancer cohort (R. P. Rull et al., 2009) suggested a non-significantly elevated risk of ALL with triazine exposure, however lack of exposure refinement including information on critical windows of effect in development result in uncertainties that preclude the Agency from reaching any definitive conclusions based upon this study. Maternal hormone levels have been associated with childhood leukemias in some studies (Adami, 2002); however, a potential role for atrazine in the alteration of maternal hormone and subsequent childhood leukemia risk is not available. Overall, this database is insufficient to inform whether a causal association may exist. Studies of a link between atrazine or triazine with colorectal cancer do not present compelling evidence of an association at this time, overall this database is weak. The magnitude and direction of the measured associations were not strong, exposure-response trends were lacking, and the number of cases was often small (Beane Freeman et al., 2011; Hoar et al., 1985; Lee et al., 2007; P. K. Mills & Yang, 2007; J. A. Van Leeuwen et al., 1999). Concerning glioma (brain cancer), a population-based case-control study in the upper Midwest of the U.S., suggest no association between atrazine exposure and glioma in men (Ruder et al., 2004) or women (Carreon et al., 2005), however under-ascertainment among the elderly given difficulty in diagnosing this cancer in the elderly population, and other sources of information bias may have influenced this result. EPA notes the recent AHS cohort study observed no link between atrazine use and brain cancer among occupationally exposed men and women (Beane Freeman et al., 2011). Overall, the studies of an association between either digestive system tumors, or brain gliomas are limited in number, and reflect uncertainties that hinder use in making causal inference. At this time, the database lacks evidence of a link between atrazine and these types of cancers. To be exhaustive in the consideration of all the available evidence, EPA considered all available evidence presented in other AHS studies which have estimated the association Page 70 of 184 between atrazine exposure and several other anatomical cancer sites not yet discussed. Overall, no additional evidence of an association was identified across these atrazinecancer risk estimates. Specifically, these include: lung, pancreatic, cutaneous melanoma, oral, esophageal, kidney, larynx, and liver cancer (Beane Freeman et al., 2011; J. Rusiecki et al., 2004). As already stated, results of animal bioassays did not indicate increased incidence of any tumors, other than mammary which is discussed above. 3.3.2 EPA Conclusion As a result of extensive investigations within both experimental and observational studies, the cancer database does not strongly suggest a link between atrazine exposure and increased cancer incidences in the human population. Recent cohort analyses among an occupationally exposed population of pesticide applicators (AHS) indicated no evidence of an association with leukemias, or lymphomas including several sub-types of lymphoma. Researchers observed no increased incidence in these tumors in experimental studies. The AHS reports were also strongly supportive of the conclusion of no association between atrazine and prostate cancer, as a result of studies with sufficient statistical power (1β>0.85) to detect an association, no other peer-reviewed papers were identified since the last EPA evaluation. The database for breast cancer is small, and, at this time, is not strongly supportive of a link with atrazine given limitations of study design, sample size, and lack of exposure-response information. EPA determined that the mode of action through with atrazine influences an increased incidence of mammary gland tumors in animals is not operative in humans. Indications of a role of atrazine in ovarian cancer in epidemiology evaluations is based on a small number of studies, some of which reported positive association (Beane Freeman et al., 2011; Donna et al., 1984; Donna et al., 1989; H. A. Young, P. K. Mills, D. G. Riordan, & R. D. Cress, 2005). However, available information concerning atrazine’s known mode of action, decrease LH surge and exposure to estrogen/ovulation, suggest a relation is not likely. There is only one study available which estimated an association between atrazine and thyroid cancer. While statistical results are somewhat suggestive of an association, the exposure-response trend was not monotonic, based on a small number of exposed cases, and not yet replicated in other populations. Other atrazine cancer epidemiology studies were not suggestive of an association including cancers of the digestive tract (colorectal, pancreas), brain, and lung, pancreatic, cutaneous melanoma, oral, esophageal, kidney, larynx, and liver cancer. The data regarding pediatric cancers are insufficient to draw a conclusion. As noted several time above, there were no other animal tumors identified in the atrazine bioassays submitted to EPA. While some epidemiology studies are weakly suggestive of an association between atrazine exposure and cancer incidence in the human population, the preponderance of the different lines of evidence – including both experimental and epidemiology studies – does not substantiate a role of atrazine on human carcinogenesis. Thus, the weight of the evidence supports that atrazine is not likely to be carcinogenic in the human population. Page 71 of 184 4. PROPOSED UPDATES TO THE DOSE-RESPONSE ASSESSMENT 4.1 Benchmark Dose Analysis Detailed information about the BMD analysis summarized below can be found in Appendix C. Numerous scientific peer review panels over the last decade have supported the Agency’s application of the BMD approach as a scientifically supportable method for deriving points of departures (PoDs) in human health risk assessment, and as an improvement over the historically applied approach of using no-observed-adverse-effect levels (NOAELs) or lowest-observed-adverse-effect-levels (LOAELs). The NOAEL/LOAEL approach does not account for the variability and uncertainty in the experimental results, which are due to characteristics of the study design, such as dose selection, dose spacing, and sample size. With the BMD approach, all the dose response data are used to derive a PoD. Moreover, the response level used for setting regulatory limits can vary based on the chemical and/or type of toxic effect and for a given chemical, all available studies can be evaluated using a common response level which aids interpretation (USEPA, 2000). In recent months, the Agency pursued performing BMD analysis on a variety of endpoints (attenuation of LH, prostatitis, delayed PPS, delayed vaginal opening). However, after an initial scoping exercise, it became clear that the LH data were the only datasets with sufficient dose-response data to support empirical modeling. For the other endpoints, toxicological effects have only been observed at 1) only at very high doses; 2) only at single dose levels; or 3) exhibit flat dose-response curves. As such, the only BMD analyses provided here and in Appendix C relate to LH attenuation. Although, it would be preferred to model additional endpoints the other outcomes reported following atrazine exposure are less sensitive than the LH data. As such, the Agency is focusing on the data which are most sensitive and thus most protective of human health. Generally, EPA calculates both the BMD (central estimate) and the BMDL (the BMDL corresponds to the 95% lower bound on dose). As a matter of science policy, EPA uses the BMDL as the point of departure. A key consideration when performing BMD modeling is determining the response level—known as the benchmark response (BMR). In the case of continuous endpoints, like LH attenuation, the BMR most often represents an X% change from background levels (or untreated controls). Typically, the BMR is selected on the basis of a combination of biological (mode of action, quantitative link between key events, historical/concurrent controls) and statistical considerations (sample size, variability, etc). However, in the absence of information concerning the level of response (or % change) associated with an adverse effect, the Agency’s draft BMD guidance suggests that the BMD and BMDL corresponding to a change in the mean response equal to one standard deviation from the control mean be used as the BMR. In the case of atrazine, the level of attenuation of the LH surge considered to be adverse is a function of Page 72 of 184 several factors including but not limited to the life-stage and functional outcomes under consideration (e.g., estrous cyclicity disruptions in rats). Moreover, the differences in reproductive cycles/aging between rodents and humans add an additional level of complexity to establishing a specific BMR value. These differences notwithstanding, in general, the reproductive physiology among mammals is remarkably similar making the use of the LH surge as a critical effect of concern appropriate. This perturbation of the LH surge is the cornerstone of the cascade of events leading to the adverse reproductive outcomes attributed to atrazine exposure. Activation of the HPG axis resulting in the pulsatile secretion of GnRH and LH is critical to puberty onset. For instance, decreased LH during puberty would lead to insufficient stimulation of the gonads to reduce the circulating hormone levels needed for development of sex accessory tissues in males and females. Moreover, researchers have found that disruption of GnRH release and the ensuing dampening of the LH surge can lead to delays in vaginal opening and preputial separation. Based on the available data, attenuation of the LH surge continues to be the most sensitive endpoint in the developmental/reproductive toxicity profile for atrazine. Attenuation of the LH surge occurs at doses that are either comparable to or ≈ 4- to 30-fold lower than those eliciting other developmental/reproductive effects (e.g., delayed ossification, estrous cyclicity disruptions, delays in sexual maturation, and prostatitis). Taking into account these challenges in establishing a response level, a BMR of one standard deviation from the control mean was chosen as the suggested choice by the EPA BMDS 2000 guidance whenever the biologically significant degree of change in a response is not clearly defined. There are various options for selection of the appropriate dose metric for atrazine including administered dose and average daily steady state AUC for plasma triazines (i.e., atrazine and its metabolites). The average daily steady state AUC dose metric may be an appropriate internal dose metric to better characterize the link between atrazine exposure and attenuation of the pre-ovulatory LH surge since it considers the contributions of all the relevant triazine species. BMD modeling was conducted using the administered atrazine dose for each of the studies identified below and using the corresponding AUC metric for the study selected as the most robust for deriving a point of departure (4-day exposure data from Cooper et al., 2010). AUC estimates were derived as indicated below using the corresponding linear expression that relates this dose metric to administered doses of atrazine. Y = 13.0x -8.57 Where Y is the daily steady state estimate and x is the administered dose of atrazine 4.1.1 Identification of Critical Studies Several oral toxicity studies in the rat provided LH data that were considered for doseresponse modeling. These include Morseth et. al. (1996a & b), Minnema (2001; 2002) and Cooper et al. (2010). Recent data from the NHEERL provided the most robust LH data in terms of dose selection (number of dose levels - particularly low dose range - and the Page 73 of 184 spacing between dose levels) and variability of the data. The study design addressed the low-dose region of the dose-response curve and exhibited less data variability (i.e., smaller standard deviations). A one month oral study by Morseth et al . (1996a) established a LOAEL of 40 mg/kg/day and a NOAEL of 5 mg/kg/day for an effect on LH surge. A limitation of the study is the large variability of the data (as reflected by standard deviations). A subsequent 6 month study (Morseth et al . 1996b) established a LOAEL of 3.65 mg/kg/day for attenuation of LH surge; the NOAEL was 1.8 mg/kg/day. The data from this study also indicated large variability. The NOAEL from this study was the basis for the point of departure used to establish the chronic reference dose (cRfD) in the EPA’s 2003 human health risk assessment for atrazine. Minnema (2001) established a LOAEL of 40 mg/kg/day for attenuation of LH surge in rats treated with atrazine for 1 month. The NOAEL was 5 mg/kg/day. These data also displayed variability. In a follow-up 6 month dietary study (Minnema 2002), atrazine (1.8, 3.4, 4.9, and 29.1 mg/kg/day) was without effect on LH surge in ovariectomized female SD rats. Confidence in this study is low; BMD/BMDL values were not calculated for Minnena (2002). Based on historical control data, the induction of LH surge in control animals seems to be suboptimal. Rats exhibited an infection that may have impacted LH surge response. In addition, estradiol data were not submitted to evaluate LH induction response. In a recent study by Coder (2011), a decreased estradiol induced LH surge was observed in female SD rats following oral gavage administration of atrazine (0. 6.5, 50, or 100 mg/kgday) for 4 days. However, no statistically acceptable BMD/BMDL values were obtained with the LH surge data from this study. The Coder (2011) study is limited by the poor selection of doses in the low dose region of the response curve. 4.1.2 Methods & Results EPA’s Benchmark Dose Software (BMDS) version 2.1.2 was utilized to estimate BMD estimates from the available LH studies. All models pertinent for continuous data were evaluated including exponential, Hill, power, polynomial and linear. Criteria used to assess the best fit included statistical (goodness-of-fit) values, model criteria (Akaike Information Criteria; AIC), BMD/BMDL ratios, and visual inspection. The most reliable model fits and BMD and BMDL estimates were obtained from the 4-day data from Cooper et al. (2010). In contrast to the NHEERL dataset, less reliable model fits and failure of some models to compute BMD/BMDL values were encountered with the other available studies that had more data variability and less well-defined dose selection. There is uncertainty in the BMD/BMDL estimates from these other studies. The table below Page 74 of 184 summarizes the results of BMD analyses of the available LH studies and the level of confidence in the estimates. Detailed results are provided in Appendix C. Table 1: Summary of Administered Dose BMD analyses LH Studies BMD BMDL NHEERL 4 day (oral gavage) 4.92 2.42 Minnema 1 Month (oral gavage) 653.3 245.1 Morseth 1 Month (oral gavage) 523.7 174.7 Morseth 6 Month (dietary) 36.2 17.1 4.30 3.40 Coder 4 day (oral gavage) Confidence High – robust dataset (less data variability; well-defined dose selection); describes LH response very well. Low – data variability; both BMD and BMDL estimates outside experimental range; large BMD/BMDL ratio; failure of some exponential models to compute estimates. Low – data variability; BMD estimate outside experimental range; large BMD/BMDL ratio; failure of some exponential models to compute estimates. Low – data variability;BMD estimate outside experimental range; failure of some exponential models to compute estimates. Low – No statistically acceptable BMD/BMDL values For the Cooper et al. (2010) 4-day data, EPA calculated the BMD/BMDL for the administered dose and internal dose metrics. When administered dose was used as the dose metric, the exponential model (subset model 4) provided the best fit to the data, and based on statistical (goodness-of-fit values), modeling criteria (Akaike Information Criteria; AIC values), and visual inspection, BMD and BMDL estimates were 4.92 and 2.42 mg/kg/day, respectively (see Cooper et al. 2010) BMD output and summaries provided in Appendix C). The BMD and BMDL estimates described the LH surge data very well. BMD analyses were also carried out with the corresponding average steady state plasma levels and average daily steady state AUC. The Hill model provided the best fits for both Page 75 of 184 dose metrics. The following table summarizes BMD and BMDL estimates corresponding to administered dose and AUC as the dose metrics. Table 2: BMD Modeling Results with average steady state plasma triazines and AUC estimates BMD and BMDL estimates (BMR = 1SD from control mean) BMD BMDL SS levels (mg/L) of Total Triazine Equivalents Daily AUCss (mg/L-hr) 2.41 1.03 57.9 24.7 BMDL values of 1.03 mg/L and 24.7 (mg/L-hr) were estimated for steady state levels of plasma triazines and daily AUCss, respectively. These BMDL estimates would be particularly useful for comparison to drinking water monitoring estimates. Exponential Model 4 with 0.95 Confidence Level Exponential 25 Mean Response 20 15 10 5 BMDL 0 BMD 10 20 30 40 dose 50 60 70 12:24 05/04 2011 Figure 4: EPA BMDS Exponential Model (Constant Variance; BMR = 1 SD from control mean) Results for Attenuation of LH Surge in Female Rats Administered Atrazine by Gavage - NHEERL Data. Page 76 of 184 Exponential Model 5 with 0.95 Confidence Level Exponential 25 Mean Response 20 15 10 5 BMDL 0 BMD 2 4 6 8 10 12 dose 08:46 05/09 2011 Figure 5: EPA BMDS Hill Model (Constant Variance; BMR = 1 SD from control mean) Results for “Area Under the Triazine Equivalent Plasma-Concentration Time Curve” as the Dose Metric NHEERL Data (Attenuation of LH Surge in Female Rats Administered Atrazine by Gavage). Page 77 of 184 Exponential Model 5 with 0.95 Confidence Level Exponential 25 Mean Response 20 15 10 5 BMDL 0 BMD 50 100 150 dose 200 250 300 09:12 05/09 2011 Figure 6: EPA BMDS Hill Model (Constant Variance; BMR = 1 SD from control mean) Results for “Daily Steady State AUC for Triazine Equivalents” as the Dose Metric - NHEERL Data (Attenuation of LH Surge in Female Rats Administered Atrazine by Gavage). 4.1.3 Proposed Point of Departure Plasma levels of atrazine and its metabolites appear to reach pseudo steady state at or near four days of exposure in the adult rat. Specifically, after four days of exposure, the plasma levels of total triazine equivalents plateau during periods of repeated dosing. Similarly, data from multiple laboratories ranging in duration from four days up to six months of exposure show that attenuation of LH is fairly constant at a given dose and suggest PD steady state. As such, given both PK and PD steady state, it is prudent to select the most robust study from those studies where steady state has been reached (i.e., any study four days or longer). Data from Cooper et al. (2010) contain multiple doses across a broad range and provide comparatively low variability. As such, the data from Cooper et al. (2010) are proposed as the most appropriate for deriving a PoD. Specifically, the BMDL of 2.56 mg/kg/day (based on AUC dose metrics) is proposed as a point of departure (PoD) for atrazine risk assessment for repeated exposure scenarios. Page 78 of 184 Possible approaches for extrapolating this animal PoD to humans is discussed in Section 8 below. 4.2 Pharmacokinetic Analysis 4.2.1 Background Characterization of the pharmacokinetics and internal dosimetry of atrazine and its metabolites represents a critical step for elucidating the link between exposure and attenuation of the pre-ovulatory LH-surge for the application of a mode of action approach to risk assessment. LH attenuation resulting from atrazine exposure has been identified as an early key event in the progression towards apical toxicities (for example, disruption of ovarian function including estrous cyclicity and delays in puberty onset), However, the temporal relationship between atrazine exposure and LH attenuation has not been fully characterized.. One of the information gaps noted at the September 2010 SAP meeting was the lack of detailed pharmacokinetic information for atrazine and its individual metabolites (FIFRA 2010c). Pharmacokinetic information is particularly important since atrazine as a parent chemical is short-lived and its chloro-s-triazine metabolites are presumed to be responsible for most, if not all, of the neuroendocrine mode of action effects including LH attenuation. The current understanding of atrazine metabolism is that upon entering the body via the oral route of administration, it is quickly and extensively metabolized to the monodealkylated chloro-s-triazine metabolites deethylatrazine (DEA) and deisopropylatrazine (DIA). A second round of CYP-450 mediated dealkylation results in diaminochlorotriazine (DACT), the most abundant metabolite of atrazine frequently observed in animal toxicity studies and human biomonitoring (depicted in Figure 7). Atrazine can also undergo dechlorination to form glutathione conjugates. This metabolic pathway, however, is assumed to be more of a detoxification pathway since the chlorinated triazine ring is considered to be the toxicologically important moiety. It has been postulated, based on urine radioactivity percentages, that glutathione conjugation may represent as much as 30% of atrazine metabolism (Timchalk et al. 1990). Page 79 of 184 Glutathione conjugation Atrazine Dealkylation Dealkylation CYP450 Deethylatrazine (DEA) Deisopropylatrazine (DIA) CYP450 Dealkylation Dealkylation Diaminochlorotriazine (DACT) Figure 7: Metabolic Scheme for Atrazine The following section is an update to the pharmacokinetic information evaluated by the Agency for atrazine and its metabolites that now includes different species including rats, monkeys, and humans. This information is being used to propose an interim pharmacokinetic modeling approach that relates orally administered doses of atrazine to plasma triazine levels given that a thorough review and evaluation of a recently submitted PBPK model by Syngenta has not been completed. 4.2.2 Updates on the Pharmacokinetics of Atrazine and Its Metabolites in the Rat Additional details of the pharmacokinetics of atrazine and its metabolites are emerging based on the need to characterize the temporal relationship between atrazine exposure and attenuation of the LH surge. In a recent effort aimed at characterizing the pharmacokinetic behavior of atrazine and its metabolites, Syngenta in collaboration with Page 80 of 184 WIL Research Laboratories, LLC, embarked on a time-intensive in vivo pharmacokinetic study with female Sprague Dawley rats following oral administration of atrazine (Coder 2011). The study consisted of two modes of oral administration: once daily bolus dosing via oral gavage for 4 days or distributed dosing for 4 days through the diet. Particular attention was given to capturing the elusive short in vivo plasma half-life of atrazine following oral administration since previous studies have not sampled plasma early enough post-dosing to properly characterize it. In the new study, plasma was sampled and analyzed for atrazine, DEA, DIA, and DACT as early as 5 minutes following oral gavage dosing. The treatment duration of 4 days was selected based on the critical study that established the LOAEL for LH attenuation at 3.12 mg/kg bw/day over a 4 day period (Cooper et al. 2010). Also included in the pharmacokinetic study was a 4-day washout period during which dosing with atrazine did not occur in order to characterize the clearance of all 4 plasma chloro-s-triazines (i.e., atrazine, DEA, DIA, and DACT) resulting from once daily dosing for 4 days. It should be noted that a mass balance analysis of the administered atrazine dose was not performed in the study at any of the levels tested. 4.2.3 Oral Gavage Dosing of Sprague Dawley Rats with Atrazine The results from oral gavage dosing indicate that atrazine is rapidly absorbed and cleared from rat plasma at all doses tested. Figure 8 shows the 24-h plasma profile of atrazine following oral gavage dosing at 3, 10, and 50 mg/kg bw atrazine. Maximum plasma levels (Cmax) were reached as early as 20 minutes following dosing at 3 and 10 mg/kg bw atrazine and much earlier at 10 minutes following dosing with the higher dose of 50 mg/kg bw atrazine. The increase in Cmax was less than proportional with respect to dose, increasing by only 2.8- and 1.7-fold as compared to 3.3-(3 to 10 mg/kg bw) and 5-fold (10 to 50 mg/kg bw) increases in dose levels, respectively. In contrast, the area under the plasma concentration-time curve (AUC) estimated from zero to infinity increased more directly in proportion with dose (Table 3). When plasma levels during the elimination phase were plotted on a semi-logarithmic scale (i.e., lnCp) versus time, a significant linear correlation was observed for all doses of atrazine tested. A plasma elimination half-life as short as 44 minutes was estimated for the low dose of 3 mg/kg bw but was noted to increase with dose to 1.2 and 2.9 hours at 10 and 50 mg/kg bw atrazine, respectively (Table 3). Page 81 of 184 40 3 mg/kg bw/day ATRA Plasma levels(µg/L) 35 10 mg/kg bw/day ATRA 30 50 mg/kg bw/day ATRA 25 20 15 10 5 0 0 1 2 3 4 5 6 7 8 10 12 14 16 18 20 22 24 time (hr) Figure 8: Plasma profile of atrazine for 24 hours following oral gavage dosing of rats with 3, 10, or 50 mg/kg bw atrazine. Each time point represents average of six measurements ± SD. Data is from Coder 2011. The plasma profiles of the mono-dealkylated metabolites DEA and DIA were found to parallel that of atrazine. Figure 9 shows the individual plasma profile of DEA or DIA in comparison to that of atrazine at the low dose of 3 mg/kg bw atrazine. Several features can be noted from Figure 9. First, the same time to maximum plasma levels (i.e., tmax) of 20 minutes as atrazine is observed for DEA and DIA. Such parallel absorption profiles into plasma for these two metabolites is indicative of a first pass effect, that is, a fraction of atrazine is metabolized to DEA and DIA in either the GI tract or liver prior to reaching the systemic circulation. Second, although plasma levels differ in magnitude, the same generic elimination profile as atrazine was observed for DEA, and DIA (see elimination decay after tmax in Figure 9).Therefore, the plasma clearance of atrazine does not appear to be due to metabolism to either DEA or DIA since its plasma profile does not decrease in correspondence to the appearance of either of these two species. Since atrazine is known to be short-lived as a parent chemical and barely detected in urine, if at all, (Ross et al. 2009; Catenacci et al. 1990; Ikonen et al., 1988), its fast disappearance from plasma may be due to metabolism to species other than DEA and DIA. Page 82 of 184 Table 3: Pharmacokinetic parameters estimated for atrazine and its chlorinated metabolites DEA, DIA, and DACT following oral gavage dosing of rats with 3, 10, and 50 mg/kg bw atrazine. Data analyzed is from Coder 2011. ORAL GAVAGE DOSING ATRAZINE AUC Cmax (µg*hr/L) (µg/L) 0→∞ 6.43 ± 1.42 6.41 18.04 ± 6.38 24.87 31.07 ± 10.31 109.6 DEA 58.17 ± 7.40 153.5 141.00 ± 22.61 465.3 219.67 ± 71.50 2277 DIA 220.67 ± 43.03 370.3 614.17 ± 93.20 1336.0 1066.0 ± 419.0 7255.0 DACT Atrazine dose (mg/kg bw/day) tmax 3.0 10.0 50.0 20 min 20 min 10 min 3.0 10.0 50.0 20 min 1 hr 45 min 3.0 10.0 50.0 20 min 1 hr 45 min 3.0 2 hr 682.00 ± 95.01 10.0 6 hr 50.0 8 hr kel (/hr) Elimination Half-life 0.941 ± 0.069 0.569 ± 0.047 0.238 ± 0.013 44.2 min 1.22 hr 2.91 hr 0.240 ± 0.009 0.0928 ± 0.0034 0.0678 ± 0.0035 3.40 hr 7.50 hr 10.22 hr 0.564 ± 0.030 0.2425 ± 0.0146 0.176 ± 0.0065 1.20 hr 2.86 hr 3.95 hr 8909.2 0.0873 ± 0.006 7.90 hr 2125.00 ± 398.00 29492.0 0.0845 ± 0.0070 8.20 hr 8321.67 ± 1630.57 149124.9 0.0752 ± 0.0102 9.22 hr Page 83 of 184 Plasma Levels ( µg/L) Atrazine DEA 60 40 20 10.0 7.5 5.0 2.5 0.0 0 1 2 3 4 5 6 7 8 7 8 time (hr) Plasma Levels ( µg/L) 300 Atrazine DIA 200 100 10 8 6 4 2 0 0 1 2 3 4 5 6 time (hr) Figure 9: The parallel plasma profiles of atrazine versus DEA or DIA following oral gavage dosing of rats with 3 mg/kg bw atrazine. Data is from Coder 2011. Oral gavage dosing with the higher doses of 10 and 50 mg/kg bw atrazine resulted in much longer tmax values of 60 and 45 minutes, respectively, for both DEA and DIA while that of atrazine remained at 20 minutes for 10 mg/kg bw and even decreased to 10 minutes at 50 mg/kg bw (Table 3). This observation suggests that the increases in tmax for DEA and DIA are likely due to saturation of first pass metabolism resulting in prolonged absorption of these two species into plasma. Limited solubility of DEA and DIA produced in the GI tract at the dose levels tested is a less likely possibility since a biphasic absorption phase in the plasma profile would be observed in such case as previously reported by McMullin and coworkers (McMullin et al. 2007). Although comparable tmax values were exhibited by DEA and DIA, plasma levels (Cmax and AUC) were several folds higher for DIA (Table 3). As with atrazine, both DEA and DIA Page 84 of 184 Plasma Levels (µg/L) exhibited fast linear elimination kinetics. At the low dose of 3 mg/kg bw atrazine, plasma elimination half-life estimates were 3.4 and 1.2 hours for DEA and DIA, respectively. Dose dependent decreases in plasma elimination were also noted for DEA and DIA. Plasma halflife estimates for DEA increased to 7.5 hr (~2.2 fold decrease) and 10.2 hr (~ 3-fold decrease) at 10 and 50 mg/kg bw atrazine, respectively. Comparable decreases in plasma clearance were seen for DIA with plasma half-life increasing to 2.9 (~2.4-fold clearance decrease) and 3.9 hrs (~3.2-fold clearance decrease) at 10 mg/kg and 50 mg/kg bw atrazine, respectively. The only metabolite that did not exhibit dose-dependent decreases in plasma clearance was DACT for which a plasma elimination half-life was consistently estimated at 8-9 hours (Table 3). DACT was also the predominant species detected in rat plasma with estimated AUC values 20-24 fold higher than those estimated for DIA and even higher for DEA. A much longer tmax of 2 hr was seen for DACT at 3 mg/kg bw atrazine while plasma levels of DIA and DEA (primarily DIA) were seen decreasing (Figure 10). Peak plasma levels of DACT did not decrease much until about 14 hours post-dosing when plasma levels of DEA and DIA had been significantly depleted. Such observation is consistent with DACT being formed from the precursor metabolites DEA and DIA after these species have entered the systemic circulation. The increases in plasma half-life at 10 and 50 mg/kg bw atrazine observed with DEA and DIA (Table 3) are likely due to saturation of the metabolic systems responsible for their metabolism to DACT. The lack of dosedependent changes in the clearance of DACT is consistent with this species being a terminal metabolite and/or the lack of saturation for further metabolism at the dose levels tested. Atrazine DEA DIA DACT 800 600 400 250 200 150 100 50 0 0.00 0.25 0.50 0.75 1.00 2 4 6 8 10 12 14 16 18 20 22 24 time (hr) Figure 10: 24-h plasma profile of atrazine and its chlorinated metabolites DEA, DIA, and DACT following oral gavage dosing of rats with 3 mg/kg bw atrazine. Each time point represents average of six measurements ± SD. Data is from Coder 2011. Page 85 of 184 Repeated daily dosing with atrazine did not change the pharmacokinetic behavior of any of the species analyzed (i.e., atrazine, DEA, DIA, or DACT) as compared to single dose behavior. The entire 4-day treatment/4-day washout plasma profile at 3 mg/kg bw atrazine is shown in Figure 11. The most abundant species was DACT followed by DIA, DEA, and atrazine as the least abundant. DACT was also the species with the slowest clearance, followed by DEA, DIA, and atrazine. With the exception of DACT for which a bit of accumulation is observed due to sustained peak plasma levels, all 4 species analyzed can be expected to be cleared from plasma by 24 hours, constituting a repeated episodic pharmacokinetic profile upon repeated (once per day) daily dosing. In summary, atrazine as a parent chemical exhibited very fast pharmacokinetics with a tmax as short as 20 minutes and a plasma half-life of 44 minutes at the low dose of 3 mg/kg bw. The tmax information for atrazine in comparison to those for DEA and DIA is consistent with the formation of these two metabolites being due to first pass metabolism. The parallel plasma elimination profiles of DEA and DIA relative to that of atrazine do not support the fast plasma clearance of atrazine being due to metabolism to DEA and DIA. Without mass balance in a pharmacokinetic study, it is difficult to disregard other metabolites as a possibility to explain the fast plasma clearance of atrazine since urinary excretion of intact atrazine is likely to be minimal. 3 mg/kg bw/day atrazine: oral gavage dosing Plasma Levels (µg/L) 800 Atrazine DEA DIA DACT 600 400 200 0 0 20 40 60 80 100 120 140 160 180 time (hr) Figure 11: Plasma profile of atrazine, DEA, DIA, and DACT from repeated once a day oral gavage dosing of rats with 3 mg/kg bw atrazine for 4 days. Values represent the mean of six replicates ± SD. Data is from Coder 2011. Page 86 of 184 4.2.4 Dietary Exposures of Sprague Dawley Rats to Atrazine Dosing with atrazine through the diet resulted in plasma levels for all 4 monitored chlorotriazines (i.e., atrazine, DEA, DIA, and DACT) being more variable and overall much lower as compared to oral gavage dosing at comparable dose levels. Figure 12 shows the plasma profile of all 4 species at the equivalent dietary dose level of 3 mg/kg/day. Maximum plasma levels of all species analyzed were reached much later at between 12.5 and 20 hours, reflecting the consumption of diet by rodents during the night. Only the plasma profile of DACT can be deciphered with the high level of variation observed. Plasma levels of DACT seem to reach a more defined pseudo steady state level by the fourth day of daily dosing, probably reflecting the more frequent dosing with atrazine through the diet and the longer half-life of this compound as compared to the other metabolites. A similar behavior was observed at the higher dose levels of 10 and 50 mg/kg bw atrazine (data not shown). 3 mg/kg bw/day atrazine: dietary dosing Plasma Levels (µg/L) 500 ATRA DEA DIA DACT 400 300 200 100 0 0 24 48 72 96 time (hr) 120 144 168 192 Figure 12: Plasma profile of atrazine, DEA, DIA, and DACT from dietary dosing of rats with atrazine at a nominal dose of 3 mg/kg bw/day atrazine for four days. Data at each time point represents measurements in 6 animals. Data points are not connected to help elucidate pattern. Data is from Coder 2011. The fluctuating behavior of plasma levels resulting from dietary dosing made analysis difficult and thus only the elimination phase of each triazine species was analyzed for comparison to oral gavage dosing. In the case of atrazine, no analysis was performed due to plasma levels being too low and highly variable for any informative analysis. The results for DEA, DIA, and DACT indicate, as with oral gavage dosing, that these species exhibit Page 87 of 184 linear elimination kinetics (assessed by plotting lnCp vs. time). As listed in Table 4, the estimated first order elimination rate constants are very comparable to those estimated via oral gavage dosing. The estimated plasma half-life of DEA of 3.7 hr is very comparable to that of 3.4 hr estimated via oral gavage. Greater differences were observed at 10 and 50 mg/kg bw/day, but with no clear pattern since a 1.2-fold slower clearance was observed at 10 mg/kg bw but an increase of about the same magnitude was observed at 50 mg/kg bw (Table 4). In the case of DIA, with the exception of 3 mg/kg bw atrazine for which an approximately 2-fold decrease in clearance was observed (plasma half-life of 3 hr versus 1.2 hr via oral gavage dosing), the results with 10 and 50 mg/kg bw atrazine were comparable to those estimated via oral gavage dosing. Plasma clearance estimates were particularly consistent for DACT as compared to oral gavage dosing. Table 4: Pharmacokinetic Elimination Parameters estimated for atrazine and its chlorinated metabolites DEA, DIA, and DACT following dosing of rats with atrazine through the diet at nominal doses of 3, 10, and 50 mg/kg bw. Data is from Coder 2011. Dose (mg/kg bw/day) 3.0 10.0 50.0 3.0 10.0 50.0 3.0 10.0 50.0 Dietary Dosing with Atrazine Elimination Rate Elimination Plasma -1 Constant (hr ) Half-life (hr) DEA 0.187 ± 0.012 3.70 0.0761 ± 0.004 9.10 0.104 ± 0.005 6.65 DIA 0.231 ± 0.016 0.248 ± 0.015 0.243 ± 0.013 DACT 0.0897 ± 0.0019 0.0729 ± 0.0021 0.0734 ± 0.0018 3.00 2.80 2.85 7.73 9.51 9.44 4.2.5 Plasma Clearance Discrepancy Based on Studies with Radiolabeled Atrazine Analysis of the new pharmacokinetic dataset submitted to the Agency in which atrazine, DEA, DIA, and DACT are monitored in rat plasma following once daily oral dosing with atrazine indicates a plasma clearance discrepancy as compared to studies performed with radiolabeled atrazine. None of the 4 species monitored in the new study exhibits a plasma clearance that is comparable to that estimated based on radiolabeled atrazine. All four species (i.e., atrazine, DEA, DIA, and DACT) can be expected based on their estimated plasma half-lives to be cleared from plasma by 24 hours, constituting an episodic pharmacokinetic profile that contrasts the bioaccumulative profile observed based on Page 88 of 184 radiolabeled atrazine (Thede 1987). Pharmacokinetic studies with radiolabeled atrazine have been carried out with a metabolically stable [14C]-radiolabeled triazine ring that allows for the study of the overall disposition of all species associated with the radiolabel (Thede 1987; Paul et al 1993). In general, radiolabel disposition studies achieve a high degree of mass balance (some very close to 100%) due to minimal sample preparation being required for quantification of radioactivity. At the September 2010 SAP meeting, the Thede 1987 study was presented as a key pharmacokinetic study for estimating plasma AUC based on radiolabeled triazine equivalents. The study consisted of repeated once daily dosing of young female Sprague Dawley rats with a wide range of radiolabeled atrazine doses (1-100 mg/kg bw/day). The resulting plasma profile (shown in Figure 13) clearly shows accumulation of the radiolabel continuing until the fourth day of daily dosing when pseudo steady state plasma levels seem to be reached (Thede 1987). Given that the LOAEL for LH attenuation was also observed in rats after 4 days of once daily dosing with atrazine (Cooper et al. 2010), the bioaccumulative pharmacokinetic profile resulting from daily dosing with radiolabeled atrazine, not the episodic profile observed with atrazine, DEA, DIA, or DACT, is consistent with the temporality of the LH attenuation endpoint. The plasma elimination from the Thede 1987 study was noted to be linear with a consistent elimination rate constant of 0.010-0.015 hr-1 (plasma half-life of 53.3 - 69.3 hrs) for the 1, 3, 7, and 10 mg/kg bw/day dose groups (Table 5) that indicates accumulation upon repeated once daily dosing with radiolabeled atrazine. At the higher doses of 50 and 100 mg/kg bw/day atrazine, the elimination rate constant was substantially higher (>20-fold) compared to the other dose groups at about 0.3 hr-1. The nature for the difference is not known, but it may be reflective of the limited time points examined, particularly at these high dose levels. Page 89 of 184 Figure 13: Plasma profiles of radiolabeled triazine equivalents resulting from repeated once daily dosing of rats with radiolabeled atrazine at 1, 3, 7, 10, 50 and 100 mg/kg bw atrazine. Data is from Thede 1987. Table 5: First order elimination rate constants estimated from plasma data in the Thede 1987 study involving repeated once daily dosing of rats with radiolabeled atrazine (Thede 1987). -1 ATRA dose (mg/kg/day) Elimination rate constant (hr ) 1 0.010 3 0.015 7 0.013 10 0.014 50 0.376 100 0.308 4.2.6 Additional Rat Radiolabeled Atrazine Studies Evaluated for Plasma Clearance One of the limitations of the Thede 1987 study noted at the September 2010 SAP meeting was the limited number of plasma data points for characterizing the elimination phase of the radiolabel (FIFRA SAP 2010b). The elimination phase consisted of plasma data at only 3 time points corresponding to one animal of each dose group (i.e., 1, 3 ,7 , and 10 mg/kg bw). In essence, the elimination rate constant of 0.010-0.015 hr-1 was estimated with 12 Page 90 of 184 data points from 4 individual animals. Recognizing that radiolabeled atrazine studies offer at this time the best information to inform internal dosimetry and risk assessment, the Agency embarked on an effort to gather more evidence for the plasma clearance obtained from analyzing the plasma data from the Thede 1987 study. Thus, other rat radiolabeled atrazine pharmacokinetic studies regardless of dose, sex, or dosing regimen (single or repeated) were gathered for evaluation. Paul et al (1993) In the first study by Paul et al. 1993, single oral gavage doses of 1 or 100 mg/kg bw radiolabeled atrazine were administered to groups of three male and female SD rats and the resulting levels of radioactivity in plasma and other tissues were measured at different times post-dosing (Paul et al. 1993). At both administered dose levels, plasma measurements were limited to male rats and at the following time points: 2, 48, 168, and 336 hrs for 1 mg/kg bw and 24, 72, 168, and 336 hrs for 100 mg/kg bw. As seen in Figure 14, the plasma elimination phase at both dose levels exhibited linear behavior when plotted on a semi-logarithmic plot versus time. The resultant elimination rate constants were 0.013 and 0.014 hr-1 which are consistent with those estimated for the 1-10 mg/kg bw/day dose groups from Thede 1987 study (Thede 1987). It should be noted that the plasma analysis of the 100 mg/kg bw/day dose group in the Thede study resulted in a much higher clearance rate as compared to those obtained from Paul et al. 1993. The nature for discrepancy is unknown. Other tissues were also examined for radioactivity clearance in the Paul et al. 1993 study including the brain, lungs, liver, kidneys, and others. Reported clearance values for all tissues examined were much greater than 24 hours, estimated at as a range of 59-300 hours (Paul et al. 1993) (data not shown). Page 91 of 184 C- atrazine 0.6 0.4 Paul et al. 1993 0.2 0.0 0 48 96 144 192 240 288 336 Single dose of 100 mg/kg bw Plasma radiolabeled equivalents (ppm) 14 14 C-atrazine 25 20 15 Paul et al. 1993 10 5 0 0 24 48 72 96 120 144 168 192 216 240 264 288 312 336 time (hr) time (hr) ln (Plasma radiolabeled equivalents) (ppm) Single dose of 1 mg/kg bw 14 Single dose of 100 mg/kg bw C-atrazine 0 Slope = - 0.013 ± 0.002 R2 = 0.74 -2 -4 -6 Paul et al. 1993 -8 0 48 96 144 192 time (hr) 240 288 336 ln (Plasma radiolabeled equivalents) (ppm) Plasma radiolabeled Equivalents (ppm) Single dose of 1 mg/kg bw 4 14 C-atrazine Slope = - 0.014 ± 0.001 R2 = 0.94 2 Paul et al. 1993 0 48 -2 96 144 192 240 288 336 384 time (hr) -4 Figure 14: Plasma profile and analysis of elimination of a single oral gavage dose of 1 and 100 mg/kg bw radiolabeled atrazine in rats. Data is from Paul et al. 1993. Simoneaux 1985 In another study with radiolabeled atrazine by Simoneaux 1985, thirty six male rats (3 per time point) were given oral doses of either 0.4 or 4.0 mg/kg bw 14C-atrazine for seven days and sacrificed at 5, 7, 9, 10, 14, and 18 days post-dosing at which point plasma was sampled for radioactivity (Simoneaux 1985). The plasma profiles are given in Figure 15. Similar to the other two studies already discussed, analysis of the elimination phase at both dose levels revealed linear behavior when plotted on a semi-logarithmic scale versus time. When the resultant rate constants (estimated as per day from the reported plasma data) are converted to per hr, the resultant values of 0.013 and 0.014 hr-1 are remarkably consistent with those estimated from the Thede 1987 and Paul et al. 1993 studies. In summary, one of the caveats of the plasma clearance analysis from the Thede 1987 study has been addressed by the evaluation of two additional rat pharmacokinetic studies with radiolabeled atrazine. The two additional studies were carried out in different laboratories with different sexes and dosing regimens (single and repeated dosing) and along with the Thede 1987 study support an elimination half-life in the order of 53.3 - 69.3 Page 92 of 184 hrs (kel of 0.013-0.015 hr-1), indicative of bioaccumulation upon repeated daily dosing with atrazine. C-Atrazine Repeated dosing with 4.0 mg/kg bw 0.10 0.08 0.06 0.04 Simoneaux 1985 0.02 0.00 0 2 4 6 8 10 12 14 18 16 time (days) Ln (Plasma radiolabeled equivalents) (ppm) Elimination Phase Analysis: 0.4 mg/kg bw/day 14 0 10 12 14 16 Simoneaux 1985 -3 -4 -5 -6 18 time (days) -2 slope = -0.236 ± 0.031 day = -0.010 ± 0.001 hr -1 -1 14 C-Atrazine 1.5 1.0 0.5 Simoneaux 1985 0.0 0 2 4 6 8 10 12 14 16 18 time (days) Elimination Phase Analysis: 4.0 mg/kg bw/day C-atrazine 1 -1 Plasma radiolabeled equivalents (ppm) 14 Ln (Plasma radiolabeled equivalents) (ppm) Plasma radiolabeled equivalents (ppm) Repeated dosing with 0.4 mg/kg bw 14 C-atrazine 1 0 10 -1 12 14 16 18 time (days) Simoneaux 1985 -2 -3 -4 slope = -0.258 ± 0.019 day-1 = -0.011 ± 0.008 hr -1 Figure 15: Plasma profile and analysis of elimination of repeated oral gavage dosing with either 0.4 or 4 mg/kg bw 14C- atrazine in rats for seven days. Data is from Simoneaux 1985. 4.2.7 Pharmacokinetic Studies with Radiolabeled Atrazine in Monkeys All of the pharmacokinetic studies discussed thus far were perfomed in rats. In an effort to investigate the pharmacokinetics of radiolabeled atrazine in non-human primates, Hui et al., 2011 dosed adult rhesus monkeys with either 1, 10, and 100 mg of [14C]- radiolabeled atrazine via oral gavage and sampled plasma for radioactivity at various time points postdosing (Hui et al. 2011). An intraveneously administered dose of 0.25 mg of [14C]-atrazine was also included in the study to estimate the oral bioavailability of atrazine. As with other radiolabel studies already discussed, a high degree of mass balance of over 95% of the administered dose was achieved. As with other species, urinary excretion was the primary route of body excretion followed by fecal excretion. An oral bioavailability greater than 60% was estimated for atrazine. Analysis of plasma radioactivity levels indicates that there were dose-dependent changes in absorption across the different doses of atrazine tested. Page 93 of 184 However, as was the case in rats, plasma clearance was linear. In fact, the plasma profiles for the three different doses could be reproduced with a one comparment linear model with optimized parameters (i.e., volume of distribution and elimination rate constant). Estimates of the steady state volume of distribution via non-compartmental analysis were about the same for the three dose groups given the variability reported around the mean values. An average value of about 3.4 L/kg bw was estimated for the three doses tested. Moreover, as similarly discussed for rat studies, both plasma Cmax and AUC increased with dose, but AUC notably scaled more directly with dose (Hui et al. 2011). The estimated plasma elimination rate constants along with other pharmacokinetic parameters are listed in Table 6 below. It should be noted that the estimated monkey plasma elimination rate constant of 0.03-0.06 hr-1 is only 2-4 fold greater than the estimated rat value of about 0.013 hr-1 with considerable variability of as much as 50% (Table 6). The corresponding plasma half-life in monkeys is estimated to be on average about 20 hours. Thus, bioaccumulation can also be expected to occur in monkeys upon repeated daily dosing with atrazine.Thus, there is a remarkable consistency in the clearance across different doses and species (rats and monkeys) resulting from the linear pharmacokinetic behavior exhibited by radiolabeled atrazine. Table 6: Estimated plasma pharmacokinetic parameters in monkeys following oral gavage dosing 14 with 1, 10, and 100 mg [ C]-atrazine. Data is from Hui et al. 2011. 14 C-Atrazine Dose (mg) 1 10 100 Cmax (mg/L) 0.05 ± 0.02 0.32 ± 0.02 3.04 ± 0.71 AUC 0→∞ ((mg/L)*hr) 1.61 ± 0.35 13.42 ± 3.58 151 ± 24 Vdss (L/kg bw) kel -1 (hr ) 2.427 ± 0.397 4.413 ± 0.781 3.255 ± 0.597 0.04 ± 0.03 0.06 ± 0.03 0.03 ± 0.00 t1/2 (hr) 20 ± 8 16 ± 9 24 ± 1 4.2.8 Pharmacokinetic Studies with Atrazine in Humans As with the majority of environmental chemicals, human information on the pharmacokinetics of atrazine is very limited. In one of the few controlled human studies available, atrazine was administered orally to six volunteers (bw range of 74 to 117 kg) via a single oral dose of 0.1 mg/kg bw (Davidson 1985). Urine was collected for each subject for up to 168 hours post-dosing while blood samples were obtained from a single subject at various time points starting at 2 hrs post-dosing. Atrazine, DEA, DIA and DACT were measured in blood and urine samples using analytical methods with 0.005 ppm minimum detection levels. Page 94 of 184 4.2.8.1 Human Blood Pharmacokinetic Information on Atrazine and its Chloro-s-triazine Metabolites It should be noted that the information reported for blood samples refers to whole blood rather than plasma. This is particularly important in comparing results to rat findings since accumulation of up to 1.5% of administered radiolabel has been reported in rat red blood cells (Thede 1987; McMullin 2003). Although the human blood information reported may be limited (from a single subject), at minimum, pharmacokinetic behavior can still be compared for consistency across different species. From the blood information reported, the most important finding from a species extrapolation perspective is the linear first order behavior exhibited by DIA and DACT which has been consistently observed in all animal species discussed thus far. Peak blood levels of DIA were reported to be reached within 2 hours and decreased rapidly thereafter with an apparent linear elimination rate constant of 0.25 hr-1 (half-life of 2.8 hr) (Table 7) which is only about 2.2-fold lower than that estimated for rats at 3.0 mg/kg bw atrazine (Table 3). The rate of appearance of DACT in human blood was noted to be slower than DIA with a tmax of about 5 hr, consistent with oxidative metabolism from the precursor metabolites DEA and DIA. A first order elimination rate constant of 0.039 hr-1 (half-life of 17.5 hr) was estimated for DACT which is about 2-fold lower than the rat value of about 0.08 hr-1 (half-life of 8-9 hours) estimated for rats (Table 3). Atrazine and DEA were detected in human blood at levels below quantitation at the administered dose level of 0.1 mg/kg bw atrazine. Such finding is consistent with rat findings in that these two plasma chlorotriazines represent the least abundant out of the 4 chlorotriazines monitored in plasma. Moreover, atrazine would be barely detected at 2 hr (the first time point analyzed) if the rat plasma half-life of about 44 minutes at 3 mg/kg bw happens to be similar or even shorter in humans. Thus, the human blood information provided in the Davidson 1987 report is at minimum qualitatively similar to the pharmacokinetics of atrazine and DEA observed in rats and other species. From a species extrapolation perspective, it should be underscored that DIA and DACT exhibited linear elimination kinetics that is about 2-3 fold lower than that observed in rats (Table 3). When the elimination rate constant of about 0.013 hr-1 estimated from the Thede 1987 study and supported by other rat radiolabeled atrazine studies is allometrically scaled for a human female with a bw of 60 kg (for example), the resulting kel of 0.0033 hr-1 is less than 4-fold lower. Thus, allometric scaling of adult rat plasma triazine linear elimination kinetics represents a reasonable approach for estimating the elimination kinetics in adult humans. 4.2.8.2 Human Urine Pharmacokinetic Information on Atrazine and its Chloro-s-triazine Metabolites Most of the information reported in the Davidson 1985 report is based on urinary levels of DEA, DIA, and DACT; atrazine was not detected in urine (Davidson 1987). The urinary excretion information for these three chloro-s-triazine metabolites were pooled and analyzed as atrazine equivalents for pharmacokinetic behavior. The kinetics of each Page 95 of 184 subject was found to be adequately described by a one-compartment linear model of urinary excretion. An estimated first order elimination rate constant of 0.065 hr-1 was estimated based on renal excretion as compared to 0.039 hr-1 estimated for DACT from blood information from a single subject (Table 7). Given the limited information and experimental variability, the blood and urinary clearance results were deemed by the study author to be comparable (Davidson 1987). By far, however, the most important result reported from the urinary excretion information was a very significant mass balance discrepancy when only atrazine, DEA, DIA, and DACT were monitored. Only a total of 14.5% of the atrazine dose administered was accounted for when only these four metabolites were monitored. It should be noted that urinary excretion is well-established as the primary route of excretion of atrazine (i.e., if present at all) and its metabolites (DEA, DIA, and DACT) and as also supported by the rapid absorption from the GI tract and peak plasma levels being reached as early as 44 minutes post-dose. Thus, biliary/fecal excretion is unlikely to explain the magnitude of the mass balance discrepancy reported in the study. The most likely possibility as reported by the study author was that a large portion of atrazine metabolites have not been identified. Several metabolites including conjugates of glutathione, glucoronide and sulfate were reported as possibilities (Davidson 1987). Such conclusion is consistent with the notion that the metabolism of atrazine has not been completely characterized. Table 7: Estimated Pharmacokinetic Parameters from Human blood and urine a Chlorotriazine Atrazine DEA DIA DACT Whole Blood Elimination rate constant -1 (hr ) 0.25 0.039 Urine tmax (hr) b b ≤2 ~5 Atrazine DEA DIA DACT - Elimination Half-life (hr) 2.8 17.8 b - 0.29 0.30 0.060 a information was from only one human subject b Below quantitation levels c Only the α phase is reported since the β phase was considered minor. Page 96 of 184 c 2.4 ± 1.0 c 2.3 ± 0.5 11.5 ± 0.4 4.2.9 Updates on the Metabolism of Atrazine Focusing on DEA, DIA, and DACT as the major toxicologically active metabolites of atrazine’s neuroendocrine mode of action is a reasonable approach given the evidence to date, assuming that the remaining metabolites are toxicologically inactive dechlorinated glutathione conjugates. Recent reports have identified novel reactive metabolites of atrazine from in vitro human liver microsomal preparations. In one of the most recent studies by Leblanc et al. 2011, a new untargeted sensitive analytical chemistry method was used to identify novel metabolites of atrazine (Leblanc et al. 2011). A new phase I metabolic scheme for atrazine was proposed based on the formation of a newly identified N-oxide metabolite which undergoes hydrolysis to form a reactive electrophilic imine species which can conjugate with cellular nucleophiles like glutathione (GSH). As indicated in the updated metabolic scheme in Figure 16, nitrogen hydroxylation and subsequent imine formation takes place at the ethyl side chain on the triazine ring. In another recent study by Joo et al. 2011, 1-hydroxyisopropyl atrazine and 2-hydroxyethylatrazine were identified along with DEA and DIA as the major metabolites from human liver microsomal incubations (Joo et al. 2011). These metabolites are also included in the updated metabolic scheme in Figure 16. Although the N-oxide, imine, and chlorinated glutathione conjugate metabolites as well as 1-hydroxyisopropylatrazine and 2-hydroxyethylatrazine depicted in Figure 16 were identified using in vitro human liver microsomal preparations, at minimum, these in vitro studies raise questions about the potential formation for these additional metabolites in vivo which may explain the plasma clearance of atrazine which according to the plasma profiles in Figure 9 does not seem to be directly due to either DEA or DIA. It should be noted that the new metabolites in Figure 16 remain chlorinated and thus may be assumed to be relevant in LH attenuation if the chlorinated triazine ring is indeed the important moiety for such effect. Page 97 of 184 GSH 2-hydroxyethylatrazine 1-hydroxyisopropylatrazine Chlorinated conjugate Imine GSH N-oxide Non-chlorinated Conjugate Atrazine Deisopropylatrazine (DIA) Deethylatrazine (DEA) Diaminochlorotriazine (DACT) Figure 16: Updated metabolic profile for atrazine. Newly identified metabolites are above the dashed line for comparison to Figure 7. New metabolites are from Leblanc et al. 2011 and Joo et al. 2011. 4.2.10 Summary of Pharmacokinetic Information Atrazine and its metabolites exhibit linear kinetic behavior when monitored in plasma individually (atrazine, DEA, DIA, and DACT) or as a whole through the use of radiolabeled triazine ring across a wide range of orally administered doses and different species including rats, monkeys, and humans. When only DEA, DIA, and DACT are individually monitored in rats, the species with the slowest clearance is DACT with an estimated plasma elimination half-life of 8-9 hr. With a plasma half-life of 8-9 hrs, a female rat being dosed daily can be expected to clear most, if not all, of the chemical by 24 hours, resulting in a repeated episodic plasma profile such as that described for DACT in Figure 11 following daily gavage dosing with atrazine. Thus, if only atrazine, DEA, and DIA, and DACT are monitored, little or no plasma bioaccumulation can be expected to occur upon repeated daily dosing. Moreover, if the pharmacokinetics of these 4 species are pooled and described as molar chlorotriazine equivalents in a one-compartment linear model, Page 98 of 184 clearance would be dominated by DACT as the most abundant and slowly cleared metabolite. Indeed, McMullin et al. 2003 reported a one-compartment linear model based on atrazine, DEA, DIA, and DACT as chlorotriazine equivalents with an estimated plasma elimination rate constant of 0.08 hr-1 (t1/2 =8.8 hr) which is in excellent agreement with the values estimated for DACT at the three atrazine dose levels of 3, 10, and 50 mg/kg bw in Table 3 (McMullin 2003). In contrast, when the pharmacokinetics of atrazine are monitored in rats via a radiolabeled triazine ring, a much longer plasma half-life in the range of 53-69 hours is observed across different studies and doses, suggestive of bioaccumulation upon repeated daily dosing. Bioaccumulation of plasma radiolabeled triazine equivalents is clear from available repeated daily dosing pharmacokinetic studies including Thede 1987 (Figure 13) and Simoneaux 1985 (Figure 15) . Thus, a plasma clearance discrepancy exists between radiolabeled triazine studies and those that only monitor atrazine, DEA, DIA, and DACT. The discrepancy may be due to unidentified/unmonitored metabolites of atrazine in vivo such as those reported from recent in vitro metabolism studies (Figure 16) (Leblanc et al. 2011 and Joo et al. 2011). It should be noted that glutathione conjugates formed via substitution of the chlorine atom on the triazine ring are unlikely to explain the clearance discrepancy since these are presumed to be cleared rapidly, a typical consequence of phase II metabolism, and have been estimated to represent about 30% of the administered atrazine dose (Timchalk et al., 1990 and McMullin et al. 2007). Because of this plasma clearance discrepancy, the Agency is relying on studies based on radiolabeled atriazine on the basis that such studies account for known and unknown metabolites of atrazine and achieve a high degree of mass balance, although the toxicological relevance of these metabolites related to atrazine’s neuroendocrine mode of action is unclear. 4.2.11 Progress on Physiologically Based Pharmacokinetic Modeling Efforts In the case of chemicals like atrazine with a very short in vivo plasma half-life and toxicologically active metabolites, pharmacokinetic modeling is necessary for the application of a mode of action approach to risk assessment. Some of the current pharmacokinetic modeling needs for atrazine include the estimation of relevant internal measures of parent chemical and toxicologically active metabolites (related to the endocrine mode of action), reliable means for extrapolating such internal measures across different species and life-stages, routes, and conditions of exposure, and most importantly for exposure assessment, means for reconstructing the human ingested dose through drinking water. The preferred modeling approach for doing the aforementioned extrapolations and dose reconstructions involves a properly calibrated and evaluated PBPK model. Previous efforts to develop a PBPK model for atrazine and its metabolites have been hampered in part by sub-optimal pharmacokinetic data for model parameterization. The integrated composite PBPK model for atrazine and its chloro-s-triazine metabolites developed by McMullin and co-workers was parameterized with plasma data obtained from oral gavage dosing of rats with atrazine, DEA, DIA, or DACT at molar equivalent levels of Page 99 of 184 150 mg/kg bw atrazine (McMullin et al. 2007). At such high dose levels, limited solubility in the gastrointestinal tract and continuing absorption of more slowly dissolving chemical resulted in complex plasma profiles difficult to interpret for model parameterization, particularly if oral uptake is not adequately described in the model. The final composite model could not simulate the pharmacokinetics of atrazine or its metabolites, resulting in overprediction of plasma levels of atrazine while underpredicting plasma levels of its metabolites. In addition to inadequate data for model parameterization, the model lacked postulated target tissues such as the brain and was limited to liver and rest of body compartments (McMullin et al. 2007) Some of the shortcomings of the McMullin 2007 PBPK model have been recently addressed in a PBPK model recently submitted to the Agency by Syngenta in collaboration with the Hamner Institutes (Campbell et al. 2011). Although the Agency has not performed a thorough review and evaluation of the model, a preliminary review indicates that the original McMullin 2007 model (McMullin et al. 2007) has been expanded to include compartments of suspected target tissues including mammary gland, hypothalamus, pituitary, testes/ovaries, and adrenals. The oral uptake of atrazine has also been refined by removing a saturable uptake description and replacing it with separate compartments for insoluble/bound and free fractions of atrazine as well as describing the potential slow release of bound atrazine. New empirically determined tissue:blood partition coefficients and hepatic in vitro metabolic rates were also incorporated. Most importantly, however, was the model calibration performed with the new in vivo time-intensive rat pharmacokinetic study described above where the chloro-s-triazine metabolites DEA, DIA, and DACT were individually monitored in female SD rats following dosing with atrazine via the oral route of administration (Coder 2011). The Agency is still evaluating the newly improved PBPK and will consider using it after it has undergone a thorough review and evaluation process. In the meantime, for purpose of this integrated toxicity-exposure assessment, an interim pharmacokinetic modeling approach is presented below. 4.2.12 Proposed Interim Pharmacokinetic Modeling Approach Non-compartmental pharmacokinetic analysis was introduced at the September 2010 meeting as a reliable way of estimating plasma AUC of radiolabeled triazine equivalents as measure of internal exposure (FIFRA SAP 2010c). Daily pseudo steady state AUC estimates were obtained from the Thede 1987 study for internal dose response analysis associated with LH attenuation. Non-compartmental analysis works best when linear elimination kinetics is observed in the plasma profile; that is, when elimination proceeds at a rate that is proportional to the remaining plasma concentration. From a modeling perspective, the most attractive feature of linear pharmacokinetics is its predictability based on a constant plasma half-life. In the case of atrazine and its metabolites, linear elimination kinetics is the most consistent feature observed when monitored individually (i.e., atrazine, DEA, DIA, and DACT) or as a whole in radiolabel studies. Thus, the behavior of plasma triazines may be described by a series of linear processes collapsed into a simplified Page 100 of 184 deterministic one-compartment model as indicated in Figure 13. The oral uptake of atrazine administered orally can be reasonably assumed to be fast and complete given its estimated short tmax of 20 minutes or less (Table 3). The first order elimination rate constant (kel) indicated in the model is representative of all clearance processes for plasma triazines that consists primarily of urinary and biliary/fecal excretion. Metabolism is not reflected in kel since the radiolabeled triazine ring is metabolically stable and thus, although metabolism is important in the clearance of individual species including atrazine, DEA, DIA, and DACT, it does not affect the clearance of the triazine ring. Deterministic models such as the one compartment linear model proposed for plasma triazines differ from statistical/probabilistic modeling approaches in that their parameters are independently estimated rather than fitted to a particular dataset. In the case of kel in Figure 17, an average value of 0.013hr-1 estimated for rats across different studies and doses of atrazine is most appropriate for relating pharmacokinetic behavior to LH attenuation in rats. Oral dose of Atrazine (mg) Fast/complete absorption Plasma Triazines (Cp, mg/L) Vdss (L) kel (hr-1) Triazines eliminated Figure 17: Proposed Interim pharmacokinetic model for atrazine and its metabolites Another feature of the behavior of plasma triazines as it relates to the toxicological endpoint of concern (i.e., LH attenuation) is that pseudo steady state is reached (or nearly reached) at the time when LH attenuation has also been observed. The condition of steady state plasma levels for LH attenuation further supports the use of a one-compartment model which assumes by definition that plasma and tissues are at least at equilibrium. Thus, plasma triazines and the two parameters (Vd and kel) that govern their behavior in the one compartment linear model are presumed to represent steady state values. In the case of kel, it was already demonstrated from analyzing plasma data from the Paul et al. 1993 study that its values does not change in single dose versus repeated dosing studies. There are several ways of estimating Vdss based on the available information. First, noncompartmental analysis offers a way of estimating Vdss through an area analysis as depicted by equation 1. AUCM refers to the area under the first moment curve and is calculated in a similar way as AUC except that it is based on the product of plasma concentration and time (Benet 1979). Using the plasma data from Paul et al. 1993 study at 100 mg/kg bw atrazine (Paul et al. 1993), a Vdss estimate of 6.55 L/kg bw was obtained. Page 101 of 184 Detailed calculations are given in Appendix A.9. The high dose of 100 mg/kg bw atrazine was selected because plasma levels are well-defined at the later time points at this dose level as compared to 1 mg/kg bw (Paul et al. 1993). AUCM (mg*hr2/L) 0 (1) Vdss (L/kg bw) = Dose (mg/kg bw) * ∞ (AUC mg*hr/L)2 0 ∞ In an attempt to evaluate the one-compartment linear model in Figure 17, the corresponding equation that relates atrazine dose rate to steady state plasma triazine levels (i.e., equation 2) was used to predict the average pseudo steady state plasma radiolabeled equivalents reported in the Thede 1987 study. Using a Vdss of 6.55 L/kg bw and kel of 0.013 hr-1, both of which were estimated independently from the single dose plasma data from the Paul et al. 1993 study using non-compartmental pharmacokinetic approaches, equation 2 yielded predictions of steady state plasma triazines that are very consistent with those from the Thede 1987 study (Table 8). Table 8: Evaluation of the one-compartment linear model for relating an atrazine dose to steady state plasma triazines animal 11 animal 21 2 Atrazine dose rate (mg/kg/day) ave. steady state plasma levels (mg/L) ave. steady state plasma levels (mg/L)2 Predicted value 3 1 0.56 0.62 0.49 3 1.88 2.16 1.47 7 3.84 3.45 3.43 10 5.11 4.83 4.89 50 22.80 25.61 24.47 100 53.10 56.84 48.93 1 Data is from Thede 1987 2 average steady state plasma levels were estimated by taking the mean value of plasma levels from day 4 through 8, i.e., when pseudo steady state was reached and maintained. 2 Using the one-compartment linear model expression that relates dose rate to steady state plasma levels (Equation 2) Equation (2) can also be rearranged to estimate Vdss based on the plasma levels from the Thede 1987 study (Table 5). An average Vdss value of 5.71 ± 0.65 L/kg bw for atrazine doses of 1-10 mg/kg bw/day was obtained. Thus, this second approach gave a Vdss estimate that is consistent with that obtained through non-compartmental analysis. However, the Vdss estimate of 6.55 L/kg bw estimated from the Paul et al. 1993 study was selected since it is based on an independent dataset that reasonably reproduces the average pseudo steady plasma triazine levels (Css) at all doses tested in the Thede 1987 study. Page 102 of 184 (2) Css (mg/L) = Dose rate (mg/kg bw/day) Vdss (L/kg bw) * kel (hr-1) * 24hr/day It should be noted that the estimated value of 6.55 L/kg bw for Vdss is comparable to the range estimated for the monkey of 2.427 - 4.413 L/kg bw (Table 6). Thus, there is reasonable consistency across different rat studies (Paul et al. 1993 and Thede 1987) and species (rats and monkeys) for which there is plasma triazine information from dosing orally with radiolabeled atriazine. The one compartment linear model depicted in Figure 17 represents a reasonable approach at this time for informing the pharmacokinetic behavior and internal dosimetry of atrazine and its metabolites given that the recent PBPK model submitted by Syngenta is still being reviewed. However, the Agency is cognizant of the emerging science with the ultimate goal of having a well-calibrated and evaluated PBPK model for atrazines and its metabolites. 4.2.13 Implications for Water Monitoring Frequency The ongoing re-evaluation of the health effects associated with atrazine includes the drinking water monitoring frequency currently based on the 2003 risk assessment involving 90-day rolling averages of atrazine and its metabolites DEA, DIA, and DACT. Ideally the frequency of drinking water monitoring should be reflective of the temporal pattern of the toxicological endpoint of concern used in risk assessment, i.e., the duration of exposure needed to elicit the toxicological endpoint. Given that the state of science has changed for atrazine, the Agency is in position to re-evaluate current risk estimates based on the current water monitoring frequency. The temporal pattern of LH attenuation is related to the pharmacokinetic behavior and internal dosimetry of all relevant plasma triazines. There are two features in the pharmacokinetic behavior of triazines that can be used to inform the frequency for water monitoring frequency. First, the linear behavior that can be reasonably described by the simplified one compartment linear model in Figure 17. The analytical solution for a one-compartment linear model relates plasma levels of triazines to administered doses of atrazine (Equation 3). The linear behavior of plasma triazines can be reasonably expected to remain linear at lower doses of atrazines due to an even lower probability of saturating ADME processes. Second, upon repeated dosing with atrazine, plasma triazines reach (or nearly reach) pseudo steady state when LH attenuation is observed. The steady state condition indicates that absorption and distribution processes are offset by clearance processes and that while brain (a postulated site of action for LH attenuation) and plasma triazine levels may be not equal, they will be at equilibrium, making plasma triazine levels a reasonable surrogate for internal dose response analysis. Page 103 of 184 4.2.13.1 Rat Forward Internal Dosimetry for LH Attenuation The current understanding of the linkage between pharmacokinetics and LH attenuation following atrazine exposure in rat studies is depicted in Figure 18. An oral gavage dose of atrazine reasonably assumed to be completely absorbed from the GI tract (step 1) will be distributed in a body volume of distribution to give plasma levels of triazines (step 2) which can be cleared by the body with an overall first order rate constant (kel) that accounts for all plasma clearance processes (primarily urinary and biliary/fecal excretion). Upon continued exposure to atrazine (step 3), plasma levels of triazines will accumulate until steady state is reached (or nearly reached) when optimal LH attenuation is observed. The proposed critical duration is the time to reach steady state plasma triazines. It should be noted that other metrics such as the time above a critical threshold of plasma levels have also been considered. Based on the currently available information where 4 days is the shortest exposure duration evaluated (Cooper et al. 2010) and the remarkably similar LOAELs reported for studies of different durations (as short as 4 days and as long as 6 months), the Agency is currently presenting steady state plasma levels in its derivation of an internal dose metric associated with attenuation of the LH surge. As additional information becomes available and as the Agency completes its review of the recently submitted Syngenta’s PBPK model , the Agency will consider other internal dose metrics and adapt appropriately. RAT: FORWARD DOSIMETRY FOR LH ATTENUATION INPUT Atrazine Oral gavage dose (mg/day) (1) 100% absorption Absorbed Dose (mg/day) (2) Body volume of distribution (Vd, L) Plasma triazines (Cp, mg/L) kel (hr-1) clearance Continued exposure (3) Elimination OUTPUT bioaccumulation Steady state (Vdss, L) Steady state plasma triazines (Cpss, mg/L) (4) Duration Daily steady state AUC ((mg/L)*hr) Optimal LH attenuation Figure 18: Proposed linkage between atrazine exposure and optimal LH attenuation in rat studies Page 104 of 184 4.2.13.2 Human Reverse Dosimetry from Rat Internal Dosimetry for LH Attenuation The starting point for the human situation is the assumption of pharmacokinetic equivalency. In other words, if the same plasma triazine AUC as the rat is ever reached in humans, an equivalent LH attenuating effect can be reasonably expected (step 1 in Figure 19) (US EPA 2006). Notably, AUC is the product of levels times duration (i.e., “how much and for how long”). Therefore, a given plasma AUC may be “matched” by high levels and short duration or alternatively by moderate levels for a longer duration. The same one compartment linear model expression but with allometrically scaled human parameters (equation 4) can be used to match the rat AUC PoD associated with LH attenuation. These types of forward and reverse calculations provide the basis for species extrapolation in risk assessment particularly when a pharmacokinetic model is available. In the case of atrazine, these calculations can be performed using simplified empirical calculations based on the consistent linear pharmacokinetic behavior of triazines observed across different studies, doses of atrazine, and species including humans. (3) (Dose rate) rat Vdss rat * kel rat PoD Daily Plasma AUC (Dose rate) human = Associated with = Vdss human * kel human LH attenuation In contrast to animal toxicity studies where the frequency and level of dosing are controlled, the dose rate in humans is determined primarily by the water consumption rate and the chlorotriazine content of the water (step 3 in Figure 19), both of which are variable. In particular, water levels of chlorotriazines can be highly variable (See hypothetical water chemograph in Figure 20) due to being dependent on numerous factors including application rate, proximity to growing field, and time of season. Nonetheless, an estimate of the equivalent human time to reach steady state can be obtained via allometric scaling of the rat elimination rate constant and assuming that it would take between 3-5 plasma halflives to reach steady state. A rat kel of 0.010 -0.013 hr-1 allometrically scaled for a human body weight of 60 kg results in kel of approximately 0.00343 -0.00471 hr-1and a plasma half-life estimate of about 6 - 8 days. Thus, an equivalent human time to reach steady state plasma triazine levels based on a constant exposure level and frequency can be estimated to be between 21-30 days. The highly variable levels of atrazine and its metabolites in drinking water can be analyzed through the use of an integrated average daily level of exposure given the predicted duration of concern of 21-30 days. The observation that LH attenuation is driven by sustained rather than acute exposures to atrazine further supports the use of an integrated average daily dose metric. Rolling averages can be used to “carry over” exposures from previous days and avoid gaps in the exposure estimates. Page 105 of 184 HUMANS: PROPOSED REVERSE DOSIMETRY FOR LH ATTENUATION Rat Daily Steady State AUC (mg/L*hr) LH attenuation ≅ (1) Human average daily AUC ((mg/L)* hr) duration Average daily plasma concentration (Cpave, mg/L)) Equivalent time to steady state (Vdss, L) (2) Bioaccumulation Continued exposure Plasma triazine concentration (Cp, mg/L) kel (hr-1) * clearance Elimination Body volume of distribution (Vd, L) * Absorbed dose (mg) (3) Water Consumption (L/day) 100% absorption Chlorotriazines in water (µg/L) Figure 19: Proposed linkages between atrazine exposure and LH attenuation in adult humans. The asterisks indicate the parameters that were allometrically scaled from adult rats. Hypothetical Water Chemograph Triazine Levels 50 40 30 20 10 0 Critical duration Time Figure 20: Hypothetical water chemograph showing the typical variation in chlorotriazine levels as a function of time. Page 106 of 184 4.2.14 Calculating Average Chlorotriazine Levels in Drinking Water for Human Plasma Triazine AUC Estimates At the April 2010 Science Advisory Panel (SAP) meeting, the panel recommended that the Agency consider a drinking water monitoring approach based on the area under the concentration-time curve (AUC) for atrazine and its chlorinated metabolites that can be used in conjunction with an integrated internal dose response assessment (FIFRA SAPa). Based on the currently available data, the Agency is presenting the use of pseudo steady state plasma triazine levels to derive a plasma PoD AUC estimate associated with LH attenuation. This value can be compared to a daily average AUC for drinking water (AUCwater) for a given duration of concern on the basis that humans are unlikely to reach steady state plasma levels of triazines as in rat studies, (Step 2 in Figure 19). AUCwater can be estimated using the trapezoid rule whereby the area for a given exposure duration would be divided into individual trapezoids as dictated by the water sampling frequency. The sum of the individual trapezoids would be AUCwater. In the case the water sampling frequency allows the estimation of a single trapezoid, AUCwater may be under- or overestimated depending on the shape of the individual trapezoid as depicted in Figure 21A. In essence, the more frequent the water sampling, the better the estimation of AUCwater (Figure 21B). Time-weighted average levels of chlorotriazines in water can be estimated from AUCwater according to equation 4. Rolling averages can be computed to “carry over” exposures from previous days and avoid gaps in the estimates. In combination, equations 5 and 6 offers an approach for estimating a human average daily AUC for plasma triazines that accounts for an integrated dose rate for a duration of concern, body distribution and plasma clearance information. The resulting human plasma AUC can be compared to the rat PoD AUC estimate for a margin of exposure (MOE) calculation that is based on internal rather than external measures of exposure (Equation 6). B Chlorotriazine Levels Chlorotriazine Levels A time time Figure 21: The trapezoid rule for estimating the area under a water chemograph for a given duration of exposure. Page 107 of 184 (4) [atrazine]ave (mg/L) = AUCwater ((mg/L)*hr) for given duration # days * 24 hrs From water chemograph (5) Human average daily plasma AUC = From human consumption information [triazine water levels]ave (µg/L) * water consumption (L/day) kel (hr-1) * Vdss (L) Allometrically scaled from rat values (6) MOE = Rat PoD BMDL plasma AUC Human plasma AU C Page 108 of 184 5. SCIENTIFIC CONSIDERATIONS IN POTENTIAL SENSITIVITY OF INFANTS & CHILDREN 5.1 Background The FFDCA, as amended by the FQPA (1996), requires the Agency to give special attention to the potential risk to infants and children. Specifically, FQPA instructs EPA, in making its “reasonable certainty of no harm” finding, that in “the case of threshold effects, an additional tenfold margin of safety for the pesticide chemical residue and other sources of exposure shall be applied for infants and children to take into account potential pre- and post-natal toxicity and completeness of data with respect to exposure and toxicity to infants and children.” Section 408 (b)(2)(C) further states that “the Administrator may use a different margin of safety for the pesticide chemical residue only if, on the basis of reliable data, such margin will be safe for infants and children.” This additional margin of safety is referred to as the “FQPA Safety Factor.” In this section, the Agency describes the state of the science with respect to the potential for pre- and post-natal toxicity and the completeness of data with respect to toxicity to infants and children. This discussion focuses on experimental toxicology relevant for considering pre- and post-natal toxicity. Although exposure considerations are a component of the FQPA Safety Factor analysis, this section of the draft issue paper does not include drinking water exposure. Drinking water exposure and statistical approaches for assessing monitoring frequency are covered in Section 6. Several key experimental toxicology studies important for the FQPA analysis have been conducted since the September 2010 SAP. These include studies evaluating 1) the potential hormonal changes and outcomes from gestational exposure to male and female rats; 2) behavioral changes in male rats (e.g., rough and tumble play) following gestational exposure (GD 14-21) and 3) other possible latent effects manifested in adults from gestational and lactational exposure. Although additional experimental toxicology studies are still on-going to better characterize the potential adverse health outcomes resulting from atrazine exposure (including the duration of exposure that may lead to an adverse health outcome), the available data do not indicate that pre- and/or post-natal exposure leads to increased sensitivity in the young relative to the attenuation of the LH surge that serves as the basis for the atrazine risk assessment. The Agency is evaluating potential approaches for assessing drinking water monitoring strategies. The uncertainties and confidence in the performance of water monitoring sampling strategies to potentially capture atrazine levels exceeding the Agency’s level of concern are an important consideration in the determination of the FQPA Safety Factor. Section 6 of this issue paper will provide examples of how these uncertainties might be addressed by evaluating the performance of different sampling strategies and some accompanying models for their interpretation. As such, this draft issue paper does not propose a new FQPA Safety Factor. Instead, the Agency will solicit comment from the SAP on 1) the current state of the Page 109 of 184 science with respect to assessing potential sensitivity to the young and 2) the overall approach for determining the FQPA Safety Factor. 5.2 Experimental Toxicology 5.2.1 Gestational Exposure Recently, EPA’s ORD conducted a series of experiments to further evaluate the impact of gestational exposure on reproductive development in both males and females. These studies in conjunction with those evaluated by the Agency in preparation for the April and September 2010 SAPs provide insight into the overall pattern of toxicity associated with atrazine exposure during this lifestage not only in terms of the dose levels but also the critical durations of exposure capable of eliciting an adverse effect. With respect to toxicity outcomes following gestational exposure, Fraites et al. (2011) did not observe effects on male reproductive development or the androgen-dependent endpoints measured in the study after in utero exposure during gestation (GD 14-21) including (i) testosterone production at birth and on PND 59, (ii) rough and tumble play behavior, (iii) AGD and PPS, and (iv) androgen-dependent organ weights at doses as high as 100 mg/kg/day. This is consistent with the findings reported by Rayner et al. (2007) who observed no change in the timing of male puberty, but did report a higher incidence in prostatitis at 100 mg/kg/day. In contrast, Rosenberg et al. (2008) reported delays in PPS at 50 mg/kg/day. Another high dose effect reported after gestational exposure to atrazine is a delay in mammary gland development of female offspring (Rayner et al., 2005, 2007). This effect, however, was not replicated by Davis et al. (2011) at doses as high as 100 mg/kg/day when evaluated either using a subjective scoring approach (as described by Rayner and coworkers) or a morphometric analysis. Table 9 summarizes the toxicological data available on atrazine after gestational exposure. Detailed reviews for all the studies in the table are part of Appendix A. Page 110 of 184 Table 9: Summary of available atrazine experimental toxicology studies from the gestational period AUTHOR (YR) Rosenberg (2008) Rayner et al. (2005) Rayner et al. (2005) Rayner et al. (2007) EXPOSURE NOAEL/LOAEL mg/kg/day Gestation GD14parturition 10/50 GD 13-19 NA/100 GD 17-19 NA/100 GD 15-19 NA/100 Fraites et al. (2011) GD 14-21 20/100 Davis et al. (2011) GD 14-21 20/100 EFFECT Delayed PPS Delay in mammary gland development Delay in mammary gland development Prostatitis Decreased pup weight and pup viability (PND 14)† Decreased pup weight and pup viability (PND 04). Delay in VO † Decreased maternal body weight was also observed at the 100 mg/kg/day dose level 5.2.2 Multi-Lifestage Exposure A multifaceted study evaluating the impact of atrazine exposure across several lifestages has been submitted by the atrazine registrant – Syngenta. The purpose of this study was to evaluate the effects of atrazine on sexual maturation, estrous cyclicity, and the LH surge in Sprague-Dawley [Crl:CD(SD)] rats following at atrazine doses of 0, 6.5, 25 or 50 mg/kg/day administered via gavage. Animals were divided into two cohorts. Cohort I animals were exposed from the time of conception through sexual maturation (VO) or PND 133. Half of the animals dosed until PND 133 were evaluated after a 20 days recovery period (PND 134-154) while the remainder of the animals were evaluated on PND 133. Cohort II animals were exposed only post-natally from PND 21 through sexual maturation (VO + 5 days) or for 14 days as adults of reproductive age. Table 10 (excerpted from MRID 48381001) describes the dosing regimen for these animals. Page 111 of 184 Table 10: Dosing regimen Multi-Lifestage Study Cohort I animals (all subsets) exposed to 50 mg/kg/day atrazine exhibited a 1.4-2.3 day delay in VO (mean = 1.6 day delay). Unlike the findings reported by several investigators (Foradori et al., 2009; Cooper et al., 2007; Morseth et al., 1996, Davis et al., 2011)), no LH surge attenuation was observed at any dose level. Corticosterone and estrous cyclicity also appeared to be unaffected by atrazine exposure. The study author, however, acknowledged that the corticosterone measures were not obtained at the optimal time and this may account for the discrepancy between the observations in this study and findings previously reported by other researchers including Laws et al. (2009), Fraites et al. (2009), Morseth et al. (1996), and Cooper et al.(1996). In the case of Cohort II animals, no effect on VO was noted in animals whose exposure began on PND21 (prior to the typical age of VO) at any dose level. LH surge and corticosterone levels also appeared to be unaffected in this cohort (all subsets). A unique feature of this study is that it is the only one available in the atrazine database that evaluates the impact of atrazine exposure through multiple lifestages. The preponderance of the data available to date supports the conclusion that LH surge attenuation occurs at doses lower than those used in this study. The reason for the inconsistency between this study and the body of peer reviewed literature evaluating the impact of atrazine exposure across different life stages over the last 15 years is unclear. A detailed review of this study is included in Appendix A of this document. 5.3. Summary During the September 2010 SAP meeting, the Panel discussed several studies evaluating the impact of atrazine exposure during development. At that time, the Agency concluded – and the SAP concurred – that the available data did not identify a unique quantitative susceptibility in the developing organism. None of the available studies with atrazine evaluating the rats exposed during gestation, lactation or the peri-pubertal periods have Page 112 of 184 shown effects at doses lower than those eliciting the LH surge attenuation in adult female rats after 4 days of exposure. Three additional studies evaluating the effect of atrazine exposure across lifestages have become available within the last few months. These studies reinforce the conclusions reached during the September SAP meeting since all of the effects observed in the young in these set of studies occurred at doses ≈ 25 times higher than the dose EPA is proposing to use (BMDL of 2.56 mg/kg/day) as the PoD for human health risk assessment (derived from LH data collected after four days of exposure to adult females). In making determinations regarding the FQPA Safety Factor, not only should potential lifestage sensitivity be considered but the need to ensure the exposure assessment will not underestimate risks during development. During the September 2010 SAP meeting the Agency proposed a water monitoring frequency between a few days and four weeks based on durations of exposure that may result in adverse human health outcomes. In its December 2010 report, the SAP commented that based on the assumptions and extrapolations to the proposed critical window of exposure and the overall uncertainties these assumptions introduce, “the imprecision in the Agency’s proposed sampling frequency seems justified. This may be about as precise an estimate as can be obtained when starting with the experimental animal data and the exposure requirements for LH surge suppression…” Based on the SAP’s feedback, the Agency will continue to use attenuation of the LH surge as the critical effect for the atrazine risk assessment and will conduct an analysis of the water monitoring frequency ranging from a few days to 4 weeks. Section 7 of this draft issue paper will present various approaches to assess the inherent uncertainties associated with various drinking water monitoring strategies. Page 113 of 184 6.EVALUATING ATRAZINE DRINKING WATER MONITORING DATA FOR USE IN HUMAN HEALTH ASSESSMENTS 6.1 Introduction The 2003 risk assessment for atrazine (USEPA, 2003) determined that LH surge attenuation was the most sensitive toxicological endpoint for the human health risk assessment associated with exposure to atrazine and other chloro-triazines. Based on the best available information at the time, the USEPA identified a 90-day critical duration of exposure, based on results showing LH surge attenuation following dietary dosing in a 6 month toxicity study and corresponding to the period within the year during which most elevated concentrations of atrazine are likely to occur in drinking water. As a condition of reregistration, the USEPA required the registrant, Syngenta, to monitor vulnerable community water systems (CWS) to ensure that 90-day average concentrations did not exceed the toxicological level of concern (37.5 µg/L total chloro-triazines) for that duration. Syngenta proposed monitoring weekly during the use/ runoff season and bi-weekly during the rest of the year to estimate total chloro-triazine (TCT) exposure over a 90-day time frame for both source (raw) and treated (finished) waters at vulnerable CWS in the atrazine use area. Any CWS that exceeded annual mean concentrations of 1.6 µg/L atrazine based on quarterly Safe Drinking Water Act (SDWA) sampling was included in this more intensively sampled Atrazine Monitoring Program (AMP). Since 2004, the AMP has monitored TCTs in water at up to 166 CWS. No 90-day average concentration has exceeded the toxicological level of concern in that time. New data related to the hazard potential and dose-response for atrazine resulted in this reevaluation of human health effects for atrazine. Those studies suggest a shorter duration of exposure may be appropriate for the toxicological endpoint of concern. With a shorter duration of concern, the USEPA needs to consider how well the existing AMP monitoring design characterizes shorter durations of exposure and whether the monitoring program needs to include more frequent sampling or other means to augment existing data. In the April and September 2010 FIFRA SAP consultations on atrazine, the USEPA reviewed methods for designing monitoring studies to capture exposures of concern and approaches for analyzing and interpreting existing monitoring data and characterizing the uncertainties in those estimates for use in human health risk assessments (USEPA, 2010a and 2010b). The SAP commented that the toxicological exposure timeframe of interest will define the importance of peaks and determine the most useful approaches both for designing a monitoring study and for evaluating the utility of existing monitoring studies (FIFRA SAP, 2010a). However, the SAP noted that, in light of uncertainty over a critical exposure period ranging from a few days to a few weeks, the best course of action may be attempting “to capture the pattern of atrazine concentrations in the source water of each CWS based on the characteristics of that particular water system, as opposed to a onesize-fits-all approach” (FIFRA SAP, 2010b). Page 114 of 184 Identifying monitoring sites and determining the sampling frequency require careful consideration to avoid underestimating true atrazine concentrations (FIFRA SAP, 2010b). A predictive model can be used to target sites that merit the most detailed monitoring and target the time periods when atrazine is most likely to be present (FIFRA SAP, 2010a, 2010b). The Panel recommended that the USEPA give thought to “using the simulation models and CWS characterizations as part of the monitoring process. In particular, it is feasible that models will eventually be accurate enough to provide predictions of atrazine concentrations in source waters to a CWS for the coming crop season. Instead of requiring a CWS to collect and analyze water samples in their output stream (drinking water) at some predefined frequency (e.g., daily or weekly in the case of some sites), it should be possible to use the models to facilitate targeting sampling to periods of time most likely to experience an exceedance.” (FIFRA SAP, 2010b) To estimate exposures from monitoring data sampled at intervals, particularly where peak concentrations over shorter durations are important, the USEPA needs methods that can predict values that may be greater than those sampled (FIFRA SAP, 2010a). The SAP recommended combining regression-based models, such as the United States Geological Survey’s (USGS) Watershed Regression on Pesticides (WARP), with either a deterministic model or with a statistical approach using kriging or random-function models (FIFRA SAP, 2010b). In preparation for the upcoming registration review for atrazine scheduled to begin in 2013, the USEPA has focused on methods for characterizing the uncertainties resulting from the existing AMP monitoring study design. Such methods include providing the uncertainty bounds in estimated exposures based on how well the sampling design captures the temporal variability of pesticide concentrations in water at locations where concentrations are expected to be greatest. Additional uncertainties related to the number of years of monitoring and site locations are also considered. While the discussion in this section focuses on estimating atrazine concentrations, the human health assessment will be based on TCTs. In most surface water sources, the elevated exposures will be dominated by atrazine; in some instances, simazine or chloro-triazine degradates of atrazine may also occur. 6.2 Key Questions to Address in Analyzing Monitoring Data Pesticide occurrence in water is the result of a complex interaction of pesticide properties, use patterns, application methods and timing, weather, watershed characteristics (e.g., soil, slope, land cover type), and hydrology (Gilliom et al., 2006). While pesticide concentration patterns resulting from this interaction are difficult to adequately characterize both geographically and temporally with monitoring (ILSI, 1999; USEPA, 2000; Stone and Gilliom, 2009), an understanding of the factors affecting pesticide distributions in water can be used to predict pesticide occurrence in place and time and to characterize the degree of uncertainty resulting from a monitoring study design for use in exposure estimates. Page 115 of 184 For a pesticide with an established use pattern, such as atrazine, the variability of pesticide occurrence in surface waters can be described by: • The spatial pattern related to pesticide use, crop and management practices, soil and hydrologic vulnerabilities, and rainfall distribution; • Seasonal and daily temporal patterns at any given location related to intensity and timing of pesticide applications and coincidence of rainfall events; • A temporal correlation in which pesticide concentrations found on one day are related to concentrations occurring on the previous and following days; • Year-to-year temporal patterns at any given location reflecting changes in cropping and pesticide use as well as variations in rainfall from year to year. A well-designed surface water monitoring study takes into account both spatial and temporal patterns of exposure, focusing sampling on when and where the pesticide is used. Considerations for sampling frequency include not only the duration of toxicological concern, but also the pesticide properties and use patterns, expected interactions of the pesticide with the environment, and the nature of the water body and watershed. The USEPA considers the following questions in determining the extent to which monitoring data can be incorporated in exposure assessments: • Given monitoring of a specified sampling frequency, what can we say about the actual structure (magnitude, frequency, duration, time series distribution) of the concentrations at that site in a given year? This relates to the short-term temporal variability and associated temporal correlations in pesticide occurrence in water. • What can we say about the nature of pesticide occurrence at the given site in other years? How much variability in magnitude and duration of exposures can we expect from year to year? This relates to longer-term temporal variability. • What can we say about pesticide exposures at other sites based on the results at one or a few sites? This relates to the spatial variability of pesticide concentrations in water. • Given the spatial variability in pesticide occurrence, how do we link monitoring data with sound statistical designs? How do we deal with non-random monitoring designs, such as the current atrazine CWS monitoring program? • How do we combine monitoring results across datasets, taking into account such factors as watershed sizes, water body types, climate/weather differences, and different study purposes and designs? Page 116 of 184 • How do we best analyze datasets with censored data (i.e., below the limit of quantification)? This question spans a number of issues, including timing (were samples taken when the pesticide was most likely to occur in water?), spatial distribution (were samples taken from areas where the pesticide was used?), and analytical methods (are the analytical methods sufficient to detect the pesticide at concentrations of concern?). Determining the reason for non-detects can be useful in interpreting the extent of risk both spatially and temporally and can aid in the development of risk mitigation practices, if necessary. • How do we best integrate modeling and monitoring data to provide water exposure estimates? Each question addresses aspects of the uncertainties in monitoring analyses that need to be considered in the context of the risk assessment. The recent series of SAP consultations on human health and drinking water for atrazine have focused on addressing the short-term temporal variability based on monitoring sampling frequency. Section 6.3 evaluates a cross-section of approaches for evaluating monitoring data and uncertainties, ranging from conventional statistical analyses to mechanistic models and combinations in between. However, in determining the uncertainty in monitoring data, the USEPA must also consider year-to-year variability. How many years of monitoring are sufficient to characterize the likely range in exposures resulting from year-to-year variability? Section 6.4 briefly explores ways of using monitoring and modeling together to estimate the potential long-term range pesticide concentrations at a given location. The USEPA typically addresses the spatial variability in pesticide concentrations in water by focusing on vulnerable water bodies that are expected to have the greatest exposures. For the atrazine monitoring program, CWS were considered vulnerable based on quarterly compliance monitoring data. Section 6.5 will provide a brief discussion on spatial variability and sound statistical designs, as well as potential applications of lessons learned regarding watershed characteristics that contribute to higher atrazine concentrations from a separate assessment of atrazine monitoring for ecological exposures (USEPA, 2007, 2009a, and 2009b). 6.3 Short-term Temporal Variability: Daily Variations within a Given Year In a given year, pesticide concentrations in surface water bodies generally occur at low or undetectable levels except during the active use period, where short-term seasonal pulses of higher concentrations are related to the timing of pesticide use and rainfall patterns. Such short-term pulses in pesticide concentrations range from less than a day to several weeks, depending on the interaction of pesticide properties, use practices and patterns, watershed characteristics, weather, and management practices. Atrazine, with relatively widespread use and properties favorable to pesticide transport (mobility and persistence), Page 117 of 184 may be detected year-round, with periods of high concentrations ranging from days to weeks during the spring and summer. Consequently, monitoring data for atrazine provides a good test for evaluating alternative methods for data analysis, as trends in time and space can be observed above the level of quantification. Within this seasonal pattern, pesticide concentrations in water show levels of temporal autocorrelation (serial correlation) that are not well accounted for with traditional statistical methods (Crawford, 2004). Pesticides such as atrazine show a serial correlation in which the probability of detecting the pesticide in water on a particular day is greater when the pesticide was detected on the previous day. Sampling designs and methods of interpreting monitoring data need to account for the temporal co-dependence of pesticide concentrations in water: the exposures measured on one day are related to the exposures measured on preceding and subsequent days. The extent to which monitoring studies capture this short-term variability depends on how frequently and how targeted to the use season the sampling events occur. The AMP study focused weekly samples during the anticipated high use/runoff period to estimate 90-day average concentrations of concern. To estimate shorter duration exposures from noncontinuously-sampled monitoring data, particularly where peak concentrations may become more important, the USEPA needs methods that can predict values that may be greater than those sampled (FIFRA SAP, 2010a). The Agency considered several approaches for characterizing time series distributions and uncertainties based on sampling design, beginning with statistical simulations and moving on to geostatistical approaches, combinations of WARP with statistical models, and deterministic models. While the emphasis is on analyzing the CWS monitoring data collected for atrazine, such approaches could be applied to other monitoring datasets and other pesticides. The USEPA used a limited number of intensive monitoring datasets to provide a proof-ofconcept for the approaches presented below. A more complete analysis, involving the full set of intensively sampled datasets, grouped by watershed area, water body type, and geography, could be used for estimating uncertainty factors and time series for the weekly CWS monitoring data collected for the AMP. 6.3.1 General Approach for Assessing Uncertainties in Monitoring Related to Short-term Temporal Variability To assess how well monitoring at a specified sampling frequency estimates the actual magnitude, frequency, and duration of the concentrations at that site, we must consider daily variations in pesticide concentrations and the temporal auto-correlation in those concentrations. The USEPA is considering whether shorter durations of exposure (ranging from 4 to 28 days) are more appropriate for assessing atrazine effects on humans. The Page 118 of 184 question applied to atrazine is how well exposure estimates based on 7-day sampling intervals during the high use season reflect actual exposures for 4-, 14-, or 28-day toxicity durations of concern, as well as for the current 90-day duration. In this analysis, the USEPA also evaluated uncertainty in estimates based on different sampling intervals (ranging from 4 to 28 days). The SAP emphasized that evaluations of different exposure estimation methods need to be benchmarked against intensive, preferably daily, monitoring data for an adequately representative range of sites (FIFRA SAP, 2010a and 2010b). While the AMP covers the geographic range of intensive atrazine use, the weekly sampling interval is not sufficient because it provides a “biased representation” of actual concentration profiles (FIFRA SAP, 2010a). The USEPA proposed using the more intensively sampled monitoring data from Heidelberg University’s National Center for Water Quality Research (NCWQR) and the more frequently-sampled headwater streams from Syngenta’s Atrazine Ecological Exposure Monitoring Program (AEEMP) to evaluate sampling frequency and exposure estimation methods (USEPA, 2010a and 2010b). The NCWQR (formerly known as the Heidelberg College Water Quality Laboratory) has intensively sampled a number of pesticides, including atrazine, on selected Ohio rivers since the 1980s (NCWQR, 2010). The Heidelberg dataset reflects frequent sampling over many years in a geographically limited area. Syngenta’s AEEMP includes a subset of sites with intensive atrazine measurements using an autosampler for a number of headwater watersheds across the atrazine use area. While this dataset adds to the geographic diversity in monitoring, it represents watersheds that are smaller than most watersheds that supply CWS. Thus, generalizations from these datasets need to be done in the context of both geographic distribution and representative watershed sizes. With this context in mind, these two datasets do provide a basis for evaluating the ability of data analysis methods to evaluate the uncertainty in monitoring datasets and their ability to predict the likelihood of pesticide concentration excursions outside the range of the measured data in between sampling events. In keeping with SAP recommendations, the USEPA will attempt to match the intensively-sampled sites to CWS sites based on water body and watershed characteristics to the extent possible (FIFRA SAP, 2010b). Syngenta submitted a report summarizing a plan to monitor daily during the use season at six CWS that are included in the AMP (Merritt, 2011). The six CWS – three in Kansas, two in Indiana, and one in Ohio – draw water from single sources of flowing water. This monitoring data will provide more intensive sampling of source and treated water samples at CWS drawing from flowing waters that can be used to further evaluate the approaches proposed in this section. Analogous daily monitoring data do not exist for reservoirs or other static water bodies. A different temporal exposure profile would be expected from reservoirs, which integrate pesticide loading from their watersheds and have slower flowthrough, resulting in a greater residence time and smaller temporal variations than flowing water bodies (Blomquist et al., 2001). The SAP suggested collecting more intensivelyPage 119 of 184 sampled data for other representative water bodies, such as reservoirs (FIFRA SAP, 2010b). An effort to target CWS drawing from a single reservoir water source similar to that described in Merritt (2011) could be completed in time for the 2013 registration review, providing additional data to analyze uncertainties in sampling frequencies for static water bodies. Given the uncertainty in defining the duration of toxicological concern, the USEPA plans to use a multi-step approach for analyzing the uncertainty in exposure estimates from monitoring data. The first step is a statistical analysis of monitoring frequency by simulating different sampling frequencies using the intensively sampled datasets described above. This approach, described in Section 6.3.2, addresses the question: Given a sampling frequency of X (samples per month or average interval between samples) and an exposure duration of concern of Y (4 to 28) days, what is the relative performance of each sampling strategy (X), based on measures representing the differences of its estimates (Y) from the known true values? These differences may be used to quantify the uncertainty from sampling frequency. A bias or uncertainty factor generated from the analysis could be applied to bracket the exposure estimates from the weekly CWS samples for comparison against the concentration and duration of concern. If the bracketed exposure estimates exceed the concentration of toxicological concern for the specific duration of concern, further analysis using geostatistical methods, hybrid geostatistical-WARP, or deterministic modeling approaches described in Sections 6.3.3 – 6.3.5 may be needed. The case study (Section 7) illustrates how a time series estimated from weekly monitoring using geostatistical models could be incorporated into the dietary exposure assessment. 6.3.2 Comparative Analysis of Monitoring Frequency Performance To evaluate the relative performance of different monitoring frequencies for estimating exposure windows of specific durations, the USEPA performed a simulation analysis on two example sites representing two different watershed areas for which daily or near daily concentration measurements are available. The AEEMP MO-01 data from 2007 and the NCWQR Maumee River data from 1995 will be used in example analyses (see Appendix D.1 for the measured atrazine concentrations for each of these data sets). From the daily to near daily concentrations measured at from April through August/ September in each year, the USEPA divided the data into sampling intervals of 4, 7, 14, or 28 days and sampled at random from each interval 10,000 times. The resulting compiled chemographs, derived from the known true profile, represented trials that reproduced sampling within 4-day, 7-day, 14-day (biweekly), or 28-day (monthly) windows. Intervening values were estimated for each resulting chemograph by linear interpolation, with daily, 4day, 14-day, 28-day, and 90-day running averages calculated for each sampling. The Page 120 of 184 maximum value for each of these running averages was compiled into a distribution, resulting in five distributions of 10,000 maxima for each estimated average. Percentiles of those distributions of maximum values (Tables 11 and 12) are compared against the site’s true maximum value for each measurement. Tables 11 and 12 provide the results for each example site, along with a simple ‘bias’ factor for assessing the varying average duration estimates (peak, 4-, 14-, 28-, and 90-day averaging periods) based on the 4-, 7-, 14-, and 28-day sampling intervals. The multiplicative ‘bias’ factor in the tables is simply the ratio of the true maximum to the 5th percentile maximum for each estimate derived from the 10,000 runs. For example, with 4day sampling intervals, estimates of the 1-day peak can be nearly 4 times lower than the true peak concentration. Sampling within a 2-week window can return estimates of the 4day average that are over 10 times lower than the true 4-day average. (Minus signs on the factors simply note an underestimation of true values.) The bias factors in Tables 11 and 12 assume that the sampled concentrations reflect an average across the day. This assumption only holds true for autosampler data collected for MO-01, which integrated time sips across a 24-hour period. Grab samples collected during a single time period within a day have additional uncertainty related to intra-day variability in pesticide concentrations in the water body. This intra-day variability is not reflected in the analysis below. The bias factors for estimating daily concentrations run from about 1.5X lower to more than 50X lower for the 2007 MO-01 data, increasing as the sampling interval increases. Sampling within every 4 or 7 day period underestimates the true average, but not by as large a factor as with biweekly or monthly sampling intervals. Page 121 of 184 Table 11: Summary of 10,000 simulations of monitoring estimates at different sampling intervals for the AEEMP MO-01 site in 2007. th th th th Averag -ing Period Sampling Interval True Max (ppb) Lowest Max (ppb) 5 %ile Max (ppb) 10 %ile Max (ppb) Bias Factor th (5 %ile) 50 %ile Max (ppb) 95 %ile Max (ppb) Daily Peak 4 days 7 days 14 days 28 days 4 days 7 days 14 days 28days 4 days 7 days 14 days 28days 4 days 7 days 14 days 28 days 4 days 7 days 14 days 28 days 91.6 91.6 91.6 91.6 59.7 59.7 59.7 59.7 27.0 27.0 27.0 27.0 15.1 15.1 15.1 15.1 8.1 8.1 8.1 8.1 23.2 17.2 2.6 0.97 18.9 15.6 2.5 1.0 11.2 11.4 2.2 1.0 6.7 7.2 1.7 1.0 4.3 4.0 1.3 0.8 23.4 19.0 6.2 1.57 22.8 17.4 5.9 1.6 15.3 12.4 4.8 1.5 8.9 7.7 3.5 1.4 5.5 4.9 2.1 1.2 26.2 23.2 8.1 2.58 23.1 20.3 7.6 2.5 16.6 13.1 6.4 2.4 9.8 8.2 4.8 2.2 5.8 5.1 2.8 1.5 -3.9 -4.8 -14.8 -58 -2.6 -3.4 -10.1 >-37 -1.8 -2.2 -5.6 -18.0 -1.7 -2.0 -4.3 -10.8 -1.5 -1.7 -3.9 -6.7 71.8 33.3 23.2 14.26 57.5 29.4 21.7 14.0 29.6 20.7 17.8 13.2 16.8 13.2 12.6 11.9 7.9 7.2 7.7 5.5 >91 >91 >91 71.8 76.4 82.2 85.0 69.4 41.9 56.0 69.3 63.0 23.1 34.3 46.8 54.3 10.8 14.4 17.9 23.7 4-day avg. 14day avg 28day avg. 90day avg. In contrast, data for the Maumee River generally show much lower bias factors compared with those for MO-01 (Table 12). Page 122 of 184 Table 12: Summary of 10,000 simulations of monitoring estimates at different sampling intervals for the NCWQR Maumee River site in 1995. th th th th Averag -ing Period Sampling Interval True Max (ppb) Lowest Max (ppb) 5 %ile Max (ppb) 10 %ile Max (ppb) Bias Factor th (5 %ile) 50 %ile Max (ppb) 95 %ile Max (ppb) Daily Peak 4 days 7 days 14 days 28 days 4 days 7 days 14 days 28days 4 days 7 days 14 days 28days 4 days 7 days 14 days 28 days 4 days 7 days 14 days 28 days 14.1 14.1 14.1 14.1 11.7 11.7 11.7 11.7 7.8 7.8 7.8 7.8 6.3 6.3 6.3 6.3 4.4 4.4 4.4 4.4 8.9 5.7 4.6 3.4 7.9 5.3 4.5 3.3 5.7 4.4 4.3 3.1 5.3 4.4 3.9 2.8 3.7 3.1 2.5 1.8 8.9 6.2 4.7 4.4 8.1 6.0 4.7 4.4 6.3 5.4 4.7 4.3 5.7 5.1 4.4 4.1 4.0 3.7 3.3 3.0 8.9 7.4 5.1 4.5 8.3 7.0 5.0 4.5 6.4 5.9 4.9 4.4 5.8 5.3 4.6 4.2 4.1 3.8 3.4 3.1 -1.6 -2.3 -3.0 -3.2 -1.4 -2.0 -2.5 -2.7 -1.2 -1.4 -1.7 -1.8 -1.1 -1.2 -1.4 -1.5 -1.1 -1.2 -1.3 -1.5 10.6 9.9 7.6 5.7 9.8 9.0 7.4 5.6 7.3 7.3 6.9 5.4 6.2 6.3 6.1 5.1 4.4 4.3 4.2 3.9 >14.03 >14.03 >14.0 12.3 >13.3 13.2 13.3 12.0 9.5 10.8 11.3 11.1 7.1 7.9 9.1 9.9 4.7 5.1 5.7 6.4 4-day avg. 14day avg 28day avg. 90day avg. In these examples, as in all sites, the differences are directly related to the nature of the data and the magnitude of difference between what may be a short-lived peak and the concentrations preceding and following it. Concentrations at Maumee are lower overall, thus making the magnitude of potential bias smaller with regard to missed peaks. In addition, if the duration of peak values extends over the course of a week instead of for a day or two, then sampling within a 4-day window is more likely to ‘pick up’ some portion of the elevated concentration profile. For averages of longer duration, potentially missed peaks obviously have less impact on the estimates. These differences in the nature and magnitude of the profiles can be attributed, at least in part, to differences in watershed size and hydrology but may also reflect differences in atrazine use intensity in the watersheds. Based on these two examples only, an uncertainty factor for a 7-day sampling frequency, as used for the CWS in the AMP, might be in the range of 2-4X for estimating a 4-day average concentration and 2X for estimating a 14- or 28-day average concentration. Using these examples for the 90-day average period from the 2003 assessment (USEPA, 2003), Page 123 of 184 the bias factor would be less than a factor of 2 for 7-day sampling intervals. Differences are expected to exist for sampling from larger watersheds and from reservoirs, and for other conditions that a larger, representative sample of watersheds might show. Bias factors may tend to be slightly lower for reservoirs, for example, based on both differences in size and on the ‘smoothing out’ we would expect to see as water from streams/rivers reach the larger static water body. They may be higher where sampling is primarily from small rivers. The USEPA’s planned analysis of intensive monitoring representative of a range of watersheds and water bodies will likely result in a smaller uncertainty factor for CWS that are supplied by reservoirs with a large watershed and a larger uncertainty factor for CWS supplied by flowing waters with smaller watersheds. Syngenta has submitted a report using this basic methodology, but covering a larger sample of sites and years. In that report, Mosquin et al ( 2011b) examined data from three different sources: (1) one Community Water System (St. Louis/St. Charles, 1993-98 and 2000) with results for treated (“finished”) water; (2) four rivers/streams from the NCWQR data set (Maumee River, Sandusky River, Honey Creek, and Rock Creek) for the years 1993-2008; and (3) sites from the Atrazine Ecological Exposure Monitoring Program (AEEMP) with autosampler data (11 sites from 2009, and 23 sites from 2010). Simulating sampling frequencies from 2 days to 7 days on the near-daily data, Mosquin et al (2011b) calculated 1-day (peak), 2-day, 3-day, 4-day, 7-day, and 28-day average concentrations. Performance was measured by examining % error, % bias, and % relative mean squared error (RMSE). The authors calculated exact measures of performance for systematic sampling, in addition to performing the stratified random sampling as described above, running 1,000 simulations of each combination of sampling interval and average duration. For 7-day sampling intervals, Mosquin et al (2011b) found up to -77% error for estimating daily peaks and -67% error for estimating 4-day averages. They did not investigate sampling periods of longer than 7 days, where performance would be expected to decline. It is not always clear how the percentiles were compiled, however, and results for the AEEMP sites have been presented as a group. In the AEEMP, particular sites and years can have fairly high values that would be missed in certain instances by 7-day grab samples. Bias factors corresponding to these error estimates, as a result, are expected to be much greater when individual sites and years are examined. The USEPA has presented results for two example sites to demonstrate both some general concepts as well as the difference between results for two particular site/types, and it is clear that the discussion can be greatly enhanced by the addition of more data. The Agency believes, however, that a more helpful approach would be to present summary statistics on the low end percentiles of the simulated runs by site and by year, rather than to present percentiles of the grouped data. Important differences exist among sites and from Page 124 of 184 one year to the next at the same site (see Section 6.4). A protective approach will highlight these differences where they exist. 6.3.3 Geostatistical Analysis and Stochastic Application The September 2010 SAP noted that combining a deterministic model, such as PRZM or USGS’s SEAWAVE-Q (used in Sullivan et al, 2009), with a regression-based model, such as WARP, offers “the most promising approach to deal with sparse data in flowing water” (FIFRA SAP, 2010b). However, because of the difficulty in implementing these methods, the SAP suggested that it may be easier to use WARP in combination with a statistical approach such as kriging (FIFRA SAP, 2010b). This section describes two approaches that build on geostatistical analysis and stochastic simulations: (1) conditional stochastic simulations from variography with kriging and (2) using soft (indirect) data when monitoring data are sparse. This second approach involves regressions with WARP combined with variography and stochastic simulations. The full analysis can be found in Appendix D.2 (Geostatistical Analysis and Conditional Stochastic Simulation Application). The geostatistical modeling approach selects a stationary random function model to describe the relationship of variance as a function of distance or time (Isaaks and Srivastava, 1989). This relationship can be quantified using a variogram, correlogram, or covariance function. For a pesticide time series, this nearest-neighbor modeling approach assumes that concentrations in close proximity in time are more similar than concentrations farther apart in time. Applying a random function model to pesticide time series generally requires a condition of stationarity: that is, an estimate has an equal probability of occurring regardless of location in space and time (Isaaks and Srivastava, 1989). This may not hold for an entire year of pesticide concentrations due to seasonal trends, but may be applicable to specific times during the pesticide use season. To address stationarity concerns associated with seasonality of pesticide concentrations in water, the USEPA targeted the monitoring analysis from the earliest planting date to approximately 100 days post-planting. This represents the time with the highest magnitude and frequency of atrazine detections. The USEPA chose two intensively sampled monitoring datasets – NCWQR Maumee River monitoring for 1995 and AEEMP monitoring for MO-01 in 2007 – to evaluate the approach for different years and hydrologic characteristics (watershed size and stream flow characteristics). We simulated 4-, 7-, 14-, and 28-day sampling intervals with the monitored time series. Sample sets with less than 15 points – the 14- and 28-day sampling intervals – provided the poorest variograms for predicting daily pesticide concentrations from kriging. In those instances, we first used WARP to estimate the percentiles in the time series and evaluated options for fitting the merged WARP and sampled data in a new time series. Figure 22 illustrates the analysis strategy used for time series modeling from the sampled monitoring data. Page 125 of 184 DETERMINE APPROPRIATE POINT ESTIMATION PROCEDURE FOR TIME SERIES INFILL TIME SERIES WITH MODEL ESTIMATED VALUES NO ARE THERE MORE THAN 15 DATA POINTS IN TIME SERIES? YES PROCEED WITH VARIOGRAM ANALYSIS PROCEED WITH 1DIMENSIONAL ORDINARY KRIGING MERGE MONITORING DATA USE WARP MODEL TO ESTIMATE PERCENTILES IN TIME SERIES USE COVARIATE VARIABLE TO ESTIMATE CONCENTRATION ARRANGE PERCENTILES OF FLOW DATA WITH MERGE DATA PERCENTILES SEQUENTIAL STOCHASTIC SIMULATION- GENERATE EQUALLY PROBABLE REALIZATIONS OF TIME SERIES Figure 22: Analysis Strategy for Modeling Pesticide Time Series Concentrations from Monitoring Data. In the initial variogram analysis, spherical or gaussian models fit the 1995 Maumee River and MO-01 2007 atrazine data simulated with 4-, 7-, and 14-day sampling intervals (Table 13). Although the extent of autocorrelation, as depicted by the range, was highly variable among the case study data sets, longer ranges (52-80 days) were found in the 1995 Maumee River data. This phenomenon may be associated with the larger watershed size and daily flow characteristics. While kriged time series provided reasonable prediction of Page 126 of 184 missing data for the 4- and 7-day sample intervals (Figure 23), reliable variograms could not be estimated for the 14- and 28-day sample intervals from either monitoring dataset. Table 13: Variogram Models Using the NCWQR Maumee River 1995 and AEEMP MO-01 2007 Time Series with Different Sampling Frequencies. Data Sampling Interval Number of samples Maumee Actual (daily) 100 River, 1995 4-day 25 (days 1007-day 15 200) 14-day 8 28-day 4 6 MO-01, 2007 4-day 27 (days 917-day6 15 6 190), 14-day 8 6 complete 28-day 4 time series5 Nugget1 (µg/L2) Model2 Sill3 (µg/L2) 0.10 0.00 0.00 0.30 2.2 0.00 0.00 0.50 3.80 Spherical 0.97 Spherical 3.70 Spherical 3.20 Gaussian 3.20 No temporal correlation Spherical 3.60 Spherical 3.80 Spherical 3.80 No temporal correlation Range4 (Days) 16 52 61 80 27 27 32 1 The nugget describes the amount of variance resulting from random processes without temporal correlation; it is associated with sampling error, sampling technique, or inherent variation in daily pesticide concentrations. 2 Variogram models provide a description of variance as a function of lag distance or lag time 3 intervals. The Spherical model is defined as the variance at lag(h)=C*(1.5*h/a-0.58(h/a) )+nugget. 2 2 The Gaussian model is defined as the variance at lag(h)=C*(1-exp(-(3h) /a )+nugget. This model does not represent the Gaussian distribution model. Parameters in the variogram model: h = lag time or lag distance 2 C = sill, the variance for random process (ug/L ) a = range in days or distance for correlation of variance 3 The sill is the variance value at which the variogram levels off (see Figure 23 for an illustration). 4 The range is the lag time at which the variogram reaches the sill. It represents the period in which monitoring values are autocorrelated. 5 Grab samples collected every 4 days from 4/01/07 to 7/19/07. Complete time series includes autosampler data collected on most days in between the grab sample dates. Appendix 7.A also includes an analysis of the MO-01 2007 monitoring using only the 4-day grab samples. The models based on the daily sampling series had greater means, maximums, and variances than the models based on the 4-day grab samples, likely because the 4-day samples missed the maximum concentration detected in the daily samples. 6 Sampling intervals estimated from kriged samples. Page 127 of 184 Figure 23: Variograms for 4-day and 7-day Sampling Frequency for the Maumee and MO-01 2007 data: a.) 4-day 1995 Maumee; b.) 7-day 1995 Maumee; c.) 4-day grab MO-01 2007; and d.) 7-day grab MO-01 2007. We used conditional stochastic simulations to assess the uncertainty associated with estimated concentrations by running 1,000 equally-probable realizations of time series with the same temporal autocorrelation and descriptive statistics as the kriged time series. A subsampling of realizations for the 4-day and 7-day sampling derived time series provide a reasonable description of temporal behavior and descriptive statistics of atrazine concentration for the Maumee 1995 site from Julian day 100 to 200 (Figure 24a and b). Although a variogram was modeled for the 14-day sample data (Figure 24c), insufficient data precluded providing a reliable estimate of the temporal structure and descriptive statistic of atrazine at the Maumee River site. Because a variogram could not be constructed from 4 data points for the 28 day sample data, we tried to use daily flow to approximate the temporal structure of atrazine concentrations (Figure 24d). As expected, the estimate of peak concentrations decreased and the uncertainty around the time series estimations increased as the sampling interval increased; beyond 14-day intervals, the number of data points was insufficient for reliable estimates. Page 128 of 184 (a) 4-day intervals 16 Estimate Mean (SD) Actual Data 14 12 Atrazine Conc (ug/L) Atrazine Conc (ug/L) Estimate Mean (SD) Actual Data 14 12 10 8 6 10 8 6 4 4 2 2 0 0 100 120 140 160 180 200 220 100 (c) 14-day intervals Julian Day 16 120 140 160 180 200 220 (d) 28-day intervals Julian Day 16 Estimate Mean (SD) Actual Data 14 Estimate Mean (SD) Actual Data 14 12 Atrazine Conc (ug/L) 12 Atrazine Conc (ug/L) (b) 7-day intervals 16 10 8 6 10 8 6 4 4 2 2 0 0 100 120 140 160 180 200 220 100 120 140 160 180 200 220 Julian Day Julian Day Figure 24: Representative realizations of kriged atrazine time series from (a) 4-day, (b) 7-day, (c) 14day, and (d) 28-day sampling of the NCWQR Maumee River 1995 data. Similar patterns are evident in the conditional simulations for the MO-01 2007 monitoring data, although the realizations underestimated the actual peak concentrations in the 4- and 7-day sampling simulations (Figure 25). Although the temporal pattern for the 14-day sampling interval for MO-01-2007 did not exactly match the actual data, the conditional simulations were capable of identifying the temporal occurrence of the peaks (Figure 25c), in contrast to the 14-day simulations for the Maumee River data (Figure 24c). The simulated sampling intervals for the 2007 MO-01 data set did not capture the highest atrazine peak that occurred on Julian days 152-154 (Figure 25). This results in an underestimate of the true maximum concentrations for the subsequent conditional simulations. While a different sampling of the daily concentrations from MO-01 might capture the true peak, this particular sampling was used in Section 6.3.6 as an evaluation of how much of an impact missing the true peak concentration might have on estimating longer duration rolling average concentrations with conditional simulations of the variogram models. Page 129 of 184 Figure 25: Representative realizations of kriged atrazine time series from (a) 4-day, (b) 7-day, (c) 14day, and (d) 28-day sampling of the MO-01 2007 data. With fewer data points resulting from less frequent monitoring samples, the uncertainty in predicting daily atrazine concentrations increases. In the absence of data, the incorporation of additional information may assist in predicting a daily time series of pesticide concentrations. One approach involves the use of a covariate, such as average daily flow, to estimate atrazine concentrations between monitoring points. Another approach is to merge monitoring data with WARP estimates of the annual distribution of pesticide concentrations. This approach requires additional steps to derive a time series from the distribution. Appendix D.2 analyzes two covariate approaches – one using a co-occurring pesticide and one using stream flow data. Covariate approaches using other pesticides will suffer from the same sampling frequency issues as atrazine monitoring. One option may be to explore the use of more frequently monitored nutrient concentrations, such as nitrate as a covariate. While average daily stream flow data may be a useful covariate for estimating Page 130 of 184 atrazine concentrations in flowing water bodies, the relationship between flow and pesticide concentrations is also dependent on the timing of application as well as the watershed size. Analyses of the NCWQR 1995 Maumee River and the AEEMP 2007 MO-O1 monitoring data sets found significant regressions between atrazine concentrations and average daily flow rate within certain time segments of the data sets. The timing of atrazine application is a critical piece of information in determining the correct time interval. The highest concentrations in a time series are generally found in the first runoff events after pesticide application (USEPA, 2007, 2009a). Additionally, high average daily flow events in the absence of atrazine prior to atrazine application or after extensive degradation of atrazine are expected to impact the slope of the regression equation. Exploiting any relationship of flow and atrazine concentration requires knowledge of the atrazine application dates within a watershed. Unfortunately, data on the date of pesticide applications in a watershed are not readily available. However, since atrazine application dates will likely parallel corn planting dates in most areas, any variable influencing corn planting dates, such as air temperature, soil temperature, or soil moisture conditions, may be an appropriate predictor for estimating atrazine application timing. Appendix D.2 presents a proof-of-concept approach for integrating sparse monitoring data with WARP predictions of distribution percentiles into a reliable concentration time series. WARP provides estimates of the 5th, 10th, 15th, 25th, 50th, 75th, 85th, 90th and 95th percentiles and maximum concentration across the entire year (Stone and GIlliom, 2009). The WARP percentile estimates and the available monitoring data are merged into a single dataset. The newly generated distribution still needs to be sequenced in a time series. One possible procedure for sequencing is relating atrazine concentrations to flow. As previously discussed, a significant regression equation exists for atrazine and flow in the 120 -200 day samples in the Maumee River data. The flow was weighted to ensure that flow events immediately after pesticide application have more influence on atrazine runoff. The weighting factors for flow were derived using the first-order degradation model: A=A0e-k*time where A0 = 1 lb/A k = degradation rate (0.01152 days-1) time = days The actual daily flow data from Day 120 to Day 200 were multiplied by the daily flow weighting factor. The flow and atrazine data were separately ranked according to percentiles in the distribution and then matched to create a time series. The newly sequenced time series were used for variogram analysis and kriging. In the absence of watershed- and yearPage 131 of 184 specific WARP parameters for the 1995 Maumee River monitoring data, the USEPA used surrogate inputs from two Ohio sites (OH-01-2004 and OH-03-2005) included in the AEEMP study to evaluate how well the approach could identify the times of peak concentrations. The merged WARP data predicted for OH-01-2004 had a much higher predicted maximum atrazine concentration than OH-03-2005. This difference is mainly attributed to the difference in May and June precipitation. However, the two simulated time series illustrate that the temporal occurrence of the atrazine peak mimics the actual time series. This situation suggests that weighted flow may be a reasonable predictor of temporal occurrence daily atrazine concentrations for location where flow data is available. 6.3.4 Mechanistic Approach: PRZM Hybrid Model Syngenta submitted a proposal for developing a PRZM Hybrid model to augment measured monitoring data by estimating possible runoff events that may result in increased concentrations on unsampled days (Miller et al, 2011). The model would use local rainfall and crop planting data for the year of monitoring and watershed-specific soil, cropping, and hydrologic properties to simulate the environmental fate and transport of atrazine at the watershed scale. For each watershed, daily atrazine concentrations will be estimated from PRZM-generated atrazine flux and runoff values for each soil map unit that are then aggregated to the watershed based on area-weighted map unit concentrations and adjusted by the percent of corn and sorghum in the watershed (Miller et al, 2011). Miller et al (2011) propose combining the resulting PRZM time series with monitoring data to estimate potential concentrations on unsampled dates between measured values. This approach appears to be similar to a method Syngenta explored to estimate atrazine concentrations between 4-day sampling events for the ecological exposure monitoring program (Snyder et al, 2007). The estimation approach may not be suitable for larger watersheds where the time of travel for pesticide loads is greater than one day without accounting for the lag time. Although no results were available in time to be reviewed for this document, this approach may provide an option for estimating pesticide time series where available monitoring data are too sparse to estimate daily concentrations with other statistical methods. Miller et al (2011) plan to explore distribution matching to calibrate PRZM concentrations to measured data. However, in instances where the monitoring data are sparse, distributions based on measured data are likely to underestimate the true distribution of pesticide concentrations in water. An alternate approach may be to match PRZM model distributions with percentile distributions estimated by WARP. Page 132 of 184 6.3.5 Watershed Atrazine Analysis Model In addition to the hybrid PRZM modeling approach, the USEPA briefly evaluated a simple watershed-scale mass balance model (see Appendix D.3). This mass balance model uses an iterative fitting method to optimize the parameter estimates of a simplified watershed mass balance model for watersheds currently monitored for atrazine. Potentially, spatial variation in watershed parameters from sampled watersheds could be used to predict parameters for non-sampled watersheds located between the sampled watersheds, which in turn could be used to predict daily concentrations for those non-sampled watersheds. If a simple model could be found that fits the data well, it would provide a method based on mass balance principles for predictions of “non-sampled” dates (i.e., the concentration of atrazine rises based on the amount of atrazine applied to the watershed and falls based on the amount degraded and discharged from the watershed). However, the more complex such a model needs to be, the more data would be needed to provide assurance that proper calibration has been achieved. Because larger watersheds are generally more complex than smaller watersheds (greater variation in application dates, rainfall events differ across a large watershed, greater variation in travel times to deliver atrazine to the monitoring point), a simpler model may not be possible to provide enough assurance of proper calibration at CWS with larger catchment areas that have a limited numbers of samples. The USEPA conducted a preliminary evaluation of the mass balance approach on selected AEEMP monitoring sites (Appendix D.3). We used simple mass balance models to estimate the contribution of atrazine to the watershed from each individual storm event greater than 0.4 cm rainfall based on the load of atrazine discharged from the watershed. Discharge was calculated by iteratively fitting daily predictions of atrazine concentration to the less-frequently measured (monitoring) atrazine concentrations. While the simple watershed mass balance models were often able to fit the variation in atrazine concentrations (r2s often exceeding 0.8), the mass of atrazine discharged often differed by an order of magnitude in either direction from estimates of atrazine applied to the watershed based on estimates of the watershed area planted in corn and sorghum and local estimates of atrazine use rates (documented in USEPA, 2009a and 2009b). The “imbalance” issues may result in part from apparent anomalies or errors in the flow records. However, the imbalance problem is likely due primarily to the coarseness of the model temporal resolution; i.e., the time-step (see Appendix D.3 for a discussion on this). The discrepancy between expected maximum mass applied and expected loads may be because the watershed effluent flow and atrazine concentrations are quite dynamic in these small watersheds. Therefore a grab sample with a high concentration in a relatively low flow condition can be collected early on a day with rising flow conditions and falling concentrations. Because flow appears in the data set as average daily flow, the high Page 133 of 184 concentration paired with a large average daily flow results in a load for a particular day that exceeds maximum expected mass of atrazine applied to the watershed. For this method to be useful, the model would need a better method of deriving a flowweighted atrazine concentration. One method would be to tie atrazine concentrations measured in a particular day to the flow data at the sub-daily intervals. Typically, USGS collects flow data at automated gages at 15 minute intervals and then calculates an average daily flow from the 15 minute interval data. 6.3.6 Generating Atrazine Time Series from Weekly CWS Monitoring for Use in Dietary Exposure Assessments Sections 6.3.2 – 6.3.5 evaluated approaches that could be applied to weekly monitoring from the CWS to generate estimates of TCT exposures for comparison to a human health endpoint of concern with a duration that may range from 4 to 28 days. The analysis of monitoring frequencies in Section 6.3.2 indicates that the bias factors from 7-day sampling intervals would likely range from 2 to 4 for a 4-day average duration to 2 for a 28-day duration. While the September 2010 SAP recommended using WARP in combination with geostatistical or deterministic models to estimate atrazine concentrations for drinking water exposures, the analysis in Section 6.3.3 suggests that a reasonable time series can be approximated from the 7-day sampling frequencies using only kriging and stochastic simulations. We evaluated this for the case study (Section 7), using the variogram models for the 1995 Maumee River and 2007 MO-01 data to generate 1,000 daily time series realizations from conditional stochastic simulations (Figures 26 and 27). The actual daily time series for each data set in black. The 95th and 5th percentiles of the stochastic simulations are shown as red dashes, with the shaded area representing a 90 percent probability range for the time series concentrations based on the simulations. In the absence of a full daily time series, the stochastic simulations of weekly monitoring would characterize the likely range in exposures over time (shown as the grade shading between the red-dashed lines in Figures 26 and 27). The simulations of the daily time series from the variogram models underestimated the actual peak concentrations (Figures 26a and 27a), suggesting a bias factor may still be needed for estimates that require a peak exposure. However, the 95th percentile of the rolling average concentrations from the simulations provided a reasonable approximation of the actual rolling-average concentrations (Figures 26b-d and 27b-d). Page 134 of 184 16 Sim 95th %ile Sim 5th %ile Actual 12 10 8 6 4 4-Day Averages 12 10 8 6 4 2 2 0 0 91 16 91 98 105 112 119 126 133 140 147 154 161 168 175 182 189 Julian Day Sim 95th %ile 14-Day Averages Sim 5th %ile Actual 16 12 10 8 6 4 98 105 112 119 126 133 140 147 154 161 168 175 182 189 Julian Day Sim 95th %ile Sim 5th %ile Actual 14 Atrazine Concentration, ug/L 14 Atrazine Concentration, ug/L Sim 95th %ile Sim 5th %ile Actual 14 Atrazine Concentration, ug/L Atrazine Concentration, ug/L 14 16 Daily Time Series 28-Day Averages 12 10 8 6 4 2 2 0 0 91 98 105 112 119 126 133 140 147 154 161 168 175 182 189 Julian Day 91 98 105 112 119 126 133 140 147 154 161 168 175 182 189 Julian Day Figure 26: Estimates of daily and rolling-average atrazine concentrations from the actual time series (black line) and 5th and 95th percentiles (dashed red lines) of realizations of kriged time series from 7-day sampling of the Maumee River 1995 data. Page 135 of 184 100 Sim 95th %ile Sim 5th %ile Actual 90 70 60 50 40 30 20 10 80 4-Day Averages 70 60 50 40 30 20 10 0 0 91 100 98 105 112 119 126 133 140 147 154 161 168 175 182 189 Julian Day Sim 95th %ile Sim 5th %ile Actual 90 91 100 14-Day Averages 80 70 60 50 40 30 20 10 98 105 112 119 126 133 140 147 154 161 168 175 182 189 Julian Day Sim 95th %ile Sim 5th %ile Actual 90 Atrazine Concentration, ug/L 80 Atrazine Concentration, ug/L Sim 95th %ile Sim 5th %ile Actual 90 Atrazine Concentration, ug/L Atrazine Concentration, ug/L 80 100 Daily Time Series 28-Day Averages 70 60 50 40 30 20 10 0 0 91 98 105 112 119 126 133 140 147 154 161 168 175 182 189 Julian Day 91 98 105 112 119 126 133 140 147 154 161 168 175 182 189 Julian Day Figure 27: Estimates of daily and rolling-average atrazine concentrations from from the actual time series (black line) and 5th and 95th percentiles (dashed red lines) of realizations of kriged time series from 7-day sampling of the MO-01 2007 data. While the CWS monitoring data for the AMP are sampled at 7-day intervals, other monitoring data for atrazine and for other pesticides are often sampled less frequently. For those monitoring data sets with sampling intervals of longer than 7 days, time series and rolling-average concentration estimations will require a more complicated approach, combining monitoring data with WARP percentile estimates and using kriging and conditional simulations to generate a time series for exposure estimates: • • • • Generate a WARP-predicted distribution for the site, using watershed data specific for the CWS or monitoring location Merge monitoring data with WARP distributions by ranking flow within a determined time frame in which flow and concentrations are correlated and matching predicted concentration percentiles with flow percentiles Generate a variogram of simulated time series with nugget, range, and sill Use kriging to interpolate the time series Page 136 of 184 • • Generate stochastic simulations of the time series for uncertainty analysis Generate realizations of time series and bracket the estimated time series based on an upper and lower percentile of realizations. These steps are illustrated in Appendix D.2. 6.4 Year-to-Year Temporal Variability Uncertainty factor considerations based on monitoring design must account not only for uncertainty in sampling frequencies in capturing short-term temporal variability, but also for the year-to-year variability at the same site. During registration review, the USEPA will look at the multi-year history of CWS monitoring for atrazine to characterize the expected range in concentrations from year to year. In any given location, the magnitude and duration of pesticide concentrations in water vary from year to year, depending on such factors as the magnitude, timing, and frequency of pesticide application, the timing and intensity of rainfall events post-application, soil moisture conditions at the time of application and rainfall, and the type, timing, and extent of land management practices in place (Gilliom et al., 2006). Atrazine occurrence in water follows a seasonal pattern that corresponds to seasonal usage and weather patterns, with larger spikes in concentration related to pre-emergent spring applications that are followed by periods of runoff-producing rainfall. For the most part, atrazine has had fairly consistent usage (average application rate, percent of crop treated) over time. Changes in the relative proportion of the watershed in corn and sorghum as a result of crop rotations and commodity demands may occur, particularly in smaller watersheds, but can be accounted for with yearly cropping data. However, much of the year-to-year variations in atrazine concentrations will likely result from yearly differences in rainfall during the atrazine use season. An analysis of the influence of rainfall on atrazine concentrations in the AEEMP study suggests that the relationship between atrazine concentrations in streams and precipitation is a function of the distribution and timing of the rainfall in relation to application rather than to the amount of rainfall in a given month or season (USEPA, 2007 and 2009a). In evaluating a monitoring study, the USEPA has to take into consideration where the exposure estimates from the time period of sampling fall into the extended period of pesticide use and how protective the exposure estimates from a monitoring study are in relation to that longer use period. For atrazine, the key issue in characterizing the potential year-to-year variability in atrazine concentrations in water relates to how long sampling should continue at individual CWSs before concluding, with a reasonable certainty of no harm, that TCT concentrations in drinking water do not exceed the level of toxicological concern. Page 137 of 184 An analysis of the year-to-year variations in TCT concentrations for the CWS included in the AMP indicates that the ratio of lowest to highest maximum yearly TCT detections increases with increasing number of years of monitoring (Table 14). This would be expected since longer study durations increase the likelihood of capturing wider ranges in likely exposures. Table 14: Comparison of lowest and highest yearly maximum detections for CWS in the AMP based on number of years sampled. No. of Yrs. Sampled No. of CWS Sampled Ratio of the Lowest to Highest Annual Maximum Detections Minimum Median 90th Maximum Percentile Total Chloro-Triazines Measured in Source Waters 2-3 years 30 1.1 2.9 5.6 11.3 4-5 years 13 1.7 3.0 11.8 25.4 6 years 102 1.6 4.7 19.9 57.4 Total Chloro-Triazines Measured in Treated Waters 2-3 years 30 1.3 2.9 6.4 15.8 4-5 years 13 2.1 6.1 13.2 15.7 6 years 102 1.5 4.2 13.2 49.0 Mosquin et al (2011a) collected atrazine monitoring data from different datasets at across a range in years (1993-2009). The authors considered quarterly compliance monitoring at CWS for the Safe Drinking Water Act (SDWA) as well as more intensive monitoring with up to weekly sampling for the AMP and earlier atrazine Voluntary Monitoring Program (VMP). The more intensively sampled AMP and VMP datasets offer the potential to evaluate the range in maximum yearly atrazine detections over time for those CWS that were included in both studies. Even with the intensity of sampling in the VMP and AMP datasets (4 per month), a number of actual peak concentrations are expected to be underestimated for each CWS sampling year by a factor that depends on the sampling frequency (see Section 6.3.2). The PRZM Hybrid model described by Miller et al (2011), or a variation of the modeling approach, may provide an alternative approach to estimating the potential year-to-year variability in atrazine concentrations in water as a result of variations in rainfall. The USEPA has used PRZM to generate multiple years of exposure estimates using daily rainfall and temperature data collected from meteorological stations across the country. A hybrid modeling approach would entail using site- and year-specific weather, cropping, soil, hydrology, and pesticide usage information to calibrate the estimated daily PRZM concentrations with measured monitoring data. Subsequent years of weather data could then be modeled to characterize the range in potential exposures as a result of variability in rainfall patterns from year to year. Such an approach would involve substantial data Page 138 of 184 collection and model calibration against available monitoring in order to characterize the year-to-year variability in atrazine concentrations in water. The USGS WARP model could also be used to characterize the range in year-to-year variability in atrazine/TCT concentrations. This approach would involve varying the temporal parameters – May-June precipitation, atrazine use intensity – within expected ranges to characterize the extent of year-to-year variations. Because atrazine use intensity, which reflects both application rates and area grown in treated crops in the watershed, is a major driver in the WARP model (Stone and Gilliom, 2009), the error in estimating this parameter may be the dominant component in the variability in the WARP output. 6.5 Spatial Variability The applicability of any pesticide monitoring data beyond the sites for which they are collected depends on the extent to which we can place the monitoring data in a broader spatial context. Spatial statistical designs that are tiered by relative watershed vulnerability assist in placing monitoring results in context with other sites in the pesticide use area (see USEPA, 2007, for a spatially-balanced design used to select monitoring sites representing vulnerable watersheds for the AEEMP). In the case of the AMP monitoring study, CWS were not selected randomly or with a statistical design, but were identified as vulnerable based on results of quarterly compliance monitoring. Thus, the results are applicable to this population of CWS. To the extent that we can relate the CWS sites to vulnerable tiers (such as those estimated with WARP), we may be able to make inferences about other CWS or about the population of CWS. While the geographic distribution of pesticide use is the major predictive variable for pesticide occurrence in surface water on a broad geographic scale (Gilliom et al., 2006; Stone and Gilliom, 2009), watershed characteristics such as soil susceptibility to runoff and/or erosion, weather patterns, land management practices, and hydrology can be important factors driving pesticide occurrence on a local scale (USEPA, 2009a and 2009b; FIFRA SAP, 2009). The CWS included in the AMP occur in the more vulnerable watershed tiers identified by WARP (Figure 28; USEPA, 2010b). While this provides confirmation that the CWS included in the AMP are located in the most vulnerable atrazine use areas, it does not necessarily provide a basis from which to draw inferences regarding other CWS within the same vulnerable areas that did not meet the monitoring threshold. As a part of registration review, the USEPA will evaluate the entire population of CWS within the most vulnerable atrazine use area to determine whether additional CWS should be included in the AMP. The USEPA (2009a and 2009b) watershed-scale vulnerability assessment for atrazine in the AEEMP focused on headwater watersheds, but may have potential applications for better identifying vulnerable CWS that should be included for more intensive monitoring. Page 139 of 184 Figure 28: Location of CWS in the AMP in relation to vulnerable watersheds identified by WARP based on atrazine use on corn and sorghum. 6.6 Summary This section focused on methods for evaluating how well the existing AMP drinking water monitoring study characterizes the uncertainties in temporal and spatial variability of TCT concentrations in water. The analysis needs to address uncertainties resulting from (a) how well the monitoring sampling frequency captures daily and seasonal variations in pesticide concentrations in surface water, (b) year-to-year variations in magnitude and duration of concentrations, and (c) spatial patterns of pesticide occurrence in surface water. Estimating atrazine/TCT concentrations for various durations (ranging from 4 to 28 days) in a given year at a given CWS involves reconstructing the exposure pattern from weekly snapshots taken of the actual time series. The degree of uncertainty in those estimates depends upon the frequency of sampling (uncertainty or bias factors increase with increasing intervals), the exposure duration of concern (factors decreasing as the Page 140 of 184 timeframe of concern increases), the size of the watersheds, and the magnitude of concentration (Section 6.3.2). An analysis of two intensively sampled monitoring data sets – those with daily or near-daily sampling during the high use/runoff season – to represent larger (NCWQR monitoring for Maumee River in OH) and smaller (AEEMP monitoring in MO-01) suggests that a bias factor of at least 2 to 4X may be needed to estimate maximum exposures from 7-day sampling intervals. The actual bias factors are likely to vary depending not only on sampling intervals and exposure durations of concern but also on watershed size and water body type. As noted in the previous SAPs on atrazine (USEPA, 2010a and 2010b), a combination of monitoring and statistical/modeling approaches can provide an approximation of time series distributions for use in drinking water exposure assessments. While the SAP recommended a combination of WARP with deterministic models or kriging, the USEPA’s analysis described in Section 6.3.3 found that a reasonable estimate of 4- to 28-day rolling average concentrations can be approximated from 7-day sampling frequencies using kriging and stochastic simulations, with upper and lower percentiles from the conditional simulations providing a characterization of uncertainty. The USEPA used the 95th and 5th percentiles for uncertainty bounds in the case study described in Section 7. For less-frequent sampling (in most instances, 14-day sampling intervals is a more common lower bound of sampling intervals), the bias factors will be much greater. Where estimated exposures from monitoring adjusted to account for the bias factor overlap with the toxicity concentration of concern, exposure approximations become more complicated, involving not only geostatistical simulations but WARP and/or deterministic modeling. In determining the uncertainty in monitoring data, the USEPA must also consider year-toyear variability. Approaches described in section 6.4 include extending the analysis of yearto-year variations within the CWS monitoring time frame using a PRZM hybrid approach or WARP. The applicability of the monitoring results in the AMP beyond the CWS included in the study depends on the extent to which we can place the data in a broader spatial context. For the AMP, CWS were considered vulnerable based on quarterly compliance monitoring data. The USEPA (2009a and 2009b) watershed-scale vulnerability assessment for atrazine in the AEEMP may provide additional context in better interpreting the CWS results and identifying vulnerable CWS that should be included for more intensive monitoring. Page 141 of 184 7. CASE STUDY: IMPLICATIONS OF MOA & TOXICITY PROFILE ON WATER MONITORING 7.1 Introduction In the 2003 human health risk assessment, a duration of 90 days was used for doing exposure assessment of trichlorotriazines (i.e., atrazine and its metabolites) for comparison to an administered dose no-observed adverse effect level (NOAEL) for attenuation of the LH surge in a 6-month rat study (IRED 2003). Syngenta proposed a weekly sampling frequency of selected CWS to characterize TCT concentrations over the critical exposure duration of 90 days. The current scientific re-evaluation of the health effects of atrazine has identified shorter atrazine exposure durations that can potentially result in LH attenuation in rats. According to the critical study by Cooper et al. 2010, 4 days of repeated once daily dosing with atrazine (the minimum duration evaluated thus far) beginning on the day of proestrous is enough to induce a statistically significant attenuation of the LH surge in rats (Cooper et al. 2010). Since the frequency of drinking water monitoring should be informed by the temporal relationship between exposure and the toxicological endpoint of concern (i.e., LH attenuation), shorter durations than 90 days are being considered in the current reevaluation of the health effects associated with atrazine exposure. Plasma triazines in the rat build up upon repeated once daily dosing with atrazine to reach pseudo steady state by the fourth day when attenuation of the LH surge is observed. In essence, based on the information available, LH attenuation is observed when plasma levels are sustained at a specific range. The time of 4 days to reach steady state in the rat is in the range of predictions based on linear elimination pharmacokinetics that such time should take roughly 3-5 half-lives. When the estimated rat plasma elimination rate constant for triazines is allometrically scaled for a human female body weight of 60 kg, the estimated time to reach steady state is 21-30 days (plasma half-life of 6-8 days). Similar to rat studies, this time estimate would be based on a constant dose level and frequency of water consumption, both of which are known to be variable in the human situation. Thus given the lack of human specific information that would allow a more precise estimate of the critical duration of exposure, the Agency proposes a range of exposure durations of 4 - 28 days relevant to LH attenuation. More specifically, three exposure durations consisting of 4, 14, and 28 days are being proposed to bracket the most likely human exposure durations that may result in LH attenuation based on the exposure requirements for LH attenuation in rat toxicity studies. The 28-day duration is within the range predicted by allometric scaling of rat plasma elimination pharmacokinetics for humans and is also the average duration of the human menstrual cycle. Notably, in both rats and humans, the time estimated to reach steady state coincides with the respective length of their ovarian cycles of 4 and 28 days, respectively. Page 142 of 184 The duration of 14 days is being proposed as a midpoint between the other two durations that would allow better characterization of an appropriate water monitoring frequency. Finally, the duration of 4 days was selected on the basis that the limited human information available from the Davidson 1987 study suggests that the pharmacokinetics of atrazine in humans may be somewhat similar to that observed in rats. For instance, similar to rat findings, atrazine and DEA appeared to be short-lived in human blood, not having been detected at the first time point of 2 hours (Table 7). The estimated human whole blood halflife of 2.8 hours for DIA is in the range of estimated rat plasma half-life values of 1.2 - 3.9 hours at 3, 10, and 50 mg/kg bw atrazine (Table 3). The estimated human half-life for DACT of about 11-18 hours (Table 7) is also very comparable to the rat plasma value of 89 hours (Table 3). Thus, human pharmacokinetics may not be that much different as compared to rats and justifies including 4 days as an exposure duration that should be evaluated. 7.2 Atrazine Time Series Estimates for Use in Dietary Exposure Assessments In Section 6, the USEPA evaluated methods for characterizing the uncertainty in estimating 4- to 28-day rolling average TCT concentrations in drinking water from weekly samples from CWS included in the AMP. The Agency determined that a reasonable time series can be approximated from the 7-day sampling frequencies using geostatistical methods (kriging and variogram models) and stochastic simulations (see Sections 6.3.3 and 6.3.6). For the case study, we used the NCWQR 1995 Maumee River and AEEMP 2007 MO-01 data sets to represent a range in watershed sizes for flowing water bodies and a range in atrazine concentrations. These data sets do not represent actual CWS but have been used for the monitoring analysis because they include daily or near-daily sampling during the high use/runoff season and allow for evaluation of sampling frequencies. The results presented below are for illustrative purposes only. Figures 29 and 30 (also shown as Figures 27 and 28 in Section 6) show the actual daily time series for each data set in black for comparison to the stochastic simulations of the variogram models for 7-day sampling intervals. The 95th and 5th percentiles of the stochastic simulations are shown as red dashes, with the shaded area representing a 90 percent probability range for the time series concentrations based on the simulations. The simulations of variogram models for the daily time series underestimate the actual peak concentrations (Figures 29a and 30a); for acute toxicity endpoints of concern, estimates of peak, single-day exposures might require an additional bias factor for characterization. However, the 95th percentile of the rolling average concentrations from the simulations provided a reasonable approximation of the 4- to 28-day rolling-average concentrations (Figures 29b-d and 30b-d). Page 143 of 184 16 Sim 95th %ile Sim 5th %ile Actual 12 10 8 6 4 4-Day Averages 12 10 8 6 4 2 2 0 0 91 16 91 98 105 112 119 126 133 140 147 154 161 168 175 182 189 Julian Day Sim 95th %ile 14-Day Averages Sim 5th %ile Actual 16 12 10 8 6 4 98 105 112 119 126 133 140 147 154 161 168 175 182 189 Julian Day Sim 95th %ile Sim 5th %ile Actual 14 Atrazine Concentration, ug/L 14 Atrazine Concentration, ug/L Sim 95th %ile Sim 5th %ile Actual 14 Atrazine Concentration, ug/L Atrazine Concentration, ug/L 14 16 Daily Time Series 28-Day Averages 12 10 8 6 4 2 2 0 0 91 98 105 112 119 126 133 140 147 154 161 168 175 182 189 Julian Day 91 98 105 112 119 126 133 140 147 154 161 168 175 182 189 Julian Day Figure 29: Estimates of daily and rolling-average atrazine concentrations from the actual time series (black line) and 5th and 95th percentiles (dashed red lines) of realizations of kriged time series from 7-day sampling of the Maumee River 1995 data. Page 144 of 184 100 Sim 95th %ile Sim 5th %ile Actual 90 70 60 50 40 30 20 10 80 4-Day Averages 70 60 50 40 30 20 10 0 0 91 100 98 105 112 119 126 133 140 147 154 161 168 175 182 189 Julian Day Sim 95th %ile Sim 5th %ile Actual 90 91 100 14-Day Averages 70 60 50 40 30 20 10 98 105 112 119 126 133 140 147 154 161 168 175 182 189 Julian Day Sim 95th %ile Sim 5th %ile Actual 90 80 Atrazine Concentration, ug/L 80 Atrazine Concentration, ug/L Sim 95th %ile Sim 5th %ile Actual 90 Atrazine Concentration, ug/L Atrazine Concentration, ug/L 80 100 Daily Time Series 28-Day Averages 70 60 50 40 30 20 10 0 0 91 98 105 112 119 126 133 140 147 154 161 168 175 182 189 Julian Day 91 98 105 112 119 126 133 140 147 154 161 168 175 182 189 Julian Day Figure 30: Estimates of daily and rolling-average atrazine concentrations from the actual time series (black line) and 5th and 95th percentiles (dashed red lines) of realizations of kriged time series from 7-day sampling of the MO-01 2007 data. 7.3 Estimating human plasma average daily AUC from rolling average atrazine concentrations The short in vivo half-life of atrazine, primarily due to metabolism and the intrinsic activity of metabolites, suggest that the link between atrazine exposure and attenuation of the LHsurge can be better characterized through the use of an internal dose metric. At the September 2010 SAP meeting, the Agency presented AUC estimates of plasma triazines as the basis for doing an internal dose-response assessment associated with LH attenuation in the rat. In this report, methodologies based on the pharmacokinetic behavior of plasma triazines are being proposed by the Agency for estimating the corresponding human plasma average daily AUC from water exposure estimates that can then be compared to the PoD plasma AUC for LH attenuation in the rat. The proposed approach incorporates the best information available with respect to LH attenuation and pharmacokinetics of plasma triazines. As compared to the approach used in the 2003 human health risk assessment, the proposed water monitoring approach identifies not only Page 145 of 184 the magnitude of exceedance but also the time regions in the water chemograph where those exceedances are more likely to occur. It incorporates an integrated measure of the dose intake and information on body distribution and plasma clearance. The process begins by estimating AUCwater for a given duration of concern. As previously discussed, durations of 4, 14 and 28 days are being proposed based on LH attenuation studies in rats and human information. AUCwater is calculated based on the trapezoid rule (see Figure 21). The more frequent the water sampling frequency, the better the estimate of AUCwater. The resulting AUCwater is used in the estimation of a time-weighted average daily intake of atrazine through drinking water according to Equation 4 in Section 4. The predicted AUCplasma can then be calculated through the use of equation 5 in Section 4 that accounts for dose intake, body distribution, and plasma clearance information. For the 1995 Maumee River and 2007 M0-01 case studies, the rat plasma elimination rate constant (kel) and volume of distribution (Vdss) were allometrically scaled for a human female body weight of 60 kg and a water consumption rate of 2 L/day was assumed. A body weight of 60 kg was selected for this comparative analysis on the basis that young females may be a particular susceptible population to disruptions of LH. However, the proposed approach is readily amenable to evaluations of distributions of body weights and water consumption rates. The results corresponding to the exposure estimates of the 1995 Maumee River and 2007 MO-01 monitoring data sets, sampled at weekly intervals, (Figures 29 and 30) are shown in Figures 31 and 32, respectively. For comparison purposes, the PoD BMDL rat plasma AUC with an applied 300X combined uncertainty factor (UF) is depicted in the graphs. The combined UF is hypothetical at this time but would presumably be reflective of inter- and intra-species extrapolation factors as well as any contribution for FQPA. As can be seen in Figure 31, the predicted human AUCplasma profile for the 1995 Maumee River dataset is well-below the BMDL PoD rat plasma AUC at all time points. The MOEs based on the 95th percentiles and the rat PoD plasma AUC range from 1424 at the low end to over 64,000 at the high endpoint. Thus, the exposure levels of the 1995 Maumee River dataset seem to be below the level of concern based on a 300X combined uncertainty factor. Page 146 of 184 actual (daily sampling) Sim 95%ile 0.06 Sim 5%ile 0.04 0.02 0.00 100 110 120 130 140 160 150 180 170 190 200 Julian Day Maumee 95: 14-d rolling averages 0.10 BMDL AUC/300x 0.08 actual (14-d rolling ave) Sim95%ile Sim5%ile 0.06 0.04 0.02 0.00 100 110 120 130 140 150 160 Julian Days 170 180 190 200 Predicted human plasma AUC ((mg/L)*h) BMDL AUC/300x 0.08 Predicted human plasma AUC ((mg/L)*h) Predicted human plasma AUC ((mg/L)*h) Predicted human plasma AUC ((mg/L)*h) Maumee 95: Daily Sampling 0.10 Maumee 95: 4-day rolling averages 0.10 0.08 0.06 BMDL AUC/300x actual (4-d rolling ave) Sim95%ile Sim5%ile 0.04 0.02 0.00 100 110 120 130 140 150 160 170 180 190 200 Julian Days Maumee 95: 28-d rolling averages 0.10 actual (28-d rolling ave) Sim95%ile 0.08 BMDL AUC/300x Sim5%ile 0.06 0.04 0.02 0.00 100 110 120 130 140 150 160 170 180 190 200 Jilian Days Figure 31: Estimates of rolling average daily human plasma AUC values corresponding to the actual time series (black line) and 5th and 95th percentiles (dashed red lines) for the Maumee River 1995 dataset according to Figure 29. The dashed line is for comparison to the rat plasma PoD AUC with a 300X applied uncertainty factor. Evaluation of the 2007 MO-01 dataset suggests, unlike the Maumee River 1995 dataset, that there are time periods of exceedance for a 4-day duration of exposure based on a level of concern of 300. Figure 32 shows that the rat plasma PoD AUC is exceeded for ≈ 5 days based on the 95th percentile. The time periods, but not necessarily the magnitudes of exceedance, seem to be adequately captured by 4- but not 14- or 28-day rolling averages. A 100X combined UF is also included in Figure 32 to show that the resulting human AUCplasma values would be below the level of concern in such case. In comparison to the constant dose level and dosing frequency in rat studies, TCT exposures resulting from consuming drinking water will be variable in nature (see, for example, Figures 29 and 30). While such variability would likely preclude human plasma levels from reaching steady state, the equations described in Section provide an average time-weighted estimate of TCT levels in human plasma representative of sustained levels of exposure for a given duration that can then be compared to the PoD BMDL rat plasma AUC. The key comparison, thus, would be TCT plasma levels greater than the PoD BMDL for extended periods of time, related to a given duration of potential concern. Page 147 of 184 The proposed approach that estimates internal measures of exposures in the form of human plasma AUC can provide valuable perspective for evaluating water monitoring frequencies in the characterization of exposure to pesticides like atrazine. MO-01: 4-d rollin averages BMDL AUC/100x 0.25 0.20 0.15 actual (daily sampling) Sim 95%ile Sim 5%ile 0.10 BMDL AUC/300x 0.05 0.00 100 110 120 130 140 150 160 170 180 190 200 Predicted human plasma AUC ((mg/L)*h) Pre dicte d human plasma AUC ((mg/L)*h) MO-01: daily sampling BMDL AUC/100x 0.25 0.20 0.15 actual (4-d rolling ave) Sim 95%ile Sim 5%ile 0.10 BMDL AUC/300x 0.05 0.00 100 110 120 130 140 150 160 170 180 190 200 Julian Days Julian Day MO-01: 28-d rolling averages BMDL AUC/100x 0.25 actual (14-d rolling ave) 0.20 Sim 95%ile Sim 5%ile 0.15 0.10 BMDL AUC/300x 0.05 0.00 100 110 120 130 140 150 160 Julian Days 170 180 190 200 Predicted human plasma AUC ((mg/L)*h) Predicted human plasma AUC ((mg/L)*h) MO-01: 14-d rolling averages BMDL AUC/100x 0.25 actual (28-d rolling ave) 0.20 Sim 95%ile 0.15 Sim 5%ile 0.10 BMDL AUC/300x 0.05 0.00 90 100 110 120 130 140 150 160 170 180 190 200 Julian Days Figure 32: Estimates of daily rolling-average atrazine concentrations from the actual time series (black line) and 5th and 95th percentiles (dashed red lines) of realizations of kriged time series from 7-day sampling of the MO-01 2007 data. Page 148 of 184 8. REFERENCES Adami, H. e. (2002). Textbook of Cancer Epidemiology. New York NY: Oxford University Press. Alavanja, M. C., Dosemeci, M., Samanic, C., Lubin, J., Lynch, C. F., Knott, C., Blair, A. (2004). Pesticides and lung cancer risk in the agricultural health study cohort. Am J Epidemiol, 160(9), 876-885. doi: 160/9/876 [pii] 10.1093/aje/kwh290 Alavanja, M. C., Samanic, C., Dosemeci, M., Lubin, J., Tarone, R., Lynch, C. F., . . . Blair, A. (2003). Use of agricultural pesticides and prostate cancer risk in the Agricultural Health Study cohort. Am J Epidemiol, 157(9), 800-814. Alavanja, M. C., Sandler, D. P., Lynch, C. F., Knott, C., Lubin, J. H., Tarone, R., . . . Blair, A. (2005). Cancer incidence in the agricultural health study. Scand J Work Environ Health, 31 Suppl 1, 39-45; discussion 35-37. Alavanja, M. C., Sandler, D. P., McMaster, S. B., Zahm, S. H., McDonnell, C. J., Lynch, C. F., . . . Blair, A. (1996). The Agricultural Health Study. Environ Health Perspect, 104(4), 362-369. Albanito, L., Lappano, R., Madeo, A., Chimento, A., Prossnitz, E.R., Cappello, A.R., Dolce, V., Abonante, S., Pezzi, V., & Maggiolini, M. (2008). G-protein-coupled receptor 30 and estrogen receptor-alpha are involved in the proliferative effects induced by atrazine in ovarian cancer cells. Environ Health Perspect, 116(12): 1648-55. Albanito, L., Madeo, A., Lappano, R., Vivacqua, A., Rago, V., Carpino, A., Oprea, T., Prossnitz, E., Musti, A.M., Andò, S., & Maggiolini, M. (2007). G Protein–Coupled Receptor 30 (GPR30) Mediates Gene Expression Changes and Growth Response to 17ß-Estradiol and Selective GPR30 Ligand G-1 in Ovarian Cancer Cells. Cancer Res, Feb 15; 67(4): 1859-66. Amory JK (2004) Testosterone/progestin regimens: A realistic option for male contraception? Curr Opin Invest Drugs 5:1025-1030 Andreotti, G., Freeman, L. E., Hou, L., Coble, J., Rusiecki, J., Hoppin, J. A., . . . Alavanja, M. C. (2009). Agricultural pesticide use and pancreatic cancer risk in the Agricultural Health Study Cohort. Int J Cancer, 124(10), 2495-2500. doi: 10.1002/ijc.24185 Ankley GT,_ Bennett RS, Erickson RJ, Hoff DJ, Hornung MW, Johnson RD, Mount DR, Nichols JW, Russom CL, Schmieder PK, Serrano JA, Tietge JE, Villeneuve DL. (2010). Adverse Outcome Pathways: A Conceptual Framework to Support Ecotoxicology Research and Risk Assessment. Environ Toxicol Pharmacol 29(3):730-741 Page 149 of 184 Arbuckle, T. E., Lin, Z. Q., & Mery, L. S. (2001). An exploratory analysis of the effect of pesticide exposure on the risk of spontaneous abortion in an Ontario farm population. Environmental Health Perspectives, 109(8), 851-857. Baetcke, K. P., & Dellarco, V. (2000). Atrazine: Evaluation of Carcinogenic Potential HED DOC. NO 014431: U.S. EPA. Bakke B., De Roos A.J., Barr D.B., Stewart P.A., Blair A., Freeman L.B., Lynch C.F., Allen R.H., Alavanja M.C., Vermeulen R. (2009). Exposure to atrazine and selected non-persistent pesticides among corn farmers during the growing season. J Expo Sci Environ Epidemiol., Sep; 19(6): 544-54. Banks J. A. and Freeman M. E.. The Temporal Requirement of Progesterone on Proestrus for Extinction of the Estrogen-Induced Daily Signal Controlling Luteinizing Hormone Release in the Rat. Endocrinology 102 (2): 426-432 Barr DB, Panuwet P, Nguyen JV, Udunka S, Needham LL. Assessing exposure to atrazine and its metabolites using biomonitoring. Environ Health Perspect. 2007 Oct;115(10):1474-8. PubMed PMID: 17938738; PubMed Central PMCID: PMC2022667. Barraclough CA, Camp P, Weiland N, Akabori A. (1986). Stimulatory versus Inhibitory Effects of Progesterone on Estrogen-Induced Phasic LH and Prolactin Secretion Correlated with Estrogen Nuclear and Progestin Cytosol Receptor Concentrations in Brain and Pituitary Gland. Neuroendocrinology 42:6-14 Bassil, K. L., Vakil, C., Sanborn, M., Cole, D. C., Kaur, J. S., & Kerr, K. J. (2007). Cancer health effects of pesticides - Systematic review. Canadian Family Physician, 53, 1705-1711. Beane Freeman, L. E., Rusiecki, J. A., Hoppin, J. A., Lubin, J. H., Koutros, S., Andreotti, G., . . . Alavanja, M. C. (2011). Atrazine and Cancer Incidence Among Pesticide Applicators in the Agricultural Health Study (1994-2007). Environ Health Perspect. doi: 10.1289/ehp.1103561 Belloni V, Alleva E, et al . (2007). Effects of low doses of atrazine on the neurobehavioural development of mice. Ethology Ecology & Evolution, 19: 309-22. Benachour, N., Moslemi, S., Sipahutar, H., & Seralini G.E. (2007). Cytotoxic effects and aromatase inhibition by xenobiotic endocrine disrupters alone and in combination, Toxicol Appl Pharmacol, Jul 15; 222(2): 129-40. Benet LZ, G. R. (1979) Noncompartmental Determination of the Steady State Volume of Distribution, Journal of Pharmaceutical Sciences 68. Bergendahl M, Evans WS, Veldhuis JD. (1996). Current Concepts in Ultradian Rythms of Luteininzing Hormone Secretion in the Human. Hum Reprod. Upate 2:507-518. Page 150 of 184 Betancourt, M., Resendiz, A., & Fierro, E.C. (2006). Effects of two insecticides and two herbicides on the porcine sperm motility patterns using computer-assisted semen analysis (CASA) in vitro. Reproductive Toxicology, Oct, 22(3): 508-12. Blair A, Tarone R, Sandler D, Lynch CF, Rowland A, Wintersteen W, Steen WC, Samanic C, Dosemeci M, Alavanja MC. (2002) Reliability of reporting on life-style and agricultural factors by a sample of participants in the Agricultural Health Study from Iowa. Epidemiology., Jan;13(1):94-9. PubMed PMID: 11805592. Blair, A., Tarone, R., Sandler, D., Lynch, C. F., Rowland, A., Wintersteen, W., . . . Alavanja, M. C. (2002). Reliability of reporting on life-style and agricultural factors by a sample of participants in the Agricultural Health Study from Iowa. Epidemiology, 13(1), 94-99. Blair, A., Thomas, K., Coble, J., Sandler, D. P., Hines, C. J., Lynch, C. F., . . . Lubin, J. H. (2011). Impact of pesticide exposure misclassification on estimates of relative risks in the Agricultural Health Study. Occup Environ Med. doi: oem.2010.059469 [pii] 10.1136/oem.2010.059469 Blomquist, J.D., Denis, J.M., Cowles, J.L., Hetrick, J.A., Jones, R.D. & Birchfield N.B. (2001). Pesticides in Selected Water-Supply Reservoirs and Finished Drinking Water, 1999-2000: Summary of Results from a Pilot Monitoring Program. U.S. Geological Survey Open-File Report 01-456. Published in cooperation with the U.S. Environmental Protection Agency Blomquist, J.D., Denis, J.M., Cowles, J.L., Hetrick, J.A., Jones, R.D., and Birchfield, N.B. (2001). Pesticides in Selected Water-Supply Reservoirs and Finished Drinking Water, 1999-2000: Summary of Results from a Pilot Monitoring Program. U.S. Geological Survey Open-File Report 01-456. Published in cooperation with the U.S. Environmental Protection Agency. Blondell, J., & Dellcarco, V. (October, 28, 2003). Review of Atrazine Cancer Epidemiology (DP Barcode 295200). Boobis, A.R., J.E. Doe, B. Heinrich-Hirsch et al . 2008. IPCS framework for analyzing the relevance of a noncancer mode of action for humans. Crit Rev Toxicol, 38(2):87-96. Breckenridge, C.B., Yi, K.D., & Simpkins, J.W. (2010a). Cytotoxicity of atrazine and its metabolites in vitro. Unpublished data from Syngenta Crop Protection, Inc. Breckenridge, C.B., Yi, K.D., Perez, E.J., & Simpkins, J.W. (2010b). Dose and time dependent effects of atrazine and metabolites on aromatase mRNA in multiple cell types in vitro. Unpublished data from Syngenta Breckenridge, C.B., Yi, K.D., Perez, E.J., & Simpkins, J.W. (2010c). Ex vivo assessment of the effects of atrazine or its metabolites on phosphodiesterase activity. Unpublished data from Syngenta Page 151 of 184 Brown-Grant K. and F. Naftolin (1972). Facilitation of Luteinizing Hormone Secretion in the Female Rat by Progesterone. J. Endocrinol.53:37-46 Brown TJ, Clark AS, MacLusky NJ. (1987). Regional Sex Differences in Progestin Receptor Induction in the Rat Hypothalamus: Effects of Various Doses of Estradiol Benzoate. J Neurosci 7(8):2529-2536 Brown, L. M., Blair, A., Gibson, R., Everett, G. D., Cantor, K. P., Schuman, L. M., . . . Dick, F. (1990). Pesticide exposures and other agricultural risk factors for leukemia among men in Iowa and Minnesota. Cancer Res, 50(20), 6585-6591. Brown, L. M., Burmeister, L. F., Everett, G. D., & Blair, A. (1993). Pesticide exposures and multiple myeloma in Iowa men. Cancer Causes Control, 4(2), 153-156. Burmeister, L. F. (1990). Cancer in Iowa farmers: recent results. Am J Ind Med, 18(3), 295-301. Burns, L. 2002. Exposure Analysis Modeling System (EXAMS). V2.98.04. July 18, 2002. Available through the USEPA Office of Pesticide Programs Water Models web site at http://www.epa.gov/oppefed1/models/water/. Caligaris L, Astrada JJ, Taleisnik S. (1968). Stimulating and Inhibiting Effects of Progesterone on the Release of Luteinizing Hormone. Acta Endocrinol 59:177-185 Caligaris L, Astrada JJ, Taleisnik, S. (1971). Biphasic Effect of Progesterone on the Release of Gonadotropin in Rats. Endocrinology 89:331-337. Canfield MA, Honein MA, Yuskiv N, Xing J, Mai CT, Collins JS, et al . (2006) National estimates and race/ethnic-specific variation of selected birth defects in theUnited States, 1999-2001. BirthDefects Res A Clin Mol Teratol, 76:747-56. Canfield MA, Ramadhani TA, Langlois PH, Waller DK. (2006) Residential mobility patterns and exposure misclassification in epidemiologic studies of birth defects. J Expo Sci Environ Epidemiol., Nov;16(6):538-43. Epub 2006 May 31. PubMed PMID: 16736057. Cantor, K. P., Blair, A., Everett, G., Gibson, R., Burmeister, L. F., Brown, L. M., . . . Dick, F. R. (1992). Pesticides and other agricultural risk factors for non-Hodgkin's lymphoma among men in Iowa and Minnesota. Cancer Res, 52(9), 2447-2455. Carreon, T., Butler, M. A., Ruder, A. M., Waters, M. A., Davis-King, K. E., Calvert, G. M., . . . Brain Canc Collaborative Study, G. (2005). Gliomas and farm pesticide exposure in women: The Upper Midwest Health Study. Environmental Health Perspectives, 113(5), 546-551. doi: 10.1289/ehp.7456 Page 152 of 184 Cimino-Reale, G., Ferrario, D., Casati, B., Brustio, R., Diodovich, C., Collotta, A., Vahter, M., & Gribaldo, L. (2008). Combined in utero and juvenile exposure of mice to arsenate and atrazine in drinking water modulates gene expression and clonogenicity of myeloid progenitors. Toxicol Lett, Jul 30; 180(1): 59-66. Clavel, J., Hemon, D., Mandereau, L., Delemotte, B., Severin, F., & Flandrin, G. (1996). Farming, pesticide use and hairy-cell leukemia. Scandinavian Journal of Work Environment & Health, 22(4), 285-293. Coban, A., & Filipov, N.M. (2007). Dopaminergic toxicity associated with oral exposure to the herbicide atrazine in juvenile male C57BL/6 mice. J. Neurochem, Mar, 100(5): 1177-87. Coder, P. (2010). Atrazine: An Oral (Gavage) Immunotoxicity and Hormone Evaluation Study for the Effects of Atrazine on Male Rats. Final Report. Unpublished data from Syngenta Crop Protection, Inc. Report Number WIL-639072. Study Completion Date July 13, 2010. Coder, P. (2010). Atrazine: An Oral (Gavage) Study of the Effects of Atrazine on the Estrous Cycle and the Estrogen-Induced Luteinizing Hormone Surge in Female Rats. Final Interim Report. Unpublished data from Syngenta Crop Protection, Inc. Report Number WIL-639062. Study Completion Date July 14, 2010. Coder, P. (2010). Atrazine: An Oral (Gavage) Study to Assess the Potential for Detecting the Estrogen-Induced Luteinizing Hormone Surge in Female Rats. Final Report. Unpublished data from Syngenta Crop Protection, Inc. Report Number WIL-639062. Study Completion Date July 2, 2010. Coder, P. (2010). Atrazine: An Oral Developmental Toxicity Study of Atrazine in the Rat, Including Cross-Fostering and Pair-Feeding. Final Report. January 28, 2010. Project Number: WIL/639025, T007794/08, 140/107. Unpublished study prepared by WIL Research Laboratories, Inc. 2827 p. MRID 47972801. Coder, Prägati Sawhney. (2011) An Oral (Gavage and Dietary) Study Evaluating the Pharmacokinetics of Atrazine in Female Sprague-Dawley Rats, WIL Research Laboratories, LLC Project ID 639081; Sponsor: Syngenta Crop Protection, LLC. Cooper R.L., Laws S., Das P.C., Narotsky M.G., Goldman J.M., Tyrey E.L., Stoker T.E. (2007). Atrazine and reproductive function: mode and mechanism of action studies. Birth Defects Res B Dev Reprod Toxicol, Apr; 80(2): 98-112. Cooper, R, Buckalew, A. Fraites, M., et al. (2010). Internal Report: Evaluation of Effects of the Chlorotriazine Herbicide Atrazine on the Amplitude of the Pre-ovulatory LH Surge in the Long Evans Rat. Memorandum dated July 8, 2010 from Ralph Cooper, NHEERL/ORD to Elizabeth Mendez, HED. Page 153 of 184 Cooper, R, Buckalew, A. Fraites, M., et al. (2010). Internal Report: Evaluation of Effects of the Chlorotriazine Herbicide Atrazine on the Amplitude of the Pre-ovulatory LH Surge in the Long Evans Rat. Memorandum dated July 8, 2010 from Ralph Cooper, NHEERL/ORD to Elizabeth Mendez, HED. Corrected Report. Cooper, R. L., Stoker, T. E., Tyrey, L., Goldman, J. M., and McElroy, W. K. (2000) Atrazine disrupts the hypothalamic control of pituitary-ovarian function, Toxicol Sci 53, 297-307. Cooper, R.L., Stoker, T.E., Goldman, J.M., Parrish, M.B, & Tyrey, L. (1996). Effect of atrazine on ovarian function in the rat. Reprod Toxicol, Jul-Aug; 10(4): 257-64. Cooper, R.L., Stoker, T.E., Tyrey, L., Goldman, J.M., & McElroy, W.K. (2000). Atrazine disrupts the hypothalamic control of pituitary-ovarian function. Toxicol Sci., Feb; 53(2): 297-307 Cooper, Ralph, Buckalew, Angela, Fraites, Melanie, Goldman, Jerome, Laws, Susan, Narotsky, Michael and Stoker, Tammy. (2010) Evaluating the effect of the chlorotriazine herbicide atrazine on the amplitude of the pre-ovulatory LH surge in the Long-Evans Rat, In Unpublished internal report. Crawford, C.G. 2004. Sampling Strategies for Estimating Acute and Chronic Exposures of Pesticides in Streams. Journal of the American Water Resources Association (JAWRA) 40(2):485-502. Cummings, A. M., B. E. Rhodes, et al . (2000). "Effect of atrazine on implantation and early pregnancy in 4 strains of rats." Toxicol Sci 58(1): 135-143. Dabrowski, S., Hanke, W., Polańska, K., Makowiec-Dabrowska, T., & Sobala, W. (2003). Pesticide exposure and birthweight: an epidemiological study in Central Poland. Int J Occup Med Environ Health, 16(1), 31-39. Das PC, McElroy WK, Cooper RL (2000) Differential Modulation of Catecholamines by Chlorotriazine Herbicides in Pheochromocytoma (PC12) Cells In Vitro. Tox Sci. 56:324-331 Das, P.C., McElroy, W.K. & Cooper, R.L. (2001). Alteration of Catecholamines in Pheochromocytoma (PC12) Cells in vitro by the Metabolites of Chlorotriazine Herbicide. Toxicol Sci, Jan; 59(1): 127-37. Davidson, I. W. F. (1987) Metabolism and Kinetics of Atrazine in Man, Project # 101947, Department of Physiology/Pharmacology, Bowman Gray School of Medicine Sponsor: CIBA-GEIGY Corporation, Agricultural Division, Residue Chemistry Department, Greensboro, NC 27419. Davis LK, Murr AS, Best DS, Fraites MJP, Zorrilla LM, Marotsky MG, Stoker TE, Goldman JM, Cooper RL. (2011). The Effects of Prenatal Exposure to Atrazine on Pubertal and Postnatal Page 154 of 184 Reproductive Indices in the Female Rat. Reprod. Toxicol., doi:10.1016/j.reprotox.2011.04.004 De Roos, A. J., Zahm, S. H., Cantor, K. P., Weisenburger, D. D., Holmes, F. F., Burmeister, L. F., & Blair, A. (2003). Integrative assessment of multiple pesticides as risk factors for nonHodgkin's lymphoma among men. Occupational and Environmental Medicine, 60(9). doi: e11 Dennis, L. K., Lynch, C. F., Sandler, D. P., & Alavanja, M. C. (2010). Pesticide use and cutaneous melanoma in pesticide applicators in the agricultural heath study. Environ Health Perspect, 118(6), 812-817. doi: 10.1289/ehp.0901518 Deutsch, C.V. & Journel, A.G. (1998). GSLIB: Geostatistical Software Library and User’s Guide. Second Edition. Oxford University Press (New York, Oxford). Devos S., Van Den Heuvel R., et al . (2004). Limited effect of selected organic pollutants on cytokine production by peripheral blood leukocytes. Euro Cytokine Netw, Apr-Jun; 15(2): 145-51. Devos, S., Hooghe, R., Hooghe-Peters, E.L., De Bosscher, K., Vanden Berghe, W., Haegeman, G., Staels, B., Bauer, E., Roels, F., & Hooghe, R. (2003). Inhibition of cytokine production by the herbicide atrazine: Search for nuclear receptor targets. Biochem Pharmacol, Jan 15; 65(2): 303-8. Donna, A., Betta, P. G., Robutti, F., Crosignani, P., Berrino, F., & Bellingeri, D. (1984). Ovarian mesothelial tumors and herbicides: a case-control study. Carcinogenesis, 5(7), 941-942. Donna, A., Crosignani, P., Robutti, F., Betta, P. G., Bocca, R., Mariani, N., . . . Berrino, F. (1989). Triazine herbicides and ovarian epithelial neoplasms. Scand J Work Environ Health, 15(1), 47-53. Dooley, G., Hanneman, W., Carbone, D., Legare, M., Anderson, M., & Tessari, J. (2007). Development of an immunochemistry detection method for atrazine-induced albumin adducts. Chem Res Toxicol, Jul; 20(7): 1061-6. Dooley, G., Prenni, J., Prentiss, P., Cranmer, B., Anderson, M., & Tessari, J. (2006). Identification of a novel hemoglobin adduct in Sprague-Dawly rats exposed to atrazine. Chem Res Toxicol, May; 19(5): 692-700. Dooley, G.P., Reardon, K.F., Prenni, J.E., Tjalkens, R.B., Legare, M.E., Foradori, C.D., Tessari, J.E., & Hanneman, W.H. (2008). Proteomic analysis of diaminocholorotriazine adducts in Wistar rat pituitary glands and LbetaT2 rat pituitary cells. Chem Res Toxicol, Apr, 21(4): 844-51. Page 155 of 184 DR Mann and CA Barraclough (1973). Role of Estrogen and Progesterone in Facilitating LH Release in 4-Day Cyclic Rats. Endocrinology 93(3):694-699 Ebling, F.J.P. (2005) The Neuroendocrine Timing of Puberty. Reproduction 129:675-683 Eldridge, J.C., Wetzel, L.T., & Tyrey, L. (1999). Estrous cycle patterns of Sprague-Dawley rats during acute and chronic atrazine administration. Reprod Toxicol, Nov-Dec; 13(6): 491-9. Engel, L. S., Hill, D. A., Hoppin, J. A., Lubin, J. H., Lynch, C. F., Pierce, J., . . . Alavanja, M. C. (2005). Pesticide use and breast cancer risk among farmers' wives in the agricultural health study. Am J Epidemiol, 161(2), 121-135. doi: 161/2/121 [pii] 10.1093/aje/kwi022 Enoch, R.R., Stanko, J.P., Greiner, S.N., Youngblood, G.L., Rayner, J.L., & Fenton, S.E. (2007). Mammary gland development as a sensitive end point after acute prenatal exposure to an atrazine metabolite mixture in female Long-Evans rats. Environ Health Perspect, Apr, 115(4): 541-7. EPA), U. S. E. P. A. U. S. (2003). FIFRA Scientific Advisory Panel Final Meeting Report (July 17, 2003 Meeting). Retrieved from http://www.epa.gov/scipoly/sap/meetings/2003/071703_mtg.htm#materials. EPA, U. S. (2000). FIFRA Scientific Advisory Panel (SAP) Meeting Final Report (June 27-29, 2000). Arlington, VA: Retrieved from http://www.epa.gov/scipoly/sap/meetings/2000/june27/finalatrazine.pdf. Everett JW, and Sawyer CH. (1950). A 24-Hour Periodicity in the "LH-Release Apparatus" of Female Rats, Disclosed by Barbiturate Sedation. Endocrinology 47 (3): 198-218 Experimental procedures that decreased maternal prolactin lead to impaired TIDA neuronal development (Shyr et al. 1986, Phelps et al., 2000). Falagas ME, Pitsouni EI, Malietzis GA, Pappas G. (2008) Comparison of PubMed, Scopus, Web of Science, and Google Scholar: strengths and weaknesses. FASEB J., Feb;22(2):338-42. Epub 2007 Sep 20. PubMed PMID: 17884971. Fan, W., Yanase, T., Morinaga, H., Gonda, S., Okabe, T., Nomura, M., Hayes, T., Takayanagi, R., & Nawata, H. (2007a). Herbicide atrazine activates SF-1 by direct affinity and concomitant co-activators recruitments to induce aromatase expression via promoter II. Biochemical and Biophysical Research Communications, 355(4): 1012-8. Fan, W., Yanase, T., Morinaga, H., Gondo, S., Okabe, T., Normura, M., Komatsu, T., Moroashi, K., Hayes, T.B., Takayanagi, R., & Nawata, H. (2007b). Atrazine-induced aromatase expression is SF-1 dependent: implications for endocrine disruption in wildlife and reproductive cancers in humans. Environ Health Perspect, May, 115(5): 720-7. Page 156 of 184 Farr, S., Cai, J., Savitz, D., Sandler, D., Hoppin, J., & Cooper, G. (2006). Pesticide exposure and timing of menopause: the Agricultural Health Study. Am J Epidemiol, 163(8), 731-742. doi: kwj099 [pii] 10.1093/aje/kwj099. Farr, S., Cooper, G., Cai, J., Savitz, D., & Sandler, D. (2004). Pesticide use and menstrual cycle characteristics among premenopausal women in the Agricultural Health Study. Am J Epidemiol, 160(12), 1194-1204. doi: 160/12/1194 [pii] 10.1093/aje/. Fauser, B.C., de Jong, D., Olivennes, F. et al . (2002). Endocrine Profiles after Triggering of Final Oocyte Maturation with GnRH Agonist after Cotreatment with the GnRH Antagonist Ganirelix during Ovarian Hyperstimulation for in Vitro Fertilization. J Clin Endocrinol Metab. 87(2):709-715. Feldkamp ML, Carey JC, Sadler TW. Development of gastroschisis: review of hypotheses, a novel hypothesis, and implications for research. (2007) Am J Med Genet A., Apr 1;143(7):639-52. Review. PubMed PMID: 17230493. Fenton, S.E. (2006). Endocrine-disrupting compounds and mammary gland development: Early exposure and later life consequences. Endocrinology, 147: S18-24. FIFRA SAP. (2000b). "Atrazine :Hazard and Dose-Response Assessment and Characterization." Report from the FIFRA Scientific Advisory Panel Meeting of June 27-29, 2000. FIFRA Scientific Advisory Panel, Office of Science Coordination and Policy, Office of Prevention, Pesticides and Toxic Substances, U.S. Environmental Protection Agency. Washington, DC. SAP Report 2000-05. Available: http://www.epa.gov/scipoly/sap/meetings/2000/june27/finalatrazine.pdf FIFRA Scientific Advisory Panel (FIFRA SAP). (2009). A Set of Scientific Issues Being Considered by the Environmental Protection Agency Regarding: The Ecological Significance of Atrazine Effects on Primary Producers in Surface Water Streams in the Corn and Sorghum Growing Region of the United States (Part II). SAP Minutes No. 2009-06. May 12 – 14, 2009 FIFRA Scientific Advisory Panel. Document available at http://www.epa.gov/scipoly/sap/meetings/2009/051209meeting.html FIFRA Scientific Advisory Panel (FIFRA SAP). (2010a). A Set of Scientific Issues Being Considered by the Environmental Protection Agency Regarding: Re- Evaluation of Human Health Effects of Atrazine: Review of Experimental Animal and In Vitro Studies and Drinking Water Monitoring Frequency, FIFRA Office of Science Coordination and Policy, Office of Prevention, Pesticides and Toxic Substances, U.S. Environmental Protection Agency. Available at: http://www.epa.gov/scipoly/sap/meetings/2010/042610meeting.html SAP Minutes No. 2010-04. Page 157 of 184 FIFRA Scientific Advisory Panel (FIFRA SAP). (2010b). A Set of Scientific Issues Being Considered by the Environmental Protection Agency Regarding: Re- Evaluation of Human Health Effects of Atrazine: Review of Non-cancer Effects and Drinking Water Monitoring Frequency, FIFRA Office of Science Coordination and Policy, Office of Prevention, Pesticides and Toxic Substances, U.S. Environmental Protection Agency. Available at: http://www.epa.gov/scipoly/sap/meetings/2010/042610meeting.html SAP Minutes No. 201007. FIFRA Scientific Advisory Panel (FIFRA SAP). 2009. A Set of Scientific Issues Being Considered by the Environmental Protection Agency Regarding: The Ecological Significance of Atrazine Effects on Primary Producers in Surface Water Streams in the Corn and Sorghum Growing Region of the United States (Part II). SAP Minutes No. 2010-04. FIFRA Scientific Advisory Panel, Office of Science Coordination and Policy, Office of Prevention, Pesticides and Toxic Substances, U.S. Environmental Protection Agency. Washington, DC. Available at http://www.epa.gov/scipoly/sap/meetings/2009/051209meeting.html FIFRA Scientific Advisory Panel (FIFRA SAP). 2010. Transmittal of Meeting of the FIFRA Scientific Advisory Panel Meeting on the Draft Framework and Case Studies on Atrazine, Human Incidents, and the Agricultural Health Study: Incorporation of Epidemiology and Human Incident Data in Human Health Risk Assessment. SAP Minutes No. 2010-03. FIFRA Scientific Advisory Panel, Office of Science Coordination and Policy, Office of Prevention, Pesticides and Toxic Substances, U.S. Environmental Protection Agency. Washington, DC. Available: http://www.epa.gov/scipoly/sap/meetings/2010/index.html FIFRA Scientific Advisory Panel (FIFRA SAP). 2010. A Set of Scientific Issues Being Considered by the Environmental Protection Agency Regarding: Re- Evaluation of Human Health Effects of Atrazine: Review of Experimental Animal and In Vitro Studies and Drinking Water Monitoring Frequency. SAP Minutes No. 2010-04. FIFRA Scientific Advisory Panel, Office of Science Coordination and Policy, Office of Prevention, Pesticides and Toxic Substances, U.S. Environmental Protection Agency. Washington, DC. Available: http://www.epa.gov/scipoly/sap/meetings/2010/042610meeting.html FIFRA Scientific Advisory Panel (SAP). (1988). “A set of issues being considered by the Agency in connection with the peer review classification of atrazine a class C carcinogen.” Report from the FIFRA Scientific Advisory Panel Meeting of September 14, 1988. FIFRA Scientific Advisory Panel, Office of Science Coordination and Policy, Office of Prevention, Pesticides and Toxic Substances, U.S. Environmental Protection Agency. Washington, DC. Available: http://www.epa.gov/scipoly/sap/1988 FIFRA Scientific Advisory Panel (SAP). (2000a). “Atrazine: Hazard and Dose-Response Assessment and Characterization.” Report from the FIFRA Scientific Advisory Panel Meeting of June 27-29, 2000. FIFRA Scientific Advisory Panel, Office of Science Page 158 of 184 Coordination and Policy, Office of Prevention, Pesticides and Toxic Substances, U.S. Environmental Protection Agency. Washington, DC. Available: http://www.epa.gov/scipoly/sap/tools/atozindex/atrazine.htm FIFRA Scientific Advisory Panel (SAP). (2003). “Characterization of Epidemiology Data Relating to Prostate Cancer and Exposure to Atrazine.” Report from the FIFRA Scientific Advisory Panel Meeting of July, 2003. FIFRA Scientific Advisory Panel, Office of Science Coordination and Policy, Office of Prevention, Pesticides and Toxic Substances, U.S. Environmental Protection Agency. Washington, DC. Available: http://www.epa.gov/scipoly/sap/tools/atozindex/atrazine.htm Filipov, N.M., Pinchuk, L.M., Boyd, B.L., & Crittenden P.L. (2005). Immunotoxic effects of shortterm atrazine exposure in young male C57BL/6 mice. Toxicol Sci, Aug; 86(2): 324-32. Filipov, N.M., Stewart, M.A., Carr, R.L., & Sistrunk, S.C. (2007). Dopaminergic toxicity of the herbicide atrazine in rat striatal slices. Toxicology, Mar 22; 232(1-2): 68-78. Flower, K. B., Hoppin, J. A., Lynch, C. F., Blair, A., Knott, C., Shore, D. L., & Sandler, D. P. (2004). Cancer risk and parental pesticide application in children of agricultural health study participants. Environmental Health Perspectives, 112(5), 631-635. Follicle-Stimulating Hormone Surge in Postmenopausal Women: A Simulated Ovulatory Peak. PNAS 61:529-536 Foradori, C., Hinds, L., Hanneman, W., & Handa, R. (2009a). Effects of atrazine and its withdrawal on gonadotropin-releasing hormone neuroendocrine function in the adult female Wistar rat. Biol Reprod, Dec, 81(6): 1099-105. Foradori, C., Hinds, L., Hanneman, W., Legare, M., Clay, C., & Handa, R. (2009b). Atrazine inhibits pulsatile luteinizing hormone release without altering pituitary sensitivity to a gonadotropinreleasing hormone receptor agonist in female Wistar rats. Biol Reprod, Jul; 81(1): 40-5. Fox SR, Smith MS. (1986). Changes in the Pulsatile Pattern of Luteininzing Hormone Secretion During the Rat Extrous Cycle. Endocrinology 116:1485-1492 Fraites, M.J.P., Cooper, R.L., Buckalew, A., Jayaraman, S., Mills, L., and Laws, S.C. (2009). Characterization of the Hypothalamic-Pituitary-Adrenal Axis Response to Atrazine and Metabolites in the Female Rat. Tox. Sci. 112(1):88-99. Friedmann, A.S. (2002). Atrazine inhibition of testosterone production in rat males following peripubertal exposure. Reprod. Toxicol, May-Jun; 16: 275-9. Page 159 of 184 Frydman, R., Cornel, C., de Ziegler D. et al . (1992) Spontaneous luteininzing hormone surges can be reliably prevented by the timely administration of donadotrophin releasing hormone antagonist (Nal0Glu) during the late follicular phase. Human Reproduction 7(7):930-933 Fujimoto, N. and Honda, H. (2003). Effects of environmental estrogenic compounds on growth of a transplanted estrogen responsive pituitary tumor cell line in rats. Food Chem Toxicol, Dec, 41(12): 1711-7. Fukamachi, K., Han, B.S., Kim, C.K., Takasuka, N., Matsuoka, Y., Matsuda, E., Yamasaki, T., and Tsuda, H. (2004). Possible enhancing effects of atrazine and nonylphenol on 7,2dimethylbenz[a]anthracene-induced mammary tumor development in human c-Ha-ras protooncogene transgenic rats. Cancer Sci., May; 95(5): 404-10. Gao H.B., Shan L.X., Monder C., Hardy M.P. (1996). Suppression of endogenous corticosterone levels in vivo increases the steroidogenic capacity of purified rat Leydig cells in vitro. Endocrinology, May; 137(5):1714-8. Gilliom, R.J., Barbash, J.E., Crawford, C.G., Hamilton, P.A., Martin, J.D., Nakagaki, N., Nowell, L.H., Scott, J.C., Stackelberg, P.E., Thelin, G.P., and Wolock, D.M. (2006). The quality of our Nation’s waters—Pesticides in the Nation’s streams and ground water, 1992–2001. U.S. Geological Survey Circular 1291, 172 p. Available at http://pubs.usgs.gov/circ/2005/1291/pdf/circ1291.pdf Giusi, G., Facciolo, R.M., Canoaco, M., Alleva, E., Santucci, D., Belloni, V., & Dessi-Fulgheri, F. (2006). The endocrine disruptor atrazine accounts for a dimorphic somatostatinergic neuronal expression pattern in mice. Toxicol Sci, Jan; 89(1): 257-64. Gojmerac, T., Pleadin, J., Zuric, M., Rajkovic-Janje, R., & Korsic, M. (2004). Serum luteinizing hormone response to administration of gonadotropin-releasing hormone to atrazine-treated gilts. Vet Hum Toxicol, Oct, 6(5): 245-7. Goldfein A, Monroe SE. (1994) Ovaries. In: Greenspan FS, Baxter JD, editors.Basic and clinical endocrinology 4th ed. Norwalk, CT: Appleton and Lange, . pp. 419-70. Goldman, J.M., Laws, S.C., Balchak, S.K., Cooper, R.L. and Kavlock, R.J. Endocrine disrupting chemicals: Prepubertal exposures and effects on sexual maturation and thyroid activity in the female rat. A review of the EDSTAC recommendations. Critical Reviews in Toxicology. 30: 135-196, 20 Goldner, W. S., Sandler, D. P., Yu, F., Hoppin, J. A., Kamel, F., & Levan, T. D. (2010). Pesticide use and thyroid disease among women in the Agricultural Health Study. Am J Epidemiol, 171(4), 455-464. doi: kwp404 [pii] 10.1093/aje/kwp404 Page 160 of 184 Grumbach, M.M. (2002) The Neuroendocrinology of Human Puberty Revisited. Horm. Res. 57(supp 2):2-14 Guillaume AJ , Benjamin F, Sicuranza B, Deutsch S, Spitzer M. (1995). Luteal phase defects and ectopic pregnancy. Fertility and Sterility 63(1): 30-33 H. J. Paul, J. P. D., D. Hedley. (1993) The Absorption, Distribution, Degradation, and Excretion of [U-14C]-Triazine G 30027 in the Rat, Inveresk Reseach International IRI Report # 9523; Novartis # 523-92. Hase, Y., Tatsuno, M., Nishi, T., Kabe, Y., Yamaguchi, Y., Hanada, H., Kataoka, K., Ozawa, N., Natori, M., Handa, H., & Wantabe, H. (2008). Atrazine binds to F1F0-ATP synthase and inhibits mitochondrial function in sperm. Biochem Biophys Res Commun, Feb 1; 366(1): 6672. Heidelberg University National Center for Water Quality Research (NCWQR). 2010. Ohio Tributary Monitoring Program. Information on program, publications, and water quality data retrieval can be found at http://www.heidelberg.edu/WQL. Henderson KD, Bernstein L, Henderson B, Kolonel L, Pike MC. (2008) Predictors of the timing of natural menopause in the Multiethnic Cohort Study. Am J Epidemiol. 2008 Jun 1;167(11):1287-94. Epub Mar 21. PubMed PMID: 18359953. Henderson, B. E., Ponder, B., & Ross, R. K. (2003). Hormones, Genes and Cancer. new York, NY: Oxford University Press. Heneweer, M., Van den Berg, M., & Sanderson, J.T. (2004). A comparison of human H295R and rat R2C cell lines as in vitro screening tools for effects on aromatase. Toxicol Lett, Jan 15; 146(2): 183-94. Hertig AT, and Adams EC. (1967). Studies on the Human Oocyte and its Follicle. JCB 34(2): 647675 Hertig AT, Sheldon WH. (1943). Minimum Criteria Required to Prove Prima Facie Case of Traumatic Abortion or Miscarriage. Ann Surg 117:596-606 Hessel, P. A., Kalmes, R., Smith, T. J., Lau, E., Mink, P. J., & Mandel, J. (2004). A nested casecontrol study of prostate cancer and atrazine exposure. Journal of Occupational and Environmental Medicine, 46(4), 379-385. doi: 10.1097/01.jom.0000121128.73921.a1 Hessel, P., Kalmes, R., Smith, T., Lau, E., Mink, P., & Mandel, J. (2004). A nested case-control study of prostate cancer and atrazine exposure. J Occup Environ Med, 46(4), 379-385. doi: 00043764-200404000-00011 [pii] Page 161 of 184 Higley, E.B., Newsted, J.L., Zhang, X., Giesy, J.P., & Hecker, M. (2010). Assessment of chemicals effects on aromatase activity using the H295R cell line. Environ Sci Pollut Res Int., Jan 20, Epub ahead of print Hoar Zahm, S., Weisenburger, D., Cantor, K., Holmes, F., & Blair, A. (1993). Role of the herbicide atrazine in the development of non-Hodgkin's lymphoma. Scand J Work Environ Health, 19(2), 108-114. Hoar, S. K., Blair, A., Holmes, F. F., Boysen, C. D., Robel, R. J., Hoover, R., & Fraumeni, J. F. (1986). Agricultural herbicide use and risk of lymphoma and soft-tissue sarcoma. JAMA, 256(9), 1141-1147. Hoar, S. K., Blair, A., Holmes, F. F., Boysen, C., & Robel, R. J. (1985). Herbicides and colon cancer. Lancet, 1(8440), 1277-1278. Hocking, B. (2009). Melamine-contaminated powdered formula and urolithiasis. N Engl J Med., June 18; 360(25): 2676-7. Hoehn K., Salmon A., Hohnen-Behrens C., Turner N., Hoy A., Maghzal G., Socker R., Van Remmen H., Kraegen E., Cooney G., Richardson A., James D. (2009). Insulin resistance is a cellular antioxidant defense mechanism. Proc Natl Acad Sci U S A, Oct 20 106(42):1778792. Holloway, A.C., Anger, D.A., Crankshaw, D.J., Wu, M., & Foster, W.G. (2008). Atrazine-induced changes in aromatase activity in estrogen sensitive target tissues. J. Appl. Toxicol., Apr, 28(3): 260-70. Hopenhayn-Rich, C., Stump, M. L., & Browning, S. R. (2002). Regional assessment of atrazine exposure and incidence of breast and ovarian cancers in Kentucky. Archives of Environmental Contamination and Toxicology, 42(1), 127-136. Hoppin JA, Umbach DM, London SJ, Henneberger PK, Kullman GJ, Coble J, Alavanja MC, Beane Freeman LE, Sandler DP. (2009) Pesticide use and adult-onset asthma among male farmers in the Agricultural Health Study. Eur Respir J., Dec;34(6):1296-303. Epub 2009 Jun 18. PubMed PMID: 19541724; PubMed Central PMCID: PMC2822385. Hoppin JA, Yucel F, Dosemeci M, Sandler DP. (2002) Accuracy of self-reported pesticide use duration information from licensed pesticide applicators in the Agricultural Health Study. J Expo Anal Environ Epidemiol., Sep;12(5):313-8. PubMed PMID: 12198579. Hoppin, J. A., Umbach, D. M., London, S. J., Alavanja, M. C. R., & Sandler, D. P. (2002). Chemical predictors of wheeze among farmer pesticide applicators in the agricultural health study. American Journal of Respiratory and Critical Care Medicine, 165(5), 683-689. doi: 10.1164/rccm.2106074. Page 162 of 184 Hoppin, J. A., Yucel, F., Dosemeci, M., & Sandler, D. P. (2002). Accuracy of self-reported pesticide use duration information from licensed pesticide applicators in the Agricultural Health Study. J Expo Anal Environ Epidemiol, 12(5), 313-318. doi: 10.1038/sj.jea.7500232 Hornemann et al . “Seasonality of omphalocele in Northern Germany,” (2008) [Ultraschall in Med. 29:591-595. ]. Hornemann, A., Weichert, J., Thill, M., Altgassen, C., Luedders, D., & Bohlmann, M. K. (2009). Seasonality of Omphalocele in Northern Germany. Ultraschall in Der Medizin, 30(6), 591595. doi: 10.1055/s-2008-1027896. [German] Hossain, M. M. & Filipov, N. M.(2008). Alteration of dopamine uptake into rat striatal vesicles and synaptosomes caused by an in vitro exposure to atrazine and some of its metabolites. Toxicology, Jun 3; 248(1): 52-8. Hovey, R.C., Coder, P.S., Wolf, J.C., Sielken, R.L., Tisdel, M.O., Brekenridge, C.B. (2011). Quantitative assessment of mammary gland development in female Long Evans rats following in utero exposure to atrazine. Tox Sci, 119: 380-390. Huang, Y., Batterman, S. 2003. An Extreme Value Analysis of Pollutant Concentrations in Surface Soils Due to Atmospheric Deposition. Human and Ecological Risk Assessment, 9:17291746. Hyun Joo, K. C., Ernest Hodgson. (2010) Human metabolism of atrazine, Pesticide Biochemistry and Physiology 98, 73-79. Iffy L, Wingate MB. (1970). Risks of Rhythm Method of Birth Control. J Reprod Med 5:96-102 Iffy l. (1961). Contribution to the Aetiology of Ectopic Pregnancy. BJOG: An International Journal of Obstetrics & Gynaecology 68(3):441-450 ILSI (International Life Sciences Institute), (1999). A Framework for Estimating Pesticide Concentrations in Drinking Water for Aggregate Exposure Assessments. International Life Sciences Institute Risk Science Institute Working Group Report, Washington, D.C., 61 pp. Available at http://www.ilsi.org/ResearchFoundation/Publications/Estimating%20Pesticide%20Concentra tions.pdf International Life Sciences Institute (ILSI). (1999). A Framework for Estimating Pesticide Concentrations in Drinking Water for Aggregate Exposure Assessments. International Life Sciences Institute Risk Science Institute Working Group Report, Washington, D.C., 61 pp. Available at http://www.ilsi.org/ResearchFoundation/Publications/Estimating%20Pesticide%20Concentra tions.pdf Page 163 of 184 Isaaks, E.H., and R.M. Srivastava. 1989. Applied Geostatistics. Oxford University Press (New York, Oxford). Jakobovits AA, Iffy L. (1988). Disorders of Follicular Ripening and Ovulation in Human. Acta Med Hungar 45:179-190 Jerry Campbell, M. A., Harvey Clewell. (2011) PBPK model for atrazine and chlorinated metabolites, Sponsor: Syngenta Crop Protection, LLC. Judson R.S., Houck K.A., Kavlock R.J., Knudsen T.B., Martin M.T., et al . (2009). In vitro Screening of Environmental Chemicals for Targeted Testing Prioritization - The Tox Cast Project. Environ Health Perspect, doi:10.1289/ehp.0901392 Juliani, C.C., Silva-Zacarin, E.C., Santos, D.C., & Boer, P.A.(2008). Effects of atrazine on female Wistar rats: morphological alterations in ovarian follicles and immunocytochemical labeling of 90 kDa heat shock protein. Micron, Jul, 39(5): 607-16. Kalra, S.P., & Kalra, P.S.(1983). Neural regulation of luteinizing hormone secretion in the rat. Endocr Rev., Fall; 4(4): 311-51. Kamel, A., Qian, Y., Podhorniak, L. (2010). Final Report on “Determination of Atrazine and Its Metabolites in Rat Plasma, Milk, and Tissue.” ACB Project B10-15. Memorandum dated July 13, 2010. Analytical Chemistry Branch/BEAD/OPP to Ralph Cooper, NHEERL/ORD Kandori, H., Suzuki, S., Asamoto, M., et al . (2005). Influence of atrazine administration and reduction of calorie intake on prostate carcinogenesis in probasin/SV40 T antigen transgenic rats. Cancer Sci, Apr; 96(4): 221-6. Karrow, N.A., McCay, J.A., Brown, R.D., Musgrove, D.L., Guo, T.L., Germolec, D.R., & White, K.L. Jr (2005). Oral exposure to atrazine modulates cell-mediated immune function and decreases host resistance to the B16F10 tumor model in female B6C3F1 mice. Toxicology, Apr 1; 209(1): 15-28. Kettles, M. A., Browning, S. R., Prince, T. S., & Horstman, S. W. (1997). Triazine herbicide exposure and breast cancer incidence: An ecologic study of Kentucky counties. Environmental Health Perspectives, 105(11), 1222-1227. Kettles, M., Browning, S., Prince, T., & Horstman, S. (1997). Triazine herbicide exposure and breast cancer incidence: an ecologic study of Kentucky counties. Environ Health Perspect, 105(11), 1222-1227. Kim, D., Breckenridge, C. (Syngenta) (2010). Atrazine - Derivation of the short-term inhalation NOAEL using a physiologically-based pharmacokinetic model. Unpublished data from Syngenta Page 164 of 184 Kirby E.D., Geraghty A.C., Ubuka T., Bentley G.E., Kaufer D. (2009). Stress increases putative gonadotropin inhibitory hormone and decreases luteinizing hormone in male rats. Proc Natl Acad Sci USA, 106(27): 11324-9. Kłuciński, P., Kossmann, S., Tustanowski, J., Friedek, D., & Kamińska-Kołodziej, B. (2001). Humoral and cellular immunity rates in chemical plant workers producing dust pesticides. Med Sci Monit, 7(6), 1270-1274. doi: 1270 [pii]. Kmetic, I., Gaurina Srcek, V., Silvac, I., Simic, B., Kniewald, Z., & Kniewald, J. (2008). Atrazine exposure decreases cell proliferation in Chinese Hamster Ovary (CHO-K1) cell line. Bull Environ Contam Toxicol., Aug, 81(2): 205-9. Kniewals, J., Kmetič, I., Gaurina Srček, V., Šimiċ,B., & Kniewald, Z. (2003). Cytotoxicity of Herbicide Atrazine On Chinese Hamster ovary cell Line In Monolayer. Toxicol Lett, Sept; 144 (Suppl 1):S111-S112. Kossmann, S., Konieczny, B., & Hoffmann, A. (1997). The role of respiratory muscles in the impairment of the respiratory system function in the workers of a chemical plant division producing pesticides. Przegl Lek, 54(10), 702-706. Koutros, S., Alavanja, M. C., Lubin, J. H., Sandler, D. P., Hoppin, J. A., Lynch, C. F., . . . Freeman, L. E. (2010). An update of cancer incidence in the Agricultural Health Study. J Occup Environ Med, 52(11), 1098-1105. doi: 00043764-201011000-00006 [pii] 10.1097/JOM.0b013e3181f72b7c Kyriakidis, P.C. and Journel, A.G. (2001). Stochastic modeling of atmospheric pollution: a spatial time-series framework. Part I: Methodology. Atmospheric Environment. 35, 2331-2337. Lall, U., and Sharma, A. (1996). A Nearest Neighbor Bootstrap for Resampling Hydrologic Time Series. Water Resources Research. 32(3), 679-693. Larson, S.J., C.G. Crawford, and R.J. Gilliom. (2004). Development and application of Watershed Regressions for Pesticides (WARP) for estimating atrazine concentration distributions in streams. U.S. Geological Survey Water-Resources Investigations Report 03–4047, 68 p. Available at http://pubs.usgs.gov/wri/wri034047/ Laville, N., Balaguer, P., Brion, F., Hinfray, N., Casellas, C., Porcher, J.M., & Ait-Aissa, S. (2006). Modulation of aromatase activity and mRNA by various selected pesticides in the human choriocarcinoma JEG-3 cell line. Toxicology, Nov 10; 228(1): 98-108. Laws, S. C., Ferrell, J. M., Stoker, T. E., Schmid, J. & Cooper, R. L. (2000). The Effects of Atrazine on Female Wistar Rats: An Evaluation of the Protocol for Assessing Pubertal Development and Thyroid Function. Toxicol Sci, 58: 366-76. Page 165 of 184 Laws, S.C., Ferrell, J.M., Stoker, T.E., & Cooper, R.L. (2003). Pubertal development in female Wistar rats following exposure to propazine and atrazine biotransformation by-products, diamino-S-chlorotriazine and hydroxyatrazine. Toxicol Sci, Nov; 76(1): 190-200. Laws, S.C., Hotchkiss, M., Ferrell, J., Jayaraman, S., Mills, L., Modic, W., Tinfo, N., Fraites, M., Stoker, T., & Cooper, R. (2009). Chlorotriazine Herbicides and Metabolites Activate an ACTH-Dependent Release of Corticosterone in Male Wistar Rats. Tox. Sci., 112(1): 78-87. Laws, S.C., Stoker, T.E., Ferrell, J.M., Hotchkiss, M.G., & Cooper R.L. (2007). Effects of Altered Food Intake During Pubertal Development in Male and Female Wistar Rats. Toxicol. Sci., Nov; 100(1): 194-202. LeBlanc Andre, Lekha Sleno. (2011) Atrazine Metabolite Screening in Human Microsomes: Detection of Novel Reactive Metabolites and Glutathione Adducts by LC-MS, Chemical Research in Toxicology 24, 329-339. Lee, H.K. (2004). Bayesian Nonparametrics via Neural Networks. Philadelphia: ASA/SIAM. Lee, W. J., Sandler, D. P., Blair, A., Samanic, C., Cross, A. J., & Alavanja, M. C. R. (2007). Pesticide use and colorectal cancer risk in the Agricultural Health Study. International Journal of Cancer, 121(2), 339-346. doi: 10.1002/ijc.22635 Lim, S., Ahn, S.Y., Song, I.C., Chung, M.H., Jang, H.C., Park, K.S., Lee, K., Pak, Y.K., & Lee, H.K. (2009). Chronic exposure to the herbicide, atrazine, causes mitochondrial dysfunction and insulin resistance. PLOS One, 4(4): 1-11. Lyons, W.R., Berlin, I., and Friendlander, S. (1942) Cornification of balano-preputial epithelium in normal rats and castrated rats treated with testosterone proprionate. Endocrinology, 31(6): 659-63. MacLennan, P. A., Delzell, E., Sathiakumar, N., & Myers, S. L. (2003). Mortality among triazine herbicide manufacturing workers. Journal of Toxicology and Environmental Health-Part aCurrent Issues, 66(6), 501-517. doi: 10.1080/15287390390156254 MacLennan, P. A., Delzell, E., Sathiakumar, N., Myers, S. L., Cheng, H., Grizzle, W., . . . Wu, X. C. (2002). Cancer incidence among triazine herbicide manufacturing workers. Journal of Occupational and Environmental Medicine, 44(11), 1048-1058. doi: 10.1097/01.jom.0000038322.36316.59 MacLennan, P., Delzell, E., Sathiakumar, N., & Myers, S. (2003). Mortality among triazine herbicide manufacturing workers. J Toxicol Environ Health A, 66(6), 501-517. Page 166 of 184 MacLennan, P., Delzell, E., Sathiakumar, N., Myers, S., Cheng, H., Grizzle, W., . . . Wu, X. (2002). Cancer incidence among triazine herbicide manufacturing workers. J Occup Environ Med, 44(11), 1048-1058. Malik, S.I., Terzoudi, G.I., & Pantelias, G.E. (2004). SCE analysis in G2 lymphocyte prematurely condensed chromosomes after exposure to atrazine: the non-dose-dependent increase in homologous recombinational events does not support its genotoxic mode of action. Cytogenet and Genome Research, 104(1-4): 315-9. Manske, M., Beltz, L, and Dhanwada, K. (2004). Low-Level Atrazine Exposure Decreases Cell Proliferation in Human Fibroblasts. Arch. Environ. Contam. Toxicol., 46: 438-44. Mao D., Miller, P.S., Davidson, P.C., Prenger, J.J., Zwilling, L.F., Cox, M.J., Harbourt, C.M., and Hendley, P. (2011). Methodology for Compiling a Database of Atrazine Monitoring Program Community Water System Facilities, Watersheds, and Their Characteristics. Unpublished study from Syngenta Crop Protection, Inc. MRID 48470008 Maravilla-Galvan, R., Fierro, R., Gonzalez-Marquez, H., Gomez-Arroyo, S., Jimenez, I., & Betancourt, M. (2009). Effects of atrazine and fenoxaprop-ethyl on capacitation and the acrosomal reaction in boar sperm. Int. J. Toxicol., Jan-Feb, 28(1): 24-32. Marshall, F. (1966) Marshall’s Physiology of Reproduction. Longmans, Green and Co. LTD Matsumoto AM, Karpas AE, Southworth MB, Dorsa DM, Bremner WJ (1986). Evidence for Activation of the Central Nervous System-Pituitary Mechanism for Gonadotropin Secretion at the Time of Puberty in the Male Rat. Endocrinology 119(1):362-369 Mattix, K., Winchester, P., & Scherer, L. (2007). Incidence of abdominal wall defects is related to surface water atrazine and nitrate levels. J Pediatr Surg, 42(6), 947-949. doi: S00223468(07)00056-5 [pii] 10.1016/j.jpedsurg.2007.01.027. McElroy, J. A., Gangnon, R. E., Newcomb, P. A., Kanarek, M. S., Anderson, H. A., Brook, J. V., . . . Remington, P. L. (2007). Risk of breast cancer for women living in rural areas from adult exposure to atrazine from well water in Wisconsin. Journal of Exposure Science and Environmental Epidemiology, 17(2), 207-214. doi: 10.1038/sj.jes.7500511 McElroy, J., Gangnon, R., Newcomb, P., Kanarek, M., Anderson, H., Brook, J., . . . Remington, P. (2007). Risk of breast cancer for women living in rural areas from adult exposure to atrazine from well water in Wisconsin. J Expo Sci Environ Epidemiol, 17(2), 207-214. doi: 7500511 [pii] 10.1038/sj.jes.7500511 McMullin T.S. (2003). Pharmacokinetic modeling of disposition and time-course studies with [14C]atrazine. J. Toxicol. Environ. Health, May 23;66(10):941-64. Page 167 of 184 McMullin TS, Andersen ME, Nagahara A, Lund TD, Pak T, Handa RJ, Hanneman, WH. (2004) Evidence That Atrazine and Diaminochlorotriazine Inhibit the Estrogen/Progesterone Induced Surge of Luteinizing Hormone in Female Sprague-Dawley Rats Without Changing Estrogen Receptor Action Tox Sci. 79:278-286 McMullin, T. S., Hanneman, W. H., Cranmer, B. K., Tessari, J. D., and Andersen, M. E. (2007) Oral absorption and oxidative metabolism of atrazine in rats evaluated by physiological modeling approaches, Toxicology 240, 1-14. Meczekalski b, Podfigurna-Stopa A, Warenik-Szymankiewicz Al. Riccardo Genazzani A.(2008). Functional hypothalamic amenorrhea: Current view on neuroendocrine aberrations. Gyn Endocr. 24(1):4-11 Meek, M.E., J.R. Bucher, S.M. Cohen et al . (2003). A framework for human relevance analysis of information on carcinogenic modes of action. Crit Rev Toxicol, 33(6): 591-653. Melvin M. Grumbach (2002). The Neuroendocrinology of Human Puberty Revisited. Horm Res;57(suppl 2):2–14 Merritt, A. 2011. Summary of Site Selection and Monitoring Plan for Atrazine Daily Monitoring in Selected Community Water Systems: Study Plan. Report Number: T001301-03. Task Number: T001301-03. Mikamo K (1968) Mechanism of Non-dysjunction of Meiotic Chromosomes and of Degeneration of Maturation Spindles in Eggs Affected by Intrafollicular Overripeness. Experientia 24: 75-78 Miller, P.S., Northcott, W.J., Andrus, J. M., Zwilling, L.F., Cheplick, J. M., Harbourt, C.M., and Hendley, P. (2011). Atrazine - Development of the PRZM-Hybrid Watershed Exposure Modeling System, Unpublished study from Syngenta Crop Protection, Inc. MRID 48470007 Mills, P. (1998). Correlation analysis of pesticide use data and cancer incidence rates in California counties. Arch Environ Health, 53(6), 410-413. Mills, P. K., & Yang, R. (2003). Prostate cancer risk in California farm workers. Journal of Occupational and Environmental Medicine, 45(3), 249-258. doi: 10.1097/01.jom.0000058339.05741.0c Mills, P., & Yang, R. (2006). Regression analysis of pesticide use and breast cancer incidence in California Latinas. J Environ Health, 68(6), 15-22; quiz 43-14. Minnema, D. (2001). "Special Study (non-guideline). Comparison of the LH surge in female rats administered atrazine, simazine, or diaminochlorotriazine (DACT) via gavage for one month." Covance Laboratories Inc. Vienna, VA Laboratory report number 6117-398(MRID # 45471002). Page 168 of 184 Minnema, D.J. (2002). 52-Week toxicity study of simazine, atrazine, and DACT administered in the diet to female rats. Covance Laboratories Inc., 9200 Leesburg Pike, Vienna, VA. Laboratory Identification Number 6117-399. February 21, 2002. MRID 45622309. Unpublished Mizota, K. & Ueda, H. (2006). Endocrine disrupting chemical atrazine causes degranulation through Gq/11 protein-coupled neurosteroid receptor in mast cells. Toxicol Sci., Apr, 90(2): 362-8. Modic, MW. (2004). The role of testicular aromatase in the atrazine mediated changes in estrone and estradiol in the male wistar rat. North Carolina State University. Molcrani M.C., Duval F., Crocq M.A., Bailey P., Macher J.P. (1997). HPA axis dysfunction in depression: Correlation with monoamine system abnormalities. Psychoneuroendocrinology, 22, Suppl 1: S63-8 Monosson E, Kelce WR, Lambright C, Ostby J, Gray LE Jr. (1999). Peripubertal exposure to the antiandrogenic fungicide, vinclozolin, delays puberty, inhibits the development of androgendependent tissues, and alters androgen receptor function in the male rat. Toxicol Ind Health, 15(1-2): 65-79. Morseth, S. (1996a). "Evaluation of the lutenizing hormone (LH) surge in atrazine expose female Sprague Dawley rats." Covance Laboratories Inc. Vienna, VA Laboratory report # CHV 2386-111(MRID # 43934406). Morseth, S. (1996b). "Evaluation of the lutenizing hormone (LH) surge in atrazine exposed female Sprague Dawley rats - 6 month report." Covance Laboratories Inc. Vienna, VA Laboratory study no. CHV 2386-108(MRID # 44152102). Mosquin, P., Whitmore, R., and Chen W. (2011a). Estimation of statistical confidence in upper centile concentrations of historical atrazine monitoring data of Community Water Systems. Unpublished study from Syngenta Crop Protection, Inc. MRID 48470005 Mosquin, P., Whitmore, R., and Chen W. (2011b). Impact of Alternative Monitoring Frequencies on Estimation of Atrazine Rolling Average Environmental Concentrations. Unpublished study from Syngenta Crop Protection, Inc. MRID 48470006 Mu, H., Zhang, P., & Xu, J. (2008). Testicular toxicity and mechanisms of chlorotoluron compounds in the mouse. Toxicology Mechanisms and Methods, 18(5): 399-403. Muir, K., Rattanamongkolgul, S., Smallman-Raynor, M., Thomas, M., Downer, S., & Jenkinson, C. (2004). Breast cancer incidence and its possible spatial association with pesticide application in two counties of England. Public Health, 118(7), 513-520. doi: 10.1016/j.puhe.2003.12.019 Page 169 of 184 Munger, R., Isacson, P., Hu, S., Burns, T., Hanson, J., Lynch, C. F., et al . (1997). Intrauterine growth retardation in Iowa communities with herbicide-contaminated drinking water supplies. Environmental Health Perspectives, 105(3), 308-314. Narotsky, M. G., D. S. Best, et al . (2001). "Strain comparisons of atrazine-induced pregnancy loss in the rat." Reprod Toxicol 15(1): 61-69. Ochoa-Acuña, H., & Carbajo, C. (2009). Risk of limb birth defects and mother's home proximity to cornfields. Sci Total Environ, 407(15), 4447-4451. doi: S0048-9697(09)00373-8 [pii] 10.1016/j.scitotenv.2009.04.028. Ochoa-Acuña, H., Frankenberger, J., Hahn, L., & Carbajo, C. (2009). Drinking-water herbicide exposure in Indiana and prevalence of small-for-gestational-age and preterm delivery. Environ Health Perspect, 117(10), 1619-1624. doi: 10.1289/ehp.0900784. Odell WD, Swerdloff RS. (1968). Progestogen-Induced Luteinizing and Ojeda, S.R., and Urbanski, H.F. (1994). Puberty in the rat. In E. Knobil and J.D. Neill (Eds.): The Physiology of Reproduction, 2nd ed. New York, Raven Press. 363-409 Olivennes, F., Alvarez, S., Bouchard, P. et al . (1998). The use of GnRH antagonist (Cetrorelix®) in a single dose protocol in IVF-embryo transfer: a dose finding study of 3 versus 2 mg. Human Reproduction 13(9):2411-2413 Orsi, L., Delabre, L., Monnereau, A., Delval, P., Berthou, C., Fenaux, P., . . . Clavel, J. (2009). Occupational exposure to pesticides and lymphoid neoplasms among men: results of a French case-control study. Occupational and Environmental Medicine, 66(5), 291-298. doi: 10.1136/oem.2008.040972 Padoan, S., Ribatet, M., Sisson, S. 2010. Likelihood-Based Inference for Max-Stable Processes. Journal of the American Statistical Association 105 (489): 263-277. Philippe E, Ritter J, Lefakis P, Laedlein-Greilsammer D, Itten S, Foussereau S (1970). Tubal Pregnancy, Delayed Ovulation and Anomaly of Nidation. Gynecol Obstet. 69:Suppl 5:617+. Pinchuk, L.M., Lee, S.R., & Filipov, N.M. (2007). In vitro atrazine exposure affects the phenotypic and functional maturation of dendritic cells. Toxicology and Applied Pharmacology, Sept, 223(3): 206-17. Plant, T.M.; Barker-Gibb, L.M. (2004). Neurobiological mechanisms of puberty in higher primates . Human Reproduction Upda t e 10(1): 67-77 Page 170 of 184 Pogrmic, K., Fa, S., Dakic, V., Kaisarevic, S. & Kovacevic, R. (2009). Atrazine oral exposure of peripubertal male rats downregulates steroidogenesis gene expression in Leydig cells. Toxicol. Sci., 111(1): 189-97. Prilusky J, Vermouth NT., Deis RP. (1984). A Dual Modulatory Effect of Progesterone on the LHRH-Induced LH Release. J Steroid Biochem 21(1):107-110 Pruett, S.B., Fan, R., Zheng, Q., & Schwab, C. (2009). Patterns of immunotoxicity associated with chronic as compared with acute exposures to chemical or physical stressors and their relevance with regard to the role of stress and with regard to immunotoxicity testing. Toxicological Sciences, June, 109(2): 265-75. Pruett, S.B., Fan, R., Zheng, Q., Myers, L.P., & Hebert, P.(2003). Modeling and predicting immunological effects of chemical stressors: characterization of a quantitative biomarker for immunological changes caused by atrazine and ethanol. Toxicol Sci., Oct, 75(2): 343-54. Pryor JL, Hughes C, Foster W, Hales BF, Robaire B. (2000) Critical windows of exposure for children's health: the reproductive system in animals and humans. Environ Health Perspect., Jun;108 Suppl 3:491-503. Review. PubMed PMID: 10852849; PubMed Central PMCID: PMC1637840. Pulkkinen MO, Jaakkola UM. (1989). Low serum progesterone levels and tubal dysfunction--a possible cause of ectopic pregnancy. Am J Obstet Gynecol. 161(4):934-7 Ralph Cooper, A. B., Melanie Fraites, Jerome Goldman, Susan Laws, Michael Narotsky, and Tammy Stoker. (2010) Evaluating the effect of the chlorotriazine herbicide atrazine on the amplitude of the pre-ovulatory LH surge in the Long-Evans Rat, In Unpublished internal report. Rayburn, A.L., Moody, D.D., and Freeman, J.L. (2005). Cytotoxicity of technical grade versus formulations of atrazine and acetochlor using mammalian cells. Bull. Environ. Contam. Toxicol., 75: 691-8. Rayner, J. L., Wood, C., Fenton, S. E. (2004). Exposure parameters necessary for delayed puberty and mammary gland development in Long–Evans rats exposed in utero to atrazine. Toxicol Appl Pharmacol, Feb 15; 195(1): 23-34. Rayner, J.L., Enoch, R.R., Fenton, S.E. (2005). Adverse effects of prenatal exposure to atrazine during a critical period of mammary gland growth. Tox Sci, 87: 255-266. Rayner, J.L., Wolf, D.C., Rayner, J.L., Enoch, J.L., Enoch, R.R., & Fenton, S.E. (2007). Atrazineinduced reproductive tract alterations after transplacental and/or lacational exposure in male Long-Evans rats. Toxicology and Applied Pharmacology, Feb, 218(3): 238-48. Page 171 of 184 Rhomberg, L.R., and Lewandowski, T.A. (2006). Methods for identifying a default cross-species scaling factor. Hum Ecol Risk Assess 12, 1094-1127. Rivest, .S & Rivier, C. (1995). The role of corticotropin-releasing factor and interleukin-1 in the regulation of neurons controlling reproductive functions. Endocr Rev, 16(2): 177-99. Rivest, P. & Sanderson, J. (2010). Effects of pesticides on aromatase expression in bioluminescent mice and GSK 3BETA/BETA-signalling in LNCAp human prostate cancer cells. The Toxicologist 114, 405. Rivier, C. & Rivest, S. (1991). Effect of stress on the activity of the hypothalamic-pituitary-gondal axis: peripheral and central mechanisms. Biol Reprod., October 1; 45 (4): 523-532 Rivier, C., Rivier, J., & Vale, W. (1986). Stress-induced inhibition of reproductive functions: role of endogenous corticotropin-releasing factor. Science, 231(4738): 607-9. Rodriguez, V.M., Thiruchelvam, M., & Cory-Slechta, D.A. (2005). Sustained exposure to the widely used herbicide atrazine: Altered function and loss of neurons in brain monoamine systems. Environmental Health Perspectives, Jun, 113(6): 708-15. Rooney, A.A., Matulka, R.A., & Luebke, R.W. (2003). Developmental atrazine exposure suppresses immune function in male, but not female Spragus-Dawley rats. Toxicol Sci., Dec, 76(2): 366-75. Rosenberg, B.G., Chen, H., Folmer, J., Liu, J., Papadopoulos, V., & Zirkin, B.R. (2008). Gestational exposure to atrazine: effects on the postnatal development of male offspring. J Androl., May-June; 29(3): 304-11. Ross, M.K., Jones, T.L., and Filipov, M.N. (2009). Disposition of the herbicide 2-chloro-4(ethylamino)-6-(isopropylamino)-s-triazine (atrazine) and its major metabolites in mice: a liquid chromatography/mass spectrometry analysis of urine, plasma, and tissue levels. Drug Metab. Dispos., Apr; 37(4): 776-86. Rowe, A.M., Brundage, K.M., & Barnett, J.B. (2007). In vitro atrazine exposure inhibits human natural killer cell lytic granule release. Toxicol Appl Pharmacol, Jun 1; 221(2): 179-88. Rowe, A.M., Brundage, K.M., & Barnett, J.B. (2008). Developmental immunotoxicity of atrazine in rodents. Basic Clin Pharmacol Toxicol, Feb; 102(2): 139-45. Rowe, A.M., Brundage, K.M., Schafer, R., & Barnett, J.B. (2006). Immunomodulatory effects of maternal atrazine exposure on male Balb/c mice. Toxicol Appl Pharmacol, July 1; 214(1): 69-77 Page 172 of 184 Ruder, A. M., Waters, M. A., Butler, M. A., Carreón, T., Calvert, G. M., Davis-King, K. E., . . . Group, B. C. C. S. (2004). Gliomas and farm pesticide exposure in men: the upper midwest health study. Arch Environ Health, 59(12), 650-657. Rull, R. P., Gunier, R., Von Behren, J., Hertz, A., Crouse, V., Buffler, P. A., & Reynolds, P. (2009). Residential proximity to agricultural pesticide applications and childhood acute lymphoblastic leukemia. Environmental Research, 109(7), 891-899. doi: 10.1016/j.envres.2009.07.014 Rull, R., Gunier, R., Von Behren, J., Hertz, A., Crouse, V., Buffler, P., & Reynolds, P. (2009). Residential proximity to agricultural pesticide applications and childhood acute lymphoblastic leukemia. Environ Res, 109(7), 891-899. doi: S0013-9351(09)00139-X [pii] 10.1016/j.envres.2009.07.014 Rusiecki, J. A., De Roos, A., Lee, W. J., Dosemeci, M., Lubin, J. H., Hoppin, J. A., . . . Alavanja, M. C. R. (2004). Cancer incidence among pesticide applicators exposed to atrazine in the agricultural health study. Journal of the National Cancer Institute, 96(18), 1375-1382. Rusiecki, J., De Roos, A., Lee, W., Dosemeci, M., Lubin, J., Hoppin, J., . . . Alavanja, M. (2004). Cancer incidence among pesticide applicators exposed to atrazine in the Agricultural Health Study. J Natl Cancer Inst, 96(18), 1375-1382. doi: 96/18/1375 [pii] 10.1093/jnci/djh264 Rusiecki, J.A., De Roos, A., Lee, W.J., Dosemci, M., Lubin J.H., Hoppin, J.A., Blair, A., & Alavanja, M.C. (2004). Cancer incidence among pesticide applicators exposed to atrazine in the Agricultural Health Study. J Natl Cancer Inst, Sept 15; 96(18): 1375-82. Safranski TJ, Lamberson WR, Keisler DH.(1993). Correlations among Three Measures of Puberty in Mice and Relationships with Estradiol Concentration and Ovulation. Biol Repro 48:669673 Saldana, T. M., Basso, O., Hoppin, J. A., Baird, D. D., Knott, C., Blair, A., et al . (2007). Pesticide exposure and self-reported gestational diabetes mellitus in the agricultural health study. Diabetes Care, 30(3), 529-534. doi: 10.2337/dc06-1832. Sanderson J.T., Letcher R.J., Heneweer M., Giesy J.P., van den Berg M. (2001). Effects of chloros-triazine herbicides and metabolites on aromatase activity in various human cell lines and on vitellogenin production in male carp hepatocytes. Environ health Perspect, Ovt; 109(10): 1027-31. Sanderson J.T., Seinen W., Giesy J.P., van den Berg M. (2000). 2-Chloro-s-triazine herbicides induce aromatase (CYP19) activity in H295R human adrenocortical carcinoma cells: a novel mechanisms for estrogenicity? Toxicol Sci, Mar; 54(1): 121-27. Page 173 of 184 Sanderson, J.T., Boerma, J., Gideon, W.A., Lansbergen, G.W. & van den Berg, M. (2002). Induction and inhibition of aromatase (CYP19) activity by various classes of pesticides in H295R human adrenocortical carcinoma cells. Toxicol Appl Pharmacol. Jul 1; 182: 44-54. Santen R.J., Brodie H., Simpson E.R., Siiteri P.K., Brodie A . (2009). History of Aromatase: saga of an important biological mediator and therapeutic target. Endocr Rev, Jun; 30(4): 343-75. SAP. (2010a) A Set of Scientific Issues Being Considered by the Environmental Protection Agency Regarding: Re- Evaluation of Human Health Effects of Atrazine: Review of Experimental Animal and In Vitro Studies and Drinking Water Monitoring Frequency, FIFRA Office of Science Coordination and Policy, Office of Prevention, Pesticides and Toxic Substances, U.S. Environmental Protection Agency. Available: http://www.epa.gov/scipoly/sap/meetings/2010/042610meeting.html SAP Minutes No. 201004. SAP. (2010b) A Set of Scientific Issues Being Considered by the Environmental Protection Agency Regarding: Re- Evaluation of Human Health Effects of Atrazine: Review of Non-cancer Effects and Drinking Water Monitoring Frequency, FIFRA Office of Science Coordination and Policy, Office of Prevention, Pesticides and Toxic Substances, U.S. Environmental Protection Agency. Available: http://www.epa.gov/scipoly/sap/meetings/2010/042610meeting.html SAP Minutes No. 201007. Sathiakumar, N., & Delzell, E. (1997). A review of epidemiologic studies of triazine herbicides and cancer. Critical Reviews in Toxicology, 27(6), 599-612. Sathiakumar, N., Delzell, E., & Cole, P. (1996). Mortality among workers at two triazine herbicide manufacturing plants. Am J Ind Med, 29(2), 143-151. doi: 10.1002/(SICI)10970274(199602)29:2<143::AID-AJIM4>3.0.CO;2-T [pii] 10.1002/(SICI)10970274(199602)29:2<143::AID-AJIM4>3.0.CO;2-T Sathiakumar, N., Delzell, E., & Cole, P. (1996). Mortality among workers at two triazine herbicide manufacturing plants. American Journal of Industrial Medicine, 29(2), 143-151. Sathyanarayana S, Basso O, Karr CJ, Lozano P, Alavanja M, Sandler DP, Hoppin JA. 2010. Maternal pesticide use and birth weight in the agricultural health study. J Agromedicine. 2010 Apr;15(2):127-36. Savitz, D., Arbuckle, T., Kaczor, D., & Curtis, K. (1997). Male pesticide exposure and pregnancy outcome. Am J Epidemiol, 146(12), 1025-1036. Page 174 of 184 Schultz TW. (2010). Adverse Outcome Pathways: A Way f Linking Chemical Structure to In Vivo Toxicological Hazards. In: Cronin MTD, Madden JC, eds. In silico Toxicology: Principles and Applications. Cambridge, UK: Royal Society of Chemistry. Schwab, C. L., Fan R., Zheng, Q., Myers, L.P., Hébert,P., & Pruet, S.B. (2005). Modeling and predicting stress-induced immunosuppression in mice using blood parameters. Toxicol Sci, Jan; 83(1): 101-13. Scientific Advisory Panel. (2010a) A Set of Scientific Issues Being Considered by the Environmental Protection Agency Regarding: Re- Evaluation of Human Health Effects of Atrazine: Review of Experimental Animal and In Vitro Studies and Drinking Water Monitoring Frequency, FIFRA Office of Science Coordination and Policy, Office of Prevention, Pesticides and Toxic Substances, U.S. Environmental Protection Agency. Available: http://www.epa.gov/scipoly/sap/meetings/2010/042610meeting.html SAP Minutes No. 201004. Seed, J., E.W. Carney, R.A. Corley et al . (2005). Overview: Using mode of action and lifestage information to evaluate the human relevance of animal toxicity data. Crit. Rev. Toxicol., OctNov; 35(8-9):664-72. Selye, H. (1939). Effect to adaptation to various damaging agents on the female sex organs in the rat. Endocrinology, 25(4): 615-24. Shibayama, H., Kotera, T., Shinoda, Y., Hanada, T., Kajhara, T., Ueda, M., Tamura, H., Ishibashi, S., Yamashita, Y., & Ochi, S. (2009). Collaborative work on evaluation of ovarian toxicity. 14) Two- or four-week repeated-dose studies and fertility study of atrazine in female rats. J toxicol Sci., 34: 147-55. Shyr, S.W., Crowley, W.R., and Grosvenor, C.E. (1986) Effect of Neonatal Prolactin Deficiency on Prepubertal tuberoinfundibular and tuberohypophyseal dopaminergic neuronal activity. Endocrinology 119:1217-1221 Siega-Riz AM, Herring AH, Olshan AF, Smith J, Moore C; National Birth Defects Prevention Study. (2009) The joint effects of maternal prepregnancy body mass index and age on the risk of gastroschisis. Paediatr Perinat Epidemiol., Jan;23(1):51-7. Review. PubMed PMID: 19228314. Simoneaux, B. J. (1985) Metabolism of 14-C Atrazine in Orally Dosed Rats (General Metabolism), Biochemistry Department, Agricultural Division CIBA-GEIGY Corporation Study No. ABR85104. Simpson, E.R., & Davis, S.R. (2001). Minireview: aromatase and the regulation of estrogen biosynthesis – some new perspectives. Endocrinology, Nov; 142(11): 4589-94. Page 175 of 184 Simpson, E.R., Clyne, C.D., Rubin, G., Boon, W.C., Robertson, K., Britt, K., Speed, C., & Jones, M.E. (2002). Aromatase: a brief review. Annu Rev Physiol, 64: 93-127. Singh, M., Kaur, P., Sandhir, R., & Kiran, R. (2008). Protective effects of vitamin E against atrazine-induced genotoxicity in rats. Mutat Res, Jul 31; 654(2): 145-9. Singh, M., Sandhir, R., & Kiran, R.(2008). Atrazine-induced alterations in rat erthrocyte membranes: ameliorating efect of vitamin E. J Biochem Mol Toxicol, Sept, 225: 363-9. Sisk CL, Richardson HN, Chappell PE, Levine JE. (2001). In Vivo Gonadotropin-Releasing Hormone Secretion in Female Rats during Peripubertal Development andon Proestrus. Endocrinology 142(7):2929-2936 Sliwiński, Z., Hermanowicz, A., Kossmann, S., & Hrycek, A. (1991). Neutrophil function in chemical plant workers employed in the production of dust pesticides. Pol J Occup Med Environ Health, 4(3), 241-247. Smits LJ, Jongbloet PH, Zielhuis GA. (1995). Preovulatory Overripeness of the Oocyte as a Cause of Ovarian Dysfunction in the Human Female. Med Hypotheses 45:441-448 Snyder, N. J., C.M. Harbourt, P. S. Miller, J.R. Trask, P. Hendley, and E. A. Johnston. 2007. Analysis of Chemograph Behavior Between Grab Samples - Measurement and Hybrid PRZM Approaches. Syngenta Task, T001508-03. Son, H.Y., Nishikawa, A., Okazaki, K., Lee, K., Imazawa, T., & Hirose, M. (2003). Lack of modifying effects of atrazine and/or tamoxifen on thyroid carcinogenesis in rat pretreated with N-bis(2hydroxypropyl)nitrosamine (DHPN). Food Chem Toxicol, Dec; 41(12): 1811-6. Sonich-Mullin, C., R. Fielder, J. Wiltse et al . (2001). IPCS conceptual framework for evaluating a mode of action for chemical carcinogenesis. Regul Toxicol Pharmacol, Oct; 34(2):146-52. Stoker T.E & Cooper R.L. (2007). Distribution of 14C-atrazine following an acute lactational exposure in the Wistar rat. Reprod Toxicol, Jun; 23(4):607-10. Stoker T.E., Robinette, C.L. & Cooper, R.L. (1999). Maternal exposure to atrazine during lactation suppresses suckling- induced prolactin release and results in prostatitis in the adult offspring. Toxicol Sci, 52: 68-79 Stoker, T., Zorilla, L., Strader, L. Et al. 2010. Internal Report:The Distribution of Atrazine in the Wistar Rat Following Gestational/Lactational Exposures. Memorandum dated July 14, 2010 from Tammy Stoker, NHEERL/ORD to Elizabeth Mendez, HED/OPP. Page 176 of 184 Stoker, T.E., Guidici, D.L., Laws, S.C. & Cooper, R.L. (2002). The Effects of Atrazine Metabolites on Puberty and Thyroid Function in the Male Wistar Rat: An Evaluation in the Male Pubertal Protocol. Toxicol. Sci., 67(2): 198-206. Stoker, T.E., Laws, S.C., Guidici, D., & Cooper, R.L. (2000). The effects of atrazine on puberty and thryroid function in the male wistar rat: An evaluation of a protocol for the assessment of pubertal development and thyroid function. Toxicol. Sci., 58: 50-9. Stoker, T.E., Parks, L.G., Gray, L.E., and Cooper, R.L. (2000) Effects of endocrine disrupting chemicals on puberty in the male rat: A review of the EDSTAC recommendations.Critical Reviews in Toxicology 30: 197-252, 2000. Stone, W.W., and Gilliom, R.J. (2009). Update of watershed regressions for pesticides (WARP) for predicting atrazine concentration in streams. U.S. Geological Survey Open-File Report 2009–1122. Available at http://pubs.usgs.gov/of/2009/1122/ Stone, W.W., Gilliom, R.J. & Crawford, C.G. (2008). Watershed Regressions for Pesticides (WARP) for Predicting Annual Maximum and Maximum Moving-Average Concentrations of Atrazine in Streams. U.S. Geological Survey Open-File Report 08-1186, 19 p. Available at http://pubs.usgs.gov/of/2008/1186/ Suárez, 2005. Pesticide Root Zone Model (PRZM). V3.12.2. May 12, 2005. Available through the USEPA Office of Pesticide Programs Water Models web site at http://www.epa.gov/oppefed1/models/water/. Sullivan, D.J., Vecchia, A.V., Lorenz, D.L., Gilliom, R.J., and Martin, J.D. (2009). Trends in Pesticide Concentrations in Corn-Belt Streams, 1996-2006: U.S. Geological Survey Scientific Investigations Report 2009-5132. Suzawa, M. & Ingraham, H.A. (2008). The herbicide atrazine activates endocrine gene networks via non-steroidal NR5A nuclear receptors in fish and mammalian cells. PloS ONE, 3(5): 111. Swan SH. Semen quality in fertile US men in relation to geographical area and pesticide exposure. Int J Androl. 2006 Feb;29(1):62-8; discussion 105-8. PubMed PMID: 16466525. Swan, S. H., Kruse, R. L., Liu, F., Barr, D. B., Drobnis, E. Z., Redmon, J. B., et al . (2003). Semen quality in relation to biomarkers of pesticide exposure. Environmental Health Perspectives, 111(12), 1478-1484. doi: 10.1289/ehp.6417. Tanaka, T., Kohno, H., Suzuki, R., & Sugie, S. (2004). Lack of modifying effects of an estrogenic compound atrazine on 7,12-dimethylbenz(a)anthracene-induced ovarian carcinogenesis in rats. Cancer Lett, Jul 16; 210(2): 129-37. Page 177 of 184 Terasawa El, Yeoman RR, Schultz NJ. (1984). Factors Influencing the Progesterone-Induced Luteinizing Hormone Surge in Rhesus Monkeys: Diurnal Influence and Time Interval After Estrogen Biol Reprod 31:732-741 Terasawa, E., and C.L. Nyberg. 1997. LHRH pulse generation in the monkey: in vivo and in vitro studies. In: Neural control of reproduction: physiology and behavior, K. Maeda, H. Tsukamura, and E. Yokoyama, eds., pp. 57-70. JSS Press, Tokyo. Thede. (1987) Study of 14C-atrazine dose/response relationship in the rat (general metabolism), MRID # 40431305. Thorpe, N., & Shirmohammadi, A. (2005). Herbicides and nitrates in groundwater of Maryland and childhood cancers: A geographic information systems approach. Journal of Environmental Science and Health Part C-Environmental Carcinogenesis & Ecotoxicology Reviews, 23(2), 261-278. doi: 10.1080/10590500500235001 Timchalk, C., Dryzga, M. D., Langvardt, P. W., Kastl, P. E., and Osborne, D. W. (1990) Determination of the effect of tridiphane on the pharmacokinetics of [14C]-atrazine following oral administration to male Fischer 344 rats, Toxicology 61, 27-40. Tinfo, N., & Laws, S.C. (2009). Understanding the Effects of Atrazine on Steroidogenesis in Wistar rats. Society of Toxicology, Itinerary Planner, Baltimore, MD. Trentacoste, S.V., Friedmann, A.S., Youker, R.T., Breckenridge, C.B. & Zirkin, B.R. (2001). Atrazine effects on testosterone levels and androgen-dependent reproductive organs in peripubertal male rats. J. Androl, Jan-Feb; 22(1): 142-8. Turgeon JL and Waring DW. (2000). Progesterone Regulation of the Progesterone Receptor in Rat Gonadotropes. Endocrinology 141:3422-3429. Ueda, M., Imai, T., Takizawa, T., Onodera, H., Mitsumori, K., Matsul, T., & Hirose, M. (2005). Possible enhancing effects of atrazine on growth of 7,12-dimethylbenze(a) anthraceneinduced mammary tumors in ovariectomized Sprague-Dawley rats. Cancer Sci., Jan, 96(1): 19-25. U.S. Environmental Protection Agency (USEPA). (2007). Preliminary Interpretation of the Ecological Significance of Atrazine Stream-Water Concentrations Using a StatisticallyDesigned Monitoring Program. Prepared for the December 4 – 7, 2007 FIFRA Scientific Advisory Panel. Document available at http://www.epa.gov/scipoly/sap/meetings/2007/120407_mtg.htm U.S. Environmental Protection Agency (USEPA). (2000). FIFRA Scientific Advisory Panel Briefing Document for a Consultation on: Monitoring Strategies for Pesticides in Surface Derived Page 178 of 184 Drinking Water. Prepared for June 6-9, 2000 FIFRA Scientific Advisory Panel. Document available at http://www.epa.gov/scipoly/sap/meetings/2000/june/drinkingwatersurvey.pdf U.S. Environmental Protection Agency (USEPA). (2007a). Acetochlor Registration Agreement and Monitoring Program. Documents are available through the Federal docket. Docket ID: EPA-HQ-OPP-2007-0725. Available at http://www.regulations.gov/search/Regs/home.html#docketDetail?R=EPA-HQ-OPP-20070725 U.S. Environmental Protection Agency (USEPA). (2007b). N-methyl Carbamate Revised Cumulative Risk Assessment. Available through the USEPA Office of Pesticide Programs web site at http://www.epa.gov/oppsrrd1/REDs/nmc_revised_cra.pdf U.S. Environmental Protection Agency (USEPA). (2007c). Preliminary Interpretation of the Ecological Significance of Atrazine Stream-Water Concentrations Using a StatisticallyDesigned Monitoring Program. Prepared for the December 4 - 7, 2007 FIFRA Scientific Advisory Panel. Document available at http://www.epa.gov/scipoly/sap/meetings/2007/120407_mtg.htm U.S. Environmental Protection Agency (USEPA). (2009a). The Ecological Significance of Atrazine Effects on Primary Producers in Surface Water Streams in the Corn and Sorghum Growing Region of the United States (Part II). Prepared for the May 12 - 15, 2009 FIFRA Scientific Advisory Panel. Document available at http://www.epa.gov/scipoly/sap/meetings/2009/051209meeting.html U.S. Environmental Protection Agency (USEPA). (2009a). The Ecological Significance of Atrazine Effects on Primary Producers in Surface Water Streams in the Corn and Sorghum Growing Region of the United States (Part II). Prepared for the May 12 – 15, 2009 FIFRA Scientific Advisory Panel. Document available at http://www.epa.gov/scipoly/sap/meetings/2009/051209meeting.html U.S. Environmental Protection Agency (USEPA). (2009b). The Ecological Significance of Atrazine Effects on Primary Producers in Surface Water Streams in the Corn and Sorghum Growing Region of the United States (Part II). Addendum to Section VI: Additional Analysis of the 40 AEEMP Sampling Sites to Identify Additional Areas That May Exceed the LOC. Prepared for the May 12 - 15, 2009 FIFRA Scientific Advisory Panel. Document available at http://www.epa.gov/scipoly/sap/meetings/2009/051209meeting.html U.S. Environmental Protection Agency (USEPA). (2009b). The Ecological Significance of Atrazine Effects on Primary Producers in Surface Water Streams in the Corn and Sorghum Growing Region of the United States (Part II). Addendum to Section VI: Additional Analysis of the 40 AEEMP Sampling Sites to Identify Additional Areas That May Exceed the LOC. Prepared for Page 179 of 184 the May 12 – 15, 2009 FIFRA Scientific Advisory Panel. Document available at http://www.epa.gov/scipoly/sap/meetings/2009/051209meeting.html U.S. Environmental Protection Agency (USEPA). (2010a). Re- Evaluation of Human Health Effects of Atrazine: Review of Experimental Animal and In Vitro Studies and Drinking Water Monitoring Frequency. Prepared for the April 26-29, 2010 FIFRA Scientific Advisory Panel. Document available at http://www.epa.gov/scipoly/sap/meetings/2010/042610meeting.html U.S. Environmental Protection Agency (USEPA). (2010b). Re-evaluation of the Human Health Effects of Atrazine: Review of Non-Cancer Effects and Drinking Water Monitoring Frequency. Prepared for the September 14 – 17, 2010 FIFRA Scientific Advisory Panel. Document available at http://www.epa.gov/scipoly/sap/meetings/2010/091410meeting.html U.S. Environmental Protection Agency (USEPA). 2000. Atrazine: Hazard and Dose-Response Assessment and Characterization. U.S. Environmental Protection Agency (USEPA). 2000. Progress Report on Estimating Pesticide Concentrations in Drinking Water and Assessing Water Treatment Effects on Pesticide Removal and Transformation: A Consultation. Prepared for Session VI of the September 2629, 2000 FIFRA Scientific Advisory Panel. U.S. Environmental Protection Agency (USEPA). 2001. Water Treatment Effects on Pesticide Removal and Transformation. Federal Register Docket Number: OPP-00731. Available at http://www.epa.gov/pesticides/trac/science/water_treatment.pdf U.S. Environmental Protection Agency (USEPA). 2007. Preliminary Interpretation of the Ecological Significance of Atrazine Stream-Water Concentrations Using a Statistically-Designed Monitoring Program. Prepared for the December 4 - 7, 2007 FIFRA Scientific Advisory Panel. Document available at http://www.epa.gov/scipoly/sap/meetings/2007/120407_mtg.htm U.S. Environmental Protection Agency (USEPA). 2009. The Ecological Significance of Atrazine Effects on Primary Producers in Surface Water Streams in the Corn and Sorghum Growing Region of the United States (Part II). Prepared for the May 12 - 15, 2009 FIFRA Scientific Advisory Panel. Document available at http://www.epa.gov/scipoly/sap/meetings/2009/051209meeting.html U.S. Environmental Protection Agency (USEPA). 2010. Re- Evaluation of Human Health Effects of Atrazine: Review of Experimental Animal and In Vitro Studies and Drinking Water Monitoring Frequency. Prepared for the April 26-29, 2010 FIFRA Scientific Advisory Panel. Document available at http://www.epa.gov/scipoly/sap/meetings/2010/042610meeting.html Page 180 of 184 U.S. Environmental Protection Agency. (2010). DRAFT Framework for Incorporating Human Epidemiologic & Incident Data in Health Risk Assessment (Draft dated January 7, 2010). Presented to the FIFRA Scientific Advisory Panel, February 2-4, 2010. http://www.regulations.gov (Docket EPA-HQ-OPP-2009-0851). U.S. EPA, U. (2010). FIFRA SAP Final Report (February 2-4, 2010). U.S. EPA, U. S. (2003a). Atrazine Interim Reregistration Eligibility Document (IRED). Washington D.C. U.S. EPA, U. S. (2003b). FIFRA Scientific Advisory Panel Final Meeting Report (July 17, 2003 Meeting). Retrieved from http://www.epa.gov/scipoly/sap/meetings/2003/071703_mtg.htm#materials. U.S. EPA. (2010). FIFRA SAP Final Report "Re- Evaluation of Human Health Effects of Atrazine: Review of Experimental Animal and In Vitro Studies and Drinking Water Monitoring Frequency." (April 26-29, 2010) US EPA. (1999). Guidelines for carcinogen risk assessment. Risk Assessment Forum. SAB review draft. Washington, DC: U.S. Environmental Protection Agency. www.epa.gov/ncea/raf/crasab.htm.* US EPA. (2003). Interim Re-registration Eligibility Decision (IRED) US EPA. (2003). Interim Re-registration Eligibility Decision (IRED) US EPA. (2005). Guidelines for Carcinogen Risk Assessment. U.S. Environmental Protection Agency, Risk Assessment Forum, Washington, DC. EPA/630/P-03/001F. Federal Register 70(66):17765-17817. Available at http://www.epa.gov/raf US EPA. (2006). Decision Documents for Atrazine. Atrazine IREDs (January and October 31, 2003) and Atrazine RED (April 6, 2006). EPA-HQ-OPP-2003-0367. USEPA (2006). Harmonization in interspecies extrapolation: use of body weight3/4 as the default method in derivation of the oral reference dose. External Review Draft. Office of the Science Advisor. USEPA. (2000). “Benchmark Dose Technical Guidance Document.” Draft report. Risk Assessment Forum, Office of Research and Development, U.S. Environmental Protection Agency. Washington, DC. EPA/630/R-00/001. USEPA. (2006) Approaches for the Application of Physiologically Based Pharmacokinetic (PBPK) Models and Supporting Data in Risk Assessment, National Center for Environmental Assessment, Office of Research and Development. Page 181 of 184 Van Leeuwen, J., Waltner-Toews, D., Abernathy, T., Smit, B., & Shoukri, M. (1999). Associations between stomach cancer incidence and drinking water contamination with atrazine and nitrate in Ontario (Canada) agroecosystems, 1987-1991. Int J Epidemiol, 28(5), 836-840. Vecchia A.V., Crawford, C.G. 2006. Simulation of Daily Pesticide Concentrations from Watershed Characteristics and Monthly Climatic Data: U.S. Geological Survey Scientific Investigations Report 2006-5181, 60 p. Vecchia, A.V. and Crawford, C.G. (2006). Simulation of daily pesticide concentrations from watershed characteristics and monthly climatic data: U.S. Geological Survey Scientific Investigations Report 2006-5181. Vecchia, A.V., Gilliom, R.J., Sullivan, D.J., Lorenz, D.L., Martin, J.D. 2009. Trends in Concentrations and Use of Agricultural Herbicides for Corn Belt Rivers, 1996-2006. Environ. Sci. Technol. 43, 9096-9102. Vecchia, A.V., Martin, J.D., and Gilliom, R.J. (2008). Modeling Variability and Trends in Pesticide Concentrations in Streams. Journal of the American Water Resources Association (JAWRA). 44(5), 1308-1324. Vecchia, A.V., Martin, J.D., Gilliom, R.J. 2008. Modeling Variability and Trends in Pesticide Concentrations in Streams. Journal of the American Water Resources Association (JAWRA) 44(5): 1308-1324. Veldhuis JD, Keenan DM, Pincus SM. (2008). Motivations and Methods for Analyzing Pulsatile Hormone Secretion. Endocr. Rev. 29:823-864 Victor-Costa, A.B., Carozzi Bandeira, S.M., Oliveira, A.G., Bohorquiz Mahecha, G.A. & Oliveira, C.A. (2010). Changes in testicular morphology and steroiodogenesis in adult rats exposed to atrazine. Reprod Toxicol, Jan 4. [Epub ahead of print] Doi:10.1016/j.reprotox.2009.12.006 Villanueva, C., Durand, G., Coutté, M., Chevrier, C., & Cordier, S. (2005). Atrazine in municipal drinking water and risk of low birth weight, preterm delivery, and small-for-gestational-age status. Occup Environ Med, 62(6), 400-405. doi: 62/6/400 [pii] 10.1136/oem.2004.016469. Waggoner, J. K., Kullman, G. J., Henneberger, P. K., Umbach, D. M., Blair, A., Alavanja, M. C., . . . Hoppin, J. A. (2011). Mortality in the agricultural health study, 1993-2007. Am J Epidemiol, 173(1), 71-83. doi: kwq323 [pii] 10.1093/aje/kwq323 Walford TL (1871). Retrospective Address of the Pathological Society of Reading. Br Med J 2:634 Page 182 of 184 Waller, S. A., Paul, K., Peterson, S. E., & Hitti, J. E. (2010). Agricultural-related chemical exposures, season of conception, and risk of gastroschisis in Washington State. American Journal of Obstetrics and Gynecology, 202(3). doi: 10.1016/j.ajog.2010.01.023. Watkins ML, Edmonds L, McClearn A, Mullins L, Mulinare J, Khoury M. (1996) The surveillance of birth defects: the usefulness of the revised US standard birth certificate. Am J Public Health., May;86(5):731-4. PubMed PMID: 8629729; PubMed Central PMCID: PMC1380486. Weisenburger, D. (1990). Environmental epidemiology of non-Hodgkin's lymphoma in eastern Nebraska. Am J Ind Med, 18(3), 303-305. Weselak M, Arbuckle TE, Walker MC, Krewski D. The influence of the environment and other exogenous agents on spontaneous abortion risk. (2008) J Toxicol Environ Health B Crit Rev., Mar;11(3-4):221-41. Review. PubMed PMID: 18368554 Wetzel L.T., Luempert L.G., III Breckenridge C.B., Tisdel M.O., Stevens J.T. (1994). Chronic effects of atrazine on estrus and mammary tumor formation in female Sprague-Dawley and Fisher 344 rats. J Toxicol. Environ. Health, Oct; 43(2):169-82. Whalen M.M., Loganathan B.G., Yamashita, N., Saito, T. (2003). Immunomodulation of human natural killer cell cytotoxic function by triazine and carbamate pesticides. ChemicoBiological Interactions, Jun 15; 145(3): 311-19. Whitmore, R.W., and Mosquin, P.L. (2010). Investigation of Statistical Confidence in Upper Quantiles and “Peak” Concentrations of Currently Available Atrazine Drinking Water Monitoring Data. RTI International report Submitted to the U.S. Environmental Protection Agency by Syngenta Crop Protection, Inc. MRID 47972804 Williams LJ,Kucik JE,Alverson CH, Olney RS, Correa A. (2005) Epidemiology of gastroschisis in metropolitan Atlanta, 1968 through 2000. Birth Defects Res Part A Clin Mol Teratol, 73:17783. Williams R., and Merritt A. (2011). Summary of Site Selection and Monitoring Plan for Atrazine Daily Monitoring in Selected Community Water Systems. Unpublished study from Syngenta Crop Protection, Inc. MRID 48470009 Winchester, P., Huskins, J., & Ying, J. (2009). Agrichemicals in surface water and birth defects in the United States. Acta Paediatr, 98(4), 664-669. doi: APA1207 [pii] 10.1111/j.16512227.2008.01207.x. Wingate MB, Arrata WSM, Iffy L. (1973). Delayed Ovulation Ill Med J 144:44-47 Witschi E. (1952). Overripeness of the Egg as a Cause of Twinning and Teratogenesis. Cancer Res 12:763-797 Page 183 of 184 Xiaoying Hui, R. C. W. a. H. I. M. (2011) Pharmacokinetics of [14C]-atrazine in rhesus monkeys, single-dose intravenous and oral administration, Toxicological and Enviromental Chemistry 93, 370-382. Young, H. A., Mills, P. K., Riordan, D. G., & Cress, R. D. (2005). Triazine herbicides and epithelial ovarian cancer risk in central California. Journal of Occupational and Environmental Medicine, 47(11), 1148-1156. doi: 10.1097/01.jom.0000177044.43959.e8 Zahm, S. H., Weisenburger, D. D., Saal, R. C., Vaught, J. B., Babbitt, P. A., & Blair, A. (1993). The role of agricultural pesticide use in the development of non-Hodgkin's lymphoma in women. Arch Environ Health, 48(5), 353-358. Zeljeciz D, & Garaj-Vrhovac, V. (2004a). Atrazine genotoxicity evaluation in different mouse organs by comet assay. Periodicum Biologorum, 106(2): 155-9. Zeljezic, D. & Garaj-Vrhovac, V. (2004b). Genotoxicity evaluation of pesticide formulations containing alachlor and atrazine in multiple mouse tissues (blood, kidney, liver, bone marrow, spleen) by comet assay. Neoplasma, 51(3): 198-203. Zeljezic, D., Garaj-Vrhovac, V., & Perkovic, P. (2006). Evaluation of DNA damage induced by atrazine and atrazine-based herbicide in human lymphocytes in vitro using a comet and DNA diffusion assay. Toxicol In vitro, Sept; 20(6): 923-35. Zorrilla, L.M., Gibson, E.K., Stoker (2009) Effect of Chlorotriazine Pesticides on Gonadotrophin Releasing Hormone in the Neuronal GT1-7 Cell Line and Hypothalamic Explants. Annual Meeting of the Society for Study of Reproduction, Pittsburgh, PA. Biol Reprod, 81: 638. Page 184 of 184