Transcript
P H O T O D I S C I N C . hy are out-of-specification investigations still a lead-ing cause ofwarning letters issued by FDA (1)?Many workshops,seminars,consultants,and guid-ances are available,yet difficulties and inconsistentapproaches to out-of-specification investigations persist.Numerousworking practices and various degrees ofunder-standing ofcurrent expectations in this area likely have led tomany different approaches to investigating out-of-specificationresults.For this reason,the Analytical Research and Develop-ment Steering Committee ofthe Pharmaceutical Research andManufacturers ofAmerica (PhRMA) included the topic ofcon-ductingout-of-specification investigations in its annual work-shop held in September 2000.Representatives from PhRMAmember companies met to consider this topic,share currentpractices,and agree on an acceptable analytical practice (AAP)that represents good science and conforms to current regula-tory guidelines and expectations.The output ofthat workshopis presented in this article. Background The guiding principlesfor out-of-specification investigationsare based on a legal ruling by Judge Wolin in 1993 (2) and thedraft FDA guidance that followed in September 1998 (3).JudgeWolin presided over a case brought by FDA enforcement actionagainst a generic-drug manufacturer.Among the issues thatFDA cited were averaging out-of-specification values with in-specification values to get a passing result;maintaining an inef-fective programfor process validation;discarding raw data;con-ducting multiple retests with no defined end point;performinginadequate failure investigations;and lacking method valida-tion.Judge Wolin’s responsibilitywas to listen to experts fromboth sides and craft an opinion that was based on the testimony and the law.In addition,he had to interpret the current regula-tions about good manufacturingpractices.The trial itselfgenerated 2300 pages oftestimony and 400exhibits and declarations.Judge Wolin acknowledged that thepharmaceutical industry is a business with legitimate businessmotivations.He also acknowledged the very important task thatFDA has as regulator ofthis industry.The judge found that thecompany was mired in uncertainty and conflicting opinionsabout good manufacturing practices and presented a trouble-some attitude to the court.His final decision reflects good common sense.The judgestated that the history ofthe product must be considered whenevaluating the analytical result and making a determination 40 Pharmaceutical Technology JANUARY 2002 www.pharmtech.com Investigation of Out-of-Specification Results Alex M.Hoinowski,Sol Motola,Richard J.Davis,and James V.McArdle* Alex M. Hoinowski, MS, is a senior consultant for TheQuantic Group, Ltd., Livingston, NJ. Sol Motola, PhD, is the assistant vice-president of technical and regulatoryaffairs for Wyeth Ayerst Global Pharmaceuticals, St. Davids,PA. Richard J. Davis most recently was the senior vice-president for quality assurance and regulatory compliancefor DuPont Pharmaceuticals Company, Wilmington, DE. James V. McArdle, PhD, is the vice-president of ana-lytical development and quality for Isis Pharmaceuticals, Inc.,2292 Faraday Ave., Carlsbad, CA 92008, tel. 760.931.9200,fax 760.603.4655,
[email protected],www.isip.com.*To whom all correspondence should be addressed. W Despite a multitude of educationalresources,The AnalyticalResearch and Development SteeringCommittee of the Pharmaceutical Researchand Manufacturers of America (PhRMA)included the topic of conducting out-of-specification investigations in its annualworkshop held in September 2000.Theauthors summarize the results of theworkshop in this article.Despite a multitude of educationalresources,difficulties and inconsistentapproaches to out-of-specificationinvestigations persist.The AnalyticalResearch and Development SteeringCommittee of the Pharmaceutical Researchand Manufacturers of America (PhRMA)included the topic of conducting out-of-specification investigations in its annualworkshop held in September 2000.Theauthors summarize the results of theworkshop in this article. 42 Pharmaceutical Technology JANUARY 2002 www.pharmtech.com about the release ofthe product.Ofcourse,analytical resultsthat are wrong may arise.The exact cause ofthe error often isdifficult to determine,and expecting that the cause oferrorsalways will be determined is unrealistic.The inability to iden-tifythe cause ofthe error will affect retesting procedures,butit will not affect the inquiry required ofthe initial out-of-specification result.Judge Wolin placed restrictions on the application ofoutliertests.A firm may not frequently reject results on the basis of outlier tests,and outlier tests may not be applied to results of chemical testing.In fact,Judge Wolin cited USP as governingthe use ofoutlier tests.The draft guidance does allow the useofoutlier tests for providing perspective concerning the proba-bilityofobtaining the suspect result as part ofthe overall evalu-ationofthe batch.Judge Wolin stated that the firm should have a retest policy and protocol and that the protocol must define the point atwhich testing stops and evaluation occurs.He did allow retests,but only when the laboratory investigations were completedand an analyst error documented.Retests also could be justi-fied when a formal process investigation detected no processor nonprocess-related errors and a review ofthe laboratory testwas inconclusive.He also stated that the firm could not conducttwo retests and base a release on the average ofthree results,nor could the firm simply assume a sampling or preparationerror as a justification for retesting and resampling.Judge Wolindirected that retests were to be done on the same sample,not adifferent sample,but retests might be conducted on a secondaliquot from the same portion ofsample or on a portion ofthesame larger sample previously collected for laboratory tests. Invalid testing caused by known laboratory errors An obvious error that invalidates the results may occur duringtesting.Such errors are recognized easily without a formalinvestigation.A known error caused by the analyst or by instru-ment failure does not require the performance ofa formal labo-ratoryinvestigation,whether or not data have been generated.Testing should be stopped,the supervisor notified,and the datainvalidated.Examples ofthese types oferrors include tran-scription errors,miscalculations,incomplete transfer ofmate-rial,and incorrect settings ofinstrument parameters.Theanalyst should document the error on the raw data record,andthe supervisor should initial and date the record.In the case of clerical errors such as transcription errors or miscalculations,the analyst should correct the error and continue with reviewofthe data.In the case oflaboratory errors such as an incorrectdilution or a detector set on the wrong wavelength,the analystshould invalidate the result.Testing may continue with no fur-ther consideration ofthe invalid result.Ifa laboratory investigation is initiated before the discovery of the known error,the investigation should be completed as a meansofdocumenting and addressing the error.The draft guidance fur-ther discusses the responsibilities ofthe analyst in this situation. The formal laboratory investigation The primary objective ofthe investigation ofan out-of-specification result is to determine either that an assignablecause exists or that an assignable cause cannot be identified.Anassignable cause is a documented and scientifically justifieddetermination that the discrepant result can be traced to labora-toryerror.Note that atypical results (those that are unusual onthebasis ofexperience,trending,or data review but still are withinspecification) also warrant an investigation and may require theimplementation ofsuitable corrective actions.The rest ofthis dis-cussionwill focus on out-of-specification results,but many oftheprinciples and recommended courses ofaction also will apply to the investigation ofatypical results.An out-of-specificationresult is one that falls outsideofthe test’s acceptance criteria in,for example,filed applications,approved marketing submissions,or official compendia. Phase 1 of the investigation. As emphasized in the draft guid-ance,the analyst detecting an out-of-specification result mustreport that result promptly to his or her supervisor.The analystand supervisor must then immediately conduct an investiga-tion ofthe result.Ifno obvious error is revealed,an out-of-specificationresult must be reported to the quality assuranceunit within a specified period oftime,usually one or two busi-nessdays.This report to the quality assurance unit constitutesthe beginning ofthe formal investigation for most companies.However,some companies regard the notification ofthe super-visorby the analyst as the start ofthe formal investigation.Thefirm’s standard operating procedure (SOP) should be unam-biguousabout the point at which the formal investigation begins.The analyst and the supervisor conduct phase 1 ofthe labo-ratory investigation to determine whether or not the out-of-specification result is assignable to the testing laboratory.Thispart ofthe investigation should be completed in a short andpredefined period oftime,usually one or two business days,and documented in a report ofthe investigation.Ifthis part of the laboratory investigation cannot be completed within thepredefined period oftime,then an interim report should beissued within the time limit.A checklist ofpossible sources oferror may aid this part of the investigation.Checklist items to use in reviewing the ana-lyst’s notebook include ● Was the correct amount ofsample taken? ● Were the sample dilutions correctly performed? ● Were other reagents and solutions correctly prepared andused? ● Are the calculations correct? ● Were the correct volumetric flasks and pipettes used?Similar checklists can be developed to identify sources of error in sample integrity and handling,instrument performance,method validation,analyst’s technique,the test procedure,stan-dards,and instrument output.At this early stage,conducting limited testing ofthe samplesthat led to the out-of-specification result may be helpful.Thedraft guidance states that such immediate assessments “allowmore credibility to be given to laboratory error theories.”Thislimited testing usually will not suffice to determine the accept-ability ofa batch,but it should be captured in the report aboutthe laboratory investigation.Ifthe cause ofthe out-of-specification result can be assignedto the testing laboratory,then the srcinal sample is tested again.
