Scribd
Upload
48 views5 pages

Preventing Practice of Testing Into Compliance

Uploaded by

Nilkanth Chapole
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF or read online on Scribd
48 views5 pages

Preventing Practice of Testing Into Compliance

Uploaded by

Nilkanth Chapole
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF or read online on Scribd
You are on page 1/ 5
Preventing the Practice of Testing into Compliance A Practical Solutions Guide Lynn D.Torbeck The author presents several typical scenarios of how companies test into compliance and provides solutions for preventing such practi . Lynn D. Torbeck isa statistical consultant at Torbeck land Associates, Inc., 2000 Dempster Plaza, Evanston, tk 60202, tel, 847.424.1314, woru.torbeck org. Mr. Torbeck | a member of Pharmaceutical Technology's Eatorial Advisory Boar. 138 Plamaceta Tenaloy ocroxen 2002 «sting into compliance isthe prac- tice of ignoring valid information that should be used to make deci- sions. Inits Draft Guidance on Out- o-Specfications (OOS), the Center for Drug Evaluation and Research at FDA states that “some firms have used a strat- egy of repeated testing until a passing re sult is obtained (testing into compliance), then disregarding the OOS results with- out scientific justification. Testing into compliance is objectionable under the CGMP" (1) Cleary, sucha practice is at best not scientific and at worst i fraudu- lent, illegal, and immoral. An under- standing of the several ways to test into compliance will help identify and prevent this malpractice, help auditors uncover the practice, and institute corrective ac- tion and preventive action (CAPA). This article describes 16 typical examples of testing into compliance and 5 examples of acceptable practices Situation 1:“What do you mean it failed? Send them another sample!” ‘This comment was overheard by the au- thor in 1976. The specifications were 95:0-105.0%, and the first sample result ‘was 94.9%, The second sample result was 95.296. The manager concluded, “See, 1 told you they tested it wrong, Ship it” This is exactly the undesirable situation referred to in the OOS guidance docu- ‘ment. Cleary the data are suspect. Do not release the lot on the basis of the second, value being within specification while ig- noring the first value. Solution. An investigation is required as per the OOS guidance. Consider all indi- vidual reportable values inthe final deci sion (2). The sidebar, “Reportable values” provides a definition of a reportable value Situation 2: “On the average, itisin.” (version 1) ‘A variation ofthe previously described ex- ample is to average additional reportable vvalucs to get a result that is within the specification. This action is clearly wrong. Solution. Clearly define and document the reportable value for the test method Do not average additional reportable val- ues but provide al individual reportable ‘values to quality assurance (QA) person nel for their decison. If QA has disposi- tioned the lot anda single values needed for the certificate of analysis, then these values can be averaged. Situation 3:“On the average, itisin.’ (version 2) Physically averaging samples by blending them has the same effect as averaging data from different samples. Blending OOS material with in-specification material constitute testing into compliance. Note, this approach isnot the same as in-process adjustments. Solution. Do not blend unless blending is part ofthe original validated process design Situation 4: Changing the reportable value Definition ofthe reportable value should not change from the test procedure to the validation protocol tothe final report. Re- defining the reportable value depending co the data obtained to avoid an OOS value, for example in an analysis, const ‘utes testing into compliance. Solution. The definition should not change without an investigation, a report, and perhaps even some revalidation ofthe test method. Consider conducting an oer- allinvestigation when more than 3% ofthe wow. pharmtech.com reportable values are OOS (sce Reference 2 fora discussion of reportable values). Situation 5: “The analyst didit.” Even witha fall investigation, some com- panies routinely declare thatthe OOS val- ues area result ofan analyst error with- cout documentable evidence that an error hhad actually occurred. Unidentifiablean- alyst errors are automatically suspect. Solution. Document in detail ll analyst ‘errors. The root cause of a true analyst ‘error is lack of training, Situation 6: The practice of completing “napkin data forms” In an effort to help their company meet its goals some analysts have practiced so-called napkin analysis. The analyst records the data and calculations on a scrap of paper mens M aR _ Accurate, Precise Measurements and Certifications with a Fast Turnaround ‘Afante Scale can bth supply now precision weighs and weight ses and calbrate your existing preckion standards, Weight standards and new weighs wl rooive the uimost in accuracy and NST traceabity, ng wih ‘qucktunarounds and lead tines. Catration an wraoeablecerticates both Sate asfound and as-eft weight data, uncetantes, and ater extensive testng data depending on your quality syst and testing requerents. Cee oe Crate Cen RRO mms 973-661-7090 SEOUL OUI Raa PA nee) Dna CCRC Ay ‘or napkin to sce what the results will be. If itisin specification, the analyst records the data and calculations in the laboratory notebook. fit is OOS, the analyst tosses the paper out, does not make a record, and does not inform the supervisor. Solution, Record all data and calculations inthe laboratory notebook as they ae col- Jected and made. Thisis good documenta tion. A search ofthe dustbins may uncover this practice if the analyst is careless. In ‘variation ofthis scenario, the analyst works outside of normal hours. Preventing this practice requires management oversight. Situation 7: “When in doubt, tossit out.” This situation may constitute a testing into ‘compliance tactic. Do not use an outlier rejection test for chemical methods, for content uniformity, or for homogeneity. {An outlier test can be used to reject val- tues collected for biologicals, as stated in USP (111), reference Judge Wolin in the PDA v Barr Laboratories decision. Solution, Use an outlier test as one part of the overall investigation into an OOS value. Do not reject data using only the results of an outlier test. See the OOS guid- ance for further discussion. Another gray area is what constitutes a biological asay. In such cases it may be possible to define some test methods as biological methods ‘when they should be designated as chemi- ‘cal methods. Situation 8:“It's out of control.” A variation of the outlier approach is to usea control chart or some other statisti- cal technique to reject values without an Solution. investigate and document all, suspect data. Situation “Inspect it until it passes.” Some in-process inspection schemes cor- reclly use series of 100% inspections and sampling plans to inspect for noncomply- ing materials. Visual inspection of par- enteral is one example. the decision trees aresuch that amatetial or product lot can- not be rejected, this setup can constitute the practice of testing into compliance Solution. lot must have the oppor- tunity to be rejected and fail. These sit- uations can be complicated and may warw.pharmtech.com need a detailed decision flow chart to help in finding the problem. The com- pany’s QA or quality engineering de- ppartment should develop a flow chart of the standard operating procedure to re veal this situation Situation 10:“Let Joan do it over.” Having the retest conducted by a second analyst without first finding an explana- tion about why te first analyst obtained an 008 value could be suspicious to a cynical investigator. The second analyst aay be better of worse than the first an- alyst and may obtain an in-specifiation value by“bending”the method or just by simple incompetence. Solution. Investigate and document the training of all analysts. Fiction: Al depth fiters are cected equal Facts with procedures set forth in our Drug Master Feat the US FDA. Fiction: Facts Technical service isan extremely important function t ErelAlsop. No matter where you are, you can always reach someone who can provide assistance, including the CEO. Fletion: Validation of depth fiters is unnecessary. Factz We believe that all pharmaceutical processing ters should meet predetermined specifications and quality artbutes. The ErtelAlsop Validation Guide and Certificate of Validation Conformance give you this assurance. Fiction: Depth fiter removal ratings are meaningless. Fact: locto-ot consistancy. Fiction: There's nothing new in depth fiters. Fact: Phone: 845-331-4552 > Fax: ErtelAlsop. The ACR CUR in depth filter evaluation! EreiAlsop pharmaceutieal-grade depth ftars are made in accordance Depth fter manufacturers don't offer technical service | All ErtelAlsop ratings are QA tested to specific published criteria for (Our product development is far ‘com static. Innovations include sanitary design place and frame filter presses, sanitary design lenticular format fiter cartridge housings, plus 2 wide range of validated pharmaceutieal grade fter pads Box 3950 Kingston, NY 124 39-1083 + worertlalsop.com *[email protected] 142, Phamscatilelalegy ocro9en 202 Situation 11:”It must be correct. Is within specifications.” ‘Another example from the author's expe rience involved sampling tablets. As the tablets were made, they were loaded into five-gallon plastic lined buckets. When a bce was fl, «sample was taken off the top and tested in the laboratory. Ifthe sample passed, the bucket was released. If the sample failed, the bucket was rejected. About 50% ofthe buckets failed the test. Solution, An investigation found thatall buckets were the same, There were no ‘good buckets or bad buckets. The prod: Uct had very poor content uniformity. Any sample testing was the same as tossing a coin as to whether a bucket was accepted ot rejected, Being in specification is not evidence that the values are acceptable. Situation 12:*Roll them bones.” Although gambling will seldom pay off in the casino, it can pay dividends for an un- cthical supple. Many companies use sam- pling plans to check incoming materials. "These plans are designed to accept lots that _meet the specifications a great percentage of the time. However, there is always a small risk that a good lot will be rejected. Likewise, there is low probability that a bad lot will be accepted. Some vendors play the odds on sampling plans by sim- ply reshipping and resubmitting the re- jected lot back again, without any correc- tiveaction, in hopes that it will be accepted con the second try Solution. To prevent ths situation, doc uument rejected lots so that they must be 100% inspected and/or reworked before they can be resubmitted Situation 13: “Put it in the midd In this variation of fraud, a supplier will place the best product on the top of the pallet and the worstin the center in hopes that the samples will be taken off the top of the pallet. An in-process variation of this tactics done by pulling samples from physical locations known to be good and avoiding locations that are known to pro- vvide poor results such asin blending and cleaning, This practice clearly constitutes testing into compliance. Solution. Pull truly representative sarm- ples from the lot or throughout a pro- duction run, ‘www.pharmtaen.com Situation 14: “Where did they go?” Inthe case of one laboratory, samples just conveniently disappeared. Seven days into an eight-day test, the incubation plates simply disappeared overnight without ex- planation or investigation into what had happened. Solution, Management is responsible for preventing this fraudulent practice. What's New i Situation 15: “tma little normal deviate.’ Data are not always normally distributed Nonstatisticians, and even statisticians, too often assume that data are normally distributed without frst graphing or test ing that hypothesis, Mos failure data are right skewed, as are particle sizes. DNA assay data, for example, generate a text- book log-normal distribution. Environ- in Mass Flow? www.sierrasmarttrak.com 144 Panacea Retalogy ocro0ce 200 ts Cau Bi, Maney, CAS30 Te )665.0200 Fa 1) 373-402 Er iamduoransvmertr (nea: wora.serasmartrak. com mental data typically are dramatically skewed, with zeros accounting for the ma- jority of values. Selecting an incorrect dis- tribution solely to avoid OOS results con- stitutes testing into compliance. Solution. Determine the shape of the distribution of the data before a sta cal analysis is begun, Situation 16: “ve been transformed.” As part ofa statistical analysis, various ‘mathematical transforms may be use, in- cluding taking the log or square root of the original data. This practice is usually conducted to achieve normality or equal ity of variances. Transformations are sc entfially nd regulatory defensible sta- tistical tools. However, changing the transformation to avoid OOS results eon- stitates testing into compliance Solution, Once chosen for a given situ- ation, a transformation cannot be changed ‘without an investigation, justification, and documentation by QA personnel. Acceptable practices Examples that do not constitute testing {nto compliance include the following: Increase the sample ize. Its acceptable to take a large numberof initial samples. Sample size alone is not evidence of test- {ng into compliance. Of course, all sam- ples must be tested, recorded, kept, and reported to QA personnel. Do no ‘or hide any data. Itis only common sense that a larger sample size will give better estimates and better understanding. More knowledge is better. Increase the numberof retest. It is ac- ‘ceptable to perform a large number of ad-

You might also like