Proficiency Tests John H McB Miller Laboratory Department (DLab) European Department for the Quality of Medicines Council of Europe Strasbourg, France 1| QCL Training Seminar, Tanzania | 5-7 Dec 07 Interlaboratory Studies Interlaboratory Testing Description • Proficiency testing • Continuing assessment of technical competence • Collaborative study • Validation of a specific method • Certification study • Establishing the best estimate of the time value of an analyte in a reference material • Co-operative study • Laboratory assessment of samples and methods (eg educational studies) 2| QCL Training Seminar, Tanzania | 5-7 Dec 07 Proficiency testing by interlaboratory comparisons To determine the competence of individual laboratories to perform specific tests or measurements To monitor the performance of laboratories overtime 3| QCL Training Seminar, Tanzania | 5-7 Dec 07 ISO Guide 43-1 Interlaboratory comparisons may be used to: a) determine the performance of individual labs for specific tests or measurements and to monitor labs’ continuing performance; b) identify problems in labs & initiate remedial actions which may be related to, for example, individual staff performance or calibration of instrumentation; 4| QCL Training Seminar, Tanzania | 5-7 Dec 07 ISO Guide 43-1 c) establish the effectiveness & compatibility of new test or measurement methods & similarly to monitor established methods d) provide additional confidence to lab clients; e) identify interlaboratory differences; 5| QCL Training Seminar, Tanzania | 5-7 Dec 07 ISO Guide 43-1 Interlaboratory comparisons are conducted for a number of purposes and may be used by participating laboratories and other parties eg establishment of a reference material 6| QCL Training Seminar, Tanzania | 5-7 Dec 07 ISO Guide 43-2 6. Use of results by laboratory accreditation bodies 6.1 The results from proficiency testing schemes are useful for both participating laboratories & accreditation bodies. There are, however, limitations ............. that proficiency testing alone should not be used by laboratory accreditation bodies in their accreditation processes. 7| QCL Training Seminar, Tanzania | 5-7 Dec 07 ISO Guide 43-2 8| 6.2 If a laboratory submits a result(s) which fall outside acceptance criteria for a specific scheme, a laboratory accreditation body should have procedures for acting on such results 6.3 Such procedures should include early reporting to the laboratory of its results with an invitation for the laboratory to investigate and comment on its performance. QCL Training Seminar, Tanzania | 5-7 Dec 07 Quality Manual Laboratories should have a section in their Quality Manuals: - covering participation in proficiency testing - how the results are used to demonstrate the competence of the laboratory - procedures to be followed when unsatisfactory performance is reported - records of participation in PT scheme - corrective action reports 9| QCL Training Seminar, Tanzania | 5-7 Dec 07 PT Scoring System Z-score = ct SD c = calculated reported value t = true value SD = target RSZ = Z m m = number of tests RSSZ = Z2 m 10 | QCL Training Seminar, Tanzania | 5-7 Dec 07 Assigned Value, x The assigned value ( x ) may be the “true” or the consensus value. True value may be a theoretical value or known value from “spiking’ of known quantity of a known quantity of analyte to a sample. Consensus value based on the results of the participants. The consensus value is determined by the application of robust statistics (eg median value, mean interquartile range, Huber’s robust mean), to avoid the influence of “outliers” in the overall mean. 11 | QCL Training Seminar, Tanzania | 5-7 Dec 07 Target standard deviation (TSD) This is set based on experience, reported or expected precision of the techniques used and according to fitnessfor-purpose. The TSD must be realistic. The TSD should be consistently applied from round to round for the same technique/procedure in a PT scheme so that performance can be assessed over time. 12 | QCL Training Seminar, Tanzania | 5-7 Dec 07 Ranking Z-score ≼ 2 satisfactory ≽ 2≼ 3 >3 13 | questionable unsatisfactory QCL Training Seminar, Tanzania | 5-7 Dec 07 Outliers are indicated using three test statistics: Cochran’s test for outlying variances, Grubbs’ single test for outlying means and Grubbs’ paired test for outlying means, to be applied in this order. If a laboratory is excluded, the cycle is repeated from Cochran’s test until no outliers remain. 14 | QCL Training Seminar, Tanzania | 5-7 Dec 07 CochranÕs test for out lying variances is applied as follows: Calculate for each maxö2 laboratory the within laboratory variance ö2 and then . Th is value is compared ö2 with the tabled critical va lues for the number of individual determinations r. If it exceeds the cr itical va lue, the la boratory with the maximum variance is removed and the cyc le is continued with CochranÕstest. 15 | QCL Training Seminar, Tanzania | 5-7 Dec 07 Grubbs’ single test for outlying means is applied as follows: Calculate the between laboratory standard deviation ˆ of the mean values. Calculate the standard deviations ˆ L and ˆ H excluding the lowest and the highest mean value respectively * and let ˆ * be the smallest of these two values. Calculate 1 ˆˆ and compare it with the tabled critical values. If it exceeds the critical value, the corresponding laboratory is excluded and the cycle continued with Cochran’s test 16 | QCL Training Seminar, Tanzania | 5-7 Dec 07 GrubbsÕ paired test for outlying means is applied as follows: Calculate the between laboratory standard deviation ö of the mean values. Calculate the standard deviations ö2 L , ö2 H and öLH excluding the two lowest, the two highest, and the highest and the lowest mean value respectively and let ö* be the smallest of these ö* three va lues. Calculate 1 and co mpare it with the tabled critical values. If it ö exceeds the cr itical va lue, the corresponding laboratories are excluded and the cycle continued with CochranÕstest. 17 | QCL Training Seminar, Tanzania | 5-7 Dec 07 Robust Statistics Huber’s mean for calculation of the « consensus » value is preferred to using elimination of results by tests for outliers 18 | QCL Training Seminar, Tanzania | 5-7 Dec 07 Scoring over time RSD - Used for detecting consistent bias RSSZ - Magnitude of deviations. Cancellation of significant Z-scores if opposite sign is limited 19 | QCL Training Seminar, Tanzania | 5-7 Dec 07 z The rescaled sum of squared z-scores RSSZ is calculated as RSSZ = m 2 where m is the number of squared z-scores added (one for each sample). Critical values of RSSZ depend o n m according to the table: 20 | QCL Training Seminar, Tanzania | 5-7 Dec 07 m 2 3 4 5 6 21 | ‘Doubtful’ > 3.00 > 2.60 > 2.37 > 2.21 > 2.10 ‘Unacceptable’ > 4.61 > 3.78 > 3.32 > 3.02 > 2.80 QCL Training Seminar, Tanzania | 5-7 Dec 07 Proficiency Testing Corrective actions: - PT schemes can be either mandatory or voluntary - Voluntary schemes: PT records examined by external auditors during ISO 17025 assessment - Mandatory schemes: Corrective action reports must be set within a defined time limit to the organisers for assessment and approval (or not). Failure to do so will result in a sanction 22 | QCL Training Seminar, Tanzania | 5-7 Dec 07 OMCL Network Proficiency Testing Scheme Initially open to OMCls of the European Union & other OMCLs associated with the European Pharmacopoeia (member & observer states) Now open to any lab. on a fee paying basis minimum of 4 tests/year voluntary scheme 23 | QCL Training Seminar, Tanzania | 5-7 Dec 07 WHO External Quality Assurance Assessment Scheme (EQAAS) Started in 2001 for selected regional medicines control laboratories Phase 4 began in 2007 24 | QCL Training Seminar, Tanzania | 5-7 Dec 07 PTS 20 Atomic Absorption Spectrophotometry Incorrect programming of instrument resulting in insufficient significant figures Calibration 1 significant figure 1200 ppm K 3 significant figure 1290 ppm K 11/28 questionnaire/unsatisfactory results 25 | QCL Training Seminar, Tanzania | 5-7 Dec 07 PTS 025 Titration of the conjugate acid of organic bases * Corrected for water content (not required) A lab. reported a difference of 2% in results between potentiometric & visual end-point. However, incorrect standardisation procedure was employed * Deterioration in response of the electrode 23/46 questionnaire/unsatisfactory results 26 | QCL Training Seminar, Tanzania | 5-7 Dec 07 Liquid Chromatography Assay of Indapamide 98.33} 98.64} 98.11} Mean = 98.7 RSD = 0.40 True value:- 99.75 z-score: 2.13 Repeatability of ref. sol. (n=6) :- 0.57 27 | QCL Training Seminar, Tanzania | 5-7 Dec 07 Liquid Chromatography Assay of Indapamide Mg 20.28 28 | Corrected Area Mean ²% Content 1 2 - - - - 1698 1691 1675 1668 1672 20.32 1669 1668 1643 1642 1643 20.79 1696 1701 1632 1636 1634 20.97 1716 1721 1637 1641 1639 < 1.0 98.65 99.46 20.56 1690 1696 1644 1649 1647 99.11 99.94 Ref. Sol. Test Sol. Area 1.8 98.33 99.1 QCL Training Seminar, Tanzania | 5-7 Dec 07 Phase 4 : Procedure 1 - Table 1 Raw data & scoring of participating laboratories Semi-micro determination of water 29 | QCL Training Seminar, Tanzania | 5-7 Dec 07 30 | QCL Training Seminar, Tanzania | 5-7 Dec 07 EQAS on water content by Karl Fischer Number & percentage of participating laboratories having shown satisfactory performance (z-score < 2) 31 | Water content Phase 1 (2001) Phase 3 (2005) Phase 4 (2007) No of Participants 7 33 35 No of labs with satisfactory results 4 19 23 % of labs with satisfactory results 57% 58% 66% QCL Training Seminar, Tanzania | 5-7 Dec 07 EQAAS Comments: 32 | - From the 3 participants who failed on the first exercise, 2 of them did not participate in the subsequent studies. Concerning the 3rd one their results slightly improve Phase 3 and were satisfactory in Phase 4. - From the 14 participants who failed in the second exercise, 8 of them reported satisfactory results on the 3rd one (3 did not participate and 3 didn’t show any improvement) - There doesn’t seem to be an improvement in the general trend. The overall performance of laboratories using this technique is not very satisfactory and could be improved. However, it has to be pointed out that the determination of water by Karl-Fischer is problematic even for experienced laboratories as we can see from the results reported by the laboratories (including OMCLs) participating in our regular PTS programme. QCL Training Seminar, Tanzania | 5-7 Dec 07 References International Organization for S tandardization, ISO/IEC Guide 43-1: 1997, Proficiency testing by interlaboratory comparisons Ń Part 1: Development and operation of proficiency testing schemes. International Organization for S tandardization, ISO/IEC Guide 43-2: 1997, Proficiency testing by interlaboratory comparisons Ń Part 2: Selection and us e of proficiency testing schemes by laboratory accreditation bodies. International Organization for Standardizatio n, ISO 57252-1994: Statistical methods. Association of O fficial Analytical Chemis ts, Guidelines fo r Collaborative Study Procedure to Validate Ch aracteristics of a Method of Analysis, AOAC Ar lington, (1990). The Royal Society of Chemistry, Proficiency Testing in Analytical Chemistry, Teddingto n, (1997). International Organisation of Standardisation ISO/IEC Standard ISO13528 : 2005, Statistical Methods for use i n Proficiency Testing by Interlaboratory Comparis ons. 33 | QCL Training Seminar, Tanzania | 5-7 Dec 07