Interlaboratory Coefficient Of Variation Research Articles

Evaluation of GHb assays in France by national quality control surveys.

To evaluate the state of the art concerning GHb assays through analysis of a large-scale quality control survey and to compare the results with those of previous surveys. A lyophilized hemolyzate prepared from human erythrocytes containing a physiological HbA1c level (5.5%) was sent to 3,500 French laboratories in February of 1995 and assayed as a patient's sample under routine conditions. Distribution of values was analyzed from the reported results for each method. The results were compared with the assigned value (acceptable range: +/- 20%) and with the upper value of the reference range currently used. Results were obtained from 2,674 laboratories, among which 39% used cation-exchange chromatography methods, 37.5% affinity chromatography, 16% immunological methods, and 7.5% electrophoresis. The number of laboratories using immunological methods increased from 100 to 400 between 1993 and 1995. The overall interlaboratory coefficient of variation (CV) was 20.2%, with within-method CVs ranging between 3.2 and 29.5%. Method-to-method accuracy varied dramatically, with mean HbA1c values ranging from 4.4 to 8.2%. Results from 75% of the laboratories were comprised in the acceptable range; 88% of them reported a value within the normal range of the method used. The interlaboratory variability of results illustrates the difficulties encountered by diabetologists in the follow-up of diabetic patients using results obtained from different laboratories. It demonstrates the usefulness of the internationally developed standardization process of GHb measurements and points out the need for laboratories to fulfill good practices.

Phase III Interlaboratory Study of Fetax, Part 2: Interlaboratory Validation of an Exogenous Metabolic Activation System for Frog Embryo Teratogenesis Assay-Xenopus (Fetax)

ABSTRACTInterlaboratory validation of an exogenous metabolic activation system (MAS) developed for the alternative, short-term developmental toxicity bioassay, Frog Embryo Teratogenesis Assay - Xenopus (FETAX) was performed with cyclophosphamide and caffeine. Seven study groups within six separate laboratories participated in the study in which three definitive concentration-response experiments were performed with and without the MAS in a side-by-side format for each chemical. Since both chemicals had been previously tested in FETAX, the test concentrations were provided to each laboratory prior to testing. Interlaboratory coefficient of variation (CV) values for unactivated cyclophosphamide (no MAS) were 15%, 15%, 29%, and 25% for the 96-hr LC50, 96-hr EC50 (malformation). Minimum Concentration to Inhibit Growth (MCIG), and Teratogenic Index (TI) values, respectively. Addition of the MAS increased the CV values of each endpoint at least 3.9-fold. Interlaboratory CV values for unactivated caffeine were 31%, 18%, 31%, and 46% for the 96-hr LC50, 96-hr EC50 (malformation), MCIG, and TI values, respectively. Addition of the MAS decreased the CV values of each respective endpoint by at least 1.6-fold. Results indicated that bioactivated toxicants may be prone to greater variability in response amongst laboratories than compounds, which are detoxified. Even though more variability was noted with activated cyclophosphamide, results were within interlaboratory variation expected for other aquatic-based bioassays. Thus, results from these studies warrant the continued use and further refinement of FETAX for alternative developmental toxicity assessment.

International Collaborative Study for the Calibration of a Proposed Reference Preparation for Thromboplastin, Human Recombinant, Plain

Stocks of the International Reference Preparation (IRP) for thromboplastin, human, plain, coded BCT/253 and held by the World Health Organization (WHO) are nearly exhausted and must be replaced. For practical reasons the choice of the replacement candidate was restricted to two available human recombinant preparations which were coded as X/95 and Y/95 and calibrated in an international collaborative study involving 19 laboratories from Europe, Australia, Canada and Argentina. To minimize the differences between routes of calibration, the two candidates were calibrated against the existing WHO-IRP from human, rabbit and bovine origin and the final ISI was the resultant average value. On the basis of predefined criteria (i.e., within- and between-laboratory precision of the calibration and the conformity to the calibration model), X/95 was the preferred candidate. The assigned ISI (SE of the mean) value is 0.940 (0.0060) and the interlaboratory coefficient of variation 4.7%.

Precision and sensitivity of a seven-day growth and survival toxicity test using the west coast marine mysid crustaceanHolmesimysis costata

A 7-d growth and survival toxicity test using the west coast marine mysid crustacean Holmesimysis costata (Holmes) was evaluated to determine test precision and sensitivity. The intralaboratory coefficient of variation (CV) among six reference toxicant test median lethal concentrations (LC50s) was 25%, whereas the mean intralaboratory CV for side-by-side effluent test LC50s was 7%. The mean intralaboratory CV for the concentration at which growth was inhibited by 25% (IC25) was 19%. Interlaboratory CVs for effluent LC50s averaged 14%, whereas variation among growth IC25s averaged 15%. Test precision and sensitivity compared favorably with literature values for a number of commonly used toxicity tests and chemical methods. Toxicity increased slightly with increased mysid exposure from 4 to 7 d (mean effluent LC50s of 9.9% for 4 d of exposure and 7.5% for 7 d), and more significantly from 7 to 24 d (zinc LC50 values were 50 μg/L and 7.8 μg/L for concurrent 7-d and 24-d tests). Although growth was a less sensitive endpoint than survival in tests with individual chemicals (zinc and sodium azide), growth was the more sensitive endpoint in seven of nine tests with complex effluents. Seventy-five percent of tests conducted at all participating laboratories met protocol test acceptability criteria (n = 40).

A critical review of ultra-accelerated tests for alkali-silica reactivity

A large number of ultra-accelerated test procedures, for determining the potential alkali reactivity of aggregates, have been developed, particularly in the past 15 years. An ultra-accelerated test method is defined as one which yields results within a few days or, at most, a few weeks. A number of ultra-accelerated test methods have been adopted as ‘standard tests’, but few have been adequately evaluated. The rapid globalization of the construction industry will require the harmonization of National Standard Test Methods. The major requirement of ultra-accelerated test methods is that they should correctly predict the potential reactivity of aggregates in greater than 95% of the cases. Due to the complexity and variability in the composition and grain size of aggregates, it is improbable that a single test method will be developed which would be appropriate for evaluating all types of aggregates. Another major requirement for ultra-accelerated test methods is that the interlaboratory coefficient of variation should be low, preferably less than 12%. At present, only the NBRI accelerated mortar bar method has been subject to adequate inter-laboratory evaluation. However, a more limited inter-laboratory investigation showed that the autoclave mortar bar test also shows considerable potential, as a satisfactory ultra-accelerated test method. Further refinement of the NBRI and autoclave methods is required to improve their performance with a wide variety of aggregates.

Precision of the nitocra spinipes acute toxicity test and the effect of salinity on toxicity of the reference toxicant potassium bichromate

AbstractThe precision of the 96‐h LC50 test with the eurohaline copepod Nitocra spinipes was evaluated from a data set containing 85 tests with the reference toxicant potassium bichromate conducted at 12 laboratories during a 12‐year period. The overall mean was 30 mg/L as K2Cr2O7 with a coefficient of variation (CV) of 47% (N = 85). Since salinity differed between tests (7–33 o/oo), it explained part of the variation (ca. 25%; Pearson two‐tailed test r = 0.49; p = 0.0000; N = 85). The effect of salinity on Cr(VI) toxicity was confirmed in an experiment with 4 salinities (3.4, 6.8, 14.6, and 27 o/oo) and was explained to 99% by the equation 96‐h LC50 (mg/L) = 2.58 + 1.79 × salinity (o/oo). Intralaboratory CVs for corresponding salinities ranged from 8 to 35% (repeatability). Interlaboratory CVs at two salinities were 42% (N = 11) at 7–9 o/oo and 47% (N = 4) at 15–16 o/oo. © 1993 John Wiley & Sons, Inc.

Intralaboratory Performance Requirements Necessary to Pass Proficiency Testing: CAP-1990 vs CLIA-1967 (March 14, 1990) Formats Compared

The pre-1990 College of American Pathologists' (CAP) Proficiency Testing (PT) program used a two samples per analyte/four challenges per year format with performance or pass-fail grading criteria determined by the program. On Jan. 1, 1991, the Clinical Laboratory Improvement Act of 1967 (CLIA-67) final rules (March 14, 1990) mandated a revised PT format of five samples per analyte/four challenges per year, with the regulations specifying minimum performance criteria. Extending our previous analysis, we compare the maximum permissible intralaboratory imprecision at low bias compatible with passing external PT in the former CAP and current CLIA-67 formats. If a laboratory is able to reduce its internal coefficient of variation (CV) to less than 44% of the PT criterion for each analyte, its overall chance of adverse action for any of the 27 routine chemistry analytes specified in CLIA-67 will be less than 1% in a two-year (eight PT challenges or events) period. Consideration of actual interlaboratory CVs from CAP surveys suggest that a reduction of this magnitude may be difficult for the analytes total cholesterol and blood urea nitrogen, where intralaboratory imprecision comparable with the group standard deviation (SD) from 1990 CAP surveys would yield individual adverse action (PT failure) rates of 5% and 1%, respectively. Five other analytes have CLIA-67 performance limits dangerously close to CAP interlaboratory CVs.

Stability of residual levels of polychlorinated biphenyls in cold-extracted herring oil.

The measurement of residual concentrations of polychlorinated biphenyl (PCB) in a variety of foodstuffs and other biological materials has been characterized by a large (up to 50%) interlaboratory variation (bias). An interlaboratory coefficient of variation of 14% was reported for a study in which exact methodological details and quantitative standard solutions were supplied to participants who were asked to determine the levels of a suite of six chlorobiphenyls (CB) present at trace levels in eel fat. Following the completion of the earlier study, remaining vials of fortified (approximately 1 mg Aroclor 1,254/kg herring oil) and control oils were stored and analyzed sporadically to determine if any changes in measured PCB concentrations in the oils resulted from changes in PCB and the oil during storage of approximately eight years.

Interlaboratory Comparison of Analyses for Heavy Metals in Clam Tissue

Results are reported for an interlaboratory study conducted to assess the reproducibility of analyses for lead, copper, cadmium, and zinc. The 10 participating laboratories analyzed 2 samples of freeze-dried clam tissue and 1 disguised sample of NBS Oyster Tissue. Interlaboratory variations were observed for all metals, although the methods yielded reproducible data for Cu and Zn with average interlaboratory coefficients of variation of 15 and 11%, respectively. The performance of methods used for Cd and, more so, for Pb was less than satisfactory. Cadmium levels in the 3 samples ranged from about 0.7 to 3.7 ppm with an average interlaboratory coefficient of variation (CV) of 24%. Lead levels in the 3 samples were about 0.5 ppm with an interlaboratory CV of 68%. Some laboratories' results were consistently high or low but data were insufficient to relate these trends to one particular variable. Results from this study were compared with 5 other studies reported in the literature since 1980. Coefficients of variation from all studies were comparable for samples with similar metal concentrations.

Accuracy of gas analysis in lung function laboratories.

Fifty lung function laboratories in England and Wales analysed test gas mixtures of carbon monoxide and helium. Most of them also analysed mixtures of oxygen and carbon dioxide in nitrogen. The percentage accuracy of the results was within 1% of the expected value in only 14% of determinations of carbon monoxide concentration, 28% for carbon dioxide, 37% for helium, and 48% for oxygen. The accuracy of ratios of two concentrations of helium and carbon monoxide was better than that of the individual gas samples. Overall the variation in results between laboratories was wide, the coefficient of variation ranging from about 3% for analysis of helium to 9% for carbon dioxide. This variation affected the values calculated for carbon monoxide transfer factor, where 20% were in error by more than 5%, and for the calculated value of the respiratory exchange ratio, where the interlaboratory coefficient of variation was about 10%. Errors in analysis were due to unsatisfactory calibration of analysers; five oxygen analysers had large zero errors; five carbon monoxide analysers and one helium analyser had notably curvilinear calibration curves. Insufficient information was obtained to ascertain the nature of the errors in analysis of carbon dioxide. Given the improvements in instrumentation, these results are evidence for deterioration in analytical standards in lung function laboratories from the standards of 20 years ago.

Gas Chromatographic Determination of Pentachlorophenol in Gelatin: Collaborative Study

Eleven collaborators participated in this study of a gas chromatographic method for the determination of pentachlorophenol (PCP) in gelatin. Following acid hydrolysis of a 2 g sample, PCP is extracted with hexane and partitioned into KOH solution. After reacidification, PCP is again extracted with hexane for determination by electron capture gas chromatography on a 1% SP-1240DA column. Three duplicate practice samples (0.0, 0.5, and 1.5 ppm) and 5 blind duplicate collaborative samples (0.0, 0.02, 0.1, 0.5, and 2.0 ppm) were analyzed by each collaborator. Mean recoveries of PCP in the collaborative samples ranged from 88% at the 0.02 ppm fortification level to 102% at the 0.1 ppm level; the overall mean recovery was 96%. Interlaboratory coefficients of variation ranged from 16.4% for the 0.1 ppm fortification level to 22.9% for the 0.5 ppm level; the overall interlaboratory coefficient of variation was 19.5%. The method has been adopted official first action.

Validity of the consensus mean as the target value for a small External Quality Assessment Scheme.

Data from 2 years' operation of an External Quality Assessment Scheme covering 14 analytes and with some 60 participants is presented. Following the trimming of discrepant results and statistical 'outliers' (outside the range of +/- 3 SD from the mean), there was generally close agreement between consensus mean values for a specimen analysed on different occasions, by different groups of laboratories, or when using different analytical methods. An improvement in performance, indicated by a reduction in the average inter-laboratory coefficient of variation was found for 11 of the 14 analytes over the 2-year period.

Vergleich dreier Ringversuche zur radioimmunologischen Thyrotropin-Bestimmung nach dem „Münchner Modell”

The results of a thyrotropin external quality assessment survey (EQAS) carried out in 1980 using the "Munich Model" and including a "matched reagent" experiment are presented and are compared with the results from similar surveys carried out in 1974 and 1977. Despite the fact that in 1974 only 4 different kits were used, compared with 12 in 1977 and 17 in 1980, the inter-laboratory precision has continually improved. This was reflected by the inter-laboratory coefficient of variation (CV) for all laboratories, which, for a sample with ca. 15 microU/l thyrotropin were: 65% (1974), 22% (1977) and 12% (1980), and for a sample containing ca. 5 microU/l thyrotropin: 75% (1974), 45% (1977) and 20% (1980). During the same period, the fraction of participants using commercial kits increased from 0.65 to 0.95. There was a definite trend towards shorter assays, even though with such assays there is a danger that the measured values for low thyrotropin concentrations will be too high. In spite of the fact that the "high-blank" effects have to a large extent disappeared, many kits showed an unacceptably high cross-reactivity with human chorionic gonadotropin. A fraction of 0.30 of all participants in 1974 used computer-assisted data-processing, compared with 0.78 in 1980. In 1980, the most commonly used algorithm was a spline function (0.46 of participants using a computer for data-processing). Although the number of kits has increased, the precision has improved, showing that more robust methods are now employed. This has partly come about because of the introduction of human thyrotropin-free serum as matrix for the standards (1974 - 0.16, 1980 - 0.83 of all participants). Further external quality assessment programmes are discussed for the continual monitoring of hormone assay performance.

Determination of Fatty Acid Composition by Programmed Temperature Gas Chromatography

The mixtures of known compositions of methyl esters of six fatty acids (C8, C10, C12, C14, C16, and C18) were analyzed collaboratively by programmed temperature gas chromatography. The experimental data were treated statistically to examine inter- and intralaboratory scattering. Moreover, the effect of the correction by correction factors was investigated. For the analysis of a sample containing esters which considerably differ in retention time, programmed temperature GLC was preferable to isothermal GLC in regard to the scattering of experimental values and the deviation from the known value. High program rate, such as 10°C/min, however, resulted in an increase of scattering in experimental values. The data from the laboratories, in which the intralaboratory CV (coefficient of variation) was so small as to be less than 3.0% for any five components of six in each sample, were corrected with correction factors which were calculated from the analytical values of other sample. The interlaboratory CV of the uncorrected values appreciably decreased compared with those of values obtained by all laboratories. The corrected values agreed very closely with the known values and the interlaboratory scattering of those corrected values considerably decreased.

Automated Colorimetric Determination of Phenylephrine Hydrochloride in Drug Formulations. II. Collaborative Study

Abstract The automated colorimetric method designed by Lane for the analysis of phenylephrine HCl in drug formulations has been submitted to a collaborative study. Five samples, including 2 with known interferences, were sent to 9 collaborators. The results were subjected to statistical evaluation. These show good recovery and a maximum interlaboratory coefficient of variation of 4.5%. The automated method as evaluated in this study has been adopted as official first action.

Dye Binding Method for Protein Content of Dairy Products

Abstract The Udy dye binding method for protein in milk has been under collaborative study since 1965. The test is based on the formation of an insoluble complex between Acid Orange 12 (C.I. 15970) and basic amino acid residues of the protein. A known amount of dye is added to the sample, the complex is removed, and the excess dye is measured by colorimetry. The amount of protein in the original sample is calculated from the amount of dye used in complex formation. In previous reports, it was reported that the interlaboratory coefficient of variation is 1% or less, i.e., 0.03% protein in the sample. The coefficient of variation for the Kjeldahl method is 3%. Suitable analytical control is achieved through optical determination of dye concentration in the reagents and by testing known milks. Mercuric chloride is a suitable sample preservative, but potassium dichromate decolorizes the dye. It is recommended that this method be adopted as official final action for determining protein contents of fluid milk, half-and-half, nonfat dry milk, ice cream mix, chocolate drink, and buttermilk.