PSVIII-31 Sources of variations, repeatability and reproducibility of microbial count data

The current study consists of two separate experiments to evaluate the relative contribution of systematic factors to the total variability of count data using same silage inoculant. In experiment-1, two lots of a raw material were used to prepare three batches of media. Each of three technicians weighed five independent samples per day, over 10 days and worked on their samples through dilution, plating, and counting. Each sample was grown on three plates. In experiment-2 technician effect was further evaluated at weighing-diluting, plating and counting subprocesses. Each technician weighed and diluted three independent samples per day, over three days. Samples were then shared between technicians during plating and counting. Count data were converted to log10 values and data in experiment-1 were analyzed using a random effects model that included lot, batch, technician, day, and sample within lot-batch-day-technician. Model for experiment-2 included similar effects, except technician effect was partitioned into weighing-diluting technician, plating technician, and counting technician, and there were no batch and lot effects. In experiment-1 sampling variance accounted for a relatively large proportion of the total variance (35%) followed by 33% due to technician effects. The overall repeatability standard deviation of count data was 0.08. Reproducibility standard deviation of measurements made by different technicians on the same day was 0.12. Reproducibility standard deviation of microbial count by the same technician at different days was 0.10. Contribution of sampling variance remained the highest in experiment-2 (39%). The total technician variance was mainly due to weighing-diluting technician, contributing to 31% of the total variance. The relatively large contribution of sample-to-sample variance suggests a major improvement in the precision of count data may be obtained through additional samples rather than increasing number of plates per sample. Technician training can further help reduce variation and improve consistency in sample weighing and dilution.

The ‘Horwitz ratio’—a study of the ratio between reproducibility and repeatability precisions in the analysis of foodstuffs

In the analysis of food, the ratio of reproducibility standard deviation to repeatability standard deviation is usually close to 2.0. This has implications in estimating uncertainty and detection capability.

Liquid Chromatographic Determination of Maleic Hydrazide in Technical and Formulated Products: Collaborative Study

Fourteen collaborating laboratories assayed maleic hydrazide (MH), 6-hydroxypyridazin-3(2H)-one, in technical and formulated products by reversed-phase liquid chromatography (LC) with sulfanilic acid as an internal standard. The active MH in the samples (6 lots) ranged from 16% (expressed as the potassium salt) to 98% (MH in the technical). A small amount of 1 M KOH was added to the technical MH and analytical standards to create the potassium salt of the analyte which is soluble in water. Test samples and standards were extracted with water containing the internal standard before analysis by LC on a C8 column with an ion-pairing eluting solution and UV detection at 254 nm. The concentration of MH was calculated by comparing the peak area response ratios of the analyte and the internal standard with those in the analytical standard solution. Eleven laboratories weighed each test sample twice with single analysis. Three laboratories weighed each sample once and made duplicate injections on the LC system. The data were analyzed using the 11 laboratories' results. A second data analysis was done including all laboratory results using a Youden pair approach, selecting one of 2 duplicate assay values randomly for each laboratory and sample. In the first data analysis, the repeatability standard deviation ranged from 0.07 to 1.39%; reproducibility standard deviation ranged from 0.22 to 1.39%. In the second data analysis (using all laboratory data), repeatability standard deviation ranged from 0.09 to 0.86%; reproducibility standard deviation ranged from 0.22 to 1.31%.

Determining a One-Tailed Upper Limit for Future Sample Relative Reproducibility Standard Deviations

A formula was developed to determine a one-tailed 100p% upper limit for future sample percent relative reproducibility standard deviations <inline-formula> <inline-graphic href="inline_eq1.gif"/> </inline-formula> , where sR is the sample reproducibility standard deviation, which is the square root of a linear combination of the sample repeatability variance ( <inline-formula> <inline-graphic href="inline_eq2.gif"/> </inline-formula> ) plus the sample laboratory-to-laboratory variance ( <inline-formula> <inline-graphic href="inline_eq3.gif"/> </inline-formula> ), i.e., <inline-formula> <inline-graphic href="inline_eq4.gif"/> </inline-formula> , and y is the sample mean. The future RSDR, % is expected to arise from a population of potential RSDR, % values whose true mean is <inline-formula> <inline-graphic href="inline_eq5.gif"/> </inline-formula> , where R and are the population reproducibility standard deviation and mean, respectively.

Fructans in Foods and Food Products, Ion-Exchange Chromatographic Method: Collaborative Study

Nine collaborating laboratories assayed 6 blind duplicate pairs of food samples containing the fructans inulin or oligofructose. The 6 sample pairs ranged from low (4%) to high levels (40%). Following the proposed method, the samples were treated with amyloglucosidase and inulinase enzymes and the released sugars were determined byion exchange chromatography. Repeatability standard deviation ranged from 2.9 to 5.8%; reproducibility standard deviation ranged from 4.7 to 11.1%. The ion-exchange chromatographic method for determinationof fructans in food and food products has been adopted first action by AOAC INTERNATIONAL (997.08)

Determination of Selenium in Feeds and Premixes: Collaborative Study

A total of 17 laboratories participated in a collaborative study for the determination of selenium in feeds and premixes using either a fluorometric or a continuous hydride generation atomic absorption (HGAA) method. Each collaborator analyzed 16 blind duplicate samples of feed and premixes from various feed manufacturers. The amount of Se in these materials ranged from 0.2 to 5500 μg/g. Six laboratories used only the fluorometric procedure, 8 laboratories used only the hydride generation atomic absorption procedure, and 3 laboratories used both procedures. One laboratory in the fluorometric study and 3 laboratories in the HGAA study were initially excluded because of invalid data. Poor agreement between the blind duplicates indicated probable sample interchange and/or dilution error. The data from 8 laboratories were submitted to statistical analysis, including data from 2 laboratories participating in both studies. The repeatability standard deviation (RSDr) for samples analyzed by the fluorometric procedure ranged from 5.9 to 33%, and the reproducibility standard deviation (RSDR) ranged from 12 to 33%. RSDf for samples analyzed by HGAA ranged from 2.8 to 18%, and RSDR ranged from 4.0 to 36%. Both fluorometric and continuous hydride generation atomic absorption methods for the determination of Se in feeds and premixes have been adopted first action by AOAC INTERNATIONAL.

Cylinder Plate Assay for Determining Bacitracin in Premix Feeds and Finished Feeds: Collaborative Study

A cylinder plate assay procedure was studied by 10 laboratories. For premix feeds, 3 samples of bacitracin methylene disalicylic acid and 3 samples of bacitracin zinc premixes covering the range of 10 to 50 g/lb were used. The repeatability standard deviation was 2.11, and the reproducibility standard deviation was 2.13. The average recovery of bacitracin was 101.5%. The method has been adopted official first action. For finished feeds, 6 samples of bacitracin methylene disalicylic acid and 6 samples of bacitracin zinc covering the range of 10 to 800 g/ton were used in the study. The procedure included a sample cleanup step using disposable reverse phase columns. This step appears to be the cause of the poor results reported by most collaborators. Continued study is needed to develop an acceptable method for finished feeds.

Microdiffusion and Fluoride-Specific Electrode Determination of Fluoride in Infant Foods: Collaborative Study

Twelve laboratories analyzed (1 replicate) 12 samples of infant foods – milk, pears, and peas – containing 0.2-5 ppm F. There was one laboratory outlier. Mean coefficients of variation were 7.06% for intralaboratory determination of 3 sets of blind duplicates and 21.6% for interlaboratory determination of 12 samples. Variance analysis for all samples yielded a reproducibility standard deviation of 0.41 ppm; for 3 sets of blind duplicates, repeatability standard deviation was 0.26 ppm and reproducibility standard deviation was 0.32 ppm.