Purpose: Quantitatively analyze variability of analysts, laboratories, and methods from a proficiency test assessing the level of indicator organisms in infant formula.
Methods: Commercial liquid and powdered infant formula samples (LIF, PIF) were spiked at two different inoculum levels in duplicate with a cocktail of Gram positive and negative organisms. The inoculation procedures, mixing methods, and homogeneity/stability were verified by three independent trials according to ISO 17043 and ISO 13528. Aerobic Plate Count, Total Coliforms, Fecal Coliforms, Enterobacteriaceae, and E. coli were measured by 79 analysts/59 laboratories (LIF) and 19 analysts/19 laboratories (PIF).
Results: Homogeneity and stability of distributed samples, log-normal distributions, skewness, Youden plot analysis, and Analyst/laboratory z-scores were calculated according to ISO 13528 using ProLab®. Relative repeatability was 12.78 - 81.25% (LIF) and 31.03 - 105.37% (PIF). Relative reproducibility range was 31.03 - 105.37% (LIF) and 83.24 - 175.82% (PIF). Equivalency testing was calculated using Q-Hampel method and showed Petrifilm™ and traditional plate count methods were not equivalent for aerobic counts in LIF samples with a relative bias of 70.31%. MPN and Petrifilm™ had comparable performance for total coliform, E.coli, and fecal coliform counts in LIF samples with relative biases of -7.64%, 7.81%, and -0.5%, consecutively.
Significance: This study demonstrates the value of data from large, interlaboratory studies, such as proficiency tests, can be used in a novel way to measure accuracy and precision of methods, analyst’s performance as well as overall performance of laboratories.