A Comparison of the BAX System Method to the U.S. Food and Drug Administration's Bacteriological Analytical Manual and International Organization for Standardization Reference Methods for the Detection of Salmonella in a Variety of Soy Ingredients

Abstract

The performances of two DuPont BAX System PCR assays for detecting Salmonella on a variety of low-moisture soy ingredients were evaluated against the U.S. Food and Drug Administration's Bacteriological Analytical Manual (FDA BAM) method or the International Organization for Standardization (ISO) 6579 reference method. These evaluations were conducted as a single laboratory validation at an ISO 17025 accredited third-party laboratory. Validations were conducted on five soy ingredients: isolated soy protein (ISP), soy fiber, fluid soy lecithin, deoiled soy lecithin, and soy nuggets, using a paired-study design. The ISP was analyzed as both 25- and 375-g composite test portions, whereas all other sample matrices were analyzed as 375-g composite test portions. To evaluate 25-g test portions of ISP, the test material was inoculated using Salmonella enterica subsp. enterica serovar Mbandaka (Q Laboratories isolate 11031.1). Salmonella enterica subsp. enterica serovar Tennessee (Q Laboratories isolate 11031.3) was used for all other trials. For each trial of the method comparison, 25 samples were analyzed for each matrix: 5 uninoculated controls and 20 samples inoculated at low levels (0.2 to 2 CFU per test portion) that were targeted to achieve fractionally positive results (25 to 75%). Using McNemar's chi-square analysis, no significant difference at P ≥ 0.05 (χ2 ≤ 3.84) was observed between the number of positives obtained by the BAX System and the reference methods for all five test matrices evaluated. These studies indicate that the BAX System PCR assays, in combination with the single buffered peptone water primary enrichment and subsequent brain heart infusion regrowth step, demonstrate equivalent sensitivity and robustness compared with the FDA BAM and ISO reference methods for both 25- and 375-g composite samples. Moreover, there was no observed reduction of sensitivity in the larger 375-g composite samples for all five matrices.