4. AOACRIMicroMethods-2018Awards

145

W allace et al .: J ournal of AOAC I nternational V ol . 97, N o . 3, 2014  869

on the day of initiation of analysis produced a high spike level (POD approximately 1.0 or approximately 5 CFU/test portion) and a low spike level (POD 0.25–0.75 or 0.2–2.0 CFU/test portion). Additional matrix was left uninoculated to serve as negative controls. To inoculate frankfurter samples, a pure colony of Salmonella Typhimuriumwas transferred fromTrypticase Soy agar with 5% sheep’s blood (SBA) into Brain Heart Infusion (BHI) broth and incubated at 35°C for 18–24 h. The inoculum was heat stressed in a 55°C water bath for 10 min to obtain a percent injury of approximately 70% as determined by plating onto selective Xylose Lysine Desoxycholate (XLD) agar and nonselective TSA. Four portions of equal size were inoculated drop-wise with an 18–22 h culture of the target organism, and then homogenized by hand. All four portions were combined one at a time into a single container, homogenizing the bulk material after each portion was added. The bulk lot was separated into two sampling containers and 40 samples (20 for each method) weighing 25 g each were removed from each container. Each 25 g sample was combined with 300 g uncontaminated matrix to create 325 g test portions. The remaining spiked matrix was rehomogenized by combining the material from both containers into one and mixing thoroughly for the purposes of maintaining an even distribution of the organism. To inoculate orange juice, a pure colony of Salmonella Hadar was transferred from SBA into BHI broth containing 1% glucose and incubated at 35°C for 18–24 h. This stress protocol resulted in a percent injury of approximately 60% (as determined by plating onto selective XLD agar and nonselective TSA). The inoculum was added drop-wise to a bulk quantity of orange juice to reach the desired contamination level, and then mixed to achieve equal distribution of the inoculum throughout. This spiked bulk quantity was divided into 25 mL test portions for analysis. All test portions were randomized and blind-coded by the organizing laboratory, then shipped overnight to each collaborating laboratory and maintained at 2–8°C until they were analyzed. The total hold time of samples was 48 h for frankfurters and 96 h for orange juice, including shipment time to each participating laboratory. On the first day of test sample analysis, a 5-tube, 3-level most probable number (MPN) estimation of contamination levels was conducted by the organizing laboratory using the appropriate reference method. The Least Cost Formulations, Ltd (Norfolk, VA) MPN Calculator-Version 1.6 (5) was used to determine the MPN values and 95% confidence intervals. The MPN is reported for each level of each matrix in Appendix 4, Tables 1–6 as MPN/ test portion with 95% confidence intervals. For testing frankfurters, each collaborator received 12 low-spike, 12 high-spike, and 12 uncontaminated 325 g test portions, blind-coded so that the contamination level was unknown to the collaborator. Approximately one-third to one-half of 2925 ± 58.5 mL of sterile buffered peptone water (BPW) was added to each portion, and each portion was homogenized approximately 2 min. The remainder of the Test Portion Distribution Test Portion Analysis

ground beef with soy (85% lean), beef trim, frankfurters (beef), shrimp, ground turkey, chicken wings, dried eggs (whole, powdered), shell eggs, frozen peas, orange juice, instant nonfat dry milk, ice cream (12% fat), peanut butter (52% fat), cocoa (unsweetened), white pepper, milk-based infant formula, ceramic tile, and plastic surfaces. The results obtained using the test method indicate no statistical difference with the reference method when compared to the corresponding reference method results (Appendix 3). In addition, two of the precollaborative study matrixes— frankfurters (pork plus turkey) and orange juice (pasteurized not-from-concentrate)—were evaluated in a total of 15 independent laboratories as part of the collaborative study to demonstrate repeatability and reproducibility of the internal laboratory results independent of the end user. The results obtained using the BAX System method indicate no statistical difference when compared to the corresponding reference method results. Study Design Collaborators analyzed two representative matrixes (pork and turkey frankfurters and pasteurized, not-from-concentrate orange juice without pulp), 12 replicate test portions from each of three contamination levels (low, high, and uninoculated), comparing the performance of the BAX System Real-Time PCR Assay for Salmonella to appropriate reference culture methods. A total of 15 laboratories participated in the study, with 14 laboratories reporting data for each matrix. Each collaborator received instructions for performing the study and required materials prior to the start of the study. If necessary, training on the BAX System was provided to laboratory personnel by a DuPont representative. The collaborative study was conducted in accordance with the AOAC INTERNATIONAL Methods Committee Guidelines for Validation of Microbiological Methods for Food and Environmental Surfaces , Appendix J (4). Frankfurter samples were evaluated against the USDA-FSIS MLG reference method as a paired study, as the test and reference method enrichment protocols are identical. Orange juice samples were evaluated against the FDA-BAM reference method as an unpaired study, as the BAX System method uses enrichment in proprietary media. Estimates of repeatability, reproducibility, and probability of detection (POD) were evaluated. Sample product was obtained from a local retail outlet and screened by the organizing laboratory to identify any naturally contaminating Salmonella and determine a total aerobic plate count. For each sample type, five analytical size portions (25 g for orange juice and 325 g for frankfurters) were screened for Salmonella using the appropriate reference method. Although naturally contaminated samples would have been preferred, all samples tested returned negative results for Salmonella . Therefore, each sample matrix was artificially inoculated with a different serovar of Salmonella for use in this study. Portions of each sample type were inoculated at levels that Collaborative Study Preparation of Inocula and Test Portions

03/10/2019