Submitted to: Poultry Science
Publication Type: Peer reviewed journal
Publication Acceptance Date: 5/15/2010
Publication Date: 9/1/2010
Citation: Cason Jr, J.A., Cox Jr, N.A., Buhr, R.J., Richardson, L.J. 2010. Comparison of the statistics of salmonella testing of chilled broiler chicken carcasses by whole carcass rinse and neck skin excision. Poultry Science. 89:2038-2040. Interpretive Summary: Worldwide, testing of chicken meat for Salmonella uses various sampling and lab methods, plus there are several standards for the number of samples taken and the number of positive samples allowed. International trade agreements require, however, that importing countries accept the control standards of exporting countries if test results can be shown to be equivalent. This study compared the statistical requirements for Salmonella testing in the United States and Europe by calculation of probabilities and random number simulation of test results. Looking only at the required number of samples and the number of positives allowed, chickens tested by neck skin sampling as specified under European rules (50 pools of 3 neck skins, maximum of 7 positive pools) should have a Salmonella-positive rate of about 4% in individual neck skins to have the same probability of a passing test set as chickens tested under U.S. rules (51 whole carcass rinse samples, maximum of 12 positive rinses, Salmonella-positive rate of 20%). This study did not consider differences between sampling and lab methods.
Technical Abstract: Whether a required Salmonella test series is passed or failed depends not only on the presence of the bacteria, but also on the methods for taking samples, the methods for culturing samples, and the statistics associated with the sampling plan. The pass-fail probabilities of the two-class attribute sampling plans used for testing chilled chicken carcasses in the United States and Europe were compared by calculation and simulation. Testing in the United States uses whole carcass rinses (WCR), with a maximum number of 12 positives out of 51 carcasses in a test set. Those numbers were chosen so that a plant operating with a Salmonella prevalence of 20%, the national baseline result for broiler chicken carcasses, has an approximately 80% probability of passing a test set. The European Union requires taking neck skin (NS) samples of approximately 8.3 g each from 150 carcasses, with the neck skins cultured in pools of 3 and with 7 positives as the maximum passing score for a test set of 50 composite samples. For each of these sampling plans, binomial probabilities were calculated and 100,000 complete sampling sets were simulated using a random number generator in a spreadsheet. Calculations indicated that a 20% positive rate in WCR samples was approximately equivalent to an 11.42% positive rate in composite NS samples or a 3.96% positive rate in individual NS samples within a pool of 3. With 20% as the prevalence rate, 79.3% of the simulated WCR sets passed with 12 or fewer positive carcasses per set, very near the expected 80% rate. Under simulated European conditions, a Salmonella prevalence of 3.96% in individual NS samples yielded a passing rate of 79.1%. The two sampling plans thus have roughly equivalent outcomes if WCR samples have a Salmonella-positive rate of 20% and individual NS samples have a positive rate of 3.96%. Sampling and culturing methods must also be considered in comparing the different standards for Salmonella.