Submitted to: Journal of Food Protection
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: November 29, 2004
Publication Date: February 14, 2005
Citation: Oscar, T.P. 2005. Validation of lag time and growth rate models for Salmonella Typhimurium: Acceptable Prediction Zone Method. Journal of Food Science. 70(2): M129-M137. Interpretive Summary: Mathematical models that predict the growth of bacterial pathogens, such as Salmonella, are useful tools for helping the food industry and regulatory agencies assess the microbiological safety of food. Most predictive models for bacterial pathogens were derived from kinetic data obtained in laboratory broth media. An important step in the development of any broth-based model is evaluation of the model predictions against data obtained with food. In the present study, the predictions of broth models for lag time (time for growth to start) and growth rate of Salmonella were evaluated against data obtained with sterile cooked chicken. Results indicated that the broth model for growth rate provided good predictions on chicken whereas the broth model for lag time provided fail-dangerous predictions at short lag times. A significant accomplishment in this study was the development of a new method for evaluating predictive models that involved one performance index rather than three.
Technical Abstract: Predictions of broth models for lag time (LT) and specific growth rate (SGR) of Salmonella Typhimurium as a function of previous growth pH (5.7 to 8.6), temperature (15 to 40 C) and pH (5.2 to 7.4) were compared with observed LT and SGR on sterile cooked chicken breast (pH = 6.1) and thigh (pH = 6.9) burgers to determine whether the broth models gave acceptable predictions of S. Typhimurium growth in a model food system. Experimental conditions and modeling methods were similar for broth model development and performance evaluation on chicken burgers. Relative error (RE) plots of broth model predictions were evaluated for prediction bias and accuracy and systematic prediction bias in one-step using a Safe Prediction Zone (SPZ) that extended from RE of -0.3 (fail-safe) to 0.15 (fail-dangerous). Relative errors outside the SPZ were considered overly fail-safe or overly fail-dangerous. The fraction of RE in the SPZ (RESPZ) was used as a single index of model performance and was found to be a more sensitive indicator of model performance than prediction bias and accuracy factors. Based on a summary of studies, models with RESPZ of 0.7 to 1 were considered to have acceptable performance. The broth model for LT had an RESPZ of 0.58 for breast burgers and an RESPZ of 0.67 for thigh burgers, whereas the broth model for SGR had an RESPZ of 0.83 for breast burgers and an RESPZ of 0.92 for thigh burgers. Thus, broth model predictions for growth of S. Typhimurium on chicken burgers were acceptable for SGR but not LT. Of note, broth models made overly fail-dangerous predictions at short LT.