Data-Based Dose-Response Modeling for Foodborne Pathogens. M. E. Coleman, H. M. Marks, and A. Baker, U.S. Department of Agriculture, Food Safety & Inspection Service, 1400 Independence Avenue, SW, Washington, DC 20250-3700
The development of methodology for microbial risk assessment requires in dose-response modeling consideration of the complexity of the disease triangle, which represents a multitude of possible interactions of pathogen, host, and environment necessary to produce foodborne illness. A few studies have controlled each aspect of the disease triangle. The human feeding studies provide the most direct evidence for human dose-response relationships. These studies were conducted with multiple doses of a few species, strains, and serotypes of pathogens and with healthy adult volunteers, generally small groups of male prisoners. The confidence with which derived dose-response models can be applied to other pathogens, other sub-populations of hosts, and other food vehicles or environmental conditions posed for deliberation. The datasets for Shigella, enteropathogenic E. coli, and Salmonella are discussed as examples where surrogate dose-response models are developed in the absence of specific data for the pathogens of interest. Variability and uncertainty are explicitly included as model inputs. Results from formal analysis of variance are presented for experimental and species/strain/serotype effects. Various alternative empirical models fit the data well in the area of observation. The impact of projection of alternative models to the low dose region typical of most foodborne exposures is explored. Treatment of these models is considered in light of epidemiological and mechanistic data for pathogenesis. The data analysis approach can be a useful tool in understanding how dose is related to response in a human sub-population and is necessary for making inferences in dose-response modeling for other populations of pathogens, hosts, and food vehicles. The use of a single model form to represent the disease triangle may be insufficient to incorporate the evidence into a realistic predictor of foodborne illness for the entire human population.