Purpose: In order to support quantitative risk assessment of this pathogen, meta-analytical dose-response models were developed—summarizing the relationship between numbers of ingested STEC O157:H7 cells and probability of someone falling ill as a result.
Methods: After an exhaustive literature search, 14 STEC observations from 10 outbreak studies were brought together, and four classical dose-response models were adjusted: Exponential, Beta-Poisson, Weibull-gamma, and Gompertz. Variables such as population type (normal and susceptible), food matrix (categorized as water, raw food, processed meat, and cheese), mean dose, exposed population, and number of ill cases were extracted from outbreak studies. A logistic regression model with random effects placed on food matrix and weighed according to exposed persons was adjusted to assess the effect of population susceptibility, food matrix, and dose.
Results: Exponential and Beta-Poisson dose-response models had comparable measures of Bayesian Information Criterion (BIC= -21.0 and -18.7, respectively) and fitted the outbreak data better than the Weibull-gamma and Gompertz dose-response models (BIC= -16.8 and -16.2, respectively). Gompertz model was the least adequate as it overestimated probabilities of illness at low doses. The weighted logistic regression model demonstrated that both population susceptibility (P<0.0001) and food matrix (P<0.0001) had significant impact on probability of illness. For the same doses, higher probabilities of illness were obtained for the susceptible population and water. The meta-analysis logistic model was capable of depicting a relationship between dose and probability of illness, specifically for processed meat in both normal and susceptible populations.
Significance: Understanding the dose-response relationship for this pathogen will promote increased food safety and as a result, can reduce the number of foodborne illnesses.