Thermal inactivation kinetics of foodborne viruses across food matrices: a Bayesian meta-analysis and predictive...

Human norovirus and hepatitis A virus are leading causes of foodborne illness, yet their thermal resistance across foods remains difficult to predict. We conducted a Bayesian hierarchical meta-analysis of time-series thermal inactivation data from 29 studies identified following PRISMA guidelines to quantify the effects of food matrix, virus group, and measurement method. Data from simple buffers, molluscan shellfish, dairy products, berries, and vegetables were evaluated for human enteric viruses and cultivable surrogates. Infectivity- and RT-qPCR-based measurements were analyzed separately to distinguish infectivity loss from loss of detectable genome signal. A Weibull primary inactivation model combined with a log-linear secondary temperature model had the best predictive performance for both outcomes. Food matrix was the strongest predictor of infectivity-based thermal resistance. Molluscan shellfish provided the greatest protection, with a D_60°C of 4.30 min [95% CrI: 1.86, 10.79] and a z-value of 30.10 °C [23.14, 42.04], whereas berries had the lowest z-value, 8.92 °C [8.17, 9.79]. Virus groups showed similar thermal behavior after accounting for matrix effects. In RT-qPCR data, genome signal declined slowly during heating, yielding a high global D_60°C estimate that varied little across predictors; however, z-values varied by pretreatment, matrix, and virus group. Viability pretreatments improved agreement between RT-qPCR- and infectivity-based estimates, with porcine gastric mucin providing the closest alignment. This Bayesian framework supports the prediction of enteric virus reduction across food-processing conditions to inform evidence-based food safety decisions.