Revisiting norovirus dose-response models: a meta-analysis and multi-model comparison for waterborne risk assessment

Norovirus is a leading cause of gastroenteritis worldwide, and robust dose-response models are essential for quantitative microbial risk assessment (QMRA). However, existing models are constrained by limited human challenge data and a lack of systematic multi-model comparison, particularly for GII norovirus. In this study, we conducted a meta-analysis of 408 publications and extracted 37 dose-response datasets. Five models, including three models incorporating an immunity parameter, were fitted separately to GI and GII data using maximum likelihood estimation. Apparent and consistent immunity was observed for GI (φ = 0.31-0.32), whereas susceptibility/immunity-related signals were weak for GII and model-dependent (φ = 0-0.02 in most models). The estimated one particle infection probability (P_inf(1)) varied substantially across models (0.08%-45.61%), highlighting strong structural uncertainty, particularly in the low-dose range where observed data are limited. Aggregation-related parameters were highly model-dependent, suggesting that aggregation functions as an effective fitting parameter rather than a directly measured property. The conditional illness probability was best represented by a one-inflated beta distribution for both GI and GII, with GII showing a stronger concentration near 1. Scenario-based QMRA across four exposure pathways showed that log reduction values (LRVs) to achieve acceptable safety level varied considerably across models (differences >2 log₁₀), and followed the scenario order: drinking > toothbrushing > showering > swimming. Overall, these findings demonstrate that norovirus risk estimates are highly sensitive to model structure and assumptions, underscoring the need to explicitly consider structural uncertainty in waterborne QMRA.