Simplicity vs. Complexity in the Development of Risk Models for Dose-Response Assessment

Abstract

Deliberations over chemical safety rely heavily upon the interpretation of toxicological and epidemiological evidence. Dose-response modeling plays an important role, allowing true effects to be more effectively discerned from background processes (qualitative determinations) and playing an even more instrumental role in quantitative judgments (e.g., the estimation of potency or acceptable dose). We survey five relatively distinct topic areas (ranging from the assessment of genotoxicity, to epidemiological studies of respirable particles) where dose-response models have been applied, and explore the varying degrees of complexity used in modeling. We choose a descriptive approach, presenting each topic area as a case study. The survey reveals a wide spectrum of complexity both within and between topic areas. The question of ‘what level of model detail is appropriate?’ is widely debated in various disciplines. Notably, the policy context of dose-response assessment (with its attendant real-world stakes) involves some added considerations. We therefore try to emphasize the role of each modeling application's outcome in affecting a decision process. The case studies provide a first step in identifying issues related to model parsimony that are unique to the dose-response, and policy realm.