Dose-time-response modeling of longitudinal measurements for neurotoxicity risk assessment

Neurotoxic effects are an important non-cancer endpoint in health risk assessment and environmental regulation. Neurotoxicity tests such as neurobehavioral screenings using a functional observational battery generate longitudinal dose-response data to profile neurological effects over time. Analyses of longitudinal neurotoxicological data have mostly relied on analysis of variance; explicit dose-time-response modeling has not been reported in the literature. As dose-response modeling has become an increasingly indispensible component in risk assessment as required by the use of benchmark doses, there are strong interests in and needs for appropriate dose-response models, effective model-fitting techniques, and computation methods for benchmark dose estimation. In this article we propose a family of dose-time-response models, illustrate statistical inference of these models in conjunction with random-effects to quantify inter-subject variation, and describe a procedure to profile benchmark dose across time. We illustrate the methods through a dataset from a US/EPA experiment involving the FOB tests on rats administered to a single dose of triethyl tin (TET). The results indicate that the existing functional observational battery data can be utilized for dose-response and benchmark dose analyses and the methods can be applied in general settings of neurotoxicity risk assessment. Copyright (c) 2005 John Wiley & Sons, Ltd.