The estimation of lifetime risk and average age at onset of a disease using a multivariate exponential hazard rate model

Readily grasped concepts such as lifetime risk of, and expected age at onset of, a disease, cannot be easily estimated by the relative risk methods commonly used in epidemiology. Here we develop a method for estimating these quantities with confidence intervals, where the likelihood is a multivariate function of exposure variables. The model is basically non-parametric, except for a proportional hazards assumption on exposures, and except for the youngest data-poor ages, where an exponential model in age is assumed. Tests of goodness-of-fit of the model are described, these allowing for product terms between exposures and age. Application of the method is made to the Adventist Health Study cohort, where the independent effects of several exposures on these outcomes are demonstrated.