ESTIMATING ASYMPTOMATIC DURATION IN CANCER - THE AIDS CONNECTION

Many chronic diseases, including AIDS and cancer, do not manifest them selves clinically until some time after their inception. In studies of disease natural history, the duration of the asymptomatic period is o f interest - in AIDS, to predict the epidemic's course, and in cancer, to develop efficient screening strategies. This article provides a br idge between the two fields with respect to estimation of the asymptom atic period. By adapting AIDS methodology to cancer, the article ident ifies a non-parametric method for estimating the duration of the asymp tomatic period in cancer. The method is similar to one developed by Lo uis et al. (Mathematical Biosciences, 40, 111-144 (1978)), and is desi gned to apply to data from a cohort of individuals, screened periodica lly. After reviewing the similarities and differences between the AIDS and cancer contexts, we develop an EM algorithm that, at convergence, yields a maximum or saddle point of the likelihood. We investigate th e performance of the algorithm by means of a simulation study, explore the effect of adding a smoothing step to the estimation procedure, an d adapt the method for use with a data set in which disease prevalence is low. We apply the method to data from the HIP breast cancer screen ing trial. (C) 1997 by John Wiley & Sons, Ltd.