ObjectiveThe treatment-free incubation period distribution was estimated from data obtained from a cohort of homo/bisexual men in Amsterdam.DesignParticipants in a cohort study that started in Amsterdam at the end of 1984 were seen at 3− or 6-monthly intervals. The analysis excluded time since the start of zidovudine treatment (given to 27 individuals) and no cohort member received primary Pneumocystis carinii pneumonia prophylaxis (i.e., before AIDS diagnosis) before February 1990.MethodsBoth HIV-positive (n = 269) and HIV-seroconverted (n = 79) homosexual men were included in this study. Of these, 68 developed AIDS before February 1990. The unobserved dates of seroconversion were obtained by multiple imputation. The incubation period distribution was estimated from these data by direct Kaplan—Meier analysis and by using parametric Weibull and gamma distributions (including a parameter describing the probability of never progressing to AIDS). Our results are compared with published estimates of progression to AIDS from other homo/bisexual cohort studies.ResultsBoth the Weibull and the gamma distributions provide equally good empirical descriptions of the incubation period distribution for up to 7 years postseroconversion, but the estimated gamma distribution (median, 9.2; mean, 10.2; percentage AIDS at 7 years, 33%) should be preferred beyond that time, due to a slowing of the hazard rate. There is insufficient information to be able to estimate accurately the probability of never progressing to AIDS.ConclusionsThe time-dependent pattern of HIV incidence should be considered in the analysis of prevalent cohort studies. Our results are in agreement with other homo/bisexual cohort studies, and will be valuable for future comparison with and understanding of the epidemiological consequences of clinical treatment that delays the onset of AIDS.