To study relapse into unsafe sex among homosexual men using a novel method. Time from safer to unsafe sexual behaviour was analysed using Kaplan-Meier procedures. Cox proportional hazard analysis was performed to quantify the relation between duration of safer sex and several covariates. Subjects were homosexual men enrolled in a longitudinal cohort study in Amsterdam and included in the present assessment if they had ever reported an episode of safer sexual behaviour of at least 12 months. Of the 402 participants who met the inclusion criterion, 289 (72%) reported at least one episode of unprotected anogenital intercourse during follow-up. The median time from safer to unsafe sexual behaviour was 30 months. After 88 months of follow-up, the cumulative incidence of unsafe sexual behaviour reached 88%. Younger age, HIV seropositivity and use of amyl nitrites (poppers), were related to a shorter time to unsafe sexual behaviour. Based on these findings it is possible to target relapse prevention at specific groups of homosexual men, thereby increasing programme effectiveness and efficiency.