Objective: To compare the progression and non-progression of HIV infection among 418 injecting drug users (IDU) and 422 homosexual men with documented dates of HIV seroconversion from 12 cohorts.

Methods: Seroconversion dates were calculated for each subject using a cohort-specific estimate of the cumulative HIV seroincidence over calendar time. In survival analysis, we studied the progression from seroconversion to AIDS and death by risk group. We compared non-progression between both risk groups by evaluating annual CD4 decline over the 7 years following seroconversion among AIDS-free subjects.

Results: The relative hazard (RH) of AIDS for homosexual men compared with IDU was 1.54 before, and 1.21 after, adjusting for age at seroconversion and year of seroconversion. The risk of death from any cause for homosexual men compared with IDU increased over time since seroconversion. Fifty IDU died prior to AIDS, compared with seven homosexual men (unadjusted RH for homosexual men 0.10). Ignoring this pre-AIDS mortality, the crude RH of death for homosexual men compared with IDU was 2.05. Alter adjusting for age at seroconversion and year of seroconversion in multivariate analysis, the RH became 1.42. No differences in progression between subgroups aged 24 years or older could be demonstrated, but subjects < 24 years were found to be at a decreased risk. Proportions of non-progressors based on CD4 slope > or = 0 at 7 years following seroconversion were higher for IDU than for homosexual men. No differences were found in the proportion (approximately 5%) classified as non-progressors by criteria of both slope > or = 0 and absolute CD4 counts > 500 cells x 10(6)/l, even if pre-AIDS deaths and losses to follow-up were included.

Conclusions: We found little evidence for an effect of risk group on progression and non-progression. Pre-AIDS mortality was much higher among IDU than homosexual men. Pre-AIDS mortality and a nonlinear age effect should be considered in planning interventions as well as studies comparing risk groups and modelling the epidemic.