In most clinical trials of antiretroviral therapy for patients infected with HIV, the major outcome variable has been the combined clinical endpoint of any new or recurrent AIDS defining event. We review features of combined endpoints and use data from the Terry Beirn Community Programs for Clinical Research on AIDS (CPCRA) to evaluate this outcome measure in terms of relevance, diagnostic certainty and sensitivity. We conclude that this endpoint is not relevant because: (i) the 19 different events constituting the combined endpoint are equally weighted in analyses even though they vary considerably in terms of risk of death; and (ii) events after the first are ignored, thus the event profile of patients is not taken into account in making treatment comparisons. We also conclude that power may be low with use of this endpoint if treatments under study do not have an immediate impact on disease progression, if some events which occur soon after randomization represent a disease process that has already begun to incubate, or if treatment differences for the various events constituting the combined endpoint are differentially effected by treatment. Since the ease and certainty of diagnosis of each of the 19 events also vary, we recommend that survival be the primary endpoint of antiretroviral trials, and that all opportunistic events experienced by patients, not just the first, be collected and summarized. Trial reports should include comparisons of incidence of each event by treatment group so that readers can rank events as they please. A single summary measure which considers severity and the entire event profile, as described here, would also be useful for assessing the impact of treatments on quality of life. Further research on approaches for weighting and combining multiple outcome measures is needed.