The best independent predictors of survival after AIDS diagnosis were determined in 160 homosexual men from the Amsterdam Cohort Study (1987 CDC case definition). In a Cox model, all pre-AIDS laboratory markers gathered during follow-up in the study did not independently predict survival after AIDS. A second analysis, using data available only at AIDS diagnosis, revealed that T cell reactivity after stimulation with phytohemagglutinin and monoclonal CD3 antibodies and the CD4+ cell count at diagnosis were independent predictors of survival. In an interval-based Cox model, the predictive value of pre-AIDS zidovudine use was determined. One year after AIDS diagnosis, participants with pre-AIDS zidovudine use had a worse survival probability (relative hazard = 3.99). In conclusion, the survival probability of homosexual men with AIDS can be estimate adequately using information obtained at clinical AIDS diagnosis. Analysis revealed the important predictive value of T cell function tests in vitro.