The natural history of infection with human immunodeficiency virus type 1 (HIV-1) is characterized by a relentless decline in CD4-positive lymphocytes and the ultimate development of acquired immunodeficiency syndrome (AIDS). However, variables other than the CD4-positive lymphocyte level contribute to the measurement of risk for AIDS and can be used as predictors of AIDS onset. This study was undertaken to identify factors that, independently of the CD4-positive lymphocyte level, would predict the risk of AIDS over 24 months in a cohort of HIV-1 seropositive homosexual men receiving no antiretroviral therapy. Demographic, clinical, and laboratory data from 1,325 white, HIV-1 seropositive participants in the Multicenter AIDS Cohort Study who have been studied for 4 years were analyzed with univariate and multivariate methods. To control for stage of infection, the baseline percentage of CD4-positive lymphocytes (a known marker of disease progression), and the decline of CD4-positive cells during the first 6 months of observation were used as continuous variables. The variables that were independently associated with an increased risk of developing AIDS were: low baseline CD4 percentage, decline in the CD4 percentage during the first 6 months of follow-up, the presence of serum immunoglobulin A at baseline, decrease in hemoglobin during the first 6 months of follow-up, incident fatigue, and the interaction of decline in the CD4 percentage and incident thrush. While low CD4 percentage and other variables have been previously described as prognostic markers, decline in the CD4 percentage and the interaction of that decline and incident thrush have not previously been described as being of prognostic importance. These variables and the analytic method for estimating prognosis may prove useful for selecting and evaluating antiretroviral therapy, instituting prophylactic measures against certain opportunistic infections, and recruitment into clinical trials. Because study participants received no antiretroviral prophylaxis during the period under analysis, the method could be used to estimate the prognosis for those receiving investigational treatment were they to remain untreated, effectively making any participant in a clinical trial his own untreated control.