In a cohort of 1835 homosexual men who were seropositive for human immunodeficiency virus (HIV) on entry into a prospective study, the acquired immunodeficiency syndrome (AIDS) developed in 59 during a median follow-up of 15 months. We matched 5 seropositive controls to each case according to study center and date of enrollment and performed a case-control analysis to determine factors predictive of AIDS. In a multivariate analysis, a decreased number of T helper lymphocytes, an increased number of T suppressor lymphocytes, a low level of antibody to HIV, a high titer of cytomegalovirus antibody, and a history of sex with someone in whom AIDS developed were independently associated with subsequent AIDS. Separate analyses of risk factors for Kaposi's sarcoma and opportunistic infections failed to support previously reported associations between the use of nitrites or an elevated cytomegalovirus-antibody titer and Kaposi's sarcoma. These variables may be markers rather than determinants of disease progression. A vigorous antibody response to HIV infection may confer at least temporary protection against the progression of immunodeficiency to AIDS, or a low level of antibody to HIV may reflect a later stage of infection. The increased risk associated with a history of sex with someone in whom AIDS developed may indicate earlier infection in cases or infection with a more virulent strain of HIV. These results may be useful in counseling HIV-seropositive persons and in designing studies of clinical interventions.