The prognostic value of various immunologic tests was investigated in 150 HIV-seropositive homosexual men, who were initially without HIV-related symptoms or AIDS and who were followed for a median of 12 months (range 3-28 months). The laboratory investigations included HIV antigen in serum, total lymphocyte count, T-helper (CD4) and T-cytotoxic/suppressor (CD8) counts, and lymphocyte transformation responses to the mitogens phytohemagglutinin (PHA) and pokeweed mitogen (PWM), and to antigenic extracts from Candida albicans and cytomegalovirus. 24 individuals developed HIV-related symptoms or AIDS (11 cases). All parameters except the CD8 count were of prognostic value, but a multivariate analysis of symptom-free survival showed that HIV antigenemia, a CD4 count less than 0.5 x 10(9)/l, and relative response to PWM below 25% of controls contained all the prognostic information. Individuals abnormal at entry for these 3 variables had a theoretical 36 times as high hazard of developing symptoms within the observation period as had individuals with normal parameters. There was no significant covariation between HIV antigenemia on the one hand and CD4 count and response to PWM on the other. Although, the latter 2 variables covaried, each of them provided independent information, and both were used to classify the degree of the immunodeficiency in 3 stages: Im-0 with normal values, Im-1 with one, and Im-2 with both tests abnormal. Individuals in stage Im-2 had a 10 times increased risk of developing symptoms. The immunologic staging correlated significantly with the clinical grouping (CDC criteria). This staging improved in only 1, but deteriorated in half of 36 individuals observed for at least 18 months. Thus, the staging is likely to prove useful when attempts to arrest the immunodeficiency of HIV-infected individuals has to be monitored.