The possibility that an agent in addition to human immunodeficiency virus type 1 may be involved in the etiology of Kaposi's sarcoma in acquired immunodeficiency syndrome (AIDS) patients was investigated between 1984 and 1992 in this nested case-control analysis from the Multicenter AIDS Cohort Study (MACS) of homosexual and bisexual men. A total of 316 cases of Kaposi's sarcoma were identified and compared with 510 participants with AIDS and no evidence of cancer. More of the Kaposi's sarcoma cases were from Los Angeles and used a higher number of recreational drugs. The Kaposi's sarcoma cases were also more active sexually. There was a dose-response relation between Kaposi's sarcoma and the number of sexual partners, with an odds ratio of 2 between the most and least sexually active subgroups. The odds ratio for Kaposi's sarcoma increased to 4.18 (95% confidence interval 1.29-14.1) in the presence of a history of five infections. Hepatitis and gonorrhea contributed the most to this relation. The various observed odds ratios did not change after multivariate adjustment for the other risk factors. A model was developed combining all predictive associations into a composite risk score ranging from one to 12 and based on history of infections, sexual activity, use of poppers/nitrites, and having had sexual partners from the West Coast of the United States. The subgroup with the highest scores, compared to the subgroup with the lowest score, had an odds ratio of 8.93 (95% confidence interval 3.21-30.44) for Kaposi's sarcoma. A longitudinal proportional hazards analysis among all 2,190 human immunodeficiency virus type 1-seroprevalent men at study entry, based on this risk score and CD4 cells at baseline, confirmed these findings. Identifying these specific subgroups that are at high and low risk for Kaposi's sarcoma will help future investigations to be more focused in their search for an additional etiologic factor for Kaposi's sarcoma in AIDS.