A cohort study of 214 human immunodeficiency virus (HIV)-infected patients was performed to assess the usefulness of the cytomegalovirus (CMV) antigenemia assay for predicting the occurrence of CMV disease and death. Multivariate analysis revealed that only positive baseline CMV antigenemia assays (relative risk [RR], 7.2; 95% confidence interval [CI], 3.7-14.2; P = .0001) and CD4 cell counts (RR, 0.98; 95% CI, 0.97-0.99; P = .009) were associated with CMV disease. A positive baseline CMV antigenemia assay was also associated with death by multivariate analysis (RR, 2.2; 95% CI, 1.5-3.4; P = .0003). Increasing levels of CMV antigenemia during follow-up were associated with increased risks of CMV disease and death. A positive CMV antigenemia assay that showed > 10 cells per 2 x 10(5) polymorphonuclear leukocytes during follow-up was 91% sensitive and 84% specific for predicting a diagnosis of CMV disease; the negative predictive value for this positive test was high (97%). Therefore, the CMV antigenemia assay appears to be a simple, rapid, and inexpensive test for predicting the occurrence of CMV disease and death in patients with advanced HIV infection.