Background: We studied measures of human immunodeficiency virus (HIV) replication, the viral phenotype, and immune function (CD4 cell counts) and the relation of changes in these indicators to clinical outcomes in a subgroup of patients in a controlled trial of early antiretroviral treatment for HIV, the AIDS Clinical Trials Group Study 175. Methods: The 391 subjects, each of whom entered the study with a single screening CD4 cell count of 200 to 500 per cubic millimeter, were randomly as signed to receive zidovudine alone, didanosine alone, zidovudine plus didanosine, or zidovudine plus zalcitabine. Plasma concentrations of HIV RNA were assessed in 366 subjects, and viral isolates from 332 subjects were assayed for the presence of the syncytium inducing phenotype. Results: After eight weeks, the mean (±SE) decrease from base line in the concentration of HIV RNA, expressed as the change in the base 10 log of the number of copies per milliliter, was 0.26±0.06 for patients treated with zidovudine alone, 0.65±0.07 for didanosine alone, 0.93±0.10 for zidovudine plus didanosine, and 0.89±0.06 for zidovudine plus zalcitabine (P<0.001 for each of the pairwise comparisons with zidovudine alone). Multivariate proportional-hazards models showed that higher base-line concentrations of plasma HIV RNA, less suppression of plasma HIV RNA by treatment, and the presence of the syncytium-inducing phenotype were significantly associated with an increased risk of progression to the acquired immunodeficiency syndrome and death. After adjustment for these measures of viral replication and for the viral phenotype, CD4 cell counts were not significant predictors of clinical outcome. Conclusions: Both the risk of the progression of HIV disease and the efficacy of antiretroviral therapy are strongly associated with the plasma level of HIV RNA and with the viral phenotype. The changes in the plasma concentration of HIV RNA predict the changes in CD4 cell counts and survival after treatment with reverse- transcriptase inhibitors.