Optimal timing of antiretroviral therapy (ART) initiation for HIV-infected persons remains unclear. To assess survival benefit of initiating ART at different CD4+ cell counts. Prospective observational study. U.S. clinics in the HIV Outpatient Study (HOPS). HIV-infected patients with CD4+ cell counts, plasma HIV RNA viral load, and ART use recorded from January 1994 through March 2002. Before initiation of ART, patients were grouped by their CD4+ cell counts into three subgroups: 0.201 to 0.350 x 10(9) cells/L (n = 399), 0.351 to 0.500 x 10(9) cells/L (n = 327), and 0.501 to 0.750 x 10(9) cells/L (n = 122). We compared mortality rates for each CD4+ subgroup among patients who initiated ART and patients who delayed ART until reaching a lower CD4+ subgroup. Mortality rates for 340 patients who initiated ART and 59 who delayed ART in the CD4+ subgroup of 0.201 to 0.350 x 10(9) cells/L were 15.4 and 56.4 deaths per 1000 person-years, respectively (rate ratio, 0.27 [95% CI, 0.14 to 0.55]; P < 0.001). For the CD4+ subgroup of 0.351 to 0.500 x 10(9) cells/L, mortality rates for 240 patients who initiated ART and 887 who delayed ART were 10.0 and 16.6 deaths per 1000 person-years, respectively (rate ratio, 0.61 [CI, 0.22 to 1.67]; P = 0.17). For the CD4+ subgroup of 0.501 to 0.750 x 10(9) cells/L, mortality rates in 55 patients who initiated ART and 67 who delayed ART were 7.5 and 3.1 deaths per 1000 person-years, respectively (rate ratio, 1.20 [CI, 0.17 to 8.53]; P > 0.2). Patients in the 0.201 to 0.350 x 10(9) cells/L and 0.351 to 0.500 x 10(9) cells/L CD4+ subgroups who initiated ART were more likely than those who delayed ART to achieve an undetectable HIV viral load (P = 0.03 and 0.04, respectively). Among HIV-infected persons with CD4+ cell counts of 0.201 to 0.350 x 10(9) cells/L, initiating ART is associated with reduced mortality compared with delaying such therapy. Survival benefits of earlier ART initiation (at CD4+ cell counts of 0.351 to 0.500 x 10(9) cells/L) are possible.