BACKGROUND. Cryptococcal meningitis (CM) remains a leading cause of acquired immunodeficiency syndrome-related death in sub-Saharan Africa. The timing of the initiation of antiretroviral therapy (ART) for human immunodeficiency virus (HIV)-associated CM remains uncertain. The study aimed to determine the optimal timing for initiation of ART in HIV-positive individuals with CM. METHODS. A prospective, open-label, randomized clinical trial was conducted at a tertiary teaching hospital in Zimbabwe. Participants were aged > or = 18 years, were ART naive, had received a first CM diagnosis, and were randomized to receive early ART (within 72 h after CM diagnosis) or delayed ART (after 10 weeks of treatment with fluconazole alone). Participants received 800 mg of fluconazole per day. The ART regimen used was stavudine, lamivudine, and nevirapine given twice daily. The duration of follow-up was up to 3 years. The primary end point was all-cause mortality. RESULTS. Fifty-four participants were enrolled in the study (28 in the early ART arm and 26 in the delayed ART arm). The median CD4 cell count at enrollment was 37 cells/mm(3) (interquartile range, 17-69 cells/mm(3)). The 3-year mortality rate differed significantly between the early and delayed ART groups (88% vs 54%; P < .006); the overall 3-year mortality rate was 73%. The median durations of survival were 28 days and 637 days in the early and delayed ART groups, respectively (P = .031, by log-rank test). The risk of mortality was almost 3 times as great in the early ART group versus the delayed ART group (adjusted hazard ratio, 2.85; 95% confidence interval, 1.1-7.23). The study was terminated early by the data safety monitoring committee. CONCLUSIONS. In resource-limited settings where CM management may be suboptimal, when compared with a delay of 10 weeks after a CM diagnosis, early initiation of ART results in increased mortality. Trial registration. ClinicalTrials.gov identifier: NCT00830856.