Since February 2002, the Drug Resources Enhancement against AIDS and Malnutrition Program has provided highly active antiretroviral therapy (HAART) and immunologic and virologic monitoring free of charge. We conducted a cohort study of persons infected with human immunodeficiency virus in Mozambique. Only persons treated with HAART with available CD4 cell counts at baseline and ≥ 1 CD4 cell count after HAART were included. Survival analysis was applied to evaluate the prognostic value of CD4 cell counts measured at three months. Possible confounders were considered. A total of 753 persons who started HAART included; 59% were females. Median age was 34 years (range = 16-67 years), and the median CD4 cell count at baseline was 172 cells/mm3 (interquartile range = 87-261 cells/mm3, range = 0-1,322 cells/mm3). Overall, 105 persons (14%) died. Of these persons 54 (51%) developed AIDS before they died; 25 (3%) died during the first three months. After three months of therapy, the individual median CD4 cell count change from the baseline value was +101 cells/mm3 (interquartile range = +27 to +187 cells/mm3, range = -723 to +310 cells/mm3). A median CD4 increment of 100 cells/mm3 in three months was associated with a mortality reduction of 50% compared with an increase of < 50 cells (relative hazard of death adjusted for baseline CD4 cell count = 0.54, 95% confidence interval = 0.30-0.95). A good initial response to HAART was associated with a significant reduction of mortality. This finding supports the effectiveness of HAART in resource-poor settings.