An epidemiologic study was initiated in 1987 to evaluate the long-term safety and efficacy of zidovudine in patients with advanced human immunodeficiency virus disease. Data from 886 patients with the acquired immunodeficiency syndrome (AIDS) or AIDS-related complex and CD4+ lymphocyte count less than 0.25 x 10(9)/L are reported. Eighteen-month survival was 67% for the cohort. Pretreatment factors associated with increased survival time included index diagnosis of AIDS-related complex, hematocrit of 0.35 or greater, CD4+ lymphocyte count of 0.15 x 10(9)/L or greater, high functional status, and time from diagnosis of AIDS to treatment of less than 60 days. By proportional hazards analysis, development of serious anemia was the most significant factor associated with early death. Receiving zidovudine for a high proportion of time significantly improved chances of survival even if anemia developed. Serious leukopenia occurring in 37% and serious anemia occurring in 32% of patients. Nonhematologic adverse events were uncommon and no previously unreported adverse events were seen.