BACKGROUND. Primary prophylaxis against Pneumocystis carinii pneumonia (PCP) is recommended for patients with human immunodeficiency virus (HIV) infection if their CD4 cell counts are below 200 per cubic millimeter (0.2 x 10(9) per liter). Either aerosolized pentamidine or trimethoprim-sulfamethoxazole (co-trimoxazole) is commonly prescribed for prophylaxis, but the relative efficacy and toxicity of these agents are unknown. METHODS. We conducted a multicenter trial involving 215 HIV-infected patients with no history of PCP but with CD4 cell counts below 200 per cubic millimeter. The patients were randomly assigned to one of three regimens: aerosolized pentamidine once a month, 480 mg of trimethoprim-sulfamethoxazole once a day (80 mg of trimethoprim and 400 mg of sulfamethoxazole), or 960 mg of trimethoprim-sulfamethoxazole once a day (160 mg and 800 mg, respectively). The cumulative incidence of PCP was estimated by Kaplan-Meier survival analysis. RESULTS. After a mean follow-up of 264 days, 6 of the 71 patients in the pentamidine group had a confirmed first episode of PCP (11 percent), whereas none of the 142 patients in the two trimethoprim-sulfamethoxazole groups had PCP (P = 0.002). However, adverse events that required discontinuation of the medication were much more frequent in the trimethoprim-sulfamethoxazole groups (17 and 18 patients) than in the pentamidine group (2 patients). The adverse reactions occurred significantly sooner in the group given 960 mg of trimethoprim-sulfamethoxazole than in the group given 480 mg (mean time, 16 vs. 57 days; P = 0.02). CONCLUSIONS. For patients with HIV infection, trimethoprim-sulfamethoxazole taken once a day is more effective as primary prophylaxis against PCP than aerosolized pentamidine administered once a month, although adverse drug reactions are more frequent with trimethoprim-sulfamethoxazole.