We randomized 220 HIV-1-infected subjects to receive aerosolized pentamidine (300 mg/4 weeks) or orally trimethoprim-sulfamethoxazole (320-1600 mg/day) for primary prophylaxis of Pneumocystis carinii pneumonia (PCP), and evaluated PCP and toxoplasmic encephalitis (TE) occurrence and survival. Patients developing toxicity switched to the other regimen. Analysis was on intention-to-treat. At 1 year of study, we observed in the pentamidine group a non-significant excess of PCP (4 vs. 1) and TE (7 vs. 3), and a significant increased death rate (15 vs. 2). After 2 years, no significant differences were observed: adjusted RR estimates for pentamidine vs. cotrimoxazole were 1.20 (95% CI, 0.33-4.37) for PCP (6 cases vs. 5), 1.23 (95% CI, 0.46-3.29) for TE (10 vs. 8) and 1.52 (95% CI, 0.83-2.79) for death (30 vs. 18). Crossovers were more frequent in the cotrimoxazole group (41 vs. 4, P < 0.001). Aerosolized pentamidine and cotrimoxazole were equally effective in preventing PCP, and no major differences were observed in TE occurrence and survival after 2 years follow-up.