Pneumocystis carinii pneumonia remains one of the most common opportunistic infections in patients with acquired immune deficiency syndrome (AIDS). Treatment with either intravenous pentamidine or trimethoprim-sulfamethoxazole (TMP-SMX) is frequently complicated by serious adverse reactions. This study was a prospective, blinded comparison of 600 mg/d of pentamidine as an aerosol versus 15 mg/kg/d of trimethoprim plus 75 mg/kg/d of sulfamethoxazole for patients with mild or moderately severe P. carinii pneumonia (alveolar arterial oxygen difference of less than 55 mm Hg). Of 367 participants who were randomized to receive study therapies, 287 had proven and 16 had presumed Pneumocystis pneumonia. There were 29 deaths within 35 d of study initiation: 12 in the aerosolized pentamidine group and 17 in the TMP-SMX groups (log rank p = 0.28). The difference in mortality was 3.4% (95% CI = -3.5, 10.8%). Ninety-four patients treated with aerosolized pentamidine had to have their study therapy changed because of lack of efficacy, compared with 22 patients treated with TMP-SMX (p = 0.002). In addition PaO2 improved faster in patients treated with TMP-SMX. However, aerosolized pentamidine was discontinued less often than TMP-SMX because of toxicity (9.4 versus 40% p < 0.001). Rash (0.6 versus 14.9%), nausea and vomiting (1.7 versus 12.2%), and abnormalities of liver function tests (1.7 versus 12.2%) were the most common adverse effects necessitating treatment discontinuation. During 6-mo. follow-up there was no difference in mortality.(ABSTRACT TRUNCATED AT 250 WORDS)