PURPOSE: To compare the efficacy and safety of three different doses of prophylactic aerosol pentamidine in patients with one prior episode of Pneumocystis carinii pneumonia (PCP) and the acquired immunodeficiency syndrome. PATIENTS AND METHODS: The design of the study was a double-blind, randomized, dose-comparison clinical trial conducted at 13 medical centers within the United States. In stage I of the trial, patients were randomized to receive either 5 mg, 60 mg, or 120 mg of aerosol pentamidine delivered biweekly with the Fisoneb (Fisons, Inc., Rochester, New York) ultrasonic nebulizer. After 24 weeks of therapy, patients entered stage II of the trial, where the 5-mg group was re-randomized to either the 60-mg or 120-mg group. RESULTS: One hundred seventy-five patients entered stage I of the trial and received prophylaxis for a mean of 123.6 days. Seven assigned to the 5-mg biweekly dosing schedule had a confirmed recurrence of PCP, compared with none in the 60-mg group (p = 0.007) and three in the 120-mg group (p = 0.304). During stage II of the trial, eight patients in the 60-mg group and one additional patient in the 120-mg group had recurrent PCP. After 52 weeks of observation, the likelihood of being PCP-free was 88.0% in the 60-mg group and 93% in the 120-mg group (p = 0.712). Minor adverse events related to aerosol pentamidine administration included cough, taste perversion, chest pain, bronchospasm, and dyspnea. These side effects were more common in the 60-mg and 120-mg treatment groups and resulted in withdrawal from the study by one patient. Serious events were more common after 24 weeks of therapy and included asymptomatic hypoglycemia (five), pancreatitis (two), pneumothorax (one), and extrapulmonary pneumocystosis (one). CONCLUSIONS: These results demonstrate that biweekly administration of 60 mg or 120 mg of aerosol pentamidine significantly decreases PCP recurrence when compared with a 5-mg regimen or findings in historic controls and is generally well tolerated. There is no significant difference in effect or safety between these two dosing regimens in patients followed for at least 52 weeks of therapy.