A multicenter placebo-controlled trial of early short-term high-dose methylprednisolone enrolled 78 patients with moderate to severe Pneumocystis carinii pneumonia (PCP) complicating HIV infection. The mean pressure of oxygen (PO2) at study entry was 55 mm Hg for the 71 patients who had blood gases monitored while breathing room air. Patients were randomized to receive methylprednisolone (40 mg) or placebo parenterally twice daily for 10 days, and the first dose of study medication was given within 24 h of the first dose of antimicrobial therapy for PCP. The primary end point included death, need for mechanical ventilation for > 6 days, or a partial PO2 < 70 mm Hg while breathing room air 10 days after initiation of treatment. There was no statistically significant difference in the primary end point between patients randomized to corticosteroid (CS) or placebo (PL) (p = 0.522; 95% CI = -0.30, 0.16). The incidence of superinfections during therapy or of other HIV-associated infections or malignancies in the 6 months following treatment for PCP was not significantly different between the two groups. More patients randomized to placebo had to discontinue treatment with trimethoprim-sulfamethoxazole because of hypersensitivity than those randomized to corticosteroids (p = 0.039). We conclude that addition of corticosteroids does not significantly affect the outcome of PCP in patients with HIV and a PO2 < 70 mm Hg on room air at presentation but lowers the incidence of hypersensitivity reactions to trimethoprim-sulfamethoxazole.