Cardiac involvement is an important prognostic factor in sarcoidosis, but reliable indicators of mortality risk in cardiac sarcoidosis are unstudied in a large number of patients. To determine the significant predictors of mortality and to assess the efficacy of corticosteroids, we analyzed clinical findings, treatment, and prognosis in 95 Japanese patients with cardiac sarcoidosis. Twenty of these 95 patients had cardiac sarcoidosis proven by autopsy; none of these patients had received corticosteroids. We assessed 12 clinical variables as possible predictors of mortality by Cox proportional hazards model in 75 steroid-treated patients. During the mean follow-up of 68 months, 29 patients (73%) died of congestive heart failure and 11 (27%) experienced sudden death. Kaplan-Meier survival curves showed 5-year survival rates of 75% in the steroid-treated patients and of 89% in patients with a left ventricular ejection fraction > or = 50%, whereas there was only 10% 5-year survival rate in autopsy subjects. There was no significant difference in survival curves of patients treated with a high initial dose (> 30 mg) and a low initial dose (> or = 30 mg) of prednisone. Multivariate analysis identified New York Heart Association functional class (hazard ratio 7.72 per class I increase, p = 0.0008), left ventricular end-diastolic diameter (hazard ratio 2.60/10 mm increase, p = 0.02), and sustained ventricular tachycardia (hazard ratio 7.20, p = 0.03) as independent predictors of mortality. In conclusion, the severity of heart failure was one of the most significant independent predictors of mortality for cardiac sarcoidosis. Starting corticosteroids before the occurrence of systolic dysfunction resulted in an excellent clinical outcome. A high initial dose of prednisone may not be essential for treatment of cardiac sarcoidosis.