BACKGROUND: In patients with sarcoidosis, sudden death is a leading cause of mortality, which may represent unrecognized cardiac involvement. Delayed-enhancement cardiovascular magnetic resonance (DE-CMR) can detect minute amounts of myocardial damage. We sought to compare DE-CMR with standard clinical evaluation for the identification of cardiac involvement. METHODS AND RESULTS: Eighty-one consecutive patients with biopsy-proven extracardiac sarcoidosis were prospectively recruited for a parallel and masked comparison of cardiac involvement between (1) DE-CMR and (2) standard clinical evaluation with the use of consensus criteria (modified Japanese Ministry of Health [JMH] guidelines). Standard evaluation included 12-lead ECG and at least 1 dedicated non-CMR cardiac study (echocardiography, radionuclide scintigraphy, or cardiac catheterization). Patients were followed for 21+/-8 months for major adverse events (death, defibrillator shock, or pacemaker requirement). Patients were predominantly middle-aged (46+/-11 years), female (62%), and black (73%) and had chronic sarcoidosis (median, 7 years) and preserved left ventricular ejection fraction (median, 56%). DE-CMR identified cardiac involvement in 21 patients (26%) and JMH criteria in 10 (12%, 8 overlapping), a >2-fold higher rate for DE-CMR (P=0.005). All patients with myocardial damage on DE-CMR had coronary disease excluded by x-ray angiography. Pathology evaluation in 15 patients (19%) identified 4 with cardiac sarcoidosis; all 4 were positive by DE-CMR, whereas 2 were JMH positive. On follow-up, 8 had adverse events, including 5 cardiac deaths. Patients with myocardial damage on DE-CMR had a 9-fold higher rate of adverse events and an 11.5-fold higher rate of cardiac death than patients without damage. CONCLUSIONS: In patients with sarcoidosis, DE-CMR is more than twice as sensitive for cardiac involvement as current consensus criteria. Myocardial damage detected by DE-CMR appears to be associated with future adverse events including cardiac death, but events were few, and this needs confirmation in a larger cohort.