Aims Cardiac involvement is the main determinant of poor outcomes in sarcoidosis. Right ventricular (RV) dysfunction and left ventricular (LV) late gadolinium enhancement (LGE) have been reported to be predictive of adverse outcome in non-ischaemic cardiomyopathies. The aim of our study was to determine whether delayed RV LGE with cardiovascular magnetic resonance would be predictive of adverse events in addition to LV LGE during the long-term follow-up of pulmonary sarcoidosis patients. Methods and results Eighty-four consecutive biopsy-proven pulmonary sarcoidosis patients were followed for a median of 56 months [38-74] after baseline delayed contrast-enhanced cardiac magnetic resonance. The composite primary endpoint consisted of admission for congestive heart failure, sustained ventricular tachycardia, appropriate implantable cardioverter defibrillator therapy, pacemaker implantation for high degree atrio-ventricular block, or cardiac death. The composite secondary endpoint included all-cause mortality in addition to the primary endpoint. RV and LV LGE were demonstrated in respectively 12 and 27 patients. Five of 10 events included in the primary endpoint occurred in the group with RV LGE. RV LGE, LV, or biventricular LGE yielded Cox hazard ratios of 8.71 [95% confidence interval (CI) 1.90-23.81], 9.22 (95% CI 1.96-43.45), and 12.09 (95% CI 3.43-42.68) for the composite primary endpoint. In a multivariate model, the predictive value of biventricular LGE for the composite primary and secondary endpoints was strongest. Kaplan-Meier event-free survival curves were most significant for RV LGE and biventricular LGE (log rank with P < 0.001). Conclusions Biventricular LGE at presentation is the strongest, independent predictor of adverse outcome during long-term follow-up. Asymptomatic myocardial scar <8% of LV mass carried a favourable long-term outcome.