Reproducibility of cardiac magnetic resonance imaging in patients referred for the assessment of cardiac sarcoidosis; implications for clinical practice.

Cardiac sarcoidosis (CS) is an increasingly recognized condition, but cardiac magnetic resonance (CMR) image interpretation in these patients may be challenging as findings are often non-specific. The main objective of this study was to investigate the inter-reader agreement for the overall interpretation of CMR for the diagnosis of CS in an experienced reference center and investigate factors that may lead to discrepancies between readers. Consecutive patients undergoing CMR imaging to investigate for CS were included. CMR images were independently reviewed by two readers, blinded to all clinical, imaging and demographic information. The readers classified each scan as "consistent with cardiac sarcoidosis", "not consistent with cardiac sarcoidosis" or "indeterminate". Inter-reader agreement was assessed using κ-statistics. When there was disagreement on the overall interpretation, a third reader reviewed the images. Also, two readers independently commented on the presence of edema, presence of LGE (both ventricles) and quantified the extent of left ventricular LGE. 87 patients (43 women, mean age 54.3 ± 12.2 years) were included in the study. There was agreement regarding the overall interpretation in 72 of 87 (83%) CMR scans. The κ value was 0.64, indicating moderate agreement. There was similar moderate agreement in the interpretation of LGE parameters. In an experienced referral center, we found moderate agreement between readers in the interpretation of CMR in patients with suspected CS. Physicians should be aware of this inter-observer variability in interpretation of CMR studies in patients with suspected CS.