Comparison of LGE quantitation methods in cardiac sarcoidosis to predict clinical outcomes

We sought to investigate the optimal method of quantifying late gadolinium enhancement (LGE) in cardiac sarcoidosis (CS) using cardiac magnetic resonance imaging (MRI). We retrospectively studied 53 patients with CS. LGE quantitation was performed using (a) semi-automated segmentation using Signal Threshold versus Reference Mean (STRM) cutoffs of > 2, > 3 and > 5 standard deviations (SD); (b) full-width-half-max (FWHM) method and (c) manual segmentation (MS) of affected myocardial segments. Primary outcome was a composite of cardiovascular death and ventricular tachyarrhythmia (VTA). A multivariate regression analysis was performed comparing the techniques adjusting for age, gender, NYHA class and LVEF. Mean age was 56.3 +/- 12 years, 71.6% males, 66% white. Mean LVEF was 45.1% +/- 14.7%. Over median follow-up of 28.1 months, 2 patients had cardiac death (3.7%) and 8 (15.1%) had VTA. On multivariate analysis, MS, > 2SD, > 3SD, > 5SD and FWHM had OR of 1.39 [CI 1.04-1.79], 1.09 [CI 0.99-1.21], 1.15 [CI 1.03-1.29], 1.16 [CI 1.04-1.27] and 1.08 [CI 0.96-1.21], respectively, for predicting the composite outcome. ROC curve analysis showed MS to have the highest AUC 0.89 followed by 0.81 for > 3SD and > 5SD, 0.75 for > 2SD and lowest 0.69 for FWHM method. Reproducibility was lower for manual method (ICC 0.7) than for > 3SD (ICC 0.991) and > 5SD (ICC 0.997). CS quantitation of LGE with MS or semi-automated quantitation with STRM > 3SD or > 5SD was significantly associated with the composite outcome of cardiac death and VTA. Semi-automated quantitation with STRM > 3SD provided the best combination of accuracy and reproducibility.