Cardiac magnetic resonance-based prediction of ventricular arrhythmic events in non-ischemic cardiomyopathy: derivation and external validation in the international multicenter IMPROVE-NICM Study

Background Risk stratification for sudden cardiac death (SCD) in non-ischemic cardiomyopathy (NICM) remains suboptimal. Current guidelines recommend primary prevention implantable cardioverter-defibrillator (ICD) implantation in patients with LVEF <35% (NYHA II–III) (1); however, its limited predictive accuracy (2,3) underscores the need for improved risk markers. Cardiac magnetic resonance (CMR) with late gadolinium enhancement (LGE) has emerged as a key tool for SCD risk assessment, with LGE presence, extent, and distribution providing incremental prognostic value (4-6). Advanced LGE-CMR post-processing enables detailed fibrosis characterization, including core scar, border zone (BZ), and BZ channels. Purpose To evaluate whether scar composition predicts ventricular arrhythmias (VAs) beyond conventional risk factors. We hypothesized that the composite endpoint of SCD, sustained VAs, or appropriate ICD therapy would be more frequent in patients with a higher number and greater mass of BZ channels. Methods We retrospectively analyzed a prospectively collected cohort of NICM patients undergoing LGE-CMR, irrespective of baseline LVEF. The date of first LGE-CMR served as the index date. Patients were allocated to a derivation cohort (n=495) and an external validation cohort (n=486). Fibrosis quantification included total scar, core, BZ, and BZ channel analysis. Patients were followed until the composite arrhythmic endpoint, death, or study end. NICM etiologies included dilated cardiomyopathy, non-dilated left ventricular cardiomyopathy, and myocarditis. Exclusion criteria were: (i) prior VAs; (ii) alternative structural heart disease; and (iii) inadequate CMR image quality for post-processing. Results In the derivation cohort (age 55.8±15.1 years; 67.8% male) and validation cohort, median follow-up was 25.7 months and 31.0 months, respectively. The composite endpoint occurred in 37 (7.5%) and 30 (6.5%) patients. In both cohorts, total scar, BZ mass, and core mass were significantly greater in patients experiencing events (all p<0.001). Presence of BZ channel mass >0.8 g was the strongest predictor of sustained VAs (AUC 0.83, 95% CI 0.73–0.90), outperforming symptomatic LV dysfunction (AUC 0.54, 95% CI 0.45–0.62) and septal LGE (AUC 0.69, 95% CI 0.60–0.78). Patients with BZ channel mass >0.8 g had a markedly increased risk of events (HR 19.11, 95% CI 7.38–49.46; p<0.0001). This association remained directionally consistent in the validation cohort (HR 48.28, 95% CI 6.52–357.39). Addition of BZ channel mass >0.8 g to a model including symptomatic LV dysfunction significantly improved risk discrimination (net reclassification improvement 0.55, 95% CI 0.35–0.75; p<0.001). Conclusions Among patients with NICM, scar composition, particularly BZ channel mass, shows significant, independent prognostic value for sustained VAs occurrence, outperforming existing risk stratification tools.Cumulative hazard curves for VA/SCD.