To evaluate whether cardiac magnetic resonance (CMR)-based parametric mapping and strain analysis can improve the risk-stratification for ventricular arrhythmias (VA) and sudden death (SD) in non-ischaemic cardiomyopathy (NICM). Secondary analysis of a prospective single-centre-registry (NCT02326324), including 703 consecutive NICM patients, 618 with extracellular volume (ECV) available. The combined primary endpoint included appropriate implantable cardioverter defibrillator therapies, sustained ventricular tachycardia, resuscitated cardiac arrest and SD. During a median follow-up of 21 months, 14 patients (2%) experienced the primary endpoint. Native T1 was not associated with the primary endpoint. Left ventricular global longitudinal strain lost its significant association after adjustment for left ventricular ejection fraction (LVEF). Among patients with ECV available, 11 (2%) reached the primary endpoint. Mean ECV was significantly associated with the primary endpoint and the best cut-off was 30%. ECV ≥ 30% was the strongest independent predictor of the primary endpoint (hazard ratio 14.1, P = 0.01) after adjustment for late gadolinium enhancement (LGE) and LVEF. ECV ≥ 30% discriminated the arrhythmic risk among LGE+ cases and among those with LVEF ≤ 35%. A simple clinical risk-stratification model, based on LGE, LVEF ≤ 35% and ECV ≥ 30%, achieved an excellent predictive ability (Harrell's C 0.82) and reclassified the risk of 32% of the study population as compared to LVEF ≤ 35% alone. Comprehensive CMR evaluation in NICM showed that ECV was the only parameter with an independent and strong predictive value for VA/SD, on top of LGE and LVEF. A risk-stratification model based on LGE, LVEF ≤ 35% and ECV ≥ 30% achieved an excellent predictive ability for VA/SD. UHSM CMR study (NCT02326324) https://clinicaltrials.gov/ct2/show/NCT02326324.