IntroductionGuidelines indicate primary-prevention implantable cardioverter-defibrillators (ICDs) for most patients with left ventricular ejection fraction (LVEF) <= 35%. Some patients' LVEFs improve during the life of their first ICD. In patients with recovered LVEF who never received appropriate ICD therapy, the utility of generator replacement upon battery depletion remains unclear. Here, we evaluate ICD therapy based on LVEF at the time of generator change, to educate shared decision-making regarding whether to replace the depleted ICD. MethodsWe followed patients with a primary-prevention ICD who underwent generator change. Patients who received appropriate ICD therapy for ventricular tachycardia or ventricular fibrillation (VT/VF) before generator change were excluded. The primary endpoint was appropriate ICD therapy, adjusted for the competing risk of death. ResultsAmong 951 generator changes, 423 met inclusion criteria. During 3.4 +/- 2.2 years follow-up, 78 (18%) received appropriate therapy for VT/VF. Compared to patients with recovered LVEF > 35% (n = 161 [38%]), those with LVEF <= 35% (n = 262 [62%]) were more likely to require ICD therapy (p = .002; Fine-Gray adjusted 5-year event rates: 12.7% vs. 25.0%). Receiver operating characteristic analysis revealed the optimal LVEF cutoff for VT/VF prediction to be 45%, the use of which further improved risk stratification (p < .001), with Fine-Gray adjusted 5-year rates 6.2% versus 25.1%. ConclusionFollowing ICD generator change, patients with primary-prevention ICDs and recovered LVEF have significantly lower risk of subsequent ventricular arrhythmias compared to those with persistent LVEF depression. Risk stratification at LVEF 45% offers significant additional negative predictive value over a 35% cutoff, without a significant loss in sensitivity. These data may be useful during shared decision-making at the time of ICD generator battery depletion.