BACKGROUND: Left ventricular ejection fraction (LVEF) is used to select patients for primary prevention implantable cardioverter defibrillators (ICDs). The relationship between baseline and long-term follow-up LVEF and clinical outcomes among primary prevention ICD patients remains unclear. METHODS: We studied 195 patients with a baseline LVEF ≤35% ≤6 months prior to ICD implantation and follow-up LVEF 1-3 years after ICD implantation without intervening left ventricular assist device (LVAD) or transplant. The co-primary study endpoints were: (1) a composite of time to death, LVAD, or transplant and (2) appropriate ICD therapy. We examined multivariable Cox proportional hazard models with a 3-year post-implant landmark view; the LVEF closest to the 3-year mark was considered the follow-up LVEF for analyses. Follow-up LVEF was examined using 2 definitions: (1) ≥10% improvement compared to baseline or (2) actual value of ≥40%. RESULTS: Fifty patients (26%) had a LVEF improvement of ≥10% and 44 (23%) had a follow-up LVEF ≥40%. Neither baseline nor follow-up LVEF was significantly associated with the composite endpoint. In contrast, both baseline and follow-up LVEF were associated with risk for long-term ICD therapies, whether follow-up LVEF was modeled as a ≥10% absolute improvement (baseline LVEF HR 0.87, CI 0.91-0.93, P < .001; follow-up LVEF HR 0.18, CI 0.06-0.53, P = .002) or a ≥40% follow-up value (baseline LVEF HR 0.89, CI 0.83-0.96, P = .001, follow-up LVEF HR 0.26, CI 0.08-0.87, P = .03). CONCLUSIONS: Among primary prevention ICD recipients, both baseline and follow-up LVEF were independently associated with long-term risk for appropriate ICD therapy, but they were not associated with time to the composite of LVAD, transplant, or death.