BACKGROUND: It is not known whether the benefits of sodium-glucose cotransporter 2 inhibitors in heart failure persist after years of therapy. METHODS: In the EMPEROR-Reduced (Empagliflozin Outcome Trials in Chronic Heart Failure With Reduced Ejection Fraction) and EMPEROR-Preserved (Empagliflozin Outcome Trials in Chronic Heart Failure With Preserved Ejection Fraction) trials, patients with heart failure were randomly assigned (double-blind) to placebo or empagliflozin 10 mg/day for a median of 16 and 26 months, respectively. At the end of the trials, 6799 patients (placebo 3381, empagliflozin 3418) were prospectively withdrawn from treatment in a blinded manner, and, of these, 3981 patients (placebo 2020, empagliflozin 1961) underwent prespecified in-person assessments after ≈30 days off treatment. RESULTS: From 90 days from the start of closeout to the end of double-blind treatment, the annualized risk of cardiovascular death or hospitalization for heart failure was lower in empagliflozin-treated patients than in placebo-treated patients (10.7 [95% CI, 9.0-12.6] versus 13.5 [95% CI, 11.5-15.6] events per 100 patient-years, respectively; hazard ratio 0.76 [95% CI, 0.60-0.96]). When the study drugs were withdrawn for ≈30 days, the annualized risk of cardiovascular death or hospitalization for heart failure increased in patients withdrawn from empagliflozin but not in those withdrawn from placebo (17.0 [95% CI, 12.6-22.1] versus 14.1 [95% CI, 10.1-18.8] events per 100 patient-years for empagliflozin and placebo, respectively). The hazard ratio for the change in risk in the patients withdrawn from empagliflozin was 1.75 (95% CI, 1.20-2.54), P=0.0034, whereas the change in the risk in patients withdrawn from placebo was not significant (hazard ratio 1.12 [95% CI, 0.76-1.66]); time period-by-treatment interaction, P=0.068. After withdrawal, the Kansas City Cardiomyopathy Questionnaire Clinical Summary Score declined by 1.6±0.4 in patients withdrawn from empagliflozin versus placebo (P<0.0001). Furthermore, withdrawal of empagliflozin was accompanied by increases in fasting glucose, body weight, systolic blood pressure, estimated glomerular filtration rate, N-terminal pro-hormone B-type natriuretic peptide, uric acid, and serum bicarbonate and decreases in hemoglobin and hematocrit (all P<0.01). These physiological and laboratory changes were the inverse of the effects of the drug seen at the start of the trials during the initiation of treatment (≈1-3 years earlier) in the same cohort of patients. CONCLUSIONS: These observations demonstrate a persistent effect of empagliflozin in patients with heart failure even after years of treatment, which dissipated rapidly after withdrawal of the drug. REGISTRATION: URL: https://www.clinicaltrials.gov; Unique identifiers: NCT03057977 and NCT03057951.