Erythropoietin responsiveness and its prognostic association according to sodium–glucose cotransporter 2 inhibitor use in patients with heart failure

Background Erythropoietin (EPO) responsiveness to endogenous EPO has been proposed as a key determinant of anaemia and clinical outcomes in patients with heart failure (HF). Although sodium–glucose cotransporter 2 inhibitors (SGLT2i) have been reported to enhance EPO production, it remains unclear whether SGLT2i use modifies EPO responsiveness and affects its prognostic significance in anaemic patients with HF. Purpose We sought to investigate the association between SGLT2i use and EPO responsiveness and to assess whether the prognostic relevance of EPO responsiveness differs in patients treated with and without SGLT2i. Methods In this multicentre observational study, 1084 anaemic patients with chronic HF were enrolled from 29 centres between January 2020 and August 2025 (median age 80 years; 44% female) irrespective of left ventricular ejection fraction (LVEF). Patients were stratified according to SGLT2i use (non-use, n = 777; use, n = 307). EPO responsiveness was assessed using the observed-to-predicted EPO (O/P) ratio and categorised, based on previously established criteria, as EPO-normal (0.916–1.087), EPO-deficient (< 0.916), or EPO-resistant (> 1.087). The primary outcome was a composite of all-cause death or hospitalisation for worsening HF. Results Patients treated with SGLT2i had a lower prevalence of EPO-deficiency (67% vs 76%, p = 0.001) and a higher prevalence of EPO-normal (21% vs 15%, p = 0.017) than those not receiving SGLT2i (Figure 1). In multivariable linear regression analyses, SGLT2i use was independently associated with a higher EPO O/P ratio (β = 0.035, p = 0.015) after adjustment for covariates related to the severity of HF and anaemia. During a median follow-up period of 776 days (IQR 459–1242), among patients not treated with SGLT2i, the incidence of the primary outcome was highest in the EPO-resistant group compared with the EPO-normal and EPO-deficient groups (33.3% vs 17.5% vs 19.4%, respectively; Figure 2A). In contrast, among patients treated with SGLT2i, the incidence of the primary outcome did not differ significantly among the EPO-normal, EPO-resistant, and EPO-deficient groups (18.2% vs 21.0% vs 19.4%, respectively; Figure 2B). In multivariable Cox proportional hazards analyses, EPO resistance was independently associated with the primary outcome among patients not treated with SGLT2i (hazard ratio [HR] 2.44, 95% confidence interval [CI] 1.30–4.59) adjusting for age, sex, body mass index, prior HF hospitalisation, blood urea nitrogen, haemoglobin, N-terminal pro–B-type natriuretic peptide, LVEF, and New York Heart Association functional class. However, this association was not observed among patients treated with SGLT2i (HR 0.97, 95% CI 0.35–2.71). Conclusions In patients with HF, SGLT2i use was associated with distinct patterns of EPO responsiveness. The prognostic association of EPO responsiveness differed according to SGLT2i use and was not apparent in patients treated with SGLT2i.