An individual treatment effect approach to predict response to mineralocorticoid receptor antagonists in patients with heart failure and reduced ejection fraction.

Abstract

AIMS Mineralocorticoid receptor antagonists (MRAs) are often underused in patients with heart failure (HF) and reduced ejection fraction (HFrEF). Individual treatment effect (ITE) may assist physicians in making timely decisions about which patients are the best suited for personalized therapy. We aimed at developing and validating a model to estimate ITE of MRAs in patients with HFrEF. METHODS AND RESULTS RALES and EMPHASIS-HF trials were the derivation trials used to estimate ITE of MRAs versus placebo on cardiovascular death or HF hospitalization in HFrEF over a 2-year period using counterfactual random forest method. ITE prediction models were built using linear regression and applied to the EPHESUS trial in patients with left ventricular systolic dysfunction and/or HF after myocardial infarction. In the RALES and EMPHASIS-HF trials (n = 3887), age, body weight, blood pressure, heart rate, hypertension and diabetes prevalence, stroke history, left ventricular ejection fraction, renal function, and serum sodium and potassium concentrations were identified to determine ITE scores (adjusted R2 = 0.25). As ITE scores increased, hazard ratio for treatment effect decreased from 0.82 (95% confidence interval [CI] 0.67-1.02) at ITE score 5 to 0.47 (95% CI 0.35-0.63) at ITE score 20 (p for interaction = 0.014). In the EPHESUS trial (n = 6472), a similar pattern was observed, with greater treatment effects in patients with higher ITE scores (p for interaction = 0.007). CONCLUSIONS In HFrEF across various clinical settings, our simple ITE model predicted individual responses to MRA therapy. Although treatment effects may be attenuated at lower ITE scores, point estimates with wide CIs still generally favour benefit, suggesting that these patients still benefit.