Targeting efficacy of spironolactone in patients with heart failure with preserved ejection fraction: the TOPCAT study

Abstract Aims We aimed to explore the heterogeneous treatment effects (HTEs) for spironolactone treatment in patients with Heart failure with preserved ejection fraction (HFpEF) and examine the efficacy and safety of spironolactone medication, ensuring a better individualized therapy. Methods and results We used the causal forest algorithm to discover the heterogeneous treatment effects (HTEs) from patients in the Treatment of Preserved Cardiac Function Heart Failure with an Aldosterone Antagonist (TOPCAT) trial. Cox regressions were performed to assess the hazard ratios (HRs) of spironolactone medication for cardiovascular death and drug discontinuation in each group. The causal forest model revealed three representative covariates and participants were partitioned into four subgroups which were Group 1 (baseline BMI ≤ 31.71 kg/m 2 and baseline ALP ≤ 80 U/L, n  = 759); Group 2 (BMI ≤ 31.71 kg/m 2 and ALP > 80 U/L, n  = 1088); Group 3 (BMI > 31.71 kg/m 2 , and WBC ≤ 6.6 cells/μL, n  = 633); Group 4 (BMI > 31.71 kg/m 2 and WBC > 6.6 cells/μL, n  = 832), respectively. In the four subgroups, spironolactone therapy reduced the risk of cardiovascular death in high‐risk group (Group 4) with both high BMI and WBC count (HR: 0.76; 95% CI 0.58 to 0.99; P  = 0.045) but increased the risk in low‐risk group (Group 1) with both low BMI and ALP (HR: 1.45; 95% CI 1.02 to 2.07; P  = 0.041; P for interaction = 0.020) but showed similar risk of drug discontinuation ( P for interaction = 0.498). Conclusion Our study manifested the HTEs of spironolactone in patients with HFpEF. Spironolactone treatment in HFpEF patients is feasible and effective in patients with high BMI and WBC while harmful in patients with low BMI and ALP. Machine learning model could be meaningful for improved categorization of patients with HFpEF, ensuring a better individualized therapy in the clinical setting.