Finerenone vs spironolactone in heart failure with mildly reduced or preserved ejection fraction: a real-world study

Abstract

Aims

Finerenone has shown benefit in chronic kidney disease, type 2 diabetes, and heart failure with mildly reduced or preserved ejection fraction (HFmrEF/HFpEF), but head-to-head data against spironolactone are limited. We compared 12-month effectiveness and safety after finerenone vs spironolactone initiation in heart failure with left ventricular ejection fraction ≥40%.

Methods and results

We conducted a retrospective active-comparator, new-user cohort study using the TriNetX Network from 9 July 2021 to 31 March 2025. Adults of both sexes with HFmrEF/HFpEF initiating finerenone or spironolactone were matched 1:1 by propensity score. The primary outcome was all-cause mortality or heart failure exacerbation. Secondary outcomes included all-cause mortality, heart failure exacerbation, all-cause hospitalization, and acute myocardial infarction; safety outcomes were hyperkalaemia and acute kidney injury. After matching, 1417 patients were included in each group. Finerenone initiation was associated with lower risk of the primary outcome (32.3% vs 41.7%; adjusted hazard ratio [aHR] 0.72, 95% confidence interval [CI] 0.64–0.82; P < .001). Lower risks were also observed for all-cause mortality (aHR 0.67, 95% CI 0.51–0.87), heart failure exacerbation (aHR 0.73, 95% CI 0.64–0.83), and all-cause hospitalization (aHR 0.83, 95% CI 0.76–0.91), but not acute myocardial infarction. Hyperkalaemia was less frequent with finerenone (17.9% vs. 22.8%; risk ratio 0.79, 95% CI 0.68–0.91); acute kidney injury showed only a borderline difference.

Conclusions

Finerenone initiation was associated with lower 12-month clinical event and hyperkalaemia risks than spironolactone initiation. These observational associations should be considered hypothesis-generating and require confirmation in randomized head-to-head trials.

Graphical Abstract

Graphical Abstract.

Introduction

Heart failure with mildly reduced ejection fraction (HFmrEF) and preserved ejection fraction (HFpEF) are the major heart failure (HF) phenotypes in clinical practice.^1,2 Both conditions are associated with substantial morbidity and recurrent hospitalizations, reflecting biological heterogeneity, comorbidities, and a limited evidence base for pharmacological therapy.²

HFmrEF and HFpEF are characterized by increased myocardial stiffness and impaired diastolic relaxation driven by chronic inflammation, oxidative stress, and myocardial fibrosis, which reduce ventricular compliance and increase filling pressures.³ Excess aldosterone and mineralocorticoid receptor (MR) overactivation exacerbate these processes, providing a mechanistic rationale for MR antagonism in HFmrEF/HFpEF.^4,5

The TOPCAT trial, which evaluated spironolactone, a steroidal MR antagonist (MRA), in HFpEF did not demonstrate a significant reduction in its primary composite outcome of cardiovascular death, aborted cardiac arrest, or hospitalization for heart failure, despite a 17% reduction in HF hospitalization (95% CI 0.69–0.99; P = .04).⁶ Treatment benefit was largely confined to participants in the Americas, raising concerns regarding population-specific confounding.^7,8 More recently, the SPIRIT-HF trial found no significant benefit of spironolactone on the composite of cardiovascular death or heart failure hospitalization in patients with heart failure and an ejection fraction ≥40%, although it was underpowered owing to pandemic-related discontinuation.⁹

Finerenone is a selective, non-steroidal MRA with pharmacological properties distinct from those of steroidal agents, including higher receptor selectivity, balanced cardiac and renal tissue distribution, and differential modulation of MR cofactor recruitment.¹⁰ In randomized controlled trials involving patients with chronic kidney disease (CKD) and type 2 diabetes mellitus (T2DM), finerenone significantly reduced cardiovascular and renal events.^11,12 More recently, the FINEARTS-HF trial demonstrated that finerenone reduced the composite endpoint of cardiovascular death or worsening HF in patients with HFmrEF and HFpEF.¹³ However, no randomized trial has directly compared the efficacy and safety of finerenone and spironolactone in this population.

To address this gap, we used electronic health record data from the TriNetX platform to compare cardiovascular outcomes and safety between finerenone and spironolactone initiators with HFmrEF or HFpEF in clinical practice.

Methods

Data source

This retrospective cohort study used electronic health record data from the TriNetX platform, a federated research network aggregating data from participating healthcare organizations, primarily academic medical centres, affiliated hospitals, and outpatient clinics. Data included demographics; diagnoses (ICD-10-CM); medications (RxNorm, Veterans Affairs Drug Classification, Anatomical Therapeutic Chemical system); laboratory results (LOINC); and healthcare utilization records. Analyses used the Global Collaborative Network within TriNetX, comprising data from more than 170 healthcare organizations and over 190 million patients. Data remain locally stored at participating institutions and are queried through a privacy-preserving analytics interface that returns only aggregated outputs.

Ethics statement

The study protocol (TMU-JIRB No. N202605005) was reviewed by the Taipei Medical University-Joint Institutional Review Board and designated as exempt from full review (meeting #115-05-4, 21 May 2026). The research was conducted in accordance with the Declaration of Helsinki. As de-identified TriNetX data were analysed, individual informed consent requirements were waived. Reporting follows the STROBE guideline. This study was conducted in accordance with the minimum standards of the CODE-EHR best-practice framework for the use of structured electronic health-care records in clinical research.

Study design and population

This was an active-comparator, new-user retrospective cohort study comparing initiation of finerenone vs spironolactone in patients with HFmrEF or HFpEF. The study period was 9 July 2021 to 31 March 2025. The index date was the first finerenone or spironolactone prescription during this period. Adults ≥18 years with HFmrEF or HFpEF were included. HFmrEF/HFpEF were identified using ICD-10-CM codes: a heart-failure diagnosis (I50) with a recorded echocardiographic left ventricular ejection fraction (LVEF) ≥40%, or a diastolic heart-failure code (I50.3).

To implement an incident user design, patients with documented MRA use within 30 days before the index date were excluded; this washout was considered sufficient to mitigate residual treatment effects and was consistent with prior MRA pharmacoepidemiology studies.^14,15 Additional exclusions: acute myocardial infarction (AMI), stroke, or cardiac arrest within 30 days before index; hospitalization for any cause or acute HF within 30 days before index; primary adrenal insufficiency; baseline serum potassium ≥5.0 mmol/L; estimated glomerular filtration rate (eGFR) ≤25 mL/min/1.73 m²; or concomitant strong CYP3A4 inhibitor use at baseline (Supplementary Table S1, Supplementary Figure S1).

Covariates

Baseline covariates were defined a priori and ascertained within 1 year before the index date. Covariates included demographics (including sex and race/ethnicity), comorbidities, concomitant medications, and laboratory values (Supplementary Table S2). Sex was pre-specified as a matching covariate and as a subgroup variable for the primary composite outcome; gender identity data were not available in the TriNetX platform. Race and ethnicity were classified using self-reported categories recorded by each participating institution (White, Black or African American, Asian, other race; and Hispanic/Latino vs. not Hispanic/Latino).

Outcomes and follow-up

The prespecified primary outcome was a composite of all-cause mortality or HF exacerbation, whichever occurred first. HF exacerbation was defined as a new encounter during follow-up carrying a diagnosis of acute or acute-on-chronic (systolic, diastolic, or combined) heart failure or acute pulmonary oedema, capturing acute decompensation (ICD codes I50.21, I50.23, I50.31, I50.33, I50.41, I50.43, J81). Secondary outcomes were all-cause mortality, HF exacerbation, all-cause hospitalization, and AMI. AMI was prespecified as a hard, coding-robust endpoint because non-fatal myocardial infarction is a component of the cardiovascular composite in the pivotal finerenone cardiorenal trials.^11,12 Safety outcomes were acute kidney injury (AKI) and hyperkalaemia. ICD codes for each outcome are listed in Supplementary Table S3. Follow-up began one day after the index date and continued for up to 12 months until an outcome of interest, death, loss to follow-up, or end of the study period, whichever occurred first.

Statistics and reproducibility

Categorical variables are reported as counts (%); continuous variables as means ± standard deviation. Propensity score matching (PSM) was used to create comparable cohorts. Propensity scores were estimated using logistic regression and matched 1:1 by greedy nearest-neighbour algorithm with a caliper of 0.1 pooled standard deviations. Covariate balance was assessed using standardized mean differences (SMDs), with values <0.1 indicating adequate balance.¹⁶ The primary and all secondary outcomes were assessed within a single propensity-matched cohort. Each subgroup or sensitivity analysis was constructed as a separate query with relevant restrictions due to the TriNetX platform limitation. Matched sample sizes therefore vary.

Cox proportional hazards (PH) models estimated hazard ratios (HRs) for clinical outcomes on the TriNetX platform under an intention-to-treat framework. The PH assumption was evaluated using the generalized Schoenfeld method. Event-free survival was estimated by Kaplan-Meier and compared using log-rank tests. Absolute risk reduction (ARR) and number needed to treat (NNT) were calculated for key outcomes. E-values were estimated to evaluate the influence of unmeasured confounding.¹⁷ Hyperkalaemia and AKI were categorized into three binary thresholds (≥1, ≥2, ≥6 episodes); between-group differences were assessed by Pearson chi-square test with risk ratios (RRs) and 95% confidence intervals (CIs) for finerenone relative to spironolactone.

Primary cohort construction, matching, Cox models, Kaplan-Meier analyses, and platform-based sensitivity analyses were performed in TriNetX. Aggregate-data supplementary analyses and figure generation were performed externally using Python. Statistical significance was two-sided P < .05. Figures were generated using Python 3.12.12 with Matplotlib 3.10.7, Numpy 1.26.4, and Pandas 2.3.3.

Subgroup, sensitivity, and supplementary analyses

Three supplementary aggregate-data analyses assessed robustness of the main Cox findings: (i) Monte Carlo probabilistic quantitative bias analysis for unmeasured frailty; (ii) restricted mean survival time (RMST) through 365 days, by numerical integration of TriNetX exported Kaplan-Meier survival probabilities; and (iii) an exploratory competing-mortality cumulative-incidence approximation for nonfatal outcomes (Supplementary Methods).

Prespecified subgroup analyses examined the primary composite outcome by age, sex, T2DM status, baseline renal function, LVEF category, and SGLT2 inhibitor use; these also evaluated the impact of residual post-matching imbalances in eGFR and LVEF on the consistency of findings. Sensitivity analyses included: time-stratified intervals; stepwise covariate adjustment; an adherence-restricted analysis requiring a second prescription; restriction by first documented HF diagnosis within prespecified time windows; and dose-restricted analyses limited to finerenone 10 mg and spironolactone 25 mg, ensuring a more comparable treatment contrast and reducing confounding from non-standard dosing, early titration, or drug intolerance. Osteoporosis, dermatitis, and eczema, which are not plausibly affected by either drug, were used as negative control outcomes to detect residual or unmeasured confounding.¹⁸

Results

Baseline characteristics

Between 9 July 2021 and 31 March 2025, 256 079 adults with HFmrEF or HFpEF initiating finerenone or spironolactone were identified. After exclusions, 127 198 patients remained (1428 finerenone; 125 770 spironolactone). After 1:1 PSM, 1417 patients were included in each group (Supplementary Figure S1). Prior to matching, substantial imbalances were observed for T2DM (79.7% vs 41.6%; SMD 0.847), CKD (71.3% vs 32.5%; SMD 0.844), SGLT2 inhibitor use (49.8% vs 18.7%; SMD 0.694), GLP-1 receptor agonist use (30.3% vs 6.9%; SMD 0.630), and eGFR (43.3 ± 20.7 vs 63.2 ± 27.8 mL/min/1.73 m²; SMD 0.817). After matching, cohorts were well balanced (most SMDs <0.1), except eGFR (0.102) and haemoglobin (0.116) (Figure 1; Supplementary Table S4).

Figure 1.

Covariate balance before and after propensity score matching. Standardized mean differences for 53 baseline covariates in the finerenone and spironolactone cohorts before matching (n = 1428 and n = 125 770) and after 1:1 propensity score matching (n = 1417 per group). Circles denote pre-matching values, diamonds post-matching values, and the connecting line joins the two values for each covariate. Dashed and dotted vertical lines mark thresholds of 0.10 and 0.20. Standardized mean difference is a descriptive measure of balance; no hypothesis test was applied. Abbreviations: BMI, body mass index; COPD, chronic obstructive pulmonary disease; DPP-4, dipeptidyl peptidase-4; eGFR, estimated glomerular filtration rate; GLP-1, glucagon-like peptide-1; HbA1c, glycated haemoglobin; LDL-C, low-density lipoprotein cholesterol; LVEF, left ventricular ejection fraction; NT-proBNP, N-terminal pro-B-type natriuretic peptide; PSM, propensity score matching; RAAS, renin-angiotensin-aldosterone system; SGLT2, sodium-glucose cotransporter 2; SMD, standardized mean difference.

Primary and secondary outcomes

During 12 months of follow-up, finerenone use was associated with significantly lower risks of the primary composite outcome and most secondary outcomes than spironolactone, with the exception of AMI (Figure 2; Table 1). The primary composite occurred in 457 (32.3%) finerenone and 591 (41.7%) spironolactone users (aHR 0.72, 95% CI 0.64–0.82; P < .001). The ARR was 9.46% (95% CI 5.92–12.99), NNT 11 (95% CI 8–17). The E-value was 1.91 (95% lower confidence limit 1.62). Finerenone was associated with lower risks of both all-cause mortality (aHR 0.67, 95% CI 0.51–0.87; P = .002) and HF exacerbation (aHR 0.73, 95% CI 0.64–0.83; P < .001). All-cause hospitalization occurred in 814 (57.4%) finerenone and 912 (64.4%) spironolactone users (aHR 0.83, 95% CI 0.76–0.91; P < .001). AMI occurred in 186 (13.1%) finerenone users and 223 (15.7%) spironolactone users (aHR 0.82, 95% CI 0.68–1.00; P = .05).

Figure 2.

Kaplan–Meier curves for the primary composite outcome and secondary outcomes. Event-free survival through 12 months in the 1:1 propensity score matched cohort (n = 1417 per group) for (A) the primary composite of all-cause mortality or heart failure exacerbation, (B) all-cause mortality, (C) heart failure exacerbation, (D) all-cause hospitalization, and (E) acute myocardial infarction. Solid lines denote finerenone and dashed lines spironolactone; shaded bands are pointwise 95% confidence intervals. Groups were compared by log-rank test: P < .001 (A), P = .002 (B), P < .001 (C), P < .001 (D), and P = .05 (E).

Table 1.

Comparison of primary and secondary outcomes between finerenone and spironolactone users

Outcome	Finerenone (n = 1417)	Spironolactone (n = 1417)	aHR (95% CI)	P	E-value (95% LCL)	ARR (95% CI)	NNT (95% CI)
Primary composite outcomea	457 (32.3)	591 (41.7)	0.72 (0.64–0.82)	<.001	1.91 (1.62)	9.46% (5.92–12.99)	11 (8–17)
All-cause mortality	95 (6.7)	141 (10.0)	0.67 (0.51–0.87)	.002	2.33 (1.58)	3.25% (1.22–5.28)	31 (19–82)
Heart failure exacerbation	420 (29.6)	541 (38.2)	0.73 (0.64–0.83)	<.001	1.90 (1.59)	8.54% (5.07–12.01)	12 (8–20)
All-cause hospitalization	814 (57.4)	912 (64.4)	0.83 (0.76–0.91)	<.001	1.49 (1.30)	6.92% (3.33–10.50)	15 (10–30)
Acute myocardial infarction	186 (13.1)	223 (15.7)	0.82 (0.68–1.00)	.05	1.69 (1.00)	2.61% (0.03–5.20)	39 (19–3953)

The PH assumption was satisfied for all outcomes (all global Schoenfeld P > .05).

aHR, adjusted hazard ratio; ARR, absolute risk reduction; CI, confidence interval; LCL, lower confidence limit; NNT, number needed to treat.

^aPrimary composite outcome includes all-cause mortality or heart failure exacerbation.

Subgroup and sensitivity analyses

In supplementary aggregate-data analyses, treatment associations were directionally consistent with the primary Cox models (Supplementary Figures S2–S3, Supplementary Tables S5–S7). Monte Carlo bias analysis for unmeasured frailty yielded median bias-adjusted HRs for the primary composite of 0.784 (95% simulation interval 0.680–0.907) under base-case and 0.816 (0.701–0.954) under conservative assumptions. RMST analyses showed longer 12-month event–free survival with finerenone, including a +0.98-month difference for the primary composite (approximately 30 additional event-free days). Exploratory competing-mortality cumulative incidence remained favourable for finerenone (HF exacerbation 30.4% vs 38.5%; all-cause hospitalization 58.9% vs 64.8%; both nominal exploratory P < .05). These aggregate analyses should be interpreted as supportive exploratory analyses rather than formal individual-level inference.

Subgroup analyses showed consistent associations favouring finerenone across age, sex, baseline eGFR, and SGLT2 inhibitor use (Figure 3). In HFpEF, finerenone was associated with a 25% lower risk (aHR 0.75, 95% CI 0.66–0.86; P < .001), whereas the association in HFmrEF did not reach significance (aHR 0.60, 95% CI 0.36–1.02; P = .06). Among patients with T2DM, finerenone was associated with a lower risk of the primary composite (aHR 0.70, 95% CI 0.61–0.79; P < .001); no significant association was observed in those without T2DM (aHR 0.93, 95% CI 0.59–1.49; P = .77).

Figure 3.

Adjusted hazard ratios for the primary composite outcome across prespecified subgroups. Adjusted hazard ratios (squares) with 95% confidence intervals (horizontal lines) for the composite of all-cause mortality or heart failure exacerbation, estimated by Cox proportional hazards regression within each subgroup. Propensity score matching was repeated independently within each subgroup because the analytic platform cannot subset an existing matched cohort while preserving matched pairs; matched sample sizes therefore differ between subgroups and from the main cohort, ranging from 69 to 1300 patients per group (events/matched n shown as n/N). Formal interaction tests were not performed for the same reason, so subgroup estimates are descriptive and should not be read as tests of effect modification. Abbreviations: SGLT2i, sodium-glucose cotransporter 2 inhibitor.

In a post hoc analysis restricted to patients with both T2DM and CKD, finerenone remained associated with a lower risk of the primary composite (aHR 0.61, 95% CI 0.53–0.70), and the secondary outcomes were consistent with the main analysis (Supplementary Table S8).

Sensitivity analyses supported these findings. Time-stratified analyses showed consistent favourable results across early and late follow-up (aHRs 0.66 [1 day–6 months], 0.74 [1 day–9 months], 0.72 [1 day–12 months]); when anchored at 3 months, aHRs were 0.64 (3–9 months) and 0.69 (3–12 months). A landmark analysis from 6–12 months yielded an aHR of 0.61 (95% CI 0.50–0.73) (Supplementary Table S9). Stepwise covariate adjustment yielded consistent results (Model 1 aHR 0.66; Model 2 0.64; fully adjusted main model 0.72; all P < .001) (Supplementary Table S10). Results remained robust in adherence-restricted cohorts (3–12 months: aHR 0.72, 95% CI 0.59–0.87; 6–12 months: aHR 0.73, 95% CI 0.58–0.94) (Supplementary Table S11), across HF diagnosis windows (Supplementary Table S12), and in dose-restricted analyses limited to finerenone 10 mg and spironolactone 25 mg (aHR 0.69, 95% CI 0.60–0.79; Supplementary Table S13). In a sensitivity analysis restricted to patients with a HF diagnosis (I50) and a recorded echocardiographic LVEF ≥40%, results were consistent (primary composite aHR 0.60, 95% CI 0.47–0.76; Supplementary Table S14).

Safety profile and negative control outcomes

Safety outcomes are summarized in Table 2. Hyperkalaemia (≥1 episode) occurred in 254 (17.9%) finerenone and 323 (22.8%) spironolactone patients (RR 0.79, 95% CI 0.68–0.91; P = .001), with consistent reductions for recurrent (RR 0.76, 95% CI 0.62–0.94; P = .009) and frequent episodes (RR 0.62, 95% CI 0.42–0.92; P = .02). AKI (≥1 episode) occurred in 490 (34.6%) finerenone and 542 (38.2%) spironolactone patients (RR 0.90, 95% CI 0.82–0.998; P = .04), a borderline difference. The reduction was firmer for recurrent episodes (RR 0.87, 95% CI 0.76–0.99; P = .03) but not significant for frequent episodes (RR 0.83, 95% CI 0.64–1.07; P = .14). Negative control outcomes showed no significant differences for osteoporosis (aHR 1.28, 95% CI 0.94–1.76; P = .11) or dermatitis/eczema (aHR 1.01, 95% CI 0.77–1.31; P = .96), supporting the absence of strong detectable residual bias (Supplementary Table S15).

Table 2.

Comparison of safety outcomes between finerenone and spironolactone users

Safety outcome	Finerenone (n = 1417)	Spironolactone (n = 1417)	RR (95% CI)	P
Hyperkalaemia
Patients affected (≥1 episode), n (%)	254 (17.9)	323 (22.8)	0.79 (0.68–0.91)	.001
Recurrent (≥2 episodes), n (%)	146 (10.3)	191 (13.5)	0.76 (0.62–0.94)	.009
Frequent (≥6 episodes), n (%)	40 (2.8)	64 (4.5)	0.62 (0.42–0.92)	.02
Acute kidney injury
Patients affected (≥1 episode), n (%)	490 (34.6)	542 (38.2)	0.90 (0.82–0.998)	.04
Recurrent (≥2 episodes), n (%)	318 (22.4)	367 (25.9)	0.87 (0.76–0.99)	.03
Frequent (≥6 episodes), n (%)	101 (7.1)	122 (8.6)	0.83 (0.64–1.07)	.14

CI, confidence interval; RR, risk ratio.

Discussion

In this real-world, propensity score-matched cohort of patients with HFmrEF or HFpEF, finerenone initiation was associated with a 28% lower risk of the primary composite outcome (all-cause mortality or HF exacerbation). Among the prespecified secondary outcomes, finerenone was associated with a 33% lower risk of all-cause mortality, a 27% lower risk of HF exacerbation, and a 17% lower risk of all-cause hospitalization over 12 months compared with spironolactone. Associations were consistent across multiple sensitivity, time-stratified, adherence-restricted, and dose-restricted analyses, although the effect did not reach significance in small subgroups including HFmrEF (aHR 0.60, P = .06) and patients without T2DM (aHR 0.93, P = .77), a pattern that most likely reflects limited sample size and current prescription patterns rather than absent benefit or true effect modification. Finerenone was also associated with significantly lower risk of hyperkalaemia.

In TOPCAT, spironolactone did not significantly reduce the primary composite outcome in HFpEF, although regional subgroup analyses suggested potential benefits, indicating the need for more selective MR blockade.^6–8 The pooled FIDELITY analysis further demonstrated consistent cardiovascular benefits of finerenone across a wide spectrum of renal function and cardiovascular risk in patients with CKD and T2DM.¹⁹ More recently, the FINEARTS-HF trial provided direct evidence supporting finerenone in HFmrEF/HFpEF, demonstrating reductions in worsening HF events and cardiovascular death compared with placebo.¹³

A recent individual patient-level meta-analysis of major MRA trials, including TOPCAT and FINEARTS-HF, showed that in HFmrEF/HFpEF, MRA therapy reduced the composite of cardiovascular death or HF hospitalization (HR 0.87, 95% CI 0.79–0.95), with a clearer effect on HF hospitalization than on mortality.²⁰ Our findings extend this evidence by providing real-world head-to-head comparisons favouring finerenone over spironolactone, most consistently for the composite and HF-specific outcomes.

Recent real-world studies comparing finerenone with steroidal MRAs have reported heterogeneous results. Two TriNetX-based analyses reported significant reductions in all-cause mortality, hospitalization, and worsening HF favouring finerenone, whereas a smaller HFpEF-restricted study found no significant differences for acute HF events or major adverse cardiovascular events.^21–23 In this context, the present study provides the largest drug-specific matched cohort to date in HFmrEF and HFpEF, with multiple prespecified sensitivity analyses. The consistency of treatment effects across analytical frameworks strengthens the credibility of the observed associations.

Residual confounding by indication is a central concern. During the study period, finerenone was prescribed almost exclusively to ambulatory patients with T2DM and CKD under its approved indication, whereas spironolactone was used across a much broader heart-failure population. To probe this, we repeated the comparison within finerenone's on-label population (concurrent T2DM and CKD), in which both agents are guideline-plausible. After re-matching (1014 pairs, all SMDs <0.1), the primary composite association was directionally consistent (aHR 0.61, 95% CI 0.53–0.70), arguing against the indication mismatch as its sole driver. However, the direction of any residual bias is not uniform. On measured characteristics, finerenone recipients were the higher-risk group (more T2DM and CKD, lower eGFR), which would bias the composite conservatively, whereas channelling of a newer, specialist-initiated agent to more closely monitored patients could bias the other way. The composite E-value was 1.91 (LCL 1.62), the association persisted under Monte Carlo modelling of unmeasured frailty, and prespecified negative-control outcomes were null. Together these analyses constrain, but cannot exclude, residual confounding by indication.

This caveat bears most heavily on all-cause mortality. The mortality association (aHR 0.67, 95% CI 0.51–0.87) is larger than any randomized MRA trial in this population would predict, as neither TOPCAT nor FINEARTS-HF demonstrated a significant mortality reduction.^6,13 Because PSM balances only measured covariates, unmeasured factors such as frailty, functional status, and care intensity, which were plausibly greater in the broadly treated spironolactone group, remain the most likely explanation for a mortality association of this magnitude. The signal was directionally unchanged in the on-label stratum (aHR 0.58, 95% CI 0.43–0.79), but restricting to that stratum does not remove these within-group differences. The mortality estimate must therefore be read strictly as hypothesis-generating rather than as evidence of a survival benefit.

The observed association favouring finerenone may be explained in part by its anti-inflammatory and antifibrotic effects. Preclinical studies have consistently shown that finerenone suppresses inflammatory cytokine expression, reduces macrophage infiltration, and lowers myocardial collagen deposition more effectively than steroidal MRAs. This suggests a mechanistic link to improved ventricular compliance and reduced susceptibility to recurrent congestion and decompensation.^24,25 These mechanisms provide biological plausibility but cannot, in an observational study subject to residual confounding, establish clinical superiority.

Finerenone use was also associated with significantly lower risk of hyperkalaemia and a borderline lower risk of AKI. These findings are clinically relevant because concerns regarding renal dysfunction and potassium disturbances frequently limit the initiation, uptitration, and long-term persistence of steroidal MRAs in HF, particularly among older patients and those with CKD.^26,27 The lower hyperkalaemia risk with finerenone likely reflects its non-steroidal structure and balanced cardiorenal tissue distribution, whereas spironolactone and its active metabolites concentrate in the kidney and more strongly promote potassium retention.^28,29 Less hyperkalaemia with finerenone than spironolactone was also observed in the head-to-head ARTS trial.³⁰

Limitations

This study has several limitations. First, as an observational analysis based on the TriNetX network, causal inference cannot be established, and residual confounding cannot be fully excluded despite extensive PSM and sensitivity analyses. Although matching achieved good overall balance, modest imbalances remained for selected continuous variables, including eGFR and haemoglobin, with SMDs slightly exceeding 0.10. E-value analyses indicated moderate resistance to unmeasured confounding, and results were stable across stepwise adjustment and additional sensitivity analyses. Second, classification of HFmrEF/HFpEF, exposures, and outcomes relied on diagnostic and procedural codes, introducing potential misclassification. Third, because PSM was repeated independently in each subgroup and sensitivity analysis, the study populations of those analyses are not identical to the main matched cohort. Fourth, information on medication dose, titration, and adherence was unavailable; adherence and dose-restricted sensitivity analyses partially mitigated this, yielding consistent results. Fifth, endpoints were constrained by the administrative origin of the data. Cause-specific mortality or hospitalization data and natriuretic-peptide trajectories could not be validly constructed and represent directions for future work. Finally, by design, the matched spironolactone cohort resembles finerenone initiators rather than the whole spironolactone population, so our findings apply to the finerenone-eligible, predominantly cardiorenal HFmrEF/HFpEF population and not to non-diabetic or broader spironolactone patients.

Conclusions

In this multinational active-comparator, new-user cohort of patients with heart failure and left ventricular ejection fraction ≥40%, finerenone initiation was associated with a lower 12-month risk of the composite of all-cause mortality or HF exacerbation compared with spironolactone initiation. Finerenone was also associated with lower risks of HF exacerbation, all-cause hospitalization, and fewer hyperkalaemia events, whereas the difference in acute kidney injury was modest and borderline. These findings provide real-world head-to-head comparative evidence between non-steroidal and steroidal MR antagonism in HFmrEF/HFpEF. Because residual confounding and channelling by indication cannot be fully excluded, these results should be interpreted as hypothesis-generating, warranting confirmation in prospective randomized trials.

Supplementary data

Supplementary data are available at ESC Heart Failure online.

Declarations

Disclosure of Interest

All authors declare no disclosure of interest for this contribution.

Data availability

Data supporting the findings of this study are available from the TriNetX Research Network. Access to the TriNetX platform requires institutional subscription and TriNetX approval. Due to TriNetX data-sharing policies, the authors did not have access to individual-level data; only aggregated, de-identified results were available for analysis. Researchers interested in using the platform may apply through TriNetX by providing appropriate credentials, a clearly defined research purpose, and compliance with relevant privacy regulations. Further information is available at https://trinetx.com or by contacting support@trinetx.com.

Funding

This work was supported by the National Science and Technology Council [NSTC114-2314-B-038-126 and NSTC114-2314-B-038-128]; and Taipei Medical University Hospital [114TMUH-C01-1 and 114TMUH-SP-03]. The funders had no role in the study design, data collection, analysis, interpretation, manuscript preparation, or decision to submit the manuscript.

Ethical Approval

The study protocol (TMU-JIRB No. N202605005) was reviewed by the Taipei Medical University-Joint Institutional Review Board and designated as exempt from full review; the requirement for informed consent was waived.

Pre-registered Clinical Trial Number

None supplied.

Author contributions

P.T.K.H.: conceptualization, methodology, formal analysis, writing—original draft. C.J.C.: methodology, writing—review and editing. C.L.H., H.Y.C., C.W.C., Y.C.L., C.Y.H. (Chun-Yao Huang): methodology, writing—review and editing. Y.H.S.: formal analysis, methodology. C.Y.H. (Chien-Yi Hsu): conceptualization, supervision, funding acquisition, writing—review and editing.