Section Structure
1. Highlights
2. Clinical Background
3. Study Design and Methods
4. Risk Model Components and Analytical Approach
5. Key Results
6. Clinical Interpretation
7. Strengths and Limitations
8. Implications for Practice and Research
9. Conclusion
10. Funding, Registration, and Citation
Highlights
Biomarker-driven prognostic models originally developed in EMPEROR-Preserved performed well when externally applied to FINEARTS-HF, supporting their broader transportability in patients with heart failure with mildly reduced or preserved ejection fraction.
The models identified marked heterogeneity in risk. Patients in the highest versus lowest risk quintile had substantially higher hazards for the composite of first heart failure hospitalization or cardiovascular death and for cardiovascular death alone.
Finerenone’s relative effect on first heart failure hospitalization or cardiovascular death was consistent across all risk quintiles and across the continuous spectrum of estimated baseline risk, with no evidence of treatment-effect modification.
These findings reinforce two clinically relevant points: prognosis in HFmrEF/HFpEF can be quantified with simple biomarker-based tools, and finerenone appears to confer benefit irrespective of baseline estimated risk.
Clinical Background
Heart failure with mildly reduced ejection fraction (HFmrEF) and heart failure with preserved ejection fraction (HFpEF) account for a growing proportion of the global heart failure burden. These syndromes are especially common in older adults and in patients with multimorbidity, including atrial fibrillation, diabetes, obesity, chronic kidney disease, and chronic lung disease. Despite increasing therapeutic options, HFpEF and HFmrEF remain clinically challenging because they are biologically heterogeneous and prognostically diverse.
Some patients experience relatively stable symptoms over years, whereas others progress rapidly with recurrent heart failure hospitalization and high mortality. This heterogeneity creates a practical need for risk tools that are clinically usable, biologically plausible, and externally valid. Robust risk stratification can support patient counseling, trial enrichment, follow-up intensity, and interpretation of absolute treatment benefit.
The EMPEROR-Preserved investigators previously developed prognostic models using readily available clinical variables and biomarkers, notably N-terminal pro-B-type natriuretic peptide (NT-proBNP) and high-sensitivity cardiac troponin T (hs-cTnT). Because both biomarkers reflect complementary pathobiology—hemodynamic stress and myocardial injury, respectively—the models are attractive for real-world and trial-based use.
FINEARTS-HF tested finerenone, a nonsteroidal mineralocorticoid receptor antagonist, in symptomatic patients with left ventricular ejection fraction (LVEF) 40% or greater. The current prespecified secondary analysis asked two clinically important questions: first, do the EMPEROR-Preserved risk models maintain performance in an independent trial population; and second, does baseline estimated risk modify the treatment effect of finerenone?
Study Design and Methods
This analysis was prespecified within FINEARTS-HF, a large randomized clinical trial conducted at 653 sites in 37 countries. Eligible participants were adults aged 40 years or older with symptomatic heart failure and an LVEF of 40% or greater. Patients were randomized between September 2020 and January 2023 to finerenone, titrated to 20 mg or 40 mg, or placebo.
The present analysis included 6001 patients: 3003 assigned to finerenone and 2998 assigned to placebo. The mean age was 72.0 years, 45.5% were women, and the median follow-up was 32 months, with an interquartile range of 23 to 37 months. This population is representative of contemporary HFmrEF/HFpEF trial cohorts in terms of age, symptom burden, and multimorbidity profile.
The main outcomes assessed here were based on EMPEROR-Preserved risk models developed for three endpoints: first heart failure hospitalization or cardiovascular death, cardiovascular death, and all-cause death. Estimated risk was compared against observed event rates in FINEARTS-HF. Model discrimination was assessed using Harrell C statistic, a standard metric for survival models that quantifies how well a model distinguishes between patients who do and do not experience an event earlier.
The second major objective was treatment-effect heterogeneity. Investigators evaluated whether finerenone’s efficacy differed across quintiles of estimated baseline risk and across the continuous risk distribution. This is clinically important because a constant relative effect across varying risk levels implies that absolute benefit may still be larger in higher-risk patients even without statistical interaction.
Risk Model Components and Analytical Approach
The EMPEROR-Preserved risk models are notable for relying on a concise set of variables with strong biologic relevance. Core variables included NT-proBNP, hs-cTnT, New York Heart Association functional class, history of chronic obstructive pulmonary disease, history of diabetes, and insulin use. Depending on the specific endpoint, additional variables included age, hemoglobin, albumin, duration of heart failure, time from prior heart failure hospitalization, and sodium-glucose transporter 2 inhibitor use.
This structure deserves emphasis. NT-proBNP captures chamber wall stress and congestion, while hs-cTnT detects low-level myocardial injury that is common in chronic heart failure and strongly prognostic. NYHA class summarizes symptom burden and functional limitation, and comorbidity variables account for noncardiac disease burden that heavily shapes outcomes in HFpEF. Hemoglobin and albumin likely reflect systemic illness, frailty, inflammation, and nutritional status. Together, the model moves beyond LVEF, which is often a weak discriminator of risk within the preserved-EF spectrum.
External validation in an independent dataset is a particularly strong test of model usefulness. Many risk scores perform well only in the derivation cohort. Demonstrating preserved discrimination in FINEARTS-HF therefore provides more compelling evidence that these models are not merely overfit products of the original EMPEROR-Preserved population.
Key Results
The EMPEROR-Preserved model effectively stratified patients in FINEARTS-HF according to risk. For the composite outcome of first heart failure hospitalization or cardiovascular death, the hazard ratio comparing the highest-risk quintile (Q5) with the lowest-risk quintile (Q1) was 10.49, with a 95% confidence interval of 8.14 to 13.52. For cardiovascular death, the hazard ratio for Q5 versus Q1 was even higher at 13.47, with a 95% confidence interval of 8.79 to 20.64.
These are large gradients in risk and indicate that the model meaningfully separates low-risk from high-risk patients. In practical terms, HFmrEF/HFpEF should not be viewed as a uniformly moderate-risk condition. Instead, there is a broad prognostic spectrum, and biomarker-informed modeling can identify patients who are far more vulnerable to hospitalization and cardiovascular death.
The authors report that model discrimination was good, based on Harrell C statistic analyses. While the abstract does not provide the exact C-statistic values, the description of good discrimination, together with the strong event separation across quintiles, supports the conclusion that the model retained clinically useful prognostic accuracy in this external cohort.
The central therapeutic finding was the absence of interaction between baseline risk and finerenone’s relative effect on the primary clinical outcome. For first heart failure hospitalization or cardiovascular death, the hazard ratios for finerenone versus placebo across risk quintiles were as follows: Q1, 0.93 (95% CI, 0.58-1.49); Q2, 1.04 (95% CI, 0.76-1.43); Q3, 0.82 (95% CI, 0.62-1.07); Q4, 0.81 (95% CI, 0.65-1.01); and Q5, 0.88 (95% CI, 0.74-1.05). The P value for interaction was .68.
This pattern suggests relative consistency rather than any clear attenuation or amplification of effect across risk groups. Importantly, the analysis also examined the continuous risk spectrum rather than only categorical quintiles, and the treatment effect remained uniform. That is methodologically reassuring because arbitrary category boundaries can sometimes obscure or exaggerate interactions.
Several points help interpret these results correctly. First, the confidence intervals in individual quintiles are wide, particularly in the lower-risk groups where fewer events occurred. Therefore, the apparent variation in point estimates should not be overinterpreted. Second, the lack of interaction means the proportional relative benefit of finerenone was similar across risk groups; it does not imply identical absolute benefit. Higher-risk patients generally accrue more events, so even the same relative effect may translate into greater absolute risk reduction in those patients.
The findings also support the broader concept that finerenone’s mechanism of benefit in HFmrEF/HFpEF is not confined to a narrow biologic subgroup defined by baseline risk score. As a mineralocorticoid receptor antagonist with anti-inflammatory, antifibrotic, and cardiorenal effects, finerenone may address pathophysiologic processes that are operative across a wide spectrum of clinical severity.
Clinical Interpretation
From a clinician’s perspective, this analysis has immediate relevance in two domains: prognosis and therapeutic decision-making.
For prognosis, the study shows that a relatively compact risk model anchored in NT-proBNP and hs-cTnT can distinguish patients with markedly different outcomes in HFmrEF/HFpEF. That is valuable because risk assessment in this population is often imprecise. Ejection fraction alone does not capture the burden of congestion, myocardial injury, extracardiac comorbidity, or frailty. A multimarker model offers a more granular and biologically coherent estimate of risk.
For treatment decisions, the main message is that clinicians should not assume finerenone is only useful in the sickest or least sick patients. The relative treatment effect was stable across low to high estimated risk. This finding is complementary to the contemporary heart failure treatment paradigm, where evidence-based therapies are often recommended broadly across eligible patients unless there is a clear contraindication or a demonstrated interaction with baseline disease severity.
There is also a trial-design implication. Risk models such as this can help enrich future HFpEF/HFmrEF trials with patients likely to accrue a sufficient number of events, improving statistical efficiency. At the same time, because treatment effects appear consistent across risk strata, risk enrichment should not be conflated with restricting therapy only to high-risk groups in routine care.
These data fit with the evolving view that HFpEF management is becoming increasingly phenotype-informed but should still be guided by therapies with broad evidence bases. SGLT2 inhibitors have already shown robust benefit across preserved-EF populations, and finerenone now adds to the therapeutic landscape. Prognostic tools may refine how clinicians prioritize intensity of follow-up, volume assessment, and multimorbidity management, even when they do not redefine who should receive therapy.
Strengths and Limitations
This study has several important strengths. It was prespecified, reducing concerns about purely post hoc signal-seeking. It leveraged a large, international, contemporary randomized trial with standardized outcome assessment. It also addressed external validation, which is one of the most important but often underperformed steps in risk-model evaluation. Finally, the use of both categorical quintiles and continuous-risk interaction testing makes the treatment heterogeneity analysis more robust.
There are, however, limitations. As a secondary analysis, it inherits the usual constraints of analyses not designed as the primary trial objective. The abstract does not present detailed calibration metrics, so although observed and estimated risks were compared, granular assessment of overprediction or underprediction across the full range of risk is not available from the summary alone. Biomarker-based models also depend on access to hs-cTnT and NT-proBNP assays, which may vary across health systems and can be influenced by assay platform, renal function, and acute clinical status.
Another important consideration is that trial populations, while broad, are not identical to unselected community populations. Patients enrolled in multinational randomized trials may differ from frailer individuals, those with severe competing illness, or those managed in resource-limited settings. Therefore, additional validation in routine clinical practice would be useful.
Finally, no interaction does not mean no clinical value in stratification. Risk estimation remains important for absolute risk communication, shared decision-making, and identifying patients who may derive the greatest absolute event reduction, even when the relative treatment effect is constant.
Implications for Practice and Research
The immediate practical implication is that clinicians and investigators should consider biomarker-based risk assessment in HFmrEF/HFpEF more routinely. A validated model that combines natriuretic peptides, troponin, symptom class, and comorbidity burden can provide more informative prognostic estimates than traditional clinical impression alone.
For clinical practice, several use cases are plausible. Higher-risk patients may merit closer surveillance for congestion, more frequent medication review, stronger emphasis on comorbidity control, and earlier discussion of goals of care when appropriate. Lower-risk patients may still require comprehensive treatment but can be counseled with a more individualized understanding of expected trajectory.
For research, the study supports further work in integrated risk prediction, including models that combine biomarkers with imaging, renal indices, frailty measures, and patient-reported outcomes. It also raises the possibility of applying such tools to estimate individualized absolute benefit from finerenone and other HFpEF therapies. That step would be particularly useful for precision implementation, cost-effectiveness analyses, and health-system planning.
In the guideline arena, this work is more likely to influence risk assessment than drug eligibility. The findings do not suggest that finerenone should be limited by baseline prognostic category. Rather, they support a broad treatment approach while using risk scores to inform prognostic counseling and intensity of follow-up.
Conclusion
This prespecified secondary analysis shows that the EMPEROR-Preserved risk models generalize well to the FINEARTS-HF population, providing clinically meaningful prognostic discrimination in patients with HFmrEF or HFpEF. The contrast in outcomes between the highest and lowest risk quintiles was substantial, underscoring the marked heterogeneity of these syndromes.
At the same time, baseline estimated risk did not modify the relative effect of finerenone on first heart failure hospitalization or cardiovascular death. For clinicians, the key takeaway is straightforward: use validated biomarker-based tools to understand who is at greatest risk, but do not assume that finerenone’s benefit is confined to any specific risk stratum. Prognostic stratification and therapeutic eligibility are related concepts, but this study shows they should not be conflated.
Funding, Registration, and Citation
Trial registration: ClinicalTrials.gov Identifier: NCT04435626.
Study citation: Chimura M, McDowell K, Jhund PS, Henderson AD, Claggett BL, Desai AS, Brinker M, Lay-Flurrie J, Glasauer A, Goea L, Berger M, Lam CSP, Senni M, Voors AA, Zannad F, Pitt B, Vaduganathan M, Solomon SD, McMurray JJV. EMPEROR-Preserved Risk Model and Outcomes in the FINEARTS-HF Trial: A Prespecified Secondary Analysis of FINEARTS-HF. JAMA Cardiology. 2026 May 1;11(5):416-426. PMID: 41903165.
PubMed URL: https://pubmed.ncbi.nlm.nih.gov/41903165/
References
1. Chimura M, McDowell K, Jhund PS, et al. EMPEROR-Preserved Risk Model and Outcomes in the FINEARTS-HF Trial: A Prespecified Secondary Analysis of FINEARTS-HF. JAMA Cardiology. 2026;11(5):416-426. PMID: 41903165.
2. Anker SD, Butler J, Filippatos G, et al. Empagliflozin in Heart Failure with a Preserved Ejection Fraction. New England Journal of Medicine. 2021;385(16):1451-1461.
3. Solomon SD, McMurray JJV, Claggett B, et al. Dapagliflozin in Heart Failure with Mildly Reduced or Preserved Ejection Fraction. New England Journal of Medicine. 2022;387(12):1089-1098.
4. McDonagh TA, Metra M, Adamo M, et al. 2023 Focused Update of the 2021 ESC Guidelines for the diagnosis and treatment of acute and chronic heart failure. European Heart Journal. 2023;44(37):3627-3639.
5. Heidenreich PA, Bozkurt B, Aguilar D, et al. 2022 AHA/ACC/HFSA Guideline for the Management of Heart Failure. Circulation. 2022;145(18):e895-e1032.
