Characterising the heterogeneity of heart failure with preserved ejection fraction: moving beyond subgroups and distinguishing disease from risk.

Soltani, F.; Black, N.; Bradley, J.; Priestner, L.; Ko Ko, N.; Glasse, H.; Milner, I.; Kreppel, S.; Schmitt, M.; Nucifora, G.; Petrie, MC.; Al-Mohammad, A.; McCann, GP.; Lumbers, RT.; Beezer, J.; Paton, MF.; Robinson, S.; Hyland, R.; Hartshorne-Evans, N.; Humphreys-Davies, L.; Raisi-Estabragh, Z.; Petersen, SE.; Newman, WG.; McDonagh, T.; Mayet, J.; Williams, SG.; Jenkins, DA.; Morris, AP.; Peek, N.; Miller, CA.

Characterising the heterogeneity of heart failure with preserved ejection fraction: moving beyond subgroups and distinguishing disease from risk.

All Authors

Soltani, F.

Black, N.

Bradley, J.

Priestner, L.

Ko Ko, N.

Glasse, H.

Milner, I.

Kreppel, S.

Schmitt, M.

Nucifora, G.

Show 10 more

LTHT Author

Paton, Maria

LTHT Department

Cardio-Respiratory
Cardiology

Non Medic

Highly Specialist Cardiac Physiologist

Publication Date

2026

Item Type

Journal Article

Subject

DATA INTERPRETATION, STATISTICAL , HEART FAILURE

Abstract

AIMS: Heart failure with preserved ejection fraction (HFpEF) is a heterogeneous syndrome commonly hypothesised to comprise discrete subtypes. This study aimed to characterise HFpEF heterogeneity using advanced data science techniques applied to deeply phenotyped, multi-modal cohorts. METHODS AND RESULTS: 902 patients with HF and a left ventricular ejection fraction >= 50% were prospectively recruited and underwent detailed clinical, imaging, and genetic profiling. Clustering algorithms were first applied to derive discrete subgroups. To overcome rigid subgroup assignment, DDRTree was applied to map continuous phenotypic and outcome variation onto a two-dimensional tree. Genome-wide association studies (GWAS) identified variants associated with tree dimensions. Internal validation assessed stability, and external validation was conducted in a UK Biobank HFpEF cohort (n=148). Sensitivity analyses included patients with mildly reduced ejection fraction (n=1154). Logistic regression compared HFpEF with comorbidity-matched controls (n=902). Whilst conventional clustering did not identify meaningful subgroups, DDRTree revealed a stable, interpretable tree structure, capturing continuous variation in cardiac structure and function, biomarkers and comorbidities. Over a median follow-up of 6.8 years, phenotypic complexes demonstrated differing risks of cardiovascular death, HF hospitalisation, adverse renal outcome, and infection hospitalisation. GWAS revealed a locus near CRACD associated with tree dimension one. Findings were reproducible in external and sensitivity analyses. A profile of cardiac and systemic characteristics that distinguish HFpEF from its associated comorbidities were identified. CONCLUSIONS: HFpEF risk and disease course reflect a variable interplay of pathophysiological processes in each individual. This study provides a biologically plausible framework to guide a more personalised approach to HFpEF.