CMR-derived parameters increase H2FPEF Score diagnostic performance

Thursday, February 5, 2026

10:43 AM - 10:50 AM

Location: Galapagos / Aruba / Ilhabela (2nd Lower Level)

Presenting Author(s)

Ambra Masi, MD

Cardiologist
Heart & Vessels Department, Cardiology Division, University Hospital of Lausanne
Lausanne, Vaud, Switzerland

Primary Author(s)

Ambra Masi, MD

Cardiologist
Heart & Vessels Department, Cardiology Division, University Hospital of Lausanne
Lausanne, Vaud, Switzerland

Co-Author(s)

AP

Aldostefano Porcari, MD

Cardiologist
3Centre for Diagnosis and Treatment of Cardiomyopathies, Cardiovascular Department, Azienda Sanitaria Universitaria Giuliano-Isontina (ASUGI), University of Trieste, Trieste, Italy, Italy
AR

Angela Rocca, MSc

PhD student
Lausanne University Hospital (CHUV) and University of Lausanne (UNIL)
Lausanne, Vaud, Switzerland
Ruud B. van Heeswijk, PhD, FSCMR

Senior lecturer
Lausanne University Hospital (CHUV) and University of Lausanne (UNIL)
Lausanne, Vaud, Switzerland
PM

Pierre Monney, MD

Cardiologist
University Hospital (CHUV) and University of Lausanne (UNIL), Switzerland
PA

Panagiotis Antiochos, MD

Cardiologist
Cardiac MR Center of the University Hospital Lausanne, CHUV, Switzerland
Lausanne, Vaud, Switzerland
HL

Henri Lu, MD

Cardiologist
University Hospital (CHUV) and University of Lausanne (UNIL), Switzerland
MS

Matthias Stuber, PhD

Professor/Director
CIBM/CHUV/UNIL
Lausanne, Switzerland
RH

Roger Hullin, MD

Chief Severe Heart Failure en Heart Transplantation Program
Lausanne University Hospital (CHUV)
Lausanne, Vaud, Switzerland
Juerg Schwitter, MD

Full Professor
University Hospital and University of Lausanne
Lausanne, Vaud, Switzerland

Background: Heart failure with preserved ejection fraction (HFpEF) is a heterogeneous syndrome with challenging diagnosis. Existing diagnostic scores such as H₂FPEF (1) mainly rely on clinical and echocardiographic parameters. Cardiac magnetic resonance (CMR) provides accurate myocardial tissue characterization that may support diagnostic accuracy in identifying clinical HFpEF. We assessed whether the integration of CMR-derived parameters improves the diagnostic performance of the H₂FPEF score in patients with HFpEF syndrome.

Methods: From a cohort of 19.535 CMR imaging scans (MAGNETOM, Siemens Healthineers, 1.5 et 3T) performed between 2010-2024 at our institution, we retrospectively selected 613 patients with left ventricle ejection fraction (LVEF) >50% and calculable H₂FPEF score (echo within ±3 months of CMR). Patients were classified into 2 groups according to the H₂FPEF score: low-intermediate (0–5) vs. high probability (6–9). The presence of clinical HF was identified by ICD code. Univariable and multivariable logistic regression analyses were performed to investigate parameters associated with clinical HF. Three models were created: H₂FPEF score alone (model 1), H₂FPEF score + native septal T₁ on 1.5T (model 2), H₂FPEF score + mean extracellular volume (ECV) on 1.5T (model 3). Diagnostic performance was assessed using the area under the curve (AUC). The DeLong test was used to compare the accuracy of different models. The Akaike information criterion (AIC) was calculated to measure the discriminatory ability of each model.

Results: In the overall cohort of 613 patients, 67 had clinical HF. At univariable analysis, a H₂FPEF score of 6 or higher identified an approximately 8-fold higher probability of clinical HF (OR = 8.10; 95% CI: 4.04–16.25; p< 0.001). Among the variables included in the H₂FPEF score, all were significantly associated with the diagnosis of HF, apart from BMI (Table 1). In the overall cohort LV morpho-functional variables were not associated with clinical HF (Table 1). Among patients with available T₁ mapping values at 1.5T (n = 269), only septal native T₁ (OR 1.125, 95% CI: 1.0.35–1.223, p=0.006) and mean ECV (OR 1.076, 95% CI: 1.017–1.138, p=0.011) were independently associated with clinical HF (n=34) (Table 1).

The integration of mean ECV into the H₂FPEF score significantly improved the discriminatory ability compared with a model composed of H₂FPEF score alone (AUC 0.73 vs 0.61; p=0.007; AIC: 192.6 vs 397.0) (Figure 1 and 2). The addition of septal native T₁ did not improve the diagnostic accuracy (Figure 1)

Conclusion: In patients with CMR-LVEF >50% and a high likelihood of HFpEF (H₂FPEF >5 points), both native septal T₁ mapping and mean ECV were independently associated with clinical HF. Integrating mean ECV into a model composed of the H₂FPEF-score-only significantly improved the identification of patients with clinical HF of any etiology. These findings support the role of CMR-guided tissue characterization in refining the noninvasive diagnosis of HFpEF.

Table 1. Univariable and multivariable binary logistic regression analysis for prediction of clinical HFpEF

Variables	Univariable OR (95% CI)	p-value	Multivariable OR (95% CI)	p-value	Multivariable OR (95% CI)	p-value
H2PEF Score >5	8.1 (4.0-16.2)	< 0.001	16.1 (3.8-67.3)	< 0.001	18.0 (4.4-73.2)	< 0.001
BMI >30	1.1 (0.6-2.1)	0.717
AF	9.6 (4.7-19.4)	< 0.001
PASP >35 mmHg	5.5 (3.1-9.9)	< 0.001
HTN (2 or more drugs)	2.6 (1.5- 2.9)	< 0.002
E/e’ >9	2.9 (1.3-3.9)	0.003
LVSV (1 ml)	0.9 (0.9-1.0)	0.243
LVSVi (1 ml)	0.9 (0.9-1.0)	0.619
LVEDV (1 ml)	1.0 (0.9-1.0)	0.847
LVEDVi (1 ml)	1.0 (0.9-1.0)	0.608
LVESV (1 ml)	1.0 (0.9-1.0)	0.329
Remodeling (LV mass/LVEDV)	1.1 (0.6-2.1)	0.688
Septal nT1 (10 ms)	1.1 (1.0-1.2)	0.006	1.1 (1.0-1.2)	0.021
Lateral nT1 (10 ms)	1.0 (0.9-1.0)	0.910
Mean nT1 (10 ms)	1.0 (0.9-1.0)	0.498
Mean ECV (1%)	1.1 (1.0-1.1)	0.011			1.0 (1.0-1.1)	0.010

Abbreviation BMI Body Mass Index, AF Atrial fibrillation, PASP Pulmonary Artery Systolic Pressure, HTN hypertension, LVSV left ventricular stroke volume, LVSVi left ventricular stroke volume indexed, LVEDV left ventricular end-diastolic volume, LVEDVi left ventricular end-diastolic volume indexed, LVESV left ventricular systolic volume, LVESVi left ventricular systolic volume indexed, nT1 Native T1, ECV Extra Cellular VolumeFigure 1. Receiver-operating characteristic (ROC) curves for the clinical prediction of clinical HFpEF using different models.: Model 1 (H2FPEF score alone); Model 2 (H2FPEF score plus septal native T1); Model 3 (H2FPEF score plus mean ECV) at 1.5T.
Fig 2. Comparison of Receiver-operating characteristic (ROC) curves among Model 1 et Model 3 for the prediction of clinical HFpEF.