AHA Scientific Sessions 2026 Abstract · Nov 2, 2026
Performance of HFpEF Diagnostic Criteria and Risk Scores in Patients with Diabetes and Risk Factors for Heart Failure - An Exploratory Analysis of TARTAN-HF
By Taylor-Sweet D, Petrie M, McKinley G et al.
Background: The predominant heart failure (HF) phenotype in people with diabetes is heart failure with preserved ejection fraction (HFpEF). Diagnostic and risk stratification tools for HFpEF exist; however, concordance between these scores is often limited, and their utility as screening tools remains uncertain. In the TARTAN-HF trial, we recently reported a prevalence of unrecognised HFpEF of 23% in people with diabetes and at least one additional risk factor for HF. In this exploratory analysis of the TARTAN-HF trial, we examined the prevalence of HFpEF according to available HFpEF diagnostic and risk scores. Methods: TARTAN-HF (NCT05705869) was a prospective, multicentre, unblinded, randomised controlled trial. Community-based participants, with no history of HF, aged ≥40 years, with an established diagnosis of diabetes and risk factors for HF were recruited. Consenting participants (n=706) were randomized to a screening strategy of NT-proBNP testing followed by echocardiography if NT-proBNP was elevated (≥125pg/mL) or to usual care. The primary outcome was a diagnosis of HF at 6 months. We calculated the proportion of participants in the screening arm who met criteria for definite HFpEF according to the HFpEF-ABA (score ≥90%) , H2FPEF (score ≥6 points), and HFA-PEFF (score ≥5 points) diagnostic scores. We also examined the proportion of patients who were predicted as being at a high or very high risk of the development of HF according to the WATCH-DM score (score ≥11 points). Results: The primary and secondary results of TARTAN-HF were presented at the 2026 ACC Scientific Sessions meeting. Of the 354 randomized to screening, 87 (25%) of participants had a diagnosis of HF made at screening; 81 (23%) had a diagnosis of HFpEF made according to the ESC 2021 guideline criteria. Of the 81 diagnosed with HFpEF through screening: 20 (25%) had an HFpEF-ABA probability ≥90%, 16 (20%) had an H ₂ FPEF score of ≥6 points, and 40 (50%) had a HFA-PEFF score ≥5 points (Figure) . When considering the WATCH-DM score, 68 (84%) had a score ≥11 which is associated with a high or very high risk of the development of HF. In the patients randomized to screening who had an NT-proBNP ≥125 pg/mL and LVEF ≥50% and were not diagnosed with HFpEF (n=87), 11 (13%) had an HFpEF-ABA probability ≥90%, 7 (8%) had an H ₂ FPEF score ≥6 points, and 32 (37%) had an HFA-PEFF score ≥5 points. In patients without a diagnosis of HFpEF, 35 (60%) had a WATCH-DM score of ≥11. Conclusions: In this exploratory analysis of the TARTAN-HF trial, the application of HFpEF diagnostic scores to identify definite HFpEF would have missed between 50-80% of cases detected through screening. Although the WATCH-DM score identified most patients with HFpEF, it displayed limited specificity. These results support a screening strategy based on NT-proBNP measurement, echocardiography, and structured clinical assessment in patients with diabetes at high risk of HF.