SCORE2-HF risk prediction algorithms: new models to estimate 10-year risk of incident heart failure in Europe

Abstract Background Heart failure (HF) presents a significant public health challenge, affecting over 60 million people globally, and is associated with diminished quality of life, frequent hospitalisations, substantial healthcare costs, and high mortality rates. The ESC recommends the SCORE2, SCORE2-OP and SCORE2-Diabetes CVD risk prediction algorithms to estimate individual risk and guide preventive intervention in primary care. However, these models do not estimate the risk of non-fatal HF, which could help refine targeted preventive interventions. Purpose To develop SCORE2-HF, a prediction model for estimating the 10-year risk of incident HF in European adults aged over 40 years without previous cardiovascular disease (CVD) or HF. Methods We derived SCORE2-HF using data from over 450,000 individuals with over 12,000 HF events from the UK Biobank and multiple European and North American cohorts contributing to the Emerging Risk Factors Collaboration. The models predict HF irrespective of left ventricular ejection fraction and incorporate sex-specific, competing risk-adjusted calculations based on age, smoking status, systolic blood pressure (SBP), antihypertensive treatment, body mass index (BMI), estimated glomerular filtration rate (eGFR), and diabetes (including age at diagnosis and HbA1c). Models were recalibrated to reflect HF incidence in different risk regions of Europe, estimated by combining population CVD mortality statistics from the World Health Organisation mortality database, with multipliers reflecting HF incidence derived from representative data sources across Europe. Results SCORE2-HF showed excellent calibration and discrimination when externally validated in 1.23 million individuals with 33,755 HF events from the Clinical Practice Research Datalink (CPRD, C-index 0.819 0.817-0.821) with improved calibration and similar discrimination compared to the AHA-PREVENT-HF models (C-index 0.820 0.818-0.823). SCORE2-HF risk estimations varied importantly across individuals with different levels of modifiable risk factors (Figure 1). For example, a 70-year-old non-smoker without diabetes or antihypertensive medication use, SBP=110mmHg, BMI=23Kg/m2 and eGFR=90ml/l/1.73Kg2 had an estimated 10-year risk of 5.1% (male) and 5.3% (female). In contrast, a current smoker with diabetes diagnosed at age 45, SBP=150mmHg, BMI=30Kg/m2, eGFR=60 ml/Kg2, and HbA1c=65mmol/mol had a 10-year risk of 15% (male) and 16% (female). Furthermore, SCORE2-HF provided meaningful risk stratification within existing guideline-recommended CVD risk categorisations using the SCORE2/SCORE2-OP/SCORE2-Diabetes CVD risk models, indicating complementary value in using both scores to guide preventive intervention (Figure 2). Conclusion SCORE2-HF - a new risk prediction model derived, recalibrated, and validated to estimate the 10-year risk of incident HF in European countries should enhance the identification of individuals at higher risk of developing incident HF.