https://doi.org/10.1007/s00392-025-02625-4
1Universitäres Herz- und Gefäßzentrum Hamburg Allgemeine und Interventionelle Kardiologie Hamburg, Deutschland; 2LUM University “Giuseppe Degennaro” Department of Medicine and Surgery Pozzilli, Italien; 3Rigshospitalet Department of Cardiology Kopenhagen, Dänemark; 4Finnish Institute for Health and Welfare (THL) Helsinki (Finland), Finnland; 5UHZ Klinik für Kardiologie Hamburg, Deutschland; 6Universitäres Herz- und Gefäßzentrum Hamburg Klinik für Kardiologie Hamburg, Deutschland; 7University College London London, Großbritannien
Background
The burden of heart failure is increasing in Europe as a result of the aging population and improved survival rates from cardiovascular disease and its comorbidities. This study aims to develop an easily applicable European risk score for predicting the incidence of heart failure based on clinical variables. It ought to include NT-proBNP, a routine biomarker used to diagnose and assess the severity and prognosis of heart failure.
Methods
The risk algorithm was developed in 8,239 individuals aged 25 to 74 years from the population-based prospective FINRISK study, in whom heart failure was not present at baseline. We performed multivariable model selection using LASSO penalized Cox regression to obtain 12-year absolute risk of heart failure, adjusting for classical cardiovascular risk factors, clinical variables, and most importantly, biomarkers. The score was validated in a similar cohort from Denmark (Glostrup, n=7,276) and from Italy (Moli-Sani, n= 21259).
Results
We developed two risk models. One score was based on readily available markers (age, sex, body mass index, and pulse pressure). Another, more comprehensive risk score also included heart rate, antihypertensive medication, diabetes, history of smoking, coronary artery disease, renal function, and lipid variables. Both scores included NT-proBNP.
Both models performed well with a C-statistic of 0.85, 95% confidence interval (CI) 0.84 to 0.87 and 0.84, 95% CI 0.83 to 0.86, respectively, and good calibration (P>0.05). Both models were recalibrated and validated with good model fit (Glostrup 0.85, 95% CI 0.83 to 0.87), Moli-Sani (0.8, 95% CI 0.78 to 0.81) for the simple model and Glostrup 0.87, 95% CI 0.85 to 0.89, Moli-Sani (0.82, 95% CI 0.81 to 0.84) for the extended model.
Conclusions
We present a heart failure risk score based on readily available clinical variables and NT-proBNP with good model fit in population-based cohorts in Europe. Both risk scores can predict heart failure almost equally well. They may serve as a benchmark for preventive efforts, screening, and analysis of additional risk factors and biomarkers.