Abstract 18786: A Combined Risk Score for Predicting Incident Heart Failure

Abstract

Introduction: Identifying individuals at risk for incident heart failure may open the avenue for early preventive treatment. The aim of the study was to develop a straight-forward and clinically relevant risk score for predicting incident heart failure based on major clinical variables and biomarkers.

Methods: Analyses were performed in 8,239 individuals of the population-based prospective FINRISK 1997 study, recruiting men and women aged 25-74 years. During the follow-up of 12 years, we identified 497 cases of incident heart failure. We considered major cardiovascular risk factors as clinical risk factor variables. We performed multivariate model selection via LASSO penalized Cox regression to obtain models predicting 6 and 12 years absolute risk of heart failure. The added value of the biomarkers NTproBNP, BNP, MR-proANP, MR-proADM, Cystatin C, hs Troponin I, MPO and CRP to the model was assessed by C-index and continuous net reclassification index (NRI).

Results: Besides age, sex and study region (forced into the model), BMI and pulse pressure were selected for the 6 years heart failure risk model revealing an area under the curve of 0.82. Of all biomarkers tested, NTproBNP, MR-proANP, and BNP provided the strongest significant increase in C-index (0.87, 0.86, 0.85, respectively; p<0.001), followed by MR-proADM and Cystatin C (CI=0.84, p<0.001, p=0.004). Continuous NRIs for natriuretic peptides and MR-proADM were 30% - 35% (p<0.001), whereas Cystatin C showed a NRI of 10.1% (p=0.2). Inclusion of either hs Troponin I or CRP only slightly increased the predictive power of the model. Results for the 12 year model were similar, however, with attenuated NRIs, presumably due to the increased impact of age. MR-proADM was the only biomarker adding predictive information in models already including a natriuretic peptide.

Conclusions: In a population based setting, age, sex, study region, BMI, and pulse pressure were the most powerful clinical variables for predicting incident heart failure. Biomarkers, i.e. natriuretic peptides and MR-proADM, significantly increased the predictive value of the basic model. These results will be validated within the European BiomarCaRE consortium to establish a biomarker supplemented European heart failure risk score.