Prediction of 10-year risk of incident heart failure in elderly hypertensive population: The ANBP2 study

Abstract

© American Journal of Hypertension, Ltd 2016. All rights reserved.BACKGROUND Multivariable risk prediction models consisting of routinely collected measurements can facilitate early detection and slowing of disease progression through pharmacological and nonpharmacological risk factor modifications. This study aims to develop a multivariable risk prediction model for predicting 10-year risk of incident heart failure diagnosis in elderly hypertensive population. METHODS The derivation cohort included 6083 participants aged 65 to 84 years at baseline (1995-2001) followed for a median of 10.8 years during and following the Second Australian National Blood Pressure Study (ANBP2). Cox proportional hazards models were used to develop the risk prediction models. Variables were selected using bootstrap resampling method, and Akaike and Bayesian Information Criterion and C-statistics were used to select the parsimonious model. The final model was internally validated using a bootstrapping, and its discrimination and calibration were assessed. RESULTS Incident heart failure was diagnosed in 319 (5.2%) participants. The final multivariable model included age, male sex, obesity (body mass index > 30 kg/m2), pre-existing cardiovascular disease, average visit-to-visit systolic blood pressure variation, current or past smoking. The model has C-statistics of 0.719 (95% CI: 0.705-0.748) in the derivation cohort, and 0.716 (95% CI: 0.701-0.731) after internal validation (optimism corrected). The goodness-offit test showed the model has good overall calibration (?2 = 1.78, P = 0.94). CONCLUSION The risk equation, consisting of variables readily accessible in primary and community care settings, allows reliable prediction of 10-year incident heart failure in elderly hypertensive population. Its application for the prediction of heart failure needs to be studied in the community setting to determine its utility for improving patient management and disease prevention.