P5824HYCHeF: a risk score to predict congestive heart failure incidence in hypertension. (based on 18795 individual patient data)

Objectives: Patients with hypertension present risk of heart failure development. We aimed to build here a risk score to predict congestive heart failure (CHF) incidence in these patients using data from well-done randomized controlled trials (RCTs) of anti-hypertensive agents. Method: Data of 18795 hypertensive patients was collected from four trials of INDANA database (Coope, SHEP, STOP and SYST-EUR. The whole population was randomly divided in two subsets: derivation set for model construction and the remaining for model validation with the ratio 2:1 and with similar patient characteristics. We used the Cox proportional hazards model to test the association of each covariable to CHF incidence separately (univariable analysis) and altogether (multivariable analysis) until obtaining the final model where all covariables were significant. Results: Thirty baseline factors were examined for risk of CHF incidence using a multivariable Cox model, adjusted on trials. Treatment was associated with a significant reduction of 50% of CHF incidence and had no interaction observed with trial or any other covariable. A risk score of CHF incidence in five years was built with 10 significant risk factors: treatment (HR 0.50, CI 95%, 0.36 - 0.69), age (HR 1.52, 1.3 - 1.78 by 5-year increase), serum total cholesterol (HR 0.83, 0.73 - 0.96), cigarette smoking (HR 1.74, 1.10 - 2.75), renal function (HR 1.52, 1.3 - 1.78 by 5 mL/min/1.73 m2 increase of glomerular filtration rate), body mass index (HR 1.05, 1.01 - 1.09 by 1 kg/m2 increase), history of myocardial infarction (HR 3.07, 1.76 - 5.34), history of stroke (HR 3.07, 1.76 - 5.34), diabetes baseline (HR 1.63, 1.01 - 2.63) and ratio of systolic/diastolic blood pressure (HR 1.36, 1.17 - 1.58). This model was then transformed to an integer system, with points added for each factor according to its association with CHF incidence risk.