Model for assessing cardiovascular risk in a Korean population

Abstract

BACKGROUND: A model for predicting cardiovascular disease in Asian populations is limited.

METHODS AND RESULTS: In total, 57 393 consecutive asymptomatic Korean individuals aged 30 to 80 years without a prior history of cardiovascular disease who underwent a general health examination were enrolled. Subjects were randomly classified into the train (n=45 914) and validation (n=11 479) cohorts. Thirty-one possible risk factors were assessed. The cardiovascular event was a composite of cardiovascular death, myocardial infarction, and stroke. In the train cohort, the C-index (95% confidence interval) and Akaike Information Criterion were used to develop the best-fitting prediction model. In the validation cohort, the predicted versus the observed cardiovascular event rates were compared by the C-index and Nam and D'Agostino χ(2) statistics. During a median follow-up period of 3.1 (interquartile range, 1.9-4.3) years, 458 subjects had 474 cardiovascular events. In the train cohort, the best-fitting model consisted of age, diabetes mellitus, hypertension, current smoking, family history of coronary heart disease, white blood cell, creatinine, glycohemoglobin, atrial fibrillation, blood pressure, and cholesterol (C-index =0.757 [0.726-0.788] and Akaike Information Criterion =7207). When this model was tested in the validation cohort, it performed well in terms of discrimination and calibration abilities (C-index=0.760 [0.693-0.828] and Nam and D'Agostino χ(2) statistic =0.001 for 3 years; C-index=0.782 [0.719-0.846] and Nam and D'Agostino χ(2) statistic=1.037 for 5 years).

CONCLUSIONS: A risk model based on traditional clinical and biomarkers has a feasible model performance in predicting cardiovascular events in an asymptomatic Korean population.