Advances in measuring the effect of individual predictors of cardiovascular risk: the role of reclassification measures

Abstract

Models for risk prediction are widely used in clinical practice to stratify risk and assign treatment strategies. The contribution of new biomarkers has largely been based on the area under the receiver-operating characteristic curve, but this measure can be insensitive to important changes in absolute risk. Methods based on risk stratification have recently been proposed to compare predictive models. Such methods include the reclassification calibration statistic, the net reclassification improvement, and the integrated discrimination improvement. This article demonstrates the use of reclassification measures and illustrates their performance for well-known cardiovascular risk predictors in a cohort of women. These measures are targeted at evaluating the potential of new models and markers to change risk strata and alter treatment decisions.

Figure

Plot of predicted 10-year risk (%) from models including cardiovascular risk factors but without and with systolic blood pressure (SBP) in the Women’s Health Study. Cases as of 8 years are denoted by triangles, and non-cases are denoted as circles. Solid line is line of unity; dashed lines represent risk strata cut points.