Net reclassification improvement: a link between statistics and clinical practice

Excerpt

A little over a decade ago, European [1] and US guidelines [2] on cardiovascular disease prevention first recommended the use of various global risk assessment models. These guidelines directly aid clinicians in making decisions on consultation of a healthy lifestyle and initiation of drug treatment. The recommendations to use risk scoring algorithms invigorated researchers to improve on these existing functions and thereby heralded the current upswing in risk prediction research. Ever since a plethora of additional risk factors have been proposed, ranging from simple questions on familial predisposition and laboratory measures, to state-of-the-art vascular imaging. In order to provide guidance the American Heart Association issued a comprehensive statement on the stepwise evaluation of the value of novel markers [3]. The expert panel suggested that a new marker should be prospectively associated with the outcome, it should add predictive information over established risk factors, the addition of the marker should have the potential to modify an individual’s risk sufficiently to change treatment recommendations, and finally whether this improves clinical outcomes in a cost-effective manner. …