Net Reclassification Improvement and Integrated Discrimination Improvement

The role of stress testing in the noninvasive evaluation of patients with possible or known coronary artery disease has evolved from functioning as a diagnostic tool to serving primarily as an aid in risk stratification. The results of clinical assessment and stress testing can be combined to estimate individual patient risk and guide subsequent treatment. National guidelines recommend that stress imaging preferentially be performed instead of standard exercise testing in patients with characteristics consisting of inability to exercise adequately, uninterpretable exercise ECG, or previous coronary artery disease revascularization.1,2 Aside from these patient subsets, controversy persists concerning the selection of a stress test modality attributable to uncertainty concerning the incremental prognostic value of stress imaging compared with standard exercise testing.Article see p 531For nuclear myocardial perfusion gated SPECT (MPGS), patient risk can be determined simply by dichotomizing images as normal/abnormal,3 but risk stratification can be refined by categorizing the images on the basis of summed perfusion scores.4,5 These scores include the summed stress score, which is a reflection of the extent and severity of combined infarction and ischemia, and the summed different score, an indicator of the extent and severity of ischemia.In this issue of Circulation: Cardiovascular Imaging , Candell-Riera et al6 report on the prognostic value of exercise MPGS in 5672 patients with known or suspected coronary artery disease. The authors apply the c-statistic, net reclassification improvement (NRI), and integrated discrimination improvement (IDI) to examine the incremental prognostic value of MPGS compared with clinical and exercise test variables for predicting the end points of all-cause death and the major cardiovascular end point of combined cardiovascular death and nonfatal myocardial infarction. The authors report 2 major findings: (1) predictive accuracy significantly increased by adding exercise variables to clinical variables for …