The benefits of advanced risk reclassification

Abstract

In years past, the value of detecting risk was measured largely by examining a clear separation among survival curves for key patient subgroups. In 2008, prognostic modeling was advanced with the introduction of risk reclassification analyses which allow one to calculate the incremental improvement in risk quantification for lowand high-risk individuals based on the addition of new markers to a risk-prediction model. A key clinical benefit to the net reclassification improvement (NRI) analyses is that the results estimate the proportion of lowand high-risk individuals that are reclassified when two prognostic models are compared. This approach has now been successfully applied in a number of prognostic analyses using nuclear imaging. Although most reports quantifying the overall NRI have reported that nuclear variables add considerably to the ability to accurately detect lowand high-risk individuals, in this issue of the journal, Koh and colleagues reported intriguing results from a relatively large, singlecenter database of 6702 patients referred for exercise myocardial perfusion imaging (MPI) for either suspected or known coronary artery disease (CAD). This report provided a unique approach by formulating three models including the base comparator of a clinical model which was then compared to a model adding the Duke Treadmill Score (DTS), and then a third model adding the summed difference score and a categorical post-stress left ventricular ejection fraction variables. This approach is interesting as it provides some insight into the comparative utility of exercise testing information alone compared with nuclear imaging. This analysis revealed that in patients with suspected coronary artery disease, the addition of DTS provided significant improvement in the NRI beyond clinical data alone; with an NRI of 0.12 (or 12%, = 0.019). However, in this lower-risk patient subset, MPI variables did not result in a significant improvement in the NRI calculation when added to DTS and clinical variables (P = 0.9). These results are in line with prior reports in patients with low cardiovascular risk. A prior multicenter series reported no improvement in risk estimation in women \55 years of age using pharmacologic stress myocardial perfusion Rb-82 PET. In the paper by Kohl et al, the primary endpoint for analysis was all-cause death. One may argue that more cardiacspecific outcomes (such as nonfatal MI, cardiac arrhythmias, need for percutaneous or surgical revascularization, and hospitalization for heart failure) may yield different results. These findings are similar to the results from the WOMEN (What is the Optimal Method for Ischemia Evaluation in Women) trial which demonstrated that ETT was an effective initial diagnostic test in the evaluation of low-risk, functional women with suspected CAD. In the WOMEN trial, 824 symptomatic women with suspected CAD were randomized to either ETT or exercise MPI. There was no difference in the 2-year MACE-free survival event rates between the two groups (P = 0.59). The lack of significant improvement in the NRI in lower-risk cohorts is also supported by current guidelines which recommended exercise testing (without imaging) as the index strategy for evaluation of suspected ischemic symptoms in patients with low to intermediate pretest likelihood of CAD. However, in patients with prior CAD, the Koh report noted that the NRI was 0.26 (or 26%, P 0.05) when MPI variables were added to clinical variables and DTS, a value similar to other published reports. Importantly, the clinical model only included a few Reprint requests: Leslee J. Shaw, PhD, Division of Cardiology, Emory Clinical Cardiovascular Research Institute (ECCRI), Emory University School of Medicine, 1462 Clifton Rd NE, Room 529, Atlanta, GA 30324; lshaw3@emory.edu J Nucl Cardiol 2016;23:384–6. 1071-3581/$34.00 Copyright 2015 American Society of Nuclear Cardiology.