Medicine and Health Sciences Commons^™

Net Reclassification Index: A Misleading Measure Of Prediction Improvement, Margaret Sullivan Pepe, Holly Janes, Kathleen F. Kerr, Bruce M. Psaty

UW Biostatistics Working Paper Series

The evaluation of biomarkers to improve risk prediction is a common theme in modern research. Since its introduction in 2008, the net reclassification index (NRI) (Pencina et al. 2008, Pencina et al. 2011) has gained widespread use as a measure of prediction performance with over 1,200 citations as of June 30, 2013. The NRI is considered by some to be more sensitive to clinically important changes in risk than the traditional change in the AUC (Delta AUC) statistic (Hlatky et al. 2009). Recent statistical research has raised questions, however, about the validity of conclusions based on the NRI. (Hilden and …

Net Reclassification Indices For Evaluating Risk Prediction Instruments: A Critical Review, Kathleen F. Kerr, Zheyu Wang, Holly Janes, Robyn Mcclelland, Bruce M. Psaty, Margaret S. Pepe

UW Biostatistics Working Paper Series

Background Net Reclassification Indices (NRI) have recently become popular statistics for measuring the prediction increment of new biomarkers.

Methods In this review, we examine the various types of NRI statistics and their correct interpretations. We evaluate the advantages and disadvantages of the NRI approach. For pre-defined risk categories, we relate NRI to existing measures of the prediction increment. We also consider statistical methodology for constructing confidence intervals for NRI statistics and evaluate the merits of NRI-based hypothesis testing.

Conclusions Investigators using NRI statistics should report them separately for events (cases) and nonevents (controls). When there are two risk categories, the …

The Net Reclassification Index (Nri): A Misleading Measure Of Prediction Improvement With Miscalibrated Or Overfit Models, Margaret Pepe, Jin Fang, Ziding Feng, Thomas Gerds, Jorgen Hilden

UW Biostatistics Working Paper Series

The Net Reclassification Index (NRI) is a very popular measure for evaluating the improvement in prediction performance gained by adding a marker to a set of baseline predictors. However, the statistical properties of this novel measure have not been explored in depth. We demonstrate the alarming result that the NRI statistic calculated on a large test dataset using risk models derived from a training set is likely to be positive even when the new marker has no predictive information. A related theoretical example is provided in which a miscalibrated risk model that includes an uninformative marker is proven to erroneously …