AUC as a Measure of the Probability of Benefit in the Context of Randomized Controlled Trials.

• Activity Number: 512 - Predicting and Evaluating Risk Models Within Distributions and Across Time
• Type: Contributed
• Date/Time: Wednesday, July 31, 2019 : 10:30 AM to 12:20 PM
• Sponsor: Section on Risk Analysis
• Author(s): Olga Demler*
• Companies: Harvard Medical School
• Keywords: AUC; NNT; RCT; stochastic order

Abstract

Area Under the Receiver Operating Characteristics Curve (AUC) or c-statistic is a measure commonly used to compare performance of prediction models in observational studies. Colditz, Miller, Moesteller proposed to use AUC as a measure of benefit of one treatment over another in randomized controlled trials (RCT). Kraemer, Kupfer extended the definition of the Number Needed to Treat to continuous outcomes measures and related it to AUC in RCT. Following these publications, AUC was applied to the analysis of clinical trials to measure the degree of superiority of the treatment arms. AUC was also used in meta-analysis of clinical trials to measure the probability of benefit. In this talk we will discuss the role of AUC in the analysis of randomized controlled trials. We will present several situations where the AUC provides important information for our understanding of the underlying stochastic behavior in two treatment arms. Our analysis will be based on a generalization of Efron’s effect to continuous distribution functions.