Evaluating Surrogate Endpoints: A New Trial-Level Surrogacy Metric Based on Concordance of Significance of Treatment Effects

• Activity Number: 688
• Type: Contributed
• Date/Time: Thursday, August 8, 2013 : 10:30 AM to 12:20 PM
• Sponsor: Biopharmaceutical Section
• Author(s): Qian Shi*+ and Lindsay A. Renfro and Daniel J. Sargent
• Companies: and Mayo Clinic and Mayo Clinic
• Keywords: Surrogate endpoints ; clinical trials ; trial-level surrogacy ; agreement ; concordance

Abstract

Hypothesis testing is the basis for decision-making in clinical trials, particularly from the regulatory point of view. Existing trial-level surrogacy (TLS) estimation methods do not address this concern. We propose a new TLS metric generated by assessing the agreement and concordance of conclusions based on test statistics associated with marginal treatment effect (TE) on true endpoint (T) and potential surrogate endpoint (S), denoted as zT and zS. Specifically, the intraclass correlation coefficient can be used to quantify the agreement between continuous zS and zT. To relax the linearity assumption, the area under the receiver operating characteristic curve can be used to estimate the predictive accuracy of TE significance on T based on continuous zS. Simulation studies demonstrate that these new metrics provided improved performances and were less influenced by within-trial variability, comparing to existing methods. The methods were further extended through decision curve analysis to ascertain the usefulness of the TE predictions on T based on observed zS. The proposed methods were applied to two meta-analysis databases in adjuvant and advanced colon cancers.