Resampling-Based Approach for Assessing Clinical Utility of Tumor Measurement-Based Metrics Using the RECIST 1.1 Data Warehouse

• Activity Number: 293
• Type: Contributed
• Date/Time: Tuesday, August 5, 2014 : 8:30 AM to 10:20 AM
• Sponsor: Biopharmaceutical Section
• Author(s): Ming-Wen An*+ and Sumithra Mandrekar and Yu Han and Daniel Sargent
• Companies: Vassar College and Mayo Clinic and University of Rochester and Mayo Clinic
• Keywords: RECIST ; endpoint ; trial ; clinical utility ; predictive value ; continuous

Abstract

We previously reported that absolute and relative changes in tumor measurements demonstrate potential, but not convincing, improvement over RECIST tumor response to predict overall survival (OS). We used a resampling approach to assess clinical utility of metrics to predict phase (ph) III trial outcomes from simulated ph II trials. 2000 ph II trials were simulated from 4 real ph III trials (2 positive (pos) and 2 negative (neg) for OS) using 1:1 or 2:1 randomization. Cox models (M), landmarked at 12 weeks and adjusted for average baseline tumor burden (sum of lesions/number of lesions), were fit for each ph II trial - M1: absolute change; M2: relative change; M3: RECIST response. Clinical utility was assessed by pos (neg) predictive value (PPV/NPV), i.e. probability of a pos (neg) ph II trial yielding an agent that is (in)effective in ph III trials, and prediction error. The change metrics had lower PPV/NPV than RECIST in 5 out of 6 treatment comparisons, and lower prediction error curves in 6 comparisons. However, absolute differences between metrics were negligible. Continuous change metrics had similar clinical utility as RECIST to predict ph III outcomes.