|Thursday, February 20|
|PS1 Poster Session I & Opening Mixer||
Thu, Feb 20, 5:15 PM - 6:45 PM
Statistical and data challenges with modeling continuous longitudinal tumor measurements as phase II endpoints for predicting overall survival (302789)Ming-Wen An, Vassar college
*Sumithra Jay Mandrekar, Mayo Clinic
Daniel J Sargent, Mayo Clinic
Keywords: RECIST, tumor-measurement, endpoints, prediction, c-index
Tumor response is a common Phase II cancer clinical trial endpoint, and in solid tumors, is assessed via the Response Evaluation Criteria in Solid Tumors (RECIST), which categorizes continuous tumor measurements. Some of the challenges with these data include: no baseline assessments, no post-baseline assessments, lesions not consistently followed across time, missed assessments or conflicting measurements at a given assessment, no measurements at the time of progression, measurements based only on clinical evaluations etc. We will present the work we did with identifying alternative tumor-measurement based metrics (both categorical and continuous) as Phase II endpoints using the largest database of ~8000 patients with cycle-by-cycle and lesion-by-lesion tumor measurements from 13 Phase III trials in breast, non-small cell lung, and colorectal cancers. Statistical challenges encountered with modeling these data will also be presented: landmark analysis versus time dependent models, complete case analysis versus missing data imputation, approaches for model / metric selection, and multicollinearity (Supported in part by the National Institute of Health grant: CA167326).