Assessing the impact of informative censoring on overall survival in a RWD source.

• Activity Number: 438 - Real-World Evidence in Clinical Trial: New Era of Informed Decision Making Session
• Type: Invited
• Date/Time: Wednesday, August 2, 2017 : 8:30 AM to 10:20 AM
• Sponsor: Biopharmaceutical Section
• Author(s): Bill Capra* and Gillis Carrigan and Michael Taylor and Brandon Arnieri and Melissa Curtis and Sandy Griffith and Aracelis Torres and Philip Hofmeister and Anala Gossai and Samuel Whipple and Peter Lambert
• Companies: Genentech Inc. and Genentech Inc. and Genentech Inc. and Genentech Inc. and Flatiron Health and Flatiron Health and Flatiron Health and Flatiron Health and Flatiron Health and Genentech Inc and Genentech Inc
• Keywords: RWD

Abstract

The assumption of non-informative censoring is necessary for the validity of Kaplan-Meier estimates of overall survival. The Flatiron Health (FIH) electronic health record (EHR) is an oncology database with mortality data that offers a number of research opportunities including evaluating the benefits of different treatment options. Real-world data sources such as FIH provide an opportunity to derive estimates of overall survival from populations where limited clinical trial data exist, but concerns exist regarding the level of completeness. If mortality information is missing from an EHR due to reasons related to disease progression, estimates of treatment effects on overall survival may be biased. For a random subset of 7579 patients with advanced non-small cell lung cancer (aNSCLC), the mortality data collected in the FIH EHR has been linked to the mortality data collected in the National Death Index (NDI). We present analyses that assess the presence of informative censoring on the mortality data of the FIH aNSCLC cohort, and the resulting impact on drug development decision making.