Patient-reported outcomes (PRO) in cancer trials submitted to the FDA from 2012-2015.

Abstract

e14024 Background: PRO measures are commonly assessed in cancer trials. We reviewed the PRO strategy, tools and trial designs for new drug applications (NDA) and biologics license applications (BLA) submitted to FDA over a 4 year period. Methods: A review of protocols and clinical study reports for original NDA and BLA applications submitted to the Office of Hematology and Oncology Products between 2012 and 2015 to support initial approval for adult malignant hematology and oncology conditions was done. We reviewed applications for inclusion of PRO data, trial design, type of PRO measure employed and statistical analysis methods. Results: Forty three trials were submitted to support 40 original NDA or BLA approvals targeting adult malignancies between 2012 and 2015. Of these 43 trials, 17/43 (40%) were accelerated approval, 26/43 (60%) were randomized, 17/43 (40%) were single arm and 27/43 (63%) were open label trials. Sponsors documented the incorporation of PRO assessments in 28/43 (65%) trials. For trials that included PRO assessments, 22/28 (79%) were randomized controlled trials, 6/28 (21%) were single arm and 17/28 (61%) were open label studies. The most common PRO instruments used were the EORTC-QLQ-C30 (15/28; 54%), EQ-5D (13/28; 46%) and various FACIT measures (8/28; 29%). Although 20/28 (71%) trials had PRO measures listed as a secondary endpoint, only 1 trial included PRO endpoints in the statistical testing hierarchy. Conclusions: PRO measures are often employed in randomized controlled cancer trials; however accelerated approval is common in oncology and trial designs are increasingly open label and single arm. Patient-focused drug development efforts will need to identify clinical trial objectives and analysis methods for PRO measures to describe symptoms and function that are suitable for these contexts. Descriptive PRO data on the tolerability of an anti-cancer agent may be one objective that is relevant across trial contexts. To support a claim of superiority, PRO endpoints should be adjusted for multiplicity by inclusion in the statistical hierarchy.