|
Abstract:
|
Adverse events (AEs) in cancer trials have historically been measured by clinicians on ordinal scales (none, mild, moderate, severe, and life threatening) using the National Cancer Institute's (NCI's) Common Terminology Criteria for AEs (CTCAE). While traditional CTCAE statistical analysis has primarily involved use of summary measures (e.g., maximum grade post-baseline), a variety of other approaches including cumulative incidence estimation in the presence of competing risks have been recommended. AE data collected directly from the patient using NCI's Patient-Reported Outcomes version of the CTCAE (PRO-CTCAE) are similar in format to CTCAE data; however, PRO-CTCAE data may introduce statistical challenges due to high baseline symptom rates and non-ignorable intermittent missing data. In this presentation, potential PRO-CTCAE statistical analysis approaches will be described and compared via simulation including a novel cumulative incidence model for ordinal data. Existing statistical methods for CTCAE data as well as approaches to PRO data such as generalized linear mixed modeling can be considered as candidate methodologies for the statistical analysis of PRO-CTCAE data.
|
