(C-153) Mind the Gap: a literature review to assess definitions of treatment discontinuation in oncology studies

Background: Treatment discontinuation (TD) can serve as either an endpoint or reason for censoring in oncology research. In real-world evidence (RWE) studies, TD is typically defined based on exceeding an allowable gap in treatment exposure (e.g., 90 days), combined with an approach for selecting a specific day as the TD date (i.e., before or after the gap). Methodology for selection of TD date is important as it could lead to bias from either an inappropriate choice of gap length (leading to under- or over-detection of TDs), or from ignoring person-time observed during the gap by using the pre-gap date.

Objectives: To describe how TD date and allowable gap length are operationalized in oncology studies, as well as to identify guidance from methodological literature.

Methods: We performed a targeted literature review via PubMed by searching for RWE studies and clinical trials in oncology as well as methodological papers (irrespective of indication) published anytime through Apr 2024 using terms related to TD (e.g., time to [treatment] discontinuation, time on treatment). Abstracts were screened for relevance prior to full-text review. Among RWE and trial-related eligible studies only a subset underwent full-text review, aiming to cover different terms for TD (up to 10 per category) and prioritize recent and higher quality studies and diverse cancer types. For all included studies the TD definition was abstracted, as well as any rationale and justification for specific decisions relating to allowable gap length and TD date selection approach.

Results: The search identified 248 RWE studies, 842 clinical trials and 39 methodological papers, of which 28, 19, and 15 full-text articles, respectively, were included and abstracted. Among 20 RWE studies that defined discontinuation by gap time, allowable gaps of 30-120 days were used but few justifications were provided for the choice; 8 used last administration/day of supply before the gap (40%), 3 used last administration plus cycle time (15%) as the date of TD, and 9 did not provide a specific definition (54%). Among oncology trials, no gap-based definitions of TD were reported; the most common date selection approach was the last administration of treatment. The methodology studies did not provide clear guidance to select gap length or TD date.

Conclusions: Our review demonstrates there is no consensus nor sufficient empirical guidance to define TD, including gap length and TD date selection, in oncology studies.