OBJECTIVES: In oncology trials, dose delays can uncouple dosing from nominal visits, which may impact longitudinal analyses of PRO data. This study evaluated two approaches to account for the timing of PRO assessments relative to dosing when performing analyses of oncology trial data.

METHODS: EQ-5D-3L visual analog scale (VAS) scores recorded in a phase 3 trial of nivolumab for recurrent or metastatic squamous cell carcinoma of the head and neck (CheckMate 141, NCT02105636) were aligned to dosing episodes (nominal or actual) using time-based and cycle-based strategies. With the time-based approach, contiguous windowing was used to align assessments to the nearest nominal treatment visit. The cycle-based approach assigned assessments to the nearest treatment cycle using a window of +0/-3 days. Descriptive analyses were performed to characterize trends in VAS score changes using both approaches.

RESULTS: One hundred thirty-seven subjects had >1 EQ-5D-3L assessment. For 77 (56%) of these individuals, dosing adhered to the prescribed schedule so that windowing approach did not impact analyses. However, for the remaining sixty (44%) subjects, cycles were dissociated from nominal visits with the majority (53/60, 88%) being patients with dose delays. Among these 60 subjects, 17 (28%) had ≥1 PRO assessments for which the cycle-based window excluded the nominal visit at which treatment was to be administered, 34 (57%) had ≥1 assessments not included in any cycle-based window, and the remaining 9 (15%) had ≥1 assessments meeting both of these criteria. While the choice of windowing approach impacted analytic results for nearly half of all subjects, differences were seldom large.

CONCLUSIONS: Although time-based and cycle-based windowing approaches entail different assumptions and limitations, the choice between them appears to have minimal impact on analyses of PRO data. A more important consideration is how the selected approach frames the hypotheses that can be tested.