Investigating the trends in patient-reported outcomes pre-treatment and implications to efficacy analyses: A post-hoc analysis of a cancer clinical trial.

Background: Uncertainty around key elements of an appropriate patient-reported outcome (PRO) baseline assessment introduces trial-specific variation in oncology clinical trials with a poorly understood consequence on drug evaluation decisions. This research investigated the impact of multiple pre-treatment PRO assessments and timing of assessments in a clinical trial.

Methods: A post-hoc analysis of a completed phase 3, open-label, randomized, parallel arm clinical trial in non-small cell lung cancer with two pre-treatment PRO assessments (screening and Week 1 Day 1 [W1D1]). Descriptive analyses, mixed models for repeated measures and time until definitive deterioration analyses were performed to estimate differences between treatment arms. Through model adjustments, different baseline specifications and assessment timing (pre/post-randomization) on W1D1 PROs were evaluated.

Results: Patients with both pre-treatment PRO assessments were included in the analysis (N = 535). Numerically small average change scores were observed between screening and W1D1 (mean change, 0-100 scale ranges): Chest pain (-0.94), Cough (-0.94), Dyspnea (1.27), Physical functioning (-1.19). Both pre-treatment assessments were moderately-highly correlated (r: 0.55-0.78) and no trend was found for deterioration or improvement during this period. Varying baseline definitions in the models produced slight differences in model fit but no impact on the between treatment group effect estimate. W1D1 PRO scores were not statistically influenced by assessment timing pre/post-randomization (p-values: 0.142-0.628).

Conclusion: Findings from this study question the need for multiple pre-treatment PRO assessments in oncology drug development trials and the degree of bias thought to be introduced through patient knowledge of treatment assignment. Implications for researchers are presented.