Impact of missing patient report outcomes in clinical trials for ulcerative colitis: Should we always assume treatment failure?

Abstract

Patient-reported outcomes (PRO) are key outcome measures in inflammatory bowel disease (IBD) trials. Typically, missing PRO data are imputed as treatment failures, even for patients completing trials with objective endpoints such as endoscopic data corroborating treatment effectiveness. This study investigated whether the final trial readout would change if PRO data imputation were more aligned with clinical practice. Post-hoc analyses from two negative ulcerative colitis phase 3 etrolizumab trials: LAUREL and HICKORY. Non-responder imputation (NRI), last observation carried forward (LOCF), and hybrid methods using multiple imputation were utilized to handle missing PRO data for study completers. Treatment effects (TE) on clinical remission were compared. Monte Carlo simulations were used to evaluate the performance of various imputation methods. Among completers, 15% in LAUREL and 11% in HICKORY had missing PRO data at the end of the maintenance period. The majority of patients with only PRO missing were in etrolizumab arms, especially for endoscopic responders. NRI resulted in a 3%-5% decrease in TE estimates compared to LOCF and hybrid methods. Monte Carlo simulations confirmed that NRI can underestimate TEs and reduce power. While LOCF can mitigate this bias and power loss, the hybrid method produces the most unbiased and robust results. Appropriately handling missing PRO data in patients completing treatment can reduce bias and improve the validity and reliability of efficacy analysis, potentially enhancing the ability of detecting therapeutic effects. This strategy should be considered in future clinical trials of IBD. LAUREL (NCT02165215) and HICKORY (NCT02100696).