Study Design Complexity and Participant Completion in Dietary Trials for Inflammatory Bowel Disease: A Systematic Review and Metaresearch Study

Dietary intervention trials in inflammatory bowel disease (IBD) are difficult to implement due to poor adherence, despite their increasing importance for disease management. We conducted a systematic review to evaluate completion rates of dietary intervention trials recruiting patients with IBD and assess completion rate associations with trial design features. We systematically searched MEDLINE (Ovid), Embase (Ovid) and CINAHL (EBSCO) databases on 13 May, 2024 for randomized controlled trials examining dietary or macronutrient supplementary intake effects in patients with IBD, excluding enteral and parenteral nutrition trials. The flow of participants and prespecified trial design features were extracted. Completion rates by study arm were estimated using a multilevel mixed-effects model. Covariates were assessed via metaregression and presented as a forest plot. For each study, the risk of bias was assessed using the Cochrane Collaboration appraisal tool (RoB2) for randomized trials and randomized cross-over trials. Three main risk of bias (RoB) domains (selection and detection biases) were associated with the completion rate. These were included to explore potential biases in the reported information. In total, 62 trials comprising 122 study arms and 3523 participants were included. The overall pooled completion rate was 0.84 [95% confidence interval (CI): 0.80, 0.87] with lower rates of completion in trials published within the last 10 y. Low completion rates were associated with fecal sampling (0.75, 95% CI: 0.50, 1.00), study duration of 4-8 wk (0.79, 95% CI: 0.40, 1.00), and having a low RoB in the management of missing data (0.74, 95% CI: 0.34, 1.00). Overall, the completion rates of patients with IBD participating in controlled dietary intervention trials was high, and there was <15% variation in completion rates in relation to trial design. No linear correlation with trial duration was found. The most pronounced association with low completion was a comprehensive intervention content, i.e., fecal sampling. This study was registered at PROSPERO CRD42022327783.