Bayesian prior elicitation on the efficacy of medical therapies in perianal fistulizing Crohn’s disease

Background & Aims: Robust evidence for most licensed Crohn’s disease therapies is lacking for perianal fistula outcomes due to a lack of dedicated clinical trials. This study aimed to use a Bayesian framework to determine the efficacy of medical therapies for perianal fistulizing Crohn’s disease (PFCD). 

Methods: A formal prior elicitation exercise was conducted by a group of 11 gastroenterologists and 5 statisticians. Consensus priors were developed leveraging both existing published data and clinical expertise, to determine one-year fistula remission rates for medical treatments with 5 different mechanisms of action (anti-TNF, anti-integrin, anti-IL-12/23, anti-IL-23, and JAK inhibitor). Consensus priors on efficacy of each treatment were determined relative to an elicited consensus prior for placebo control. 

Results: Consensus priors were obtained for the likelihood of fistula remission at 1 year. The prior mean, together with a 90% prior credible interval, of the one-year fistula remission rate was 0.22 (0.05, 0.46) for placebo, 0.58 (0.09, 0.96) for intravenous infliximab, 0.39 (0.06, 0.82) for adalimumab, 0.53 (0.09, 0.93) for subcutaneous infliximab, 0.24 (0.03, 0.60) for intravenous vedolizumab, 0.44 (0.05, 0.90) for upadacitinib, 0.34 (0.04, 0.77) for ustekinumab, and 0.36 (0.04, 0.82) for anti-IL-23 specific agents. Oral upadacitinib and subcutaneous infliximab demonstrated the highest probability for efficacy, alongside intravenous infliximab. 

Conclusions: We have conducted the first Bayesian prior elicitation exercise in inflammatory bowel disease. The generated priors could be used to enhance the design and analysis of clinical trials in PFCD by improving estimation of treatment efficacy, minimizing sample sizes, and potentially reducing the need for placebo control arms.