Predicting corticosteroid-free endoscopic remission with vedolizumab in ulcerative colitis

Area of Science:

• Gastroenterology
• Pharmacology
• Biostatistics

Background:

• Vedolizumab is an effective but slow-acting and expensive treatment for ulcerative colitis (UC).
• Early prediction of treatment response is crucial for optimizing patient management and resource allocation.

Purpose of the Study:

• To develop a predictive model for identifying ulcerative colitis patients likely to achieve corticosteroid-free endoscopic remission with vedolizumab.
• To enable early prediction of treatment response at a time point before the full therapeutic effect is evident.

Main Methods:

• Phase 3 clinical trial data for vedolizumab in 491 UC patients were utilized.
• Random forest machine learning models were trained and tested to predict week 52 remission.
• Models incorporated baseline data or data collected through week 6 of vedolizumab therapy.

Main Results:

• Prediction accuracy improved significantly when using week 6 data (AuROC 0.73) compared to baseline data (AuROC 0.62).
• The model correctly identified a higher proportion of remitters (59%) among predicted responders versus predicted non-responders (21%).
• Fecal Calprotectin (FCP) levels ≤234 μg/g at week 6 also showed strong predictive accuracy.

Conclusions:

• A machine learning algorithm using early (week 6) laboratory data can accurately predict long-term endoscopic remission in UC patients treated with vedolizumab.
• This predictive tool can inform clinical decisions regarding the continuation of vedolizumab therapy, especially when initial benefits are not apparent.