Search in the Abstract Database

Abstracts Search 2017

* = Presenting author

P361 Machine learning models at week 6 of vedolizumab therapy for ulcerative colitis can predict week 52 corticosteroid free endoscopic remission

Waljee A.1, Liu B.2, Sauder K.1, Zhu J.2, Govani S.1, Stidham R.W.1, Higgins P.D.*3

1University of Michigan, Internal Medicine - Gastroenterology, Ann Arbor, United States 2University of Michigan, Statistics, Ann Arbor, United States 3University of Michigan, Internal Medicine - Gastroenterology, Ann Arbor, MI, United States

Background

Vedolizumab is an effective therapy for ulcerative colitis (UC), but costly and slow to produce remission. New clinical responses continue to accumulate even after 30 weeks of therapy. Physicians, patients, and insurers want to know whether a given patient with UC will respond to vedolizumab when starting therapy, or at an early time point after starting therapy.

Methods

Through the Clinical Study Data Request Site, we obtained access to the phase 3 clinical trial data for the induction and maintenance of ulcerative colitis using vedolizumab. Random forest modeling was applied to 70% training sets and tested on 30% test sets to predict the outcome of corticosteroid-free endoscopic remission with vedolizumab at week 52. Models were constructed using baseline data, or data through week 6 of vedolizumab therapy.

Results

The original study included 895 subjects that were enrolled and included in the analysis. Subjects assigned to placebo (N=275), with missing predictor variables (N=125), or missing outcome data (N=4), were excluded. The AuROC for prediction of corticosteroid-free endoscopic remission at week 52 using baseline data was only 0.63, but was 0.73 when using data through week 6 of vedolizumab therapy. The sensitivity of this model was 72%, with a specificity of 68%. The most important predictors included fecal calprotectin at week 6, the slope of the vedolizumab level, slope of FCP, albumin, and vedolizumab level at week 6. Patients predicted to be in CS-free endoscopic remission at week 52 by the model achieved this endpoint 55% of the time, while patients predicted to fail only succeeded 19% of the time.

Conclusion

A machine learning algorithm using laboratory data through week 6 of vedolizumab therapy was able to accurately identify which UC patients would achieve corticosteroid-free endoscopic remission on vedolizumab at week 52. Application of this algorithm could have significant implications for clinical decisions on whether to continue this costly medication, whether to consider adding a co-therapy, or to change to an alternative therapy for ulcerative colitis at week 6.