Clinical remission with vedolizumab (VDZ) can be achieved in over one-third of patients with moderately to severely active Crohn's disease (CD). Identifying individuals who are more likely to achieve clinical remission could improve the overall effectiveness of VDZ. Our aim was to develop and validate a prediction model for achieving clinical remission after 26 weeks of treatment with VDZ in these patients.Methods:
Participant level data from the GEMINI II clinical trial were used for model derivation and validation. Participants were included if they: (1) completed VDZ induction therapy and (2) continued VDZ maintenance therapy. To mimic a treat-straight-through design, Week 6 responders and non-responders were both included. A random 40% sampling (n = 325, derivation cohort) was selected for model derivation after accounting for trial-specific stratification factors. Multivariate logistic regression analyses and backward model selection were performed for baseline variables and variable interactions found to be significant (P < 0.05) on univariate analyses. Adjusted odds ratio (OR) is presented for predictors included in the model, with OR >1 indicating an increased probability of achieving clinical remission at Week 26. The prediction model was then applied to the remaining 60% sampling (n = 489, validation cohort) for calibration and validation. Discrimination performance of the model is presented as area under the curve (AUC) from receiver operating characteristic curves.Results:
In the derivation cohort, CD participants were more likely to achieve clinical remission after 26 weeks of VDZ therapy if: they had no active extraintestinal manifestations at baseline (OR 1.81; 95% CI, 1.03–3.21), they had previously never been hospitalized (OR 3.14; 95% CI, 1.19–8.28) or never undergone bowel surgery (OR 1.99; 95% CI, 1.10–3.59), and they were naïve to anti-TNF therapy when starting VDZ (OR 6.16; 95% CI, 2.80–13.56). An elevated baseline CRP was incrementally associated with a reduction in the probability of achieving clinical remission in anti-TNF-naïve participants (OR 0.94 per mg/L increase; P-value < 0.05), but not anti-TNF-exposed/failure participants (OR 1.00; 95% CI, 0.98–1.02). The discrimination performance for Week 26 clinical remission was fair (AUC = 0.75). The discrimination performance of the model decreased when applied to the validation cohort (AUC = 0.69). A calibration was done within the validation cohort by adding baseline CD Activity Index (OR 0.993 per point; 95% CI, 0.989–0.997) into the model. The final discrimination performance for Week 26 clinical remission within the validation cohort was fair (AUC = 0.73).Conclusions:
We created and validated a prediction model for week 26 clinical remission with VDZ in patients with moderately to severely active CD. This model suggests that the optimal positioning of VDZ may be early in the disease course, prior to the use of anti-TNF therapy or development of CD-related complications (hospitalization or surgery). Baseline CRP did not impact clinical remission in anti-TNF exposed/failure participants; therefore, immunological mechanisms other than inflammatory burden may be responsible for the reduction in VDZ efficacy seen in participants with prior exposure/failure to anti-TNF therapy. Further studies are needed to assess the coherence of these predictors with other treatment goals (e.g., mucosal healing).