688 Probability of Vedolizumab Response as Defined by Clinical Decision Support Tool Is Associated With Lower Risk of Serious Infection in Patients With Inflammatory Bowel Disease

Abstract

INTRODUCTION: Disease activity is an independent predictor of infection in patients receiving biologics; non-response to a therapy might therefore present a higher risk of serious infection due to persistence of disease activity. We hypothesized that a clinical decision support tool (CDST) predicting lower treatment response may be predictive of a higher risk of infection. METHODS: Two validated CDSTs predicting vedolizumab (VDZ) response in ulcerative colitis (UC) and Crohn's disease (CD) were used on GEMINI phase 3 clinical trial data. After stratification by baseline variables into low, intermediate, or high probability of response to VDZ, time-to-event with Cox proportional hazard models was used to compare the risk of infection across groups and to identify independent predictors of infections among subgroups. RESULTS: 618 UC and 1023 CD patients were included. Those categorized as having low-intermediate probability of response (n = 479 UC, n = 814 CD) to VDZ were more likely to develop infectious complications while on therapy than those categorized as having high probability (n = 139 UC, n = 209 CD) response to VDZ (UC: hazard ratio [HR] 1.383, 95% confidence interval [CI] 1.024−1.868; CD: HR 1.458, 95% CI 1.124−1.893). Serious infections were observed in 1% of UC and 3% of CD patients in the high probability of response groups vs 2% of UC and 5% of CD patients in the low-intermediate groups. Among the low-intermediate probability of response UC group, independent predictors of developing infection were: baseline extraintestinal manifestations (no vs yes; HR 0.755, 95% CI 0.569−1.001); previous tumor necrosis factor antagonist exposure (no vs yes; HR 0.671, 95% CI 0.497−0.906); and baseline fecal calprotectin (log transformed; HR 1.135, 95% CI 1.016−1.268). Among the low-intermediate probability of response CD group, an independent predictor of developing infection was exposure to opioids during treatment (yes vs no; HR 1.829, 95% CI 1.469−2.277). CONCLUSION: This analysis confirms a pattern of inverse correlation between CDST-predicted efficacy and risk of infection. Data are consistent with the VDZ safety profile and suggest that infectious events seen were related to persistence of disease activity as opposed to drug-related mechanisms.