Among patients with ulcerative colitis who were treated in routine practice, a point-based clinical scoring tool predicted nonresponse to vedolizumab therapy, according to study findings published online in Clinical Gastroenterology and Hepatology.
A cutoff of 26 points or less was 93% sensitive (95% confidence interval, 79%-98%) for identifying patients who did not reach corticosteroid-free remission during 26 weeks of treatment and was 88% sensitive (95% CI, 83%-92%) for identifying patients who required colectomy, reported Parambir S. Dulai, MBBS, of the University of California, San Diego, and his associates. The tool was less reliable for predicting response to tumor necrosis factor (TNF) antagonists, indicating that it is treatment specific, they noted.
Vedolizumab, an alpha-4-beta-7 anti-integrin that restricts the migration of proinflammatory lymphocytes to the gut, can induce corticosteroid-free remissions and mucosal healing in patients with ulcerative colitis. In clinical practice, 22-week rates of clinical response and remission were approximately 51% and 30%, respectively, in a recent study. Noting the lack of real-world data on predictors of response, the researchers modeled data from 620 patients who received induction and maintenance vedolizumab during the blinded phase 3 GEMINI 1 trial. They used this model to create the clinical scoring tool, which they validated in a cohort of 322 patients with ulcerative colitis who had received vedolizumab (199 patients) or TNF antagonists (123 patients) in routine practice during the Vedolizumab for Health Outcomes in Inflammatory Bowel Diseases (VICTORY) study.
In the final multivariable model, predictors of steroid-free remission were TNF antagonist naivety, at least a 2-year history of ulcerative colitis, moderate rather than severe endoscopy activity, and baseline albumin concentration. The resulting clinical scoring tool included these four variables and multiplied them by factors ranging from 0.0647 (for baseline albumin concentration) to 0.2820 (for no prior TNF antagonist exposure). In the validation cohort, patients were categorized as “high probability” (of response) if they scored 33 points or more on the tool and “low probability” if they scored 26 points or fewer. Rates of corticosteroid-free remissions were substantially different at 32% and 12%, respectively. The tool also predicted responses to vedolizumab more accurately than it did predicted responses to TNF antagonists, indicating that it was drug specific.
In the validation cohort, 46% (10 of 22) of low- or intermediate-probability patients showed least a 50% decrease in symptom activity after their vedolizumab infusion interval was shortened to address an insufficient initial response. “However, none of the high-probability patients showed a clinical response to interval shortening,” probably because they had higher vedolizumab trough concentrations to begin with, the researchers said. They called for prospective validation of this finding, “ideally in a randomized, controlled trial setting.”
The derivation and validation cohorts differed in several important ways. The validation cohort included significantly more females, smokers, patients with moderate endoscopic disease, and patients who had failed prior TNF antagonist therapy. These patients also had a significantly higher median albumin level and a longer history of disease. “The lower bound of the confidence interval for the [model’s] performance reached 0.5, suggesting that model discrimination may not be ideal,” the researchers said. “Further validation will therefore be needed to understand external validity on additional cohorts.”
An American Gastroenterological Association Research Scholar Award supported the work. Dulai reported holding a provisional patent for the prediction model and consulting relationships and other ties to Takeda, Janssen, Pfizer, and AbbVie. His coinvestigators reported ties to numerous pharmaceutical companies.
SOURCE: Dulai PS et al. Clin Gastroenterol Hepatol. 2020 Feb 13. doi: 10.1016/j.cgh.2020.02.010.
This article originally appeared on MDedge.com, part of the Medscape Professional Network.