05 with stratification according to previous TNF antagonist status, concomitant corticosteroid use, and concomitant immunosuppressive use. The Cochran–Mantel–Haenszel chi-square P value, risk difference (primary test), and associated 2-tailed 95% confidence intervals (CIs) were determined, as were the relative risk and its 2-tailed 95% CI. Secondary analyses were performed sequentially, with a P value of .05 or less required to proceed to

testing of each subsequent outcome. Of the 6 secondary analyses, 4 (ie, 2 pairs of outcomes, each pair evaluating 1 end point for the 2 populations) involved simultaneous testing for the TNF antagonist–failure and overall populations ( Supplementary Figure 1). The Hochberg method was applied to each secondary outcome pair to maintain the overall type 1 error rate at a P value of .05 or less. A logistic regression model, including baseline

CDAI score, stratification factors, and geographic CDK activation region, was conducted as a sensitivity analysis using the chi-square test at a statistical significance level of 0.05; the chi-square P value and odds ratio, with associated 95% CIs, were determined. Analysis of covariance models of change from baseline to week t for the continuous efficacy outcome variables in the vedolizumab and placebo groups 5-FU in vitro was performed. For the prespecified exploratory analyses of TNF antagonist–naive patients and for those based on concomitant corticosteroid or immunosuppressive use, P values were determined and 95% CIs were calculated using the exact method (for categoric data with numerators ≤5) or the normal approximation. Power estimates for the primary and secondary outcomes were 91% and 81%–93%, respectively, on the basis of total sample sizes of 296 for the TNF antagonist–failure population and 396 for the overall population. A total of 660 patients were screened (Figure 2), of whom 244 were excluded because of not meeting enrollment criteria (n = 209), withdrawal of consent (n = 11), having an SAE (n = 5), having

a protocol violation (n = 1), or other/unknown reasons (n = 18). Parvulin Of 416 randomized patients, 315 (76%) had previous failure of (ie, inadequate response to, loss of response to, or intolerance of) 1 or more TNF antagonists, and 101 patients (24%) were TNF-antagonist naive. Demographic characteristics (Table 1) generally were similar between treatment groups in the TNF antagonist–failure population. Corticosteroids were the most common concomitant medications used at any time during the study (54% of patients), followed by immunosuppressives (34%) and mesalamine (31%). Previous immunosuppressive exposure was reported by 89% of patients. In the TNF antagonist–failure population, 2 or more TNF antagonists had failed in 66% of patients (44% of whom had a primary nonresponse), whereas 3 TNF antagonists had failed in 11% of patients.