Background The best option between vedolizumab and ustekinumab after anti-tumour necrosis factor (TNF) failure remains unclear in Crohn's disease. Aims To compare the short- and long-term effectiveness of vedolizumab and ustekinumab in Crohn's disease patients with prior anti-TNF exposure. Methods All Crohn's disease patients treated with ustekinumab or vedolizumab after exposure to at least one anti-TNF agent were included from two referral centres. Primary endpoint was corticosteroid-free clinical remission defined as Crohn's disease activity index <150 at week 54. Deep remission (corticosteroid-free clinical remission and faecal calprotectin <100 mu g/g) was assessed at week 14. Propensity-matched analyses were applied to make the two groups comparable. Results Overall, 312 patients (ustekinumab = 224 and vedolizumab = 88) were included. After propensity score analysis, ustekinumab was more effective to achieve corticosteroid-free clinical remission at week 54 (49.3% vs 41.2%, P = 0.04) and deep remission at Week 14 (25.9% vs 3.8%, P = 0.02) than vedolizumab. The rate of primary nonresponders (6.7% vs 14.8%, P = 0.034) and the long-term risk of drug discontinuation due to therapeutic failure (HR = 1.53 [1.04-2.07], P = 0.029) were lower in patients treated with ustekinumab compared with vedolizumab. Predictors of ustekinumab failure were complicated phenotype (odds ratio [OR] = 2.35 [1.31-4.22]; P = 0.004) and anti-TNF primary non-response (OR = 2.55 [1.27-5.12]; P = 0.008). We did not find any predictor of corticosteroid-free clinical remission in patients treated with vedolizumab. Vedolizumab was less effective than ustekinumab in patients >35 years old (OR = 0.41 [0.19-0.87]), with noncomplicated phenotype (OR=0.42 [0.18-0.96]), no prior bowel resection (OR = 0.49 [0.24-0.96]), and no steroids at baseline (OR=0.47 [0.23-0.97]). Conclusion Ustekinumab was more effective to achieve early and long-term effectiveness than vedolizumab in Crohn's disease patients who previously failed response to anti-TNF agents.