Therapeutic drug monitoring of vedolizumab therapy in inflammatory bowel disease

BACKGROUND: Therapeutic drug monitoring is effective for optimizing anti-tumor necrosis factor therapies in inflammatory bowel disease, but for vedolizumab, a gut-selective leucocyte migration inhibitor, data are scarce.

METHODS: Observational cohort study including 116 bio-experienced inflammatory bowel disease patients treated with vedolizumab for active luminal disease. Biobanked trough blood samples (n = 676) covering 96% of patients were analyzed using a drug-binding immunofluorometric assay. Steroid-free treatment outcomes were classified by clinical disease activity indices and objective findings, primarily endoscopy.

RESULTS: Patients with clinical remission to vedolizumab induction therapy (37%) had significantly higher trough levels than those without at weeks 6 (mean 34.1 vs 28.0 μg/mL, P = 0.03) and 10 (34.8 vs 27.5 μg/mL, P = 0.01). Optimal thresholds for discrimination were 32.4 μg/mL (AUCROC 0.66, P = 0.04) and 23.5 (AUCROC 0.67, P = 0.01), respectively. This positive association persisted during maintenance phase with 11.9 μg/mL (AUCROC 0.69, P < 0.01) associated with clinical remission (37%) and 15.3 (AUCROC 0.74, P < 0.001) for objective remission (46%). Stratification by temporal evolution of treatment effects revealed higher induction and maintenance vedolizumab levels in persistent and slow responders as compared to secondary or persistent failures. Pharmacokinetics was influenced by rare formation of anti-vedolizumab antibodies (2%), and to a lesser extent gender and albumin during induction, but not disease severity, concomitant steroids, or thiopurine metabolites. Switching to subcutaneous administrations resulted in 2.3-fold increase in steady-state trough levels.

CONCLUSION: Our study supports maintaining adequate drug exposure being essential for sustained positive outcomes of vedolizumab and emphasizes individualized, therapeutic drug monitoring-based treatment regimens. Controlled trials and pharmacokinetic modeling are, however, needed.