Therapeutic drug monitoring in patients with inflammatory bowel disease on ustekinumab

Abstract

Objective: Therapeutic drug monitoring is used clinically to guide anti-tumor necrosis factor (TNF) therapy for inflammatory bowel disease (IBD), but its use for ustekinumab (UST) remains unclear. This study aimed to determine predictive variables of UST levels.

Methods: In this retrospective cohort of patients with IBD, UST trough levels were drawn at maintenance dosing. Relationships between UST trough levels and demographics, therapy, and outcomes were analyzed. Machine-learning models were used to infer combinatorial traits predictive of UST levels.

Results: Altogether 177 patients with IBD on UST had a mean UST trough level of 4.742 μg/mL. The injection schedule correlated significantly (P < 0.001) with UST levels. Naiveté to anti-TNFs correlated with higher UST levels (P = 0.048). Univariate analysis revealed that higher inflammatory biomarkers significantly correlated to lower UST levels and a lower Simple Endoscopic Score to Crohn's Disease to adequate UST levels (P = 0.018). Multivariate analysis identified body mass index (BMI), previous anti-TNF failure, and laboratory flare as predictors of UST levels with an area under the receiver operating characteristic curve (AUROC) of 0.72. The UST cut-off level of 5.77 μg/mL yielded a 0.79 AUROC, 80% sensitivity, and 81% specificity for predicting endoscopic remission of Crohn's disease. For the clinical remission end-point in ulcerative colitis, UST level of 4.73 μg/mL yielded a 0.69 AUROC, 53% sensitivity, and 86% specificity.

Conclusions: Higher UST levels correlated with less disease activity. BMI was an important consideration for UST response as well. Therefore, UST dose adjustments to reach target levels may optimize response.

Keywords: biological products; drug monitoring; inflammatory bowel diseases; machine learning; ustekinumab.