REAL-WORLD VEDOLIZUMAB PHARMACOKINETIC STUDY IN CHILDREN IDENTIFIES TWO NOVEL BIOMARKERS OF DRUG CLEARANCE

Abstract

Abstract INTRODUCTION Use of vedolizumab in pediatric inflammatory bowel disease (IBD) often requires an interval intensification from standard dosing due to distinct pharmacokinetics (PK) in children. Adult population PK models found inclusion of albumin, weight, anti-TNF naivety and antibodies-to-vedolizumab (ATV) as covariates of drug clearance (CL) improved model accuracy. We aimed to build a population-PK model for children and identify novel covariates of drug CL to better account for pediatric-specific variability in vedolizumab PK. METHODS This was a prospective multicenter cohort study for pediatric IBD patients starting vedolizumab. Vedolizumab dosing regimens were managed by the treating clinician. Stool, serum and plasma were collected at standardized intervals for robust PK sampling and biomarker analysis. Population PK analysis were conducted with nonlinear mixed effects modeling (NONMEM 7.5). A 1-compartment model with proportional error model was created based on the PK data. In the final model, novel covariates were identified using a stepwise approach. A decrease in the objective function value (OFV) >3.84 was considered significant (P<0.05). RESULTS Twenty-one patients had a total of 277 infusions with 274 measured vedolizumab peak and trough concentrations. Among these, 57% were male, 76% had Crohn’s disease, and 81% had prior anti-TNF exposure. The median (IQR) age was 16 years (13-18), and median weight was 52kg (41-71). Baseline biomarkers are described in Table 1. Baseline CL was inversely correlated with albumin (R=-0.54, P= 0.012), hemoglobin (R=-0.45, P= 0.048), and platelets (R=-0.5, P= 0.026). There was a positive correlation of CL with CRP (R=0.8, P< 0.001), neutrophil CD64 (nCD64, R=0.68, P= 0.001), fecal calprotectin (FCP, R=0.63, P= 0.009), and BSA (R=0.59, P= 0.007). The final model estimated CL as 0.168 L/d (relative standard error (RSE), 13%) and volume (V) distribution as 3.97L (8%). The interindividual variability (IIV) of the final model was 62.2% (25%) of CL and 37% (8%) of V. Covariates that significantly decreased the OFV were incorporated into the final model including weight for V and albumin, ESR, and nCD64 for CL. Prior anti-TNF, immunomodulators (MTX, n=3) or ATV (n=1) did not impact CL. Baseline CL <0.161 L/d predicted an end of induction FCP remission (<250μg/g) with an area under the curve (AUC) 0.7 (sensitivity 0.8, specificity 0.75) with the final model. A cut-off of 37 μg/mL at infusion-3 and 20 μg/mL at infusion-4 predicted end of induction clinical remission (Figure 1). CONCLUSIONS We describe a novel pediatric PK model for vedolizumab that identified novel covariates, including ESR and nCD64, that may guide model-informed precision dosing. Larger cohorts should establish whether additional compartments and pediatric-specific covariates can further decrease the remaining unexplained variability.