Population pharmacokinetics of vadadustat, a hypoxia-inducible factor prolyl hydroxylase inhibitor for treatment of anemia associated with chronic kidney disease

Jessica Roberts (2), Pamela Navarro-Gonzales (1), Rebecca Humphrey (2), Sebastien Bihorel (2,3), Kevin Dykstra (1)

(1) Akebia Therapeutics, Inc., Cambridge, MA US; (2) Simulations Plus, Inc., Cognigen Division, Buffalo, NY US; (3) Regeneron Pharmaceuticals, Inc., Tarrytown, NY US

Objectives: Vadadustat (VADA) is a synthetic, orally bioavailable, small molecule inhibitor of hypoxia-inducible factor prolyl hydroxylase enzymes under development for the treatment of anemia associated with chronic kidney disease (CKD) in individuals who are non–dialysis-dependent (NDD) and those who are dialysis-dependent (DD). It has been approved for use in patients with NDD- and DD-CKD in Japan, and is under review in the US and EU. Initially, VADA is expected to be administered once daily, with the dose adjusted by the treating physician according to observed hemoglobin response.

Methods: A population pharmacokinetic (popPK) model of orally administered VADA was initially constructed from an integrated dataset based on 14 trials (Table 1) that included healthy volunteers (HV), patients with NDD-CKD, and patients with DD-CKD (951 participants overall) who were subjected to dense or sparse PK sampling following single doses or under steady-state treatment conditions. Doses ranged from 80–1200 mg for single doses and 120–900 mg for daily administration. The model was verified by predicting sparse concentration data from an additional four phase 3 trials including 3678 NDD-CKD and DD-CKD patients administered doses ranging from 150–600 mg daily. The analysis was carried out in NONMEM Version 7.3 (ICON, Ellicott City, MD US) using the FOCE estimation method with interaction and post-processing in KIWI (Simulations Plus, Cognigen Division, Buffalo, NY US).

Table 1. Description of VADA PopPK Model-Building and Verification Datasets

Results: A 2-compartment structural model with sigmoidal absorption (apparent zero-order infusion into a depot compartment, followed by first-order absorption into central compartment) described the data well. Inter-individual random effects could be identified on central clearance (CL), central volume (Vc), and absorption parameters. Covariate effects were evaluated relative to a typical patient with DD-CKD, and those identified as affecting CL and Vc are shown in Table 2 below.

The resulting model was then used to evaluate observed Phase 3 data estimating only the random effects (MAXEVAL=0), demonstrating the model’s accurate predictive ability in the broader target population, and allowing generation of individual exposure estimates to be further used in exposure-response analyses.

Table 2. Parameter Estimates for Best-Fit VADA PopPK Model

P, estimated effect parameter; ALB, serum albumin; BILI, total bilirubin; BW, body weight; CKD, chronic kidney disease; CL/F, apparent clearance; CV, coefficient of variation; DD, dialysis-dependent; eGFR, estimated glomerular filtration rate; h, hour; HV, healthy volunteers; IIV, inter-individual; popPK, population pharmacokinetics; NDD, non–dialysis-dependent; V_c/F, apparent central clearance; WT, body weight.

Conclusion: Although several covariate effects were identified as affecting exposure, none of the covariate effects affecting VADA exposure resulted in a significant enough change to require alteration in recommended dosing, primarily because individual dosing is guided by hemoglobin response, which varies substantially among patients and changes relatively slowly.

Reference: PAGE 31 (2023) Abstr 10421 [www.page-meeting.org/?abstract=10421]

Poster: Clinical Applications