Use of informative prior distributions from mepolizumab data to support depemokimab PKPD model analysis
Catalina Barceló (1), Lénaïg Tanneau (1), Chayan Acharya (1), Jakob Ribbing (1), Chiara Zecchin (2), Stein Schalkwijk (2)
(1) Pharmetheus AB, Uppsala, Sweden, (2) GlaxoSmithKline, Stevenage, United Kingdom
Introduction: Depemokimab is an anti-IL-5 monoclonal antibody that blocks binding to the IL-5 receptor, leading to a rapid reduction of the blood eosinophil count. Depemokimab showed enhanced affinity and extended half-life compared to first in class mepolizumab and it is currently in Phase 3 development in severe eosinophilic asthma, Chronic Rhynosinusitis with Nasal Polyposis (CRSwNP), Eosinophilic Granulomatosis with Polyangiitis (EGPA) and Hypereosinophilic Syndrome (HES). PK and pharmacology data were collected in a single subcutaneous ascending dose first-time in human (FTIH) study in subjects with mild to moderate asthma with blood eosinophil count ≥ 200 cells/µL at screening (NCT03287310) [1]. A mepolizumab PKPD model, developed on combined data from various eosinophilic conditions, and a wide range of baseline eosinophil counts and treatment regimens, was available.
Objectives: To characterize the PKPD relationship of depemokimab with blood eosinophil count, using informative prior distributions from a previous mepolizumab PKPD model.
Methods: Depemokimab predicted plasma concentrations for the PKPD model were derived from a PK model developed on FTIH data. The starting depemokimab PKPD model was an indirect response structure, based on the mepolizumab PKPD model.
Informative prior distributions from the mepolizumab PKPD model were used to support the estimation of selected depemokimab PKPD parameters, related to the mechanism of action, where limited information was available for depemokimab, and to allow for prospective predictions. The PRIOR normal-inverse Wishart prior (NWPRI) subroutine in NONMEM allowed to maximize the efficiency of depemokimab PKPD model development by avoiding having to fix certain parameters or having to pool mepolizumab and depemokimab data [2-3].
Using SCM+, parameters implemented with priors were evaluated, by testing whether parameters implemented with priors should be estimated with prior or independently [3-5]. Subsequently, for priors on parameters where depemokimab data was significantly different from the mepolizumab prior distribution, the prior was removed from that specific parameter. The significance levels used were the same as for covariate model building (p-value of 0.01 and 0.001 for forward selection and backward elimination, respectively). FOCE method with interaction and MATRIX=R in NONMEM 7.5 was used for model development and GOFs plots and visual predictive checks (VPCs) for model evaluation [6].
Results: Depemokimab FTIH eosinophil count data was well characterized by a placebo linear model plus an indirect response model with an inhibitory Emax function on kin. Baseline blood eosinophil count (338 cells/µL, RSE 5.3%), placebo slope (-9%/year, RSE 83.4%), depemokimab Emax (89.2%, RSE 1.1%), EC50 (151 ng/mL, RSE 14.4%), Hill coefficient (2.35, RSE 9.4%) and half-life of drug on/offset effect (20.8 hours, 7.4%) were estimated without prior support and the typical estimates were aligned with previous mepolizumab and depemokimab knowledge [1,7]. Supporting priors were included on covariate parameters such as different patient populations on baseline blood eosinophil count, on Emax and on EC50, Asian population on baseline blood eosinophil count, predicted baseline blood eosinophil count on Emax and on EC50, and body weight on Emax. As per prior implementation, HES patients showed a higher blood eosinophil count at baseline and EGPA patients showed a higher EC50, compared to other patient populations. The model-predicted baseline blood eosinophil count in the FTIH analysis population did not have sufficiently wide range of values to result in a large impact on Emax and EC50. However, other patient populations are expected to have higher baseline values, so that the impact on exposure-response is predicted to be clinically relevant when considering these. Prospective predictions need to be confirmed with additional depemokimab data.
Conclusions: This example is a successful application of the prior approach to efficiently support depemokimab PKPD model development, by leveraging previous mepolizumab knowledge.
References:
[1] Singh D., et al. “A Phase 1 study of the long-acting anti-IL-5 monoclonal antibody GSK3511294 in patients with asthma”. Br J Clin Pharmacol. 2021.
[2] Gisleskog PO, et al. “Use of Prior Information to Stabilize a Population Data Analysis”. J Pharmacokinet Pharmacodyn. 2002
[3] Chan Kwong AHP, et al. “Prior information for population pharmacokinetic and pharmacokinetic/pharmacodynamic analysis: overview and guidance with a focus on the NONMEM PRIOR subroutine”. J Pharmacokinet Pharmacodyn. 2020.
[4] Svensson RJ, Jonsson EN. “Efficient and relevant stepwise covariate model building for pharmacometrics”. CPT Pharmacometrics Syst Pharmacol. 2022.
[5] Brill MJ, et al. “Confirming model-predicted pharmacokinetic interactions between bedaquiline and lopinavir/ritonavir or nevirapine in patients with HIV and drug-resistant tuberculosis. Int J Antimicrob Agents. 2017.
[6] Beal S, Sheiner L, Boeckmann A, Bauer R. NONMEM 7.5 Users Guides. (1989–2020). ICON plc; 2020.
[7] Pouliquen IJ, et al. “Characterization of the relationship between dose and blood eosinophil response following subcutaneous administration of mepolizumab”. Int J Clin Pharmacol Ther. 2015
