Model-Informed selection of doses and sample size for a Phase 2a POC study of GSK3640254, a next generation HIV-1 maturation inhibitor
Chetan Rathi (1), Mark Johnson (2), Samit Joshi (3), Geraldine Ferron-Brady (1)
(1) GlaxoSmithKline, Collegeville, PA, USA, (2) ViiV, RTP, NC, USA, (3) ViiV, Branford, CT, USA
Introduction: GSK3640254 (MI-254) is a next generation HIV maturation inhibitor (MI) that prevents maturation of HIV-1 virions by binding near a key structural element within the Gag polyprotein that is required for virion maturation and assembly. Prior validation of this target was demonstrated with bevirimat and GSK3532795 (MI-795) which were later terminated in Phase 2a for treatment-emergent resistance and Phase 2b studies for GI intolerability respectively. MI-254 has improved coverage of baseline Gag polymorphisms over the prior developmental MIs. The safety, tolerability, and PK of MI-254 was investigated following single (1 to 700 mg) and repeated daily (50 to 320 mg for two weeks) administration in a Phase 1 First Time In Human (FTIH) study (NCT03231943).
Objectives: The aim of this model-based analysis was to select a) doses and b) cohort size for a Phase 2a POC study (NCT03784079) to evaluate short term maximum HIV-1 RNA decline in monotherapy and to characterize Dose-Exposure-Response relationship.
Methods: To predict the anticipated individual decline in HIV-1 RNA over a wide range of possible doses for the POC study, a model-informed framework was developed using the exposure-response (ER) model for the POC study for MI-795, the in vitro potency distribution for MI-254, and the PK profile of MI-254. The ER model links the individual inhibitory quotient (IQ) (trough concentration divided by the individual protein binding adjusted EC90 (PBAEC90)) to the maximum decline in HIV-1 RNA following 10 days of MI once-daily monotherapy, the primary endpoint in the POC study. This ER relationship was characterized using an Emax model composed of the following parameters: Reduction in HIV-1 RNA from baseline (E), placebo HIV-1 RNA decline (E0), maximum HIV-1 RNA decline (Emax) and IQ50 all obtained from the MI-795 model and the individual IQ for MI-254. These individual IQ were obtained by predicting MI-254 trough concentrations using a PopPK model based on the FTIH data and sampling from the distribution of MI-254 PBAEC90 obtained from in vitro studies. These in vitro PBAEC90 are considered representative of the distribution of clinical PBAEC90 values based on experience with MI-795. Reduction in HIV-1 RNA (E) was simulated with both variability and uncertainty included on PK parameters and with variability on PBAEC90 for 1000 trials with 6 and 8 subjects per cohort for a wide range of doses (5 to 200 mg QD). %Emax was summarized in terms of 2.5th, 50th and 97.5th percentiles achieved in each trial for each dose level. Finally, % Emax was summarized in terms of 2.5th, 50th and 97.5th percentiles of the 1000 median values obtained in the previous step. The FTIH PK data were analyzed using non-linear mixed effects modeling implemented in NONMEM V7.3.0. Clinical trial simulations were conducted using the mrgsolve package in R .
Results: A 2-compartment PK model with first order absorption and lag time adequately characterized the pharmacokinetics of MI-254 in healthy volunteers. The typical values of the model are as follows: clearance (CL/F, 7.99 L/h), volume of distributions (V2/F and V3/F, 197 and 81.8 L, respectively), inter-compartmental clearance (Q/F, 15.1 L/h) and first-order absorption rate constant (Ka, 0.912 h-1) with lag time (0.894 h). Inter-individual variability (IIV) for clearance, central volume of distribution and absorption rate constant ranged from 30.9 - 73.3% with uncertainty on PK and IIV parameters ranging from 5.9 - 34.6%. Based on simulated virtual trials for HIV-1 RNA decline, doses of 5 mg, 10 mg, 40mg, 100 mg and 200 mg were identified as doses providing ~30%, 50%, 80%, 90% and >=95% of maximum effect respectively. Simulations also suggested that increasing cohort size from 6 to 8 did not provide any additional benefits.
Conclusions: Model informed drug development (MIDD) approach provided an integrated pharmacometric framework in conjunction with historical clinical data to guide selection of doses and cohort size for the POC study. Simulation of virtual trials allowed a recommendation of 5 mg, 10 mg, 40 mg, 100 mg and 200 mg doses with a cohort size of 6 for the POC of GSK3640254 in HIV infected patients.
 Beal SL, Sheiner LB, Boeckmann AJ, Bauer RJ (Eds.) NONMEM User Guides. 1989-2015. Icon Development Solutions, Ellicott City, Maryland, USA.
 Baron KT, Hindmarsh AC, Petzold LR, Gillespie B & Margossian C. Simulation from ODE‐based population PK/PD and systems pharmacology models in R using mrgsolve.