Eric Salgado1, Elham Amini1, Luzelena Caro1, Xiaoning Wang1
1Gilead Sciences, Inc.
Objectives: Obeldesivir (ODV) is a mono-5’-isobutyryl ester prodrug of GS-441524. Following oral administration, ODV is extensively hydrolyzed presystemically to the parent nucleoside GS-441524, which can then enter cells where it is subsequently anabolized to the active triphosphate metabolite GS-443902. ODV has shown broad preclinical antiviral activity across a panel of viruses, including SARS-CoV-2, respiratory syncytial virus, and filoviruses [1-3]. The objectives of this work are: (1) to develop a population pharmacokinetic (popPK) model of ODV to characterize its disposition following oral administration, and (2) to identify covariates which account for interindividual variability (IIV) in ODV pharmacokinetics (PK). Methods: The safety, tolerability, and PK of ODV was evaluated in five Phase 1 trials in healthy volunteers: one first-in-human study, two drug-drug interaction studies, one study in Japanese and White participants, and one renal impairment study. Additionally, safety, PK, and efficacy of ODV were evaluated in two Phase 3 trials in COVID-19 patients: one in low-risk COVID-19 patients and one in high-risk COVID-19 patients. Various compartmental models were evaluated in addition to IIV on relevant structural parameters. A full covariate modeling approach was used to assess the statistical significance as well as the clinical relevance of key covariates of interest. The popPK model was developed using NONMEM (V.7.5.0) [4] and data transformation and visualization was done in R (V4.3.1) [5]. The popPK model was evaluated based on the drop in objective function value, the precision in parameter estimates, goodness-of-fit and visual predictive check plots. Results: The analysis dataset included 9,059 concentration observations from 1,444 participants. A three-compartment PK model with linear absorption and elimination was able to adequately describe the PK data of ODV. The delay in ODV absorption was modelled as a zero-order release into the depot compartment. The residual error model consisted of a combined additive and proportional error model, and the M3 method was used to model data that were below the limit of quantification, which comprised 19% of the observations. Population estimates of the apparent central volume (1.77 L, relative standard error [RSE] = 29%), first peripheral compartment apparent volume (81.2 L, RSE = 3%) and second peripheral compartment apparent volume (26.4 L, RSE = 6%), along with the apparent intercompartmental clearances (21.9 L/hr, RSE = 5% and 0.631 L/hr, 6%, respectively) suggest that ODV is rapidly distributed into one peripheral compartment, with a slower distribution into a second peripheral compartment. IIV was included on apparent clearance and apparent central volume with coefficients of variation of 36% and 281%, respectively. This is consistent with the observed moderate to large variability in the ODV PK data, particularly at lower concentrations among both Phase 3 studies. Study-specific Phase 3 proportional residual error was implemented and significantly improved model performance. A full covariate analysis identified renal function and patient status as statistically significant covariates on apparent clearance. When compared to healthy subjects, based on a univariate covariate assessment, low-risk and high-risk COVID-19 patients are predicted to have steady-state ODV exposures (area under the curve within a dosing interval at steady state [AUCt,ss]) that are 13% and 53% higher, respectively. When compared to subjects with normal renal function, subjects with mild, moderate, or severe renal impairment are predicted to have AUCt,ss exposures that approximately 49%, 192% and 455% higher, respectively. Non significant covariates included being of Japanese ethnicity and acid-reducing agent co-administration. Conclusions: The developed popPK model was able to adequately describe ODV PK data from Phase 1 and 3 studies. Renal function and patient status were identified as significant covariates on ODV exposure. Potential impact from these two variables should be taken into consideration for application of this model to support dose justifications in special populations or other indications (eg, filoviruses).
[1] Mackman RL, et al. Discovery of GS-5245 (obeldesivir), an oral prodrug of nucleoside GS-441524 that exhibits antiviral activity in SARS-CoV-2-infected African green monkeys. J Med Chem. 2023;66(17):11701-11717. doi: 10.1021/acs.jmedchem.3c00750. [2] Cross RW, et al. Oral obeldesivir provides postexposure protection against Marburg virus in nonhuman primates. Nat Med. 2025. doi: 10.1038/s41591-025-03496-y. [3] Pitts J, et al. Oral dosing of the nucleoside analog obeldesivir is efficacious against RSV infection in African green monkeys. bioRxiv. 2025. doi: 10.1101/2025.02.24.63997. [4] ICON Development Solutions, NONMEM®. Accessed: Mar. 1, 2024. [Online]. Available: https://www.iconplc.com/solutions/technologies/nonmem [5] R Core Team, “R: A language and environment for statistical computing.” Accessed: Mar. 1, 2024. [Online]. Available: https://www.r-project.org/
Reference: PAGE 33 (2025) Abstr 11659 [www.page-meeting.org/?abstract=11659]
Poster: Drug/Disease Modelling - Other Topics