Objective

Otenaproxesul (ATB-346), a drug that combines naproxen with a thiobenzamide antioxidant, is being developed as an NSAID that reduces gut toxicity effects. Liver toxicity signals were observed in early clinical studies, mostly after dosing ceased; liver signals appear almost exclusively in individuals with fatty liver. In order to mechanistically explain the observed toxicity and predict potentially safe dosing regimens, a quantitative systems toxicology (QST) representation of otenaproxesul, its main metabolite M25 (naproxen), and the H2S released by the thiobenzamide moiety was implemented in DILIsym v8A, a QST model of drug-induced liver injury.

Methods

Otenaproxesul and its metabolites H2S and naproxen were implemented as compounds in DILIsym v8A, a QST platform model of drug-induced liver injury. A PBPK model for otenaproxesul, naproxen, and H2S was constructed in GastroPlus 9.8 and used as an input into DILIsym. Otenaproxesul and naproxen (M25) were assessed in a series of in vitro experiments that found oxidative stress signals for naproxen; these were translated into inputs into DILIsym as well. H2S was implemented as a scavenger of both ambient and induced oxidative stress (ROS). Simulations of existing clinical trials in both normal healthy volunteers (NHV), post-menopausal women (PMW), and metabolism-associated fatty liver disease (MAFLD) patients were used as comparators for the model results; for modeling of the rebound effect hypothesis, some of these trials were used as calibration for the de-adaptation effect.

Conclusion

A QST model for otenaproxesul was successfully constructed in DILIsym and was used to explain the clinically observed post-dosing ALT elevations. The ALT elevations were hypothesized to be due not to the direct actions of naproxen itself but to the fact that the H2S scavenges enough ambient ROS to cause individuals with high baseline ROS to de-adapt to that baseline ROS, leading to a spike in ROS and liver injury when the drug is removed. This hypothesis was able to explain the observed ALT elevations, and was able to prospectively predict the safety of a novel clinical otenaproxesul dosing protocol. In addition, the model predicted that the most at-risk individuals are those with mild MAFLD-induced ROS (right); these individuals were the ones in which toxicity was observed clinically. This demonstrates the potential for QST modeling to explore novel hypotheses of toxicity and make successful predictions based on those hypotheses.

By Nader Hamzavi, Pallavi Bhargava, Vinal V. Lakhani, David Vaughan, Daniel Legault, Joseph Stauffer, Brett A. Howell, Jeffrey L. Woodhead

Fifteenth American Conference of Pharmacometrics (ACoP15), November 10-13, 2024, Phoenix, Arizona