Lin Yuan, MS: No financial relationships to disclose
Arijit Chakravarty: No relevant disclosure to display
Objectives: While the primary objective of a Phase I trial is to determine a novel therapeutic’s maximum tolerated dose, early evidence of drug effect is often desired. In this analysis, we use a Bayesian pharmacokinetic / pharmacodynamic (PK/PD) modeling framework to develop the relationship between concentrations of small molecule STING agonist SNX281 and interferon-beta (IFN-beta) release in a Phase I trial enrolling lymphoma and solid tumor patients (NCT04609579).
Methods: In prior work, we identified peak SNX281 concentrations (Cmax) as the driver of IFN-beta release in primates. This relationship between Cmax and IFN-beta induction was best described by a logistic function and serves as a prior in the present analysis. To determine each participant’s Cmax in the clinical trial, we fitted a population pharmacokinetic model using Phoenix NLME. We then used the primate PK/PD relationship as a prior to fit the human PK/PD relationship. The Bayesian model’s fit and predictive utility were assessed based on the confidence intervals for the PK/PD relationship and parameter standard errors.
Results: We demonstrate that the human Cmax-IFN-beta relationship for SNX281 is well-described by a logistic function. We achieved reliable estimates of the human maximal IFN-beta induction and the drug’s half-maximal effective concentration (EC50) based on small standard errors. The Bayesian model achieves much narrower confidence intervals compared to the same model fitted without a prior to the clinical data alone.
Conclusions: Using Bayesian techniques, we were able to establish the relationship between peak SNX281 concentrations and IFN-beta induction for SNX281. Applying Bayesian modeling techniques to Phase I PK/PD data using a preclinical prior can allow PK/PD relationships to be established with more confidence even if data is limited.
