Drug-drug interaction prediction: a Bayesian meta-analysis approach

Area of Science:

• Pharmacokinetics
• Drug Discovery
• Biostatistics

Background:

• Drug-drug interaction (DDI) prediction typically relies on individual drug pharmacokinetics (PK), but subject-specific data is often unavailable.
• Standardized reporting of mean plasma drug concentrations and standard deviations in PK studies provides a basis for alternative prediction methods.

Purpose of the Study:

• To develop an innovative DDI prediction method using a three-level hierarchical Bayesian meta-analysis model.
• To leverage routinely reported PK data for more robust DDI predictions.
• To assess the performance and biases of the proposed Bayesian meta-analysis approach in DDI prediction.

Main Methods:

• Developed a three-level hierarchical Bayesian meta-analysis model incorporating study-specific sample means, random effects across studies, and prior distributions for PK parameters.
• Implemented a Monte Carlo Markov chain (MCMC) procedure for PK parameter estimation.
• Validated the method using a ketoconazole-midazolam interaction case study and statistical simulations.

Main Results:

• The Bayesian meta-analysis model accurately predicted the ratio of area under the concentration curves (a DDI marker) with less than 5% bias.
• The 90% credible interval coverage rate closely approximated the nominal level, indicating reliable uncertainty quantification.
• Sensitivity analysis confirmed the robustness of prior distribution selections.

Conclusions:

• The proposed hierarchical Bayesian meta-analysis offers a powerful and accurate approach for DDI prediction, especially when detailed PK data is limited.
• This method enhances the reliability of DDI prediction by effectively utilizing available PK study data.
• The findings support the application of advanced statistical modeling in drug interaction research for improved drug safety.