Towards an integrated mPBPK/PD model for drug-optimization and prediction of relapse after treatment of tuberculosis in mice
Daniele Boaretti (1), Roberto Visintainer (1), Micha Levi (2), Shayne Watson (2), Luca Marchetti (1, 3), and Federico Reali (1)
(1) Fondazione The Microsoft Research - University of Trento Centre for Computational and Systems Biology (COSBI), Rovereto, Italy, (2) Bill & Melinda Gates Medical Research Institute, Cambridge, MA, USA, (3) University of Trento, Department of Cellular, Computational and Integrative Biology (CIBIO), Italy.
Objectives: The treatment of tuberculosis (TB), one of the most harmful infectious diseases killing millions of people worldwide [1], still needs to be improved to shorten its duration and reduce relapse events. In pursuit of this objective, combining experimental data from mouse model treatments with simulation tools offers a strategic approach to identifying the most effective treatments for eradicating Mycobacterium tuberculosis infection. Therefore, herein, we present the extension of a recently published minimal physiologically based pharmacokinetic (mPBPK) model [2] with the pharmacodynamics (PD) effects on bacteria populations. This enhanced model provides a mechanistic insight into a drug's bactericidal and bacteriostatic actions on various bacterial communities.
Methods: The current PD expansion of the model added one population of growing bacteria. The simulated drug exposure triggers bacteriostatic and bactericidal effects controlling the bacterial population quantified, at every simulated time point, as colony forming units (CFU). At this development stage, the model adopts literature parameter values reported previously [3], further refined by fitting literature CFU data using the Covariance Matrix Adaptation - Evolution Strategy (CMA-ES). We used the programming language R [4] and LSODA in rxode2 [5] as a numerical solver for the numerical implementation and simulation of the mPBPK-PD model. A numerical resolution of the whole system for an observation period of 200 days including natural growth of bacteria and treatment took less than 1 second on a standard computer. Additionally, we developed an R Shiny web-app interface where the drug PK/PD properties can be manipulated to inspect the behavior of the system.
Results: A mPBPK/PD modeling framework was developed leveraging a published mPBPK model and on a disease model. The modeling framework tracks the path of a drug from the intake to its bactericidal and bacteriostatic effects in the lung compartment. Informed by literature data[JS1] , this integrated modeling approach allows for evaluating the sterilizing efficacy of several single drug therapies in terms of bacterial burden reduction including rifampicin and bedaquiline. The user can simulate a custom new drug predicting its efficacy against the bacteria population by selecting bactericidal and bacteriostatic potency, different doses, frequencies of the doses, and treatment length.
Conclusions: The proposed combined mPBPK/PD approach effectively enhances the preclinical and translational stages of drug development against tuberculosis by quantifying the potency and efficacy of a single drug. Furthermore, the parameters used in the model can be linked to relevant PK/PD indices, e.g., the minimal inhibitory or bactericidal concentration (MIC and MBC). This information enables a comparison between the modes of action of the drugs against the bacteria and it augments the description of the drug kinetics from the PK indices that are only partially related to the action on the bacteria, e.g., the area under the curve (AUC). We are currently investigating the addition of a post-antibiotic effect (PAE) with a compartment to include the long-term action of the drug. This platform will be further expanded to evaluate combination therapies, enabling additional optimization of the current regimens. In this way, the platform will assist in shortening treatment duration and relapse rates by potentially indicating specific modes of action to include in new regimens.
References:
[1] W. H. Organization, Global Tuberculosis Report 2022. 2022.
[2] F. Reali et al., “A minimal PBPK model to accelerate preclinical development of drugs against tuberculosis,” Frontiers in Pharmacology, vol. 14, p. 1272091, 2024.
[3] S. G. Wicha et al., “Forecasting clinical dose-response from preclinical studies in tuberculosis research: translational predictions with rifampicin,” Clinical Pharmacology & Therapeutics, vol. 104, no. 6, pp. 1208–1218, 2018.
[4] R. C. Team, R: A Language and Environment for Statistical Computing. Vienna, Austria, 2023.
[5] W. Wang, K. Hallow, and D. James, “A tutorial on RxODE: simulating differential equation pharmacometric models in R,” CPT: pharmacometrics & systems pharmacology, vol. 5, no. 1, pp. 3–10, 2016.
