External validation of population pharmacokinetic models on 7 published models of moxifloxacin-treated tuberculosis patients
Hyunjung Lee(1), Woojin Jung(3), Sangkeun Jung(1,2), Jung-woo Chae(1,3), Radojka M. Savic(4), Hwi-Yeol Yun(1,3)
(1) Department of Bio-AI convergence, Chungnam National University, Daejeon, Republic of Korea, (2) Department of Computer Science and Engineering, Chungnam National University, Daejeon, Republic of Korea, (3) College of Pharmacy, Chungnam National University, Daejeon, Republic of Korea, (4) Department of Bioengineering and Therapeutic Sciences, University of California San Francisco, California, USA, †These authors contributed equally to this work.,*Those authors contributed equally as correspondence.
Object: Moxifloxacin has a potential role as a first-line agent in fluoroquinolone-based tuberculosis treatment and plays a central role in multidrug-resistant tuberculosis (MDR-TB). Moxifloxacin is considered a last resort if all other antibiotics fail, making it difficult to ensure patient compliance because of the need to take many drugs. Despite the fact that various population pharmacokinetic (PopPK) models of moxifloxacin for multidrug-resistant tuberculosis patients have been published, the predictive performance of the models is still restricted. There are significant covariates in the published popPK models of moxifloxacin, and factors such as HIV and body weight affect the model. Therefore, it is necessary to fully consider these factors and identify pharmacokinetics. The purpose of this study is not only to figure out the performance of a previously reported model with an external dataset but also to suggest optimal sampling schedules using restricted sampling.
Method: A systematic literature review procedure was performed through the PubMed database to select proper population models. The external dataset was consisted of five different clinical trials in MDR-TB patients. The Pop-PK model was developed using a Non-linear Mixed Effect Modeling (NONMEM, 7.5.1) program used to build Pop-PK models supported by Perl speak NONMEM (PsN 5.2.6). Prediction- and simulation- based diagnostics, and Bayesian forecasting were performed for external evaluation. Mean square root error (RMSE), mean square error (MSE), mean absolute error (MAE), mean absolute percentage error (MAPE), median prediction error (MPE), and distribution of prediction error (PE) were calculated and described to compare and test the prediction of covariates effect on published models. Moreover, the percentage of prediction error that falls within ± 20 (F20), and virtual predictive check (VPC) was used to evaluate the predictive performance of the model. When all data were classified by week, 1800 hours were divided into 7 days and designated as a subset column with a deviation of 24 hours; columns included week were numbered 1 to 11, and columns not included each time were grouped numbered 12. The model was read by sub-setting in the order of 1, 3, 5, 7, and 11 weeks, and the predictive power of the entire dataset was tested.
Result: A total of 1042 observed moxifloxacin concentrations from 113 subjects were included in the external dataset and 7 published models were taken as models for external evaluation. Prediction-based diagnostics showed that body weight potentially influenced model transferability. Simulation-based normalized prediction distribution error analyses indicated misspecification in some of the models, especially regarding variance. Bayesian forecasting demonstrated that the predictive performance of the models substantially improved with 3 prior observations. Structural models, covariates, and observation points potentially affected model predictability.
Conclusion: In evaluating and confirming the published moxifloxacin pharmacokinetic model, the structural model is a major factor influencing the model's predictability. Furthermore, Bayesian forecasting substantially improved model predictability. The predictive performance was sufficient, with a sample point of about a month for the drug.
References:
[1] Al-Shaer, M. H. et al. Fluoroquinolones in Drug-Resistant Tuberculosis: Culture Conversion and Pharmacokinetic/Pharmacodynamic Target Attainment To Guide Dose Selection. Antimicrobial agents and chemotherapy 63, (2019).
[2] Yun, H.-Y. et al. Model-Based Efficacy and Toxicity Comparisons of Moxifloxacin for Multidrug-Resistant Tuberculosis. Open Forum Infect Dis 9, ofab660 (2022).
[3] Öbrink-Hansen, K. et al. Moxifloxacin pharmacokinetic profile and efficacy evaluation in empiric treatment of community-acquired pneumonia. Antimicrob Agents Chemother 59, 2398–2404 (2015).
[4] Ginsburg, A. S. et al. Modeling in vivo pharmacokinetics and pharmacodynamics of moxifloxacin therapy for Mycobacterium tuberculosis infection by using a novel cartridge system. Antimicrob Agents Chemother 49, 853–856 (2005).
[5] Pranger, A. D. et al. Limited-sampling strategies for therapeutic drug monitoring of moxifloxacin in patients with tuberculosis. Ther Drug Monit 33, 350–354 (2011).
[6] Chang, M. J. et al. Population pharmacokinetics of moxifloxacin, cycloserine, p-aminosalicylic acid and kanamycin for the treatment of multi-drug-resistant tuberculosis. International journal of antimicrobial agents 49, 677–687 (2017).
[7] Florian, J. A., Tornøe, C. W., Brundage, R., Parekh, A. & Garnett, C. E. Population pharmacokinetic and concentration--QTc models for moxifloxacin: pooled analysis of 20 thorough QT studies. J Clin Pharmacol 51, 1152–1162 (2011).
[8] Strydom, N. et al. Tuberculosis drugs’ distribution and emergence of resistance in patient’s lung lesions: A mechanistic model and tool for regimen and dose optimization. PLoS Med 16, e1002773 (2019).
[9] Gumbo, T. et al. Selection of a moxifloxacin dose that suppresses drug resistance in Mycobacterium tuberculosis, by use of an in vitro pharmacodynamic infection model and mathematical modeling. J Infect Dis 190, 1642–1651 (2004).