Marquez-Megias, S (1), Ramón-López, A (1,2), Más-Serrano, P (1,2,3), Dokoumetzidis, A (4), DÃaz-González, M (3), Nalda-Molina, R (1, 2).
(1) Miguel Hernández University, Department of Engineering, Division of Pharmacy and Pharmaceutics, School of Pharmacy, San Juan de Alicante, Alicante, Spain, (2) Alicante Institute for Health and Biomedical Research (ISABIAL-FISABIO Foundation), Alicante, Spain. (3) General Universitary Hospital of Alicante, Pharmacy Department, Clinical Pharmacokinetics Unit, Alicante, Spain, (4) School of Pharmacy, University of Athens, Greece.
Abstract
Introduction:
Therapeutic drug monitoring plays an important role in the treatment of many diseases, in order to achieve the best therapeutic results with the lowest toxicity. One approximation would be to use empirical Bayesian estimates of the individual pharmacokinetic parameters, and use them to modify the treatment to achieve the target concentration. Thus, drug dose adaptation increasingly relies on population pharmacokinetic (PopPK) models, which can be found in literature or can be developed on house. Frequently, there are more than one model, published from different authors, for the same disease, indication and population.
Therefore, the first step would be to determine which PopPK model, among all of them, describes the studied population with less bias and imprecision. To do so, the model adequacy and the model prediction performance should be validated before using it in the clinical practice. There are numerous methods of validation for the model adequacy and predictive performance, however, to our knowledge, there has not been published any research of the performance of these methods in the same conditions as they would be used in very sparse data.
Objectives:
To evaluate the performance of different methods to assess the model adequacy and model prediction performance of PopPK models in therapeutic drug monitoring.
Methods:
In order to evaluate how the different methods would detect potential model misspecifications, two or more trough concentrations of a simple PopPK model (one-compartment IV model) of a population were simulated. Thereafter, the empirical Bayesian estimates of the pharmacokinetic parameters were estimated by using those troughs as the “a posteriori” information, and the “true” model and other models with different types of misspecifications as the “a priori” information.
Two different scenarios were created from the dataset to evaluate the model adequacy and predictive performance; Scenario 1: To evaluate the model adequacy, all the trough concentrations were included in the dataset; Scenario 2: To assess the predictive performance, only the first simulated plasma concentration were included for each patient to estimate the empirical Bayesian estimates, and then, the individual predictions that were left out were compared with the original concentrations.
The model adequacy of each model was evaluated by Predicted Corrected Visual Predictive Check (PCVPC). The predictive performance of each model was evaluated using the Bland-Altman analysis, calculating the bias and the limits of agreement (precision). The models were implemented in NONMEM® v7.4 All summary statistics, graphics and simulations were performed in R (version 3.6.1), using the RStudio interface (version 1.2.5001).
Results and Conclusions:
The different plots and results obtained by the PCVPC and the Bland-Altman analysis for each scenario were analysed and discussed to identify the impact of the different misspecifications on these analyses.
Conclusions:
The results obtained in these simulations could be helpful as a tutorial to evaluate the performance of different PopPK models for their use in therapeutic drug monitoring.
References:
[1] Post TM, Freijer JI, Ploeger BA, Danhof M. Extensions to the visual predictive check to facilitate model performance evaluation. J Pharmacokinet Pharmacodyn. 2008 Apr;35(2):185-202.
[2] Altman DG, Bland JM. Applying the right statistics: analyses of measurement studies. Ultrasound Obstet Gynecol. 2003; 22: 85–9317.
[3] Dhaese SAM, Farkas A, Colin P, Lipman J, Stove V, Verstraete AG, et al. Population pharmacokinetics and evaluation of the predictive performance of pharmacokinetic models in critically ill patients receiving continuous infusion meropenem: a comparison of eight pharmacokinetic models. J Antimicrob Chemother. 2019 Feb 1;74(2):432-441.
[4] Farkas A, Daroczi G, Villasurda P, Dolton M, Nakagaki M, Roberts JA. Comparative Evaluation of the Predictive Performances of Three Different Structural Population Pharmacokinetic Models To Predict Future Voriconazole Concentrations. Antimicrob Agents Chemother. 2016 Oct 21;60(11):6806-6812.
Reference: PAGE () Abstr 9549 [www.page-meeting.org/?abstract=9549]
Poster: Methodology - Model Evaluation