Damnjanovic Ivana(1), Anastasia Tsyplakova(2), Nikola Stefanovic(1), Tatjana Tošic(3), Aleksandra Catic-Ðordevic(1), Vangelis Karalis(2)
(1)Department of Pharmacy, Faculty of Medicine, University of Nis, Nis, Serbia (2)Department of Pharmacy, School of Health Sciences, National and Kapodistrian University of Athens, Athens, Greece (3)Clinic of Pediatric Internal Medicine, Department of Pediatric Neurology, University Clinical Center of Nis, Nis, Serbia
Introduction: Managing epilepsy in pediatric patients through combined pharmacotherapy is a challenging task for clinicians. This is due to the unpredictable efficacy of drugs, adverse effects and a lack of knowledge about optimal dosage regimens. Therapeutic drug monitoring (TDM) is a common practice as part of personalized epilepsy management, which has several clinical benefits [1]. However, there are limitations to TDM, as drug concentrations often do not correlate with improved clinical outcomes. Therefore, there is a need for practical decision-making tools to assist clinicians in providing personalized treatment. Pharmacometrics and mathematical simulation tools can be crucial in improving personalized epilepsy management for vulnerable pediatric populations [2].
Objectives:
- To describe the population pharmacokinetics (PopPK) of valproic acid (VA), lamotrigine (LTG), and levetiracetam (LEV) through non-linear mixed effects modeling in a pediatric population.
- To uncover the potential effect of covariates on the model parameters. The latter included demographic factors such as body weight (BW), age (in years), additional antiepileptic medications co-administered (therapeutic regimen), daily dose of the drug, and the presence or absence of epileptic seizures.
Methods: A 12-month prospective study was conducted at the Clinic of Pediatric Internal Medicine, Department of Pediatric Neurology, involving 71 patients aged 2-18 with diagnosed epilepsy. Patients were on dual antiepileptic therapy modalities: VA/LTG, VA/LEV, and LTG/LEV. Data were collected from medical records and in-person interviews, capturing demographic and therapy characteristics, as well as seizure control data. Blood samples were collected in the morning before the next dose.
A Pharmacokinetic analysis of the concentration-time data obtained for LEV, LTG, and VA was conducted using nonlinear mixed-effects modeling implemented in Monolix™ 2021R2. Initially, the number of compartments describing the disposition of LEV, LTG, and VA was determined, with one-, two-, and three-compartment models explored. Pharmacokinetic parameters were assumed to follow a lognormal distribution, and an exponential model was employed to represent inter-individual variability. Various error models of residual variability, including constant, proportional, and combined constant and proportional models, were assessed. Due to the availability of only one measurement (trough levels) per patient, conducting a maximum likelihood estimation of all parameters was not feasible. Instead, a stepwise procedure was adopted, initially allowing one variable to be freely estimated while fixing others to values reported in the literature or following maximum a posteriori estimation. Once a robust estimate was obtained, another pharmacokinetic parameter was allowed to vary. Different combinations of estimation sequences were tested, with the one yielding the best fitting results and physiologically sound estimates chosen. Additionally, the effect of several covariates on the model parameters was investigated, including demographic factors such as BW, age (in years), co-administered additional antiepileptic medications (therapeutic regimen), daily drug dose, and presence or absence of epileptic seizures. Covariate analyses were conducted using a combination of stepwise forward and backward selection. Continuous factors were examined using allometric or linear relationships, either untransformed or centered on their mean value. The final model was chosen based on statistical and graphical goodness-of-fit criteria, facilitating the identification of potential biases or structural model issues. The precision of parameter estimates was assessed using relative standard errors (RSE%). Models were compared using numerical statistical significance criteria such as log-likelihood, Akaike, and Bayesian information criteria, particularly for selecting non-hierarchical models. Graphical evaluation of predicted versus observed values, weighted residuals versus concentrations or time, and individual fits was performed to assess goodness of fit. Visual predictive check charts, constructed using 1,000 Monte Carlo simulations and 90% confidence intervals, were also used to evaluate the model’s predictive performance, stability, and robustness.
Results: TDM of antiepileptic drugs plays a crucial role in the decision-making process for optimizing epilepsy treatment, especially in pediatric patients. Our study revealed that a high percentage of measured concentrations fell within the reference range: 93.55% for VA, 86.27% for LTG, and 68.97% for LEV. These concentrations corresponded to administered doses consistent with leading recommendations, and no seizures were reported during the study period [3]. Notably, over 30% of LEV concentrations were outside the therapeutic range (12-46 mg/L), underscoring the importance of clinical pharmacokinetic considerations. The highest percentage of concentrations within the reference range was observed for VA (93.55%), the most commonly prescribed antiepileptic drug subject to routine TDM, unlike LTG and LEV, for which routine monitoring was not recommended in clinical practice.
The results from PopPK modeling have shown that a one-compartment model with first-order absorption and elimination best describes the pharmacokinetics of Lev, LTG, and VA. Regarding levetiracetam, two statistically significant (p > 0.001) covariates were identified during stepwise covariate modeling: BW on V and Cl. Secondly, for the lamotrigine model, three statistically significant (p < 0.001) covariates were identified: body weight on the apparent volume of distribution (using a weight-centered individual model), daily dose on apparent Cl, and coadministration of LTG and VA on Cl. Lastly, in the case of valproate, three statistically significant covariates were found. BW on apparent V (with an allometric exponent of 0.75; p < 0.001) and daily dose on Cl (p < 0.001) were the most influential factors. Age was also found to have a significant positive impact on distribution V (p-value = 0.032).
Conclusions:
- The weight and age of children are negatively associated with LTG, LEV and VA levels.
- Concomitant administration of VA and LTG leads to increased LTG levels.
- An increase in the total daily dose of VA leads to an increase in the Cl of the drug.
- The main factor that may influence antiepileptic activity in children is the levels of antiepileptic drugs, followed by BW and age.
- Gender demonstrates no effect on epileptic activity.
- The application of popPK and Machine Learning models in therapeutics may lead to improved control of seizures and management of seizure risk through optimization of individualized dosing regimens.
References:
[1] Oliva CF, Gangi G, Marino S, et al. Single and in combination antiepileptic drug therapy in children with epilepsy: how to use it. AIMS Med Sci 2021; 8: 138–146.
[2] Lebedev G, Fartushnyi E, Fartushnyi I, et al. Technology of supporting medical decision-making using evidence-based medicine and artificial intelligence. Procedia Comput Sci 2020; 176: 1703–1712.
[3]Patsalos PN, Spencer EP, Berry DJ. Therapeutic drug monitoring of antiepileptic drugs in Epilepsy: a 2018 Update. Ther Drug Monit 2018; 40: 526–548.
Reference: PAGE 32 (2024) Abstr 10773 [www.page-meeting.org/?abstract=10773]
Poster: Drug/Disease Modelling - Other Topics