Physiologically Based Pharmacokinetic Modeling to Characterize the Disposition of Topiramate in Children with and without Obesity

Abstract

Topiramate (TPM) is used to treat seizures in adults and children ≥2 years. In children, TPM dosing is body weight based. It is unknown whether children with obesity may benefit from an alternative dosing strategy. We developed a novel physiologically based pharmacokinetic (PBPK) model for TPM that accounts for age‐ and obesity‐related changes in disposition. The model was first developed and evaluated in adults before scaling to children then finally evaluated using opportunistic pharmacokinetic data from two multicenter studies where children with and without obesity ages 2–18 years received TPM per standard of care. Simulations were performed to characterize pharmacokinetic parameters and assess the current labeled dosing regimen in a virtual population of children with obesity (≥2 years). The adult PBPK model performed well when evaluated with observed literature data, resulting in an average fold error of 0.989 across the dosing range of 25–1200 mg. The pediatric data used in model evaluation included 43 children (86% with obesity; median [range], 8 [2–18] years); 71% of 134 observed concentrations fell within the 90% prediction interval for the PBPK model. PBPK based dosing simulations revealed comparable exposure (<20% difference) in children with and without obesity receiving the Topamax product label weight‐tiered dose for monotherapy (weight tiered for 2 to <10 years and fixed dosing for 10 to ≤18 years) or weight‐based adjunctive therapy. Comparable trough concentrations between children with and without obesity suggest the current dosing recommendations, including monotherapy and adjunctive therapy, are appropriate for children with obesity.

Introduction

Topiramate (TPM) is a broad‐spectrum antiseizure medication (ASM) approved in children ≥2 years old for both (1) initial monotherapy for focal‐onset or primary generalized tonic–clonic seizures, and (2) as adjunctive therapy for focal‐onset seizures, primary generalized tonic–clonic seizures, or seizures associated with Lennox–Gastaut syndrome. ¹ Despite its established effectiveness, ∼60% of patients discontinued this drug due to adverse events or lack of seizure control in a prospective study of children and adults with epilepsy. ² There is, therefore, a need to optimize the safe and effective use of ASMs, such as TPM, by accounting for patient factors that contribute to variability in drug disposition and response.

The pharmacokinetics (PK) of TPM is linear over the clinically relevant dose range of 100 to 1200 mg, as reported in adults and children. ³ TPM is administered orally with a bioavailability that exceeds 80%. It is available as immediate‐release (IR) tablets and sprinkle capsules (Topamax) taken twice daily, extended‐release (ER) capsules (Qudexy XR and Trokendi XR) taken once daily, compounded oral suspension, ⁴ and oral solution (Eprontia). The sprinkle capsules are bioequivalent to the IR tablets, ¹ and the once‐daily ER capsules provide comparable steady‐state plasma concentrations with the twice‐daily IR formulations when administered at the same daily dose. ⁵ , ⁶ Approximately 70% to 80% of TPM dose is eliminated by renal excretion, and the observed renal clearance of TPM is much lower than the expected glomerular filtration rate (GFR)‐mediated clearance, which suggests tubular reabsorption of the drug. ¹ , ⁷ The rest of the drug is metabolized via glucuronidation, hydroxylation, and hydrolysis and these metabolites are pharmacologically inactive. ⁸

Population PK (popPK) studies (including children) demonstrate that TPM apparent clearance (CL/F) is influenced by body weight, age, and concomitant ASMs among other influencing factors. ⁹ , ¹⁰ TPM CL/F in children per kg body weight (L/h/kg) is higher than adults, which is a typical PK characteristic, explained by less than proportional change in drug elimination capacity with increased body size across the age groups of children (≥2 years) to young adults. ³ , ¹⁰ , ¹¹ The current body weight‐based dosing recommendation for TPM accounts for these PK characteristics, where younger children receive a higher weight‐based TPM dose than older children to achieve comparable exposure to adults. ¹¹ However, the PK evidence that informed the current pediatric dosing recommendations for TPM came from studies with a limited number of children with obesity. ¹² , ¹³ , ¹⁴ , ¹⁵ , ¹⁶ Therefore, it is not clear whether the current pediatric dosing regimen for TPM can be extrapolated to children with obesity.

Obesity‐related changes in body size and composition may have additional implications on the physiological processes (e.g., GFR) pertinent to drug disposition in addition to the age‐related changes in physiology which may warrant further dose adjustment in children with obesity. ¹⁷ In addition to alterations in GFR and organ size, obesity in children may be associated with renal hyperfiltration, ¹⁸ which can increase the clearance of renally eliminated drugs. Furthermore, potential changes in tubular transporter expression could modify active secretion and reabsorption processes but data to inform this is lacking. To evaluate the dosing regimen in children with obesity, the influence of age‐ and obesity‐related physiological changes on the PK of TPM must be characterized. A virtual population of children with obesity developed by Gerhart et al ¹⁹ implemented key obesity‐related physiological changes using various sources. This virtual population of children with obesity was utilized in this physiologically based pharmacokinetic (PBPK) analysis to better understand PK and dosing in this population.

PBPK modeling is a tool that incorporates a drug's physicochemical and biochemical properties to predict plasma and tissue concentration–time profiles of drugs. To our knowledge, no TPM pediatric PBPK models exist in the literature, and this modeling technique has not been used to evaluate this research question for TPM due to lack of available data. This research aims to (1) leverage data from the literature and prospective pediatric studies to characterize the disposition of TPM in children by developing a whole‐body PBPK model, and (2) understand the dosing requirement in children with obesity with the final PBPK model.

Methods

Human Subjects Protection

Data for this study were collected from two trials conducted by the Pediatric Trials Network (PTN; NICHD‐2011‐POP01, #NCT01431326; NICHD‐2015‐AED01, #NCT02993861) with approval by the Duke Health Institutional Review Board (IRB) under IRB protocol numbers Pro00029638 (POP01) and Pro0007092 (AED01). Legal representatives of all participants provided informed consent to participate, and assent was obtained when appropriate.

Data Sources for Model Development and Evaluation

Literature PK Data in Adults

To identify relevant datasets, a comprehensive literature search was performed using PubMed. The search strategy included the keywords: “pharmacokinetics,” “children,” “adults,” and “topiramate.” Search results were systematically screened to identify studies and datasets reporting PK data of TPM in adult and pediatric populations. Several adult PK studies from the literature where healthy adult subjects were administered TPM intravenously or orally as a single dose ²⁰ , ²¹ , ²² , ²³ , ²⁴ , ²⁵ , ²⁶ , ²⁷ or as multiple doses ²⁶ , ²⁸ , ²⁹ , ³⁰ , ³¹ , ³² ranging from 25 to 1200 mg were digitized (Web Plot Digitizer 4). Details of these studies, including subject demographics, are shown in Tables S1 and S2. Virtual adult populations for PBPK model development reflect the demographics (e.g., weight and age) of respective study participants and were sampled from the built‐in PK‐Sim population library. For adult multiple‐dose model evaluation, the dosing regimens reported in the respective clinical studies were implemented in the PBPK simulations.

Pediatric Data

Data were collected from two large multicenter, prospective clinical trials in children and young adults, conducted by the PTN, an NICHD‐sponsored program under the Best Pharmaceuticals for Children Act (BPCA). ³³ The BPCA authorizes research to improve the dosing, safety, and efficacy of medication use for children. In order to facilitate uptake and implementation of research performed under BPCA, the PTN disseminates results in multiple ways, including negotiation of label changes with the US Food and Drug Administration and peer‐reviewed publication of scientific manuscripts. In both clinical trials, participants received TPM as part of standard of care (SOC) therapy. In the first study, Pharmacokinetics of Understudied Drugs Administered to Children per Standard of Care (POP01, NICHD‐2011‐POP01, clinicaltrials.gov #NCT01431326), participants were included if they were younger than 21 years and administered TPM per SOC. The second study, Pharmacokinetics of Anti‐epileptic Drugs in Obese Children (AED01, NICHD‐2015‐AED01, clinicaltrials.gov #NCT02993861), was a PK and safety study exclusively for four ASMs in children and adolescents with obesity. Participants were included if they were ≥2 to <18 years of age and had obesity. A child with obesity was defined as a participant at or above the 95th percentile of the age‐to‐body mass index (PBMI) chart specified by the Center for Disease Control (CDC). ³⁴ Plasma samples were drawn according to specific schedules after the drug was administered, and PK plasma concentrations were measured at a central lab using a validated bioanalytical assay. Participants from POP01 were included in this PBPK analysis if they had at least two TPM concentration records.

PBPK Model Development

A whole‐body PBPK model was developed for TPM in adults by accounting for its physicochemical and pharmacokinetic properties using PK‐Sim and MoBi (versions 10, Open Systems Pharmacology Suite, https://www.open‐systems‐pharmacology.org/). Tissue partition coefficients and cellular permeabilities were predicted by the Rodgers and Rowland method and PK‐Sim standard method, respectively. ³⁵ Initial estimates for drug‐specific properties were obtained from the literature. ³⁶ Renal clearance (CL_R) was parameterized as the individual GFR corrected by plasma free fraction (f_{u, p}) and GFR fraction (f_GFR) of TPM as shown in Equation (1). GFR fraction (f_GFR) represents the fraction of protein‐binding‐corrected GFR that is excreted (i.e., not reabsorbed), where (1 − f_GFR) represents the reabsorbed fraction. The initial estimate of f_GFR was calculated as the ratio of (reported) observed renal clearance of TPM (CL_{R, ref}) in a reference adult to the protein‐binding‐corrected reference GFR ($\mathrm{fu},\mathrm{p}\ast \mathrm{GFRref}$) using Equation (2). Considering an observed CL_{R, ref} of 18 mL/min, an 80% plasma free fraction, and a GFR_ref of 120 mL/min, the initial value of f_GFR becomes 0.2.

$$\begin{array}{c}{\mathrm{CL}}_{\mathrm{R}}=\left(\mathrm{fu},\mathrm{p}*\mathrm{GFR}\right)*\end{array}{\mathrm{f}}_{\mathrm{GFR}}$$ (1)

$$\begin{array}{c}{\mathrm{f}}_{\mathrm{GFR}}=\frac{{\mathrm{CL}}_{{\mathrm{R}}_{\mathrm{ref}}}}{\mathrm{fu},\mathrm{p}*\mathrm{GFRref}}=0.2\end{array}$$ (2)

In human mass balance studies, TPM along with six minor metabolites and three metabolic pathways (hydroxylation, hydrolysis, and glucuronidation) were identified. ⁸ However, the enzyme isoforms responsible for TPM metabolism have not been extensively studied. ³⁷ Therefore, a generic enzyme with a reference concentration of 1 µmol/L was created in the liver compartment to account for TPM's metabolic pathway. ⁸

TPM exhibits saturable (i.e., concentration‐dependent) binding to red blood cells (RBCs), causing nonlinear distribution between whole blood and plasma. ³⁸ At low concentrations, the blood to plasma (B:P) ratio of TPM is 8, but it decreases with increasing concentration and becomes saturated at 2 at plasma concentrations of 15 µM and above. The relationship between plasma concentrations and B:P ratio ³⁸ is shown in Figure S1, where the observed data (collected from literature) were characterized by using an exponential model shown in Equation (3) (R² = 0.9911). This exponential model was incorporated in MoBi that allowed calculation of concentration‐dependent B:P ratio of TPM in the PBPK model in a user‐defined manner.

$$\begin{array}{c}B:Pratio=6.5235\ast \mathrm{Plasma}\mathrm{Concentratio}{\mathrm{n}}^{-0.484}\end{array}$$ (3)

The final adult model was developed using mean concentration–time data following a single intravenous (IV) or oral dose of TPM reported in the literature. ²² , ³⁹ Some of the parameters were optimized as needed to fit the observed data. After ensuring a robust adult IV model, empirical fit‐for‐purpose oral absorption models were developed for each TPM formulation, including IR tablets, sprinkle capsules, and oral solution. All the IV model parameters were fixed, while the absorption‐related parameters were optimized using observed oral data. ³⁹ A Weibull absorption model with no lag time was applied for IR formulations including sprinkle capsules and oral solution.

Once an acceptable adult PBPK model was developed, the model was then scaled to children 2 years and older. The drug‐specific properties remained unchanged from adults. When the model was scaled to children, the f_GFR was assumed to remain constant between adults and children (i.e., f_GFR = 0.2); however, this represents a simplifying assumption given the limited available data. Notably, there is insufficient evidence to confirm whether tubular reabsorption processes fully mature before 2 years of age, and the impact of obesity on proximal tubular reabsorption remains poorly characterized.

Virtual Populations of Children

In this model, we developed a PBPK model for TPM and performed simulations using an existing population of children with obesity as published by Gerhart et al ¹⁹ and children without obesity from the PK‐Sim library. Virtual populations of children with and without obesity were created for PK and dosing simulations. Obesity status were classified as follows: Non‐overweight: 5th ≤ PBMI ≤ 85th; Overweight: 85th ≤ PBMI < 95th; Class I obesity: Body mass index (BMI) ≥95th percentile and BMI <120% of the 95th percentile; Class II obesity: 120% ≤ BMI < 140% of the 95th percentile; Class III obesity: BMI ≥140% of the 95th percentile. ⁴⁰ For children without obesity (BMI <95th percentile), virtual individuals were resampled from the PK‐Sim population library with their default records of anatomical and physiological characteristics (e.g., organ size and perfusion, GFR, and hematocrit) in the library.

For children with obesity, extended BMI percentiles were calculated by comparing each subject's BMI to the 95th percentile BMI for their age and sex, and accordingly the obesity class (I to III) was assigned to the virtual individuals. ⁴⁰ , ⁴¹ Demographic information for the virtual populations with and without obesity is available in Table S3. ¹⁹ On average, the virtual population of children with obesity had 19% larger kidney and 18% larger liver sizes compared with children without obesity. ¹⁹ , ⁴²

PBPK Model Evaluation

The adult model was evaluated using PK data from adult studies in the literature. Model bias was assessed by calculating the average fold error (AFE) between individual predicted and observed concentrations ⁴³ (Equation 4):

$$\begin{array}{c}AFE={10}^{\frac{1}{\mathrm{n}}\sum \log \left(\frac{\mathrm{predicted}}{\mathrm{observed}}\right)}\end{array}$$ (4)

AFE <1 indicated model under‐prediction, and AFE >1 indicated over‐prediction. An AFE between 0.5 to 2 was considered acceptable. ⁴⁴ , ⁴⁵ The simulated plasma concentration versus time profiles (mean and 90% prediction interval) were also compared with the respective observed data for visual inspection of the model performance. In addition, PK parameters were extracted from published TPM adult studies and compared with PBPK‐simulated values; predictive performance was evaluated using observed‐to‐predicted ratios.

Pediatric PK information from the literature was limited and PK parameters such as clearance from Battino et al ¹¹ and Mikaeloff et al ⁴⁶ were used to compare simulated weight‐normalized clearance in children without obesity. The pediatric model was further evaluated using clinical data from AED01 and POP01 participants. Virtual study populations were created by resampling 100 individuals from the published population libraries, based on the observed characteristics (e.g., age, weight, and height) in the AED01 and POP01 studies. Due to the inter‐individual variability in dosing frequency and amount among the subjects of AED01 and POP01, it was not feasible to compare all the subject level concentration data in aggregate with a single population‐based PBPK simulation of the concentration–time profile. Therefore, “individualized virtual populations” were generated for each participant from the AED01 and POP01 datasets by sampling virtual individuals (n = 100) within ±1 year age range of the actual age of the respective study participants. PBPK simulations were generated for each “individualized virtual population,” and the 90% prediction interval was overlaid with the respective individual observed data from the two studies for visual inspection. The pediatric model was evaluated by quantifying the number of observed concentrations falling inside of the 5th to 95th percentile prediction interval of the “individualized” population simulations.

Dosing Simulations

First, PK parameters were simulated in a virtual population of children with and without obesity 2 years and older (N = 4000) receiving 400 mg/day divided into two doses. The absolute CL/F and weight‐normalized CL/F were calculated using area under the curve at steady state (AUC_ss) obtained from PK‐Sim. Apparent volume of distribution at steady state (V_ss/F) and half‐life were calculated using the R package, PKNCA (version 0.11.0). Second, another set of dosing simulations was conducted to assess whether the currently recommended dosing regimen would result in comparable exposure (<20% difference in steady‐state minimum, C_{min, ss}, and maximum, C_{max, ss}, concentration) in children and adolescents with obesity with respect to their counterparts without obesity. ⁴⁷ TPM pediatric dosing regimens from the Topamax product label were used for the simulations. ¹ For monotherapy, the following simulations were performed for children ages 2 to <10 years: 250 mg/day for children up to 11 kg, 300 mg/day for children weighing 12 to 22 kg, 350 mg/day for children weighing 23 to 38 kg, and 400 mg/day for children >38 kg. Monotherapy simulations for children 10 to ≤18 years used a 400 mg/day dose. A 5 mg/kg/day dose was used for children ages 2 to <16 years for adjunctive therapy. Total daily doses were divided into two, given 12 h apart and were capped at the maximum adult dose of 400 mg/day, as recommended in the United States Prescribing Information (USPI). A different set of virtual populations of children with and without obesity ≥2 to ≤18 years were created for each age and dose group (N = 1000) to conduct these dosing simulations.

Results

Evaluation of the Adult Model

The initial and final whole‐body PBPK model parameters for TPM are described in Table 1. The mean predictions and 90% prediction interval of concentration–time profiles for various single‐dose adult studies with doses ranging from 25 to 1200 mg are shown in Figure 1. ⁷ , ²² , ³⁸ Figure 2 shows concentration–time profiles for various multiple‐dose adult studies. ²⁶ , ²⁸ , ²⁹ , ³⁰ , ³¹ , ³² The final PBPK model predictions were able to capture the observed data well based on visual inspection of all adult studies and comparison of PK parameters in single‐dose adult studies can be found in Table S4. Single‐dose studies in adults (Figure S2) had a mean AFE of 0.989 while multiple‐dose studies had a mean AFE of 0.964, both indicating good model fit.

Table 1.

Summary of the Parameter Estimates of the Final Physiologically Based Pharmacokinetic Model of Topiramate in Adults

Parameter	Initial Estimate	Final Estimate	Reference
Distribution, metabolism, excretion
Molecular weight (g/mol)	339.36	339.36	PubChem 36
pKa	8.7	8.7	PubChem 36
Compound type	Acid	Acid	PubChem 36
LogP	0.6	0.6	PubChem 36
fu,p	0.80	0.80	Patsalos et al 62
fGFR	0.16	0.2	Optimized
B:P ratio at 1 µM a	2	6.5 b	Shank et al 38
Generic enzyme, reference concentration (µM)	1	1	Optimized
Generic CLint (mL/min)	5	5	Optimized
Absorption
Solubility (mg/mL)	9.8	9.8	PubChem 36
T50% (min)	10	10	Liang et al 63 , 64
Dissolution shape	0.8	0.7	Optimized
Specific intestinal permeability (cm/s)	‐	1.4E‐07	Optimized

Abbreviations: B:P ratio, blood to plasma ratio; CL_int, hepatic intrinsic clearance; f_GFR, the fraction of glomerular filtration rate that escapes tubular reabsorption, f_u,p, plasma free fraction; LogP, lipid–water partition coefficient; pKa, acid dissociation constant; T50%, time to dissolve 50% of formulation strength.

^a B:P ratio changes due to concentration‐dependent red blood cell partitioning of topiramate: B:P ratio = 6.5235 * Plasma Concentration ^−0.484.

^b Calculated B:P ratio using the empirical equation at a concentration of 1 µM.

Figure 1.

Observed versus simulated plasma concentration–time profiles from the final physiologically based pharmacokinetic (PBPK) model of topiramate (TPM) in adults following a single intravenous or oral administration. For panels a–i, the simulated concentrations are shown as the arithmetic plasma mean (solid blue lines), the 68% (standard deviation) interval for panels a—d, ²² , ³⁹ and the 95% prediction interval (light blue shaded areas) for panels e–i. ⁷ For panels j–l, the simulated concentrations are shown as the mean arithmetic plasma (solid blue lines) and whole blood (solid red lines) and the 95% prediction interval for plasma (light blue shaded areas) and whole blood (light red shaded areas). ³⁸ Closed circles represent the mean observed data from the literature and the error bars represent the standard deviation (if reported). IV, intravenous; PO, oral dose.

Figure 2.

Observed versus simulated plasma concentration–time profiles of topiramate from the final physiologically based pharmacokinetic (PBPK) model in adults after multiple oral dosing. ²⁸ , ²⁹ , ³⁰ , ³¹ , ³² For panels a‐f, the simulated concentrations are shown as the arithmetic mean (solid blue lines) and the 95% prediction interval (light blue shaded areas). Closed circles represent the mean observed data from the literature. BID, twice a day; QD, once a day; TPM, topiramate.

Evaluation of the Pediatric Model

PBPK‐simulated weight‐normalized clearance values in children without obesity were generally consistent with published pediatric PK studies. In children aged 2 to <6 years, the simulated mean clearance was 0.059 ± 0.017 L/h/kg, comparable to reported values from Mikaeloff et al ⁴⁶ (0.0465 ± 0.0128 L/h/kg in children aged 0.8–3.9 years) and Battino et al ¹¹ (0.0511 ± 0.024 L/h/kg in children <10 years). In older children (≥12 years), the PBPK model estimated a mean clearance of 0.0397 ± 0.0131 L/h/kg, slightly higher than the value reported by Battino et al ¹¹ in children aged 10–17 years (0.0296 ± 0.0098 L/h/kg). Overall, literature‐reported weight‐normalized clearance values across children aged 0.8–18 years aligned with the PBPK‐simulated mean clearance of 0.049 L/h/kg for children without obesity, supporting the model's ability to predict pediatric PK in non‐obese populations.

From the pooled dataset of POP01 and AED01 studies, 43 children with two or more TPM concentrations were included in this PBPK analysis, which resulted in 134 total plasma TPM concentrations available for model evaluation. Of these 43 children, 86% had obesity and the median [range] age was 8 [2–18] years. It is important to note that only six children without obesity from the dataset were available for evaluation (four participants were under 10 years old). For children with obesity, the most common formulation was tablet followed by suspension. Concomitant drugs include carbamazepine (N = 2), phenobarbital (N = 1), valproic acid (N = 4), levetiracetam (N = 6), and oxcarbazepine (N = 2). Details of the dataset used for pediatric model evaluation are presented in Table 2. In our evaluation of the pediatric model, 71% of 134 concentrations from 43 children fell within the 90% prediction interval. Observed pediatric concentrations versus mean predicted concentrations from the individualized population simulations (n = 134) are shown in Figure 3. As majority of the data fell along the line of unity, there was an overall good agreement between the observations and model predictions within the clinically relevant concentration range (Figure S3).

Table 2.

Description of the Dataset From Standard of Care Clinical Trials Used for Pediatric Model Evaluation (n = 43).

Characteristic	Children with Obesity (N = 37)	Children without Obesity (N = 6)	Total (N = 43)
Postnatal age (years)	8.38 (2.13, 18)	6.87 (2.03, 17.4)	8.03 (2.03, 18)
Male (%)	18 (48.6)	3 (50)	21 (49)
Weight (kg)	50.2 (15, 157)	20.8 (11.5, 47.6)	45.8 (11.5, 157)
Extended BMI percentile (%)	110.7 (103.8, 114.3)	84.9 (79.2, 93.0)	110.5 (79.2, 114.3)
Number of plasma concentrations	118	16	134
TPM concentration (mg/L)	6.0 (0.3, 30.1)	4.4 (1.3, 11.3)	6.0 (0.3, 30.1)
Drug formulation
Suspension	9 (24.3%)	2 (33.3%)	11 (25.6%)
Tablet	21 (56.8%)	2 (33.3%)	23 (53.5%)
Tablet—crushed	0 (0%)	2 (33.3%)	2 (4.7%)
Othera	7 (18.9%)	0 (0%)	7 (18.9%)

Abbreviations: BMI, body mass index; TPM, topiramate.

Continuous values are reported as median and range while categorical values are reported as counts.

^a Other formulations include solution, capsule, sprinkles, crushed capsules.

Figure 3.

Observed TPM concentrations versus mean PBPK predicted concentrations from the individualized populations (n = 100 for each subject) of children (n = 43) from the POP01 and AED01 studies (NCT01431326 and NCT02993861). Red points indicate cases where the observed data point fell outside the 90% prediction interval, triangles indicate data from children without obesity and circles indicate data from children with obesity, blue dashed lines indicate line of best fit and the grey shaded area represents the 95% confidence interval.

PBPK Simulations Using a Virtual Population of Children with and without Obesity

PBPK simulations revealed that weight‐normalized CL/F in children decreases with increasing obesity status (Figure 4a). Children without obesity had a higher simulated median [range] weight‐normalized CL/F (0.047 L/h/kg) [0.014–0.15 L/h/kg] than children with obesity (0.039 L/h/kg) [0.01–0.13 L/h/kg]. Simulated weight‐normalized CL/F also decreases as age increases (Figure 4b) and this age‐related decline in CL/F resulted in prolonged half‐life in older children (Table S5). Median weight‐normalized clearance in Class III obesity was 0.0273 L/h/kg versus 0.0489 L/h/kg in non‐overweight, corresponding to a 44% decrease. Simulated weight‐normalized CL/F, V_ss/F, and half‐life by age and obesity status are reported in Table S5.

Figure 4.

Simulated body weight‐normalized clearance of topiramate in virtual population of children older than 2 years stratified across obesity status (panel a), and age group and obesity status (panel b). Virtual subjects were administered 400 mg/day in two divided doses. Obesity status was classified as per the age‐to‐body mass index percentiles (PBMI) defined in the Center for Disease Control (CDC) recommended growth chart. ³⁴ Non‐overweight: 5th ≤ PBMI ≤ 85th (N = 1470 virtual subjects); Overweight: 85th ≤ PBMI < 95th (N = 530 virtual subjects); Class I obesity: BMI ≥ 95th percentile and BMI < 120% of the 95th percentile (N = 1488 virtual subjects); Class II obesity: 120% ≤ BMI < 140% of the 95th percentile (N = 377 virtual subjects); Class III obesity: BMI ≥ 140% of the 95th percentile (N = 135 virtual subjects). In panel b, “With Obesity” includes virtual subjects from Class I to III obesity status and “Without Obesity” includes virtual subjects from “Non‐overweight” and “Overweight” weight status. Boxes represent the median and interquartile range (IQR), whiskers extend to ±1.5*IQR, and black circles indicate outliers.

PBPK dosing simulations following the current Topamax product label revealed comparable peak and trough concentrations between children with and without obesity across age groups and dose regimens (Figure 5). Virtual children with and without obesity ages 2 to <10 years old who received weight‐tiered dosing for monotherapy achieved a median (Q1, Q3) trough concentration of 15.1 (12.8, 17.8) and 15 (12.7, 17.6) mg/L, respectively. Children 10 to ≤18 years with and without obesity receiving 400 mg/day for monotherapy reached median (Q1, Q3) trough concentrations of 12 (9.4, 15) and 13.5 (10.5, 16.9) mg/L, respectively. For adjunctive therapy in children with and without obesity ages 2 to ≤16 years receiving 5 mg/kg/day, a median (Q1, Q3) trough concentration of 7.9 (6.0, 10.4) and 7 (5.2, 9.3) mg/L was achieved. The complete regimens and simulated PK exposure parameters (i.e., C_{min, ss} and C_{max, ss}) can be found in Table 3. In virtual subjects without obesity ages 2 to 16 taking TPM as adjunctive therapy, 21.7% had concentrations below 5 mg/L and 1% were above 20 mg/L while 15.8% and 2.1% of virtual subjects with obesity had concentrations below 5 mg/L and above 20 mg/L, respectively. In virtual subjects without and with obesity ages 2 to <10 years taking TPM as monotherapy, 8.8% and 12.2% had concentrations above 20 mg/L, respectively. No subjects had concentrations below 5 mg/L. In virtual subjects without and with obesity ages 10 to 18 years taking TPM as monotherapy, 15% and 8.2% had concentrations above 20 mg/L, respectively. Less than 0.5% of subjects had concentrations below 5 mg/L. These results suggest that the currently recommended pediatric dosing regimens can produce comparable TPM exposures between children with and without obesity.

Figure 5.

Simulated topiramate (TPM) trough (C_{min, ss}; top panel) and peak (C_{max, ss}; bottom panel) concentrations at steady state in children with and without obesity ages 2 to 18 years following administration of the topiramate dosage recommended in the Topamax product label for monotherapy and adjunctive therapy. ¹ For children 2 to <10 years (n = 1000) receiving monotherapy, weight‐tiered dosing were as follows: children up to 11 kg received 250 mg/day divided into two doses, 12 to 22 kg received 300 mg/day divided into two doses, 23 to 38 kg received 350 mg/day divided into two doses, >38 kg received 400 mg/day divided into two doses. For children 10 to 18 years old (n = 1000) receiving monotherapy, a 400 mg/day divided into two doses were administered. For children 2 to 16 years old receiving adjunctive therapy, a 5 mg/kg/day was administered, and the dose was capped at 400 mg/day. Boxes represent the median and interquartile range (IQR) of simulated concentrations, whiskers extend to 1.5*IQR, and black circles indicate outliers. The dashed lines represent the reference concentration range for TPM at steady state (5–20 mg/L). ⁶¹

Table 3.

Comparison of Median (Q1, Q3) Simulated Topiramate Trough (C_{min, ss}) and Peak (C_{max, ss}) Concentrations at Steady State in Virtual Children with and without Obesity (N = 1000) Receiving the Monotherapy and Adjunctive Therapy Dosing Regimens Recommended in Topamax Package Insert ¹

		With Obesity		Without Obesity
Age Group (years)	Therapy Daily Dose	Cmax, ss (mg/L)	Cmin, ss (mg/L)	Cmax, ss (mg/L)	Cmin, ss (mg/L)
2 to <10	Monotherapy Weight‐tiered dosing a	23.4 (20.6, 26.1)	15.1 (12.8, 17.8)	24.7 (21.8, 27.3)	15 (12.7, 17.6)
10 to ≤18	Monotherapy 400 mg/day	15.4 (12.6, 18.7)	12 (9.4, 15)	18.2 (14.9, 22)	13.5 (10.5, 16.9)
2 to ≤16	Adjunctive therapy 5 mg/kg/day b	11 (9.0, 13.7)	7.9 (6.0, 10.4)	10 (8.4, 12.6)	7 (5.2, 9.3)

^a Weight‐tiered dosing for monotherapy in children ages 2 to <10 years old are as follows: children up to 11 kg received 250 mg/day, 12 to 22 kg received 300 mg/day, 23 to 38 kg received 350 mg/day, >38 kg received 400 mg/day; all daily doses were divided into two doses in 12 h interval.

^b Dose was capped at 400 mg/day for children weighing >80 kg since 400 mg/day is the maximum recommended dose for adults per package insert.

Discussion

Childhood obesity is a growing public health concern in the United States, affecting approximately one in five children and adolescents and can impact drug dosing due to physiological changes. ⁴⁸ TPM is a commonly used ASM in adults and children and understanding its disposition across all patient populations is important to ensure safe and effective dosing. Previous PK prediction studies in children used a popPK approach, showing variability influenced by weight, age, and concomitant ASMs. ⁹ , ¹⁰ , ⁴⁹ , ⁵⁰ While dosing recommendations of TPM are available for children 2 years of age and above; it is unclear how appropriate these recommendations are for children with obesity. This work was conducted within a model‐informed drug development (MIDD) framework consistent with principles outlined in ICH M15 guidance. ⁵¹ This study used a mechanism‐based PBPK framework and leveraged data from the literature and SOC pediatric trials to characterize TPM's PK in children with obesity and to assess whether obesity‐related physiologic changes were predicted to substantially alter exposure under currently used dosing regimens. PBPK‐based dosing simulations revealed that the TPM pediatric dosing recommendations from the Topamax product label would produce comparable drug exposure in children with and without obesity.

Key assumptions in Vd/F and CL/F were made in developing the adult TPM PBPK model. A key variable in the prediction of Vd/F was the concentration‐dependent RBC binding as reflected in the blood to plasma (B:P) ratio. Incorporating this binding phenomenon in the model was important for predicting TPM's Vd/F at clinically relevant lower doses (e.g., 2 mg/kg/day for migraine in children). ⁵² Observed and simulated mean volumes of distribution after a 50 mg IV dose in adults were comparable (1.06 vs 1.12 L/kg). ³⁹ Approximately 70% of an administered TPM dose is eliminated unchanged in the urine with evidence of renal tubular reabsorption and this was captured in our model by calculating the fraction reabsorbed. ¹ A nonspecific metabolic pathway was assumed since the enzymes responsible for TPM's metabolism are not well elucidated. In a recent TPM PBPK model developed by Chen et al to predict renal and hepatic impairment in adults, ³⁷ the authors assumed TPM to be a CYP3A4 substrate based on DDI studies involving TPM and CYP3A4 inducers like phenytoin and carbamazepine where TPM concentrations were reduced by 40%–48% when coadministered with these drugs. ⁵³ , ⁵⁴ Under these model assumptions and using adult PK data, observed and simulated mean CL/F values (1.36 vs 1.27 L/h) closely aligned, with AFE values of 0.99 and 0.96 for single and multiple doses. ⁷ , ²⁰ , ²¹ , ²² , ²³ , ²⁴ , ²⁵ , ²⁷ , ³¹ , ³⁸ , ³⁹

When scaling the model to children, drug‐specific assumptions were retained, such as TPM's concentration‐dependent RBC binding, renal tubular reabsorption, and nonspecific metabolic. System‐specific properties differed to reflect physiological differences between children with and without obesity; specifically, the virtual population of children with obesity had on average, 19% and 18% larger kidney and liver sizes, respectively, compared with children without obesity. ¹⁹ The larger kidney size could lead to increased renal function (e.g., GFR), which potentially increases the absolute clearance of TPM. Most of the points outside the 90% prediction interval (36 out of 39 points) belonged to participants with obesity while the remaining points (3 out of 39) belonged to one participant without obesity who was also taking carbamazepine, which is known to decrease TPM concentrations by 40%. ¹ Although the exact causes of overprediction in TPM concentrations in children with obesity are unclear, possible reasons include undocumented disease states or limited capture of all the effects of obesity due to knowledge gaps. Specific mechanistic hypotheses include reduced tubular reabsorption that occurs with obesity, altered carbonic anhydrase expression in RBCs, increased nonrenal clearance, and medication adherence variability.

PBPK simulations resulted in comparable exposure (<20% difference in C_{min, ss}) in children with and without obesity using current pediatric dosing regimens across age groups. Given that the exposure‐response of TPM is similar in adults and children 2 years and older with focal‐onset and primary generalized tonic–clonic seizures, ¹⁰ , ⁴⁹ , ⁵⁵ it was a reasonable assumption that the response to TPM would be similar in children with and without obesity. The comparability in exposure is attributed to the body weight‐tiered dosing approach in the TPM label, which delivers a lower mg/kg dose in older children (who have higher total body weight) compared with younger children, thereby accounting for the less‐than‐proportional changes in clearance with increasing body size. This finding parallels levetiracetam where current weight‐tiered dosing recommendations are also suitable for children with obesity. ⁴² , ⁵⁶ Although TPM concentrations have shown a general association with clinical response, this relationship is highly variable. Routine monitoring is not recommended, and the therapeutic range of 5–20 mg/L should be interpreted as a general reference.

A major limitation is the small number of subjects without obesity (n = 6) that were available for model evaluation, which may constrain the generalizability of the model to this population. Concentration–time profiles from published studies were very limited and not accessed, but simulated clearance values were compared with reported pediatric PK data, assuming the studies involved children without obesity. Target exposures were matched between groups since exposure in children without obesity is unknown, which may overlook differences in PK and pharmacodynamics. TPM has been associated with weight loss in children and various literature suggests TPM as an effective treatment for weight management. ⁵⁷ , ⁵⁸ However, in this analysis, weight reduction was not measured and was not an objective in the SOC trials. Additionally, this analysis did not evaluate pharmacodynamic effects (e.g., reduction in seizure frequency and adverse effects) since this information was not collected in the SOC clinical trials. Including pharmacodynamic endpoints in future studies would increase the clinical relevance of PBPK‐based dosing recommendations and help elucidate the relationship between TPM's reference range and efficacy. For example, Girgis et al used PK–PD modeling to predict dosing regimen in pediatric patients to achieve seizure freedom at 1 year. ¹⁰ We acknowledge that 71% of observed concentrations fell within the 90% prediction interval, indicating that the model did not fully capture the observed variability in the opportunistic clinical dataset. The prediction interval in this analysis reflects inter‐individual variability represented in the virtual pediatric population, including variability in relevant anatomical and physiological parameters. Residual unexplained variability was not modeled separately. This is a limitation of the study, as additional sources of variability, including analytical variability, uncertainty in recorded sampling times, adherence, dose administration, and other unmeasured clinical factors, may have contributed to observed concentrations falling outside the prediction interval. Finally, in our PBPK model, 31% (36 out of 118) of observed TPM concentrations from children with obesity fell outside the 90% prediction interval. This suggests that some obesity‐related physiologic processes may not be fully represented in the current PBPK framework or virtual population. Potential contributors include incomplete characterization of obesity‐associated changes in renal clearance, renal tubular handling, tissue distribution, organ blood flows, or other physiologic alterations that remain poorly defined in children with obesity. The current PBPK model provides a framework for extension to special populations, including neonates with seizures or with hypoxic–ischemic encephalopathy treated with TPM, ⁵⁹ , ⁶⁰ by incorporating age‐dependent maturation of physiological and biochemical processes. However, this application remains prospective and will require integration of ontogeny functions and external validation in these populations before conclusions can be drawn from dosing simulations.

Conclusions

A whole‐body PBPK model for TPM was successfully constructed and evaluated by leveraging adult PK data from the literature and from SOC trials in pediatric patients with and without obesity. PBPK modeling revealed that children with obesity had lower simulated weight‐normalized CL/F than children without obesity. The number of observed concentrations from children with obesity below the prediction interval may indicate systematic overprediction of exposures in this population. PBPK‐based simulations using a virtual population of children with obesity supported the use of the currently recommended pediatric dosing regimens in children with obesity. However, additional prospective clinical validation is needed before definitive dosing recommendations can be established. This work provides a future framework to further extrapolate the model to neonates and infants where data are limited and the drug is currently being used off‐label.

Author Contributions

Participated in research design: P.D.M., J.L.F., J.S., A.E., and D.G. Performed the data analysis: P.D.M. Wrote or contributed to the writing of the manuscript: P.D.M., D.G., W.J.M., C.D.H., S.L.G., M.R., J.L.F., S.J.B., J.S., A.E., R.G.G., D.K.B., J.C., K.Z. All authors read and approved the final manuscript.

Funding

The PBPK modeling analyses were funded by the Eunice Kennedy Shriver National Institute of Child Health and Human Development (NICHD) under award 5R01HD096435. The topiramate pediatric pharmacokinetic data were collected by the Pediatric Trials Network (PTN) funded under the Eunice Kennedy Shriver National Institute of Child Health and Human Development (NICHD) contract HHSN275201000003I for the Pediatric Trials Network (PI D. Benjamin). P.D.M. received support from the Eunice Kennedy Shriver National Institute of Child Health and Human Development (NICHD) of the National Institutes of Health (NIH) under Award Number 1T32HD104576. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.

Conflicts of Interest

The authors declare no conflicts of interest.

Disclosures

DG receives support from the Eunice Kennedy Shriver National Institute of Child Health and Human Development and other sponsors for drug development in adults and children (https://dcri.org/about‐us/conflict‐of‐interest/). The affiliations of PDM, JLF, and JS reflect the institution with which they were associated at the time this research was performed.

Supporting information

Supporting File 1: jcph70264‐sup‐0001‐SuppMat.pdf

Data Availability Statement

To help expand the knowledge base for pediatric medicine, the PTN is pleased to share data from its completed and published studies with interested investigators. For requests, please contact a PTN Program Manager (PTN‐Program‐Manager@dm.duke.edu).