Address correspondence to Dr. Ihab G. Girgis, Johnson & Johnson Pharmaceutical Research & Development, LLC, 920 Route 202 South, Raritan, NJ 08869, U.S.A. E-mail: IGirgis@its.JnJ.com
Purpose: To identify and validate the efficacious monotherapy dosing regimen for topiramate in children aged 2 to <10 years with newly diagnosed epilepsy using pharmacokinetic–pharmacodynamic (PK–PD) modeling and simulation bridging.
Methods: Several models were developed in pediatric and adult populations to relate steady-state trough plasma concentrations (Cmin) of topiramate to the magnitude of clinical effect in monotherapy and adjunctive settings. These models were integrated to derive and support the monotherapy dosing regimen for pediatric patients.
Key Findings: A two-compartmental population PK model with first-order absorption described the time course of topiramate Cmin as a function of dosing regimen. Disposition of topiramate was related to age, body weight, and use of various concomitant antiepileptic drugs. The PK–PD model for monotherapy indicated that the hazard of time to first seizure decreased with increasing Cmin and time since randomization. Higher baseline seizure frequency increased risk for seizures. Age did not significantly influence hazard of time to first seizure after randomization to monotherapy. For adjunctive therapy, the distribution of drug and placebo responses was not significantly different among age groups. Based on the available PK–PD modeling data, the dosing regimen expected to achieve a 65–75% seizure freedom rate after 1 year for pediatric patients age 2–10 years is approximately 6–9 mg/kg per day.
Significance: This analysis indicated no difference in PK–PD of topiramate between adult and pediatric patients. Effects of indication and body weight on PK were adequately integrated into the model, and monotherapy dosing regimens were identified for children 2–10 years of age.
Optimal control of seizures in patients with epilepsy requires determination of the dosing regimen for an individual patient that gives maximal efficacy with minimal probability of adverse effects (Beardsley et al., 1983). Recent reviews of treatment of pediatric epilepsy have focused on minimizing the number of agents used to control seizures in these patients. Monotherapy has the potential to improve adherence and decrease the risks for drug interactions and toxicity (Sander, 2004). Single-drug treatment may also decrease the overall cost of managing patients with epilepsy (Malphrus & Wilfong, 2007).
Achieving regulatory approval for a specific treatment regimen for a therapeutic agent typically requires controlled clinical trials, but several concerns exist about studies of single antiepileptic drugs (AEDs) in pediatric patients with epilepsy. These include ethical constraints (e.g., determination of who is likely to benefit from the study), difficulty in achieving consent from parents, and pain and anxiety associated with invasive sampling (requirement for multiple blood samples) (Jacqmin et al., 2005; Meibohm et al., 2005). Regulatory agencies have recognized these important barriers, and thus have instituted regulations such as the Pediatric Research Equity Act (PREA), which makes allowance for approving pediatric indications without controlled trials. Exceptions to the requirement for additional clinical studies, given proven drug safety and analysis (not discussed in this article), include similar disease course and drug effects in pediatric and adult populations; similar drug metabolism in pediatric patients and adults, as well as similar concentration–response relationships; and proven experience with an approach to expanding indications outside of additional clinical studies (U.S. Food and Drug Administration, 2003). Pharmacokinetic–pharmacodynamic (PK–PD) analysis has been employed to provide guidance for optimal dosing of AEDs in many different patient populations (Della Paschoa et al., 1998; Snodgrass & Parks, 2000; Miura, 2004); there is precedent for this approach in gaining an extension of indications for oxcarbazepine as monotherapy in children (Nedelman et al., 2007).
Topiramate is a broad-spectrum AED with established efficacy as monotherapy and adjunctive therapy (Lyseng-Williamson & Yang, 2008). It has been used safely and effectively in adult, pediatric, and infant patients (Grosso et al., 2005a,b; Glauser et al., 2007; Krakow et al., 2007; Ramsay et al., 2008). Topiramate is approved in the United States as monotherapy in patients ≥10 years of age with partial-onset or primary generalized tonic–clonic seizures (PGTCS), and as adjunctive therapy for adults and pediatric patients 2–16 years of age with partial-onset seizures or PGTCS, and in seizures associated with Lennox-Gastaut syndrome (Topamax PI, 2009).
The PK profile of topiramate is well documented. Maximum plasma concentrations (CMAX) are observed between 1.4 and 4.3 h after administration, and mean values for plasma CMAX and area under the concentration-time curve (AUC) increase linearly with dose; however, a greater-than-proportional increase in both parameters has been observed (Doose et al., 1996; Battino et al., 2005). Approximately 50% of the topiramate dose is excreted renally, and cumulative urinary excretion increases linearly and proportionally over the 200- to 1,200-mg dose range. Elimination half-life (t½) values calculated from plasma (21.5 h) and urinary data (18.5 h) are consistent and independent of dose (Doose et al., 1996; Battino et al., 2005).
On a weight basis, infants and young children have a higher topiramate clearance than adults, presumably due to age-dependent changes in the rate of drug metabolism. As a result, younger patients require higher dosages than milligram-per-kilogram dosing would indicate to achieve serum topiramate concentrations comparable with those found in older children and adults.
The objectives of this analysis were to develop a PK–PD bridging approach to assess and validate the efficacious monotherapy dosing regimen for topiramate in children 2 to <10 years of age with newly diagnosed epilepsy using the following strategy. First, PK modeling was carried out to describe the plasma concentration versus time profile of topiramate as a function of dose, age, concomitant medication, and other explanatory variables in both the adjunctive and monotherapy setting, as well as to predict plasma concentrations as a function of dosing regimen in the pediatric patient population of interest aged 2–10 years receiving monotherapy. Furthermore, the PK–PD relationship for topiramate was characterized by relating the magnitude of clinical effects to Cmin. The efficacy parameter assessed for monotherapy was the time to first seizure after randomization. For adjunctive therapy, percent reduction from baseline in seizure frequency and responder rate (≥50% reduction of baseline seizure rate) were evaluated. Because no monotherapy data were available for patients 2–5 years of age, supportive evidence was obtained by evaluating the effect of this age range on the PK–PD relationship for topiramate in the adjunctive setting. Simulations were carried out to determine dosing regimens required to achieve target seizure freedom rates.
Patients and Methods
Data from 11 double-blind studies of both adjunctive treatment and monotherapy with topiramate (Table S1) were used to construct an integrated PK dataset across these two indications, different studies, and a wide age range (2–85 years). The PK dataset contained 4,640 observations from 1,217 patients, including 751 observations from 258 pediatric patients aged 2–15 years. For patients <6 years of age, the PK dataset contained 12 patients in the adjunctive therapy studies and none who received monotherapy.
Pharmacodynamic data from eight adjunctive topiramate therapy studies in adult and pediatric patients with partial-onset seizures or PGTCS, and three double-blind topiramate monotherapy efficacy studies in newly/recently diagnosed patients with epilepsy, were included in the PD analysis (Table S1). These data were used to develop the exposure–response model on the basis of topiramate Cmin following twice-daily dosing. Table 1 further summarizes variables in the monotherapy studies relevant to the PK–PD bridging.
Table 1. Distribution of subject variables relevant for the model development, across the three monotherapy studies
Study 9 (TOPMAT-EPMN-104) (N = 252)
Study 10 (TOPMAT-EPMN-105) (N = 396)
Study 11 (TOPMAT-EPMN-106) (N = 470)
Total (N = 1,118)
Cmin, steady-state trough plasma concentration.
aIn the respective trial.
Cmin (μg/ml), n (%)
Baseline seizure frequency (per 3 months), n (%)
Age (years), n (%)
Population pharmacokinetic model
Topiramate concentration–time data were analyzed using nonlinear mixed-effects modeling as implemented in the computer program, NONMEM (Version V Level 1.1, GloboMax LLC, Hanover, MD, U.S.A.), compiled using Compaq Digital Visual Fortran (Version 6–Update A, Compaq Computer Corporation, Houston, TX, U.S.A.). The first-order conditional estimation method (FOCE) produced estimates of structural parameters, as well as estimates of interindividual and intraindividual variability. A standard modeling approach was followed (Mandema et al., 1992). Covariates included in the model were age, body weight, serum creatinine, concomitant medications, and adjuvant treatment. A stepwise covariate model-building strategy was adopted. The current population model was further expanded to characterize the population PK of topiramate in infants 1–24 months old (Nandy et al., 2010).
The PK–PD models related Cmin of topiramate to the magnitude of clinical effect in both the monotherapy and adjunctive settings. For monotherapy, the time to first seizure after randomization was evaluated. For adjunctive therapy, percent reduction in seizure frequency from baseline and responder rate (≥50% reduction of baseline seizure rate) were evaluated. Time to first seizure was selected as the end point for patients receiving monotherapy because they were newly diagnosed rather than treatment-refractory patients and because percent seizure reduction was not an appropriate end point for this group.
Model for monotherapy
Various parametric models were evaluated relating hazard (λ) for time to first seizure after randomization to Cmin, baseline seizure frequency (BS), age, and time since randomization (t). Model development was performed on the basis of stepwise inclusion of candidate explanatory variables. Models were evaluated using S-PLUS, version 7.0 (Insightful Corp., Seattle, WA, U.S.A.) and compared on the basis of the minimum value of the objective function (MOF) as obtained from the analysis results using the glm function for Generalized Linear Models in S-PLUS, as well as on the basis of the basic diagnostic plots. The final model was further validated by stepwise inclusion/deletion of the potential explanatory variables identified during the model-building process. The final model for (log-transformed) hazard in the ith subject, λi, can be described as follows:
where λ0 is the baseline hazard, t is time since randomization, and Cmin,i is the steady-state trough plasma concentration of topiramate for the ith subject, as obtained from the PK model. Several other variables reflect the inclusion criteria for baseline seizure frequency in the study patients: BS3–10,i is an indicator value (0 or 1) reflecting whether the ith subject had up to 3–10 seizures during the 3-month run-in period (BS3–10,i = 1) or not (BS3–10,i = 0); and BS>10 is a similar indicator reflecting whether the ith subject had >10 seizures during the 3-month run-in period (BS>10,i = 1) or not (BS>10,i = 0). λt, λCmin, λBS3–10, and λBS>10 are the parameters describing the relationship between log (hazard) and t, Cmin,i, BS3–10,i, and BS>10,i, respectively.
Model for adjunctive therapy
For the primary end point analysis, the efficacy response variable, Y, was used to represent the log-transformed percent reduction in seizure frequency from baseline, thus (Girgis et al., 2009):
where S is the average seizure frequency per 28 days during the double-blind phase, B is the corresponding average during the baseline phase, and
is the percent change of seizure from baseline.
Based on this, the efficacy-exposure relationship of topiramate for seizure frequency reduction in adjunctive therapy was expressed as follows (Nedelman et al., 2007):
where the individual observed efficacy response variable, Yobs, i, is expressed as a function of the individual trough plasma concentration at steady state, Cmin,i, and the subject’s baseline seizure frequency, Bi (centered around the population mean, ), and an interaction term between Bi and Cmin,i. Normal least-squares regression was used to provide estimates for β1, β2, and β3 and their significance.
The second adjunctive model describes the relationship between responder rate (≥50% reduction of baseline seizure rate) and possible explanatory variables (Cmin and age) on the basis of a logistic regression. The base model describing the probability of response (PRESP) is defined as follows (Girgis et al., 2009):
and p0 is the placebo responder rate, Cmin is the topiramate model-predicted steady-state trough concentration, and PED is the pediatric status (<16 years). EMAX is the maximum drug effect, EC50 is the value for Cmin that results in 50% of the maximum response, and PPED is the difference in placebo effect between adult and pediatric patients.
The final PK model, as shown in Fig. S1, was a two-compartment linear model with first-order absorption. The model included the effects of weight, age, and concomitant AEDs on clearance. In addition, the model took into consideration the apparent difference in baseline clearance of topiramate for treatment-naive patients versus patients previously treated with other concomitant AEDs, and the effect of weight on the central volume of distribution. The parameter values obtained from the final model are presented in Table S2.
The model confirmed the influence of covariates such as age, body weight, and concomitant medications on clearance. Patients on adjunctive treatment exhibited a clearance of topiramate about 200% higher compared with patients on monotherapy. The influence of body weight and age on clearance was more pronounced in the adjunctive therapy treatment group than in the monotherapy group. For example, for an increase in weight from 25–50 kg, the modeled clearance increased almost twofold for pediatric patients on adjunctive therapy compared with a 1.25-fold increase for pediatric patients on monotherapy.
Using the PK model, Fig. 1 shows the relationship between dose and Cmin for various ages. As a result of the nonlinear relationship between age and systemic clearance, for younger pediatric patients, a more-than-proportional correction for body weight is required to achieve similar topiramate steady-state trough concentrations.
Time to first seizure after randomization
The parameters of the final PK–PD model of topiramate for time to first seizure after randomization in monotherapy are summarized in Table 2. The results of the model validation by stepwise inclusion/deletion of the potential explanatory variables identified during the model-building process are shown in Table S3. As is evident from the table, age was carefully examined as a key explanatory variable in the model. Fig. 2 shows the goodness-of-fit plot for the final model of time to first seizure versus exposure to topiramate as monotherapy. Higher topiramate steady-state Cmin values are correlated with a lower hazard of the time to first seizure after randomization. There is also a clear relationship between the number of seizures observed during the 3-month baseline period and the hazard (instantaneous risk) of time to first seizure after treatment randomization (hazard increases at higher baseline seizure frequency). Fig. 2 shows that the effects of these explanatory variables are adequately captured in the model. There was no apparent relationship between time to first seizure after randomization and age.
Table 2. Parameter estimates of the final hazard model for time to first seizure after randomization to topiramate as monotherapy (see Equation 1)
Estimate ± SE
SE, standard error; λ0, hazard (the instantaneous risk of a first seizure after randomization to occur); λt, parameter describing the relationship between log (hazard) and t; λCMIN, parameter describing the relationship between log (hazard) and Cmin; λBS3–10, parameter describing the relationship between log (hazard) and BS3–10,i; λBS>10, parameter describing the relationship between log (hazard) and BS>10,i.
−3.130 ± 0.0919
−0.051 ± 0.0036
−0.112 ± 0.0151
1.048 ± 0.1046
2.411 ± 0.1356
Comparison of predicted versus observed seizure times
In a subsequent stage, any effect of age on the PD of topiramate monotherapy was further excluded. Predicted seizure times from the final model were compared with the observed values of time to first seizure over the time course of the three monotherapy studies in the age ranges 6–9, 10–15, and 6–15 years. The results (Fig. 3) showed that the model-predicted seizure times in the relevant age ranges were within the 95% confidence interval of the nonparametric Kaplan-Meier estimate of survival. Between 20 and 40 weeks, the fraction of patients without seizures after randomization seems to be somewhat underpredicted, particularly in 10- to 15-year-old children. However, it should be noted that the relevant clinical end point, seizure freedom after 1 year (52 weeks), is well described by the model for all age groups.
Models for adjunctive therapy and validation
The lack of difference in the PK–PD relationship for topiramate monotherapy between children aged 2 to <10 years and adolescents/adults may be due to the limited amount of clinical data in the relevant age range (see Table S1). However, if plasma levels associated with seizure control in adults during adjunctive therapy are similar to those associated with seizure control during pediatric adjunctive therapy, then this would provide indirect, supportive evidence of a similar exposure–response relationship during monotherapy between adults and children ≥10 years and pediatric patients 2 years to <10 years of age.
To test this hypothesis, the final PD model for both the pediatric and adult populations for seizure frequency reduction in adjunctive therapy was defined as follows:
Both the baseline seizure frequency and the interaction term (Equation 4) were shown to be statistically insignificant (p = 0.58 and p = 0.11, respectively). As predicted, no statistically significant differences between the two populations have been found for the placebo effect (βo) or the concentration–effect relationship (β1) estimates. The Kolmogorov-Smirnov goodness-of-fit test as well as visual inspections suggested that the distributions of the placebo response were not different in pediatric and adult patients during adjunctive therapy (p = 0.532). The final model parameter estimates and fit are shown in Table 3 and Fig. 4, respectively.
Table 3. Parameter estimates ± SE of the final topiramate model for efficacy response and for responder rate in adjunctive therapy
βo, placebo effect on Y; β1, concentration effect on Y; EMAX, maximum drug effect; EC50, value for CMIN resulting in 50% of the maximum response; P0, placebo responder rate; SE, standard error.
Parameter estimates for efficacy response
Adults (n = 663)
Pediatric patients (n = 115)
4.4469 ± 0.0313
4.4830 ± 0.0916
4.4538 ± 0.0361
−0.0627 ± 0.0097
−0.0579 ± 0.0305
−0.0628 ± 0.0092
Parameter estimates for responder rate
Estimate ± SE
−1.605 ± 0.110
1.564 ± 0.196
0.613 ± 0.415
The second adjunctive model, the PD model for the responder rate in adjunctive therapy, showed no pediatric effect (<16 years of age) on the exposure–response relationship. The final model is defined as follows:
The parameters of the final model for responder-rate end point have been summarized at the bottom of Table 3. Validation of the model is summarized in Table S4. Additional validation for all monotherapy and adjunctive models was conducted to explore the sensitivity of the model to any unobserved confounders.
Establishment of an effective monotherapy regimen for topiramate in pediatric patients
Identification of the dosing regimen for children 2 to <10 years of age was based on patients with 1–2 baseline seizures, similar to the TOPMAT-EPMN-106 study population, which was the pivotal trial supporting initial approval in the United States. Patients with only 1–2 seizures during baseline are less likely to have refractory epilepsy and more likely to benefit from AED monotherapy. The expected PD (Fig. 5) and PK profiles (Fig. 1) are integrated (Fig. 6) to establish the dosing regimen for the target response rate (expected seizure freedom rate after 1 year) in pediatric patients 2 to <10 years of age. The seizure freedom rate increases from about 50 to 82% for a dose range from 1–10 mg/kg per day. Furthermore, the seizure freedom rate for a particular dose per kilogram body weight increases with age, indicating that dose correction on the basis of body weight alone is not fully sufficient to account for differences in exposure between patients of different ages. As shown in Fig. 6, based on simulations with the PK and PK–PD models, the target seizure freedom rate of 65–75% after 1 year can be reached by patients aged 2–10 years with a dosing regimen of approximately 6–9 mg/kg per day. The model suggests that patients aged 6–10 years can achieve the seizure freedom target with doses near the lower end of the dose range, whereas younger patients may achieve the targeted seizure freedom rate with doses near the high end of the dose range.
Results from this PK–PD analysis showed that there is a clear exposure–response relationship for topiramate monotherapy with respect to the time to first seizure after randomization, and that this relationship could be best described on the basis of a linear hazard model. Most importantly, there was no effect of age on this relationship, thereby providing evidence that topiramate efficacy is not different among various age groups. The exposure–response relationship for the seizure frequency reduction with topiramate adjunctive therapy can be best described on the basis of a linear model. These results also indicated no effect of age on either the placebo effect or the concentration–effect relationship of topiramate for seizure frequency reduction. The exposure–response relationship for responder rate with topiramate as adjunctive therapy can be best described on the basis of an Emax PD model. Results also indicated the absence of an effect of age on both the placebo-effect and the concentration-effect relationships of topiramate for responder rate in the adjunctive setting.
Although both monotherapy and adjunctive data strongly indicate that there is no direct evidence of an effect of age or pediatric status on the PD characteristics of topiramate, the pharmacokinetics of topiramate are clearly dependent on age, body weight, and concomitant medication, which were all adequately captured in the population PK model. With this model, adequate predictions of drug exposure in the proposed pediatric monotherapy population (aged 2 to <10 years) were obtained. In combination with the established PD model that relates drug exposure (Cmin) to seizure freedom, a dosing regimen was proposed that is expected to result in a predefined target response rate of seizure freedom over 1 year.
The results from this modeling analysis are consistent with dose–response effects for topiramate observed in individual clinical studies. Results from a multinational, randomized, double-blind trial in adults and children (≥6 years old, weights ranging from 25 to <85 kg) with newly/recently diagnosed epilepsy who were randomized to 400 or 50 mg/day topiramate as target maintenance dosages, indicated that the probability of being seizure free at 6 months was 83% in patients randomized to 400 mg/day and 71% in those randomized to 50 mg/day (p = 0.005). Seizure-free rates at 12 months were 76% and 59%, respectively (p = 0.001) (Arroyo et al., 2005). A study of adults and children (≥3 years of age, weights ranging from 25–110 kg) randomized to 50 or 500 mg/day topiramate (25 or 200 mg/day if weight was ≤50 kg) indicated higher seizure-free rates (54% vs. 39%; p = 0.02) among those assigned to the higher doses. Results from this study also demonstrated that higher plasma topiramate concentrations were associated with increased time to first seizure (p < 0.01) (Gilliam et al., 2003). A double-blind, dose-controlled study evaluating topiramate as monotherapy in 470 patients with newly diagnosed epilepsy or epilepsy relapse in the absence of therapy indicated that 400 mg/day (8.9 ± 3.7 mg/kg per day) and 50 mg/day (1.1 ± 0.4 mg/kg per day) resulted in respective probabilities of being seizure free at 6 months of 90% and 78%. The respective values at 12 months were 85% and 62% (Glauser et al., 2007).
The integrated analysis of PD data of topiramate across adult and pediatric age ranges in both monotherapy (≥6 years old) and adjunctive therapy settings (≥2 years of age) indicates that, on the basis of available clinical trial data, there is no direct evidence of an effect of age or pediatric status on the PD characteristics of topiramate when used alone or as adjunctive therapy. The combination of this information with PK modeling results has permitted determination of steady-state Cmin values for topiramate monotherapy required to achieve seizure freedom in different age groups. Based on the available PK–PD modeling data, the dosing regimen expected to achieve a 65–75% seizure freedom rate after 1 year for pediatric patients aged 2–10 years is approximately 6–9 mg/kg per day.
The authors thank Dr. Paul Soons, Dr. An Vermeulen, Dr. Donald Heald, and Dr. Prasarn Manitpisitkul for their scientific input. Editorial support was provided by Kara Quick, ELS, at Publication CONNEXION (Newtown, PA).
The authors confirm that they have read the Journal’s position on issues involved in ethical publication and affirm that this report is consistent with those guidelines. None of the authors has any conflict of interest to disclose.