Extrapolating Efficacy of Cariprazine to Pediatric Schizophrenia and Bipolar Mania Through Population Pharmacokinetic Analysis

Abstract

The US FDA recently accepted efficacy extrapolation from adult to pediatric patients for atypical antipsychotics in the treatment of schizophrenia (SCZ) and bipolar mania (BPM), considering the similarity in both disease and exposure‐response relationships for this class of drugs. Under this pharmacokinetic (PK)‐based efficacy extrapolation framework, pediatric population PK (popPK) models for cariprazine and its two active metabolites (desmethyl cariprazine [DCAR], didesmethyl cariprazine [DDCAR]) were developed by modifying previously validated adult popPK models and using pediatric PK data from two phase I trials. Covariate effects on PK parameters in legacy adult models were retained in the pediatric models, except that body weight effect on apparent elimination clearance and apparent volume of distribution of the central compartment were re‐estimated for all analytes. The final pediatric popPK models adequately described observed plasma concentrations of all three analytes with no significant bias over age or body weight in pediatric patients. Simulated total cariprazine (sum of cariprazine, DCAR and DDCAR) exposure at 1.5‐4.5 mg once daily (QD) in adolescents (13‐17 years) with SCZ and 3.0‐4.5 mg QD in children (10‐12 years) and adolescents with BPM resulted in predicted total cariprazine exposures within established efficacious ranges in adults with these conditions. Body weight did not meaningfully affect total cariprazine exposure and analyses did not support weight‐based dosing in 10‐12 year olds. This PK analysis supported the FDA approvals of cariprazine for the treatment of SCZ and BPM in pediatric patients by identifying efficacious doses for each indication based on efficacy extrapolation.

Introduction

Schizophrenia (SCZ) and bipolar disorder are chronic psychiatric disorders associated with substantial disability. ¹ , ² Approximately 8% and 14% of people with SCZ and bipolar disorder, respectively, are diagnosed before 18 years of age. ³ Onset of these conditions during childhood and adolescence can be particularly devastating for both patients and their caregivers. Current pharmacologic treatment options for pediatric populations are limited for both conditions. ⁴ , ⁵ Further, use of on‐label second‐generation antipsychotics in children and adolescents are associated with cardiometabolic risk, including body weight gain and lipid abnormalities, which could have lasting cardiometabolic effects. ⁶ Lastly, both SCZ and bipolar disorder can become treatment resistant. ⁴ Therefore, new safe and effective medications to treat pediatric SCZ and bipolar disorder are in great need.

Cariprazine, a dopamine D₃/D₂ receptor (D₃‐preferring) and serotonin 5‐HT_1A receptor partial agonist, ⁷ is approved to treat a variety of psychiatric conditions in the US. This includes the treatment of SCZ (1.5‐6 mg daily) and bipolar disorder (both bipolar mania [BPM; 3‐6 mg daily] and bipolar depression [1.5 or 3 mg daily]) as a monotherapy (approvals received between 2015 and 2019) and major depressive disorder as an adjunctive therapy (1.5 or 3 mg daily; approval received in 2022) in adult patients. In cariprazine clinical trials, most common adverse events for adults with SCZ or BPM were extrapyramidal symptoms, akathisia, insomnia, headache, somnolence, nausea, and vomiting. ⁸ In 2025, the FDA also approved cariprazine for the treatment of pediatric SCZ (13‐17 years; 1.5‐4.5 mg daily) and BPM (10‐17 years; 3 or 4.5 mg daily). ⁸ Importantly, the overall safety profile was similar in adult and pediatric SCZ and BPM populations. ⁹

Medication approvals for use in children tend to lag behind approvals for use in adults by several years due to the need for dedicated pediatric trials. As a result, regulatory agencies have provided general guidance on extrapolating adult data to pediatric populations to optimize and streamline pediatric drug development. ¹⁰ , ¹¹ In addition, Kalaria et al. established similarity between antipsychotic exposure‐response relationships ¹² and the placebo effect ¹³ between adults and adolescents with SCZ ¹² , ¹⁴ and BPM. ¹⁴ These analyses provided the foundation to extrapolate atypical antipsychotic efficacy in adults with SCZ and BPM to a similar pediatric population, leveraging available pharmacokinetics (PK) data and modeling/simulation methods for exposure matching.

Cariprazine PK has been clinically characterized in adult ¹⁵ , ¹⁶ and pediatric ⁹ patients with SCZ or BPM. Cariprazine has two active metabolites, desmethyl cariprazine (DCAR) and didesmethyl cariprazine (DDCAR), both of which have similar binding properties and pharmacological activity as the parent drug. ⁷ , ⁸ The drug is extensively metabolized to DCAR by CYP3A4 and, to a lesser extent, CYP2D6 with negligible urinary excretion; DCAR is metabolized to DDCAR by CYP3A4. ⁸ In both adult and pediatric populations, total cariprazine (total CAR; cariprazine+DCAR+DDCAR) exposures were dose proportional, with cariprazine and DCAR plasma concentrations reaching steady state within 1‐2 weeks; DDCAR within 4‐5 weeks. ⁸ , ⁹ , ¹⁵ The half‐lives of cariprazine, DCAR, and DDCAR have been estimated at 2‐4 days, 1‐2 days, and 1‐3 weeks, respectively, and all three moieties accumulate over time. ⁸ At steady‐state, cariprazine, DCAR, and DDCAR account for approximately 28%, 8% and 64%, respectively, of total cariprazine exposure. ¹⁵

Here, we present the pharmacokinetic analyses that supported the extrapolation of efficacy of cariprazine in SCZ and BPM from adult to pediatric patients. A pediatric popPK model for cariprazine, DCAR, and DDCAR concentration was developed by modifying an established popPK model for adults with SCZ or BPM (legacy adult models) ¹⁵ using PK data from two phase 1 clinical studies. Cariprazine, DCAR, and DDCAR exposures were then simulated using this model to guide pediatric dosing recommendations, targeted to achieve comparable adult exposures at approved doses. ¹⁷

Methods

This analysis used existing clinical PK data from two phase 1 pediatric studies. The Independent Ethics Committee or Institutional Review Board at each study site approved the study protocols, informed consent forms, and recruitment materials before patient enrollment. The studies were conducted in accordance with the International Council for Harmonization guidelines, applicable regulations, and the Declaration of Helsinki. Each study was discussed with the patient and their parent or legally authorized representative and written informed consent was obtained.

Clinical Study Data

The two Phase 1 pediatric clinical studies examined cariprazine safety, tolerability, and PK. Study 1 included pediatric patients with SCZ (13‐17 years old) or BPM (10‐17 years old) and had a duration of 42 days (daily dose: 1.5 mg, 3 mg, or 4.5 mg following titration). ⁹ Blood samples for PK analyses of cariprazine, DCAR, and DDCAR were collected on Day 1 at 0 h (predose), and 2, 3, 4, 6, 8, and 24 h postdose; on Days 7, 14, 21, 28, 35, 39, 40, and 41 predose; and Day 42 predose and 2, 3, 4, 6, 8, 24, 48, 168, 336, 672, and 1008 h postdose. ⁹ Study 2 included pediatric (13‐17 years) and adult (18‐40 years) patients with SCZ (daily dose: 1.5 mg, 3 mg, or 6 mg after titration; EudraCT: 2016‐002327‐29 [unpublished data]) and had a duration of 28 days. Blood samples for PK analyses of cariprazine, DCAR, and DDCAR were collected on Day 14 at 0 h (predose); Day 21 at 0 h (predose); Day 28 at 0 h (predose) and 1, 2, 3, 4, and 8 h postdose; and Days 29, 31, 35, and 42 (24, 72, 168, and 336 h post Day 28 dose). Patients who had received at least one cariprazine dose and who had at least one measurable cariprazine plasma concentration were included in the popPK analysis. Comparison of observed cariprazine, DCAR, and DDCAR plasma concentrations in patients with SCZ and BPM justified combining the two patient populations into a single data set for pediatric popPK model development (Figure S1). The M1 imputation method was used and all measurements below the lower limits of quantitation (cariprazine: 0.02 ng/mL, DCAR: 0.02 ng/mL, DDCAR: 0.05 ng/mL) were excluded from the analysis. ¹⁸

Pediatric Population Pharmacokinetic Model

The legacy adult popPK model that described cariprazine, DCAR, and DDCAR concentrations served as the starting model for this pediatric analysis. ¹⁸ Briefly, cariprazine PK was described by a three‐compartment model with zero‐order input to a dose depot compartment followed by first‐order absorption and linear elimination, which enabled the formation of DCAR. ¹⁵ DCAR PK was characterized by a two‐compartment model followed by linear elimination serving as the input to a transit compartment, effectively delaying the formation of DDCAR via linear elimination from the transit compartment. DDCAR PK was modeled by a two‐compartment model with linear elimination. ¹⁵  Because cariprazine is metabolized to DCAR and DCAR is metabolized to DDCAR, the models for the three moieties were interconnected and developed sequentially. This adult popPK model formed the basis for establishing a pediatric popPK model. The current analysis presented here focused on confirming model adequacy with re‐estimated body weight effects on PK parameters to account for the wider age and body weight range of patients in the dataset, including children as young as 10 years old. The pediatric popPK analysis was performed in a step‐wise manner as follows:

1. Exploratory graphical analysis of cariprazine, DCAR, and DDCAR concentrations over time since last cariprazine dose

2. Performance assessment of legacy adult popPK model for cariprazine, DCAR, and DDCAR with re‐estimated body weight effects in pediatric patients with SCZ and BPM.

3. Different dosing scenarios were simulated using the final pediatric popPK model.

All PK parameters in the pediatric models were fixed to the values of the legacy adult model except for inter‐individual variabilities (IIV) for the apparent elimination clearance (CL/F), apparent volume of distribution of the central compartment (V_c/F), and absorption rate constant (k _a), which were re‐estimated. Covariates included in the legacy adult popPK models for cariprazine, DCAR, and DDCAR were retained and included body weight (cariprazine, DCAR, DDCAR models), race (cariprazine, DCAR, DDCAR models), and sex (DCAR model) for CL/F and body weight (cariprazine, DCAR, DDCAR models) and race (DCAR models) for V_c/F. However, in order to account for the wider age and body weight range of patients in the dataset, including children as young as 10 years old, the body weight effect was re‐estimated with reference to the fixed adult CL/F and V_c/F population values as follows (Equation 1):

$${\theta }_i=\mathrm{T}{\mathrm{V}}_{\mathrm{pop}.\mathrm{adult}}\times {\left(\frac{\mathrm{BW}}{79\mathrm{kg}}\right)}^{{\theta }_{\mathrm{pediatric}}}\times e^{{\eta }_i}$$ (1)

where θ_i represents either CL/F or Vc/F for the individual pediatric patient i, TV_pop.adult is the typical value of the adult population at the reference weight (79 kg), BW is the individual body weight of the pediatric patient, θ _pediatric is the estimated allometric coefficient for body weight effect, and η_i is the inter‐individual variability (IIV), normally distributed with mean 0 and variance ω ².

Population Pharmacokinetic Model Selection and Evaluation

Cariprazine was modeled first, then DCAR, and finally DDCAR. Estimated model parameters were fixed for subsequent model development (i.e. cariprazine parameters fixed in DCAR model, cariprazine and DCAR parameters fixed in DDCAR model). A model estimation routine yielded fit‐for‐purpose results that were evaluated for robustness, convergence, parameter uncertainty, condition number, and shrinkage. For acceptance, model parameter estimates had to be physiologically plausible, goodness‐of‐fit plots needed to have no systematic trends, and visual predictive checks (VPCs) had to show adequate characterization of the central tendency and observed variability.

Goodness‐of‐fit plots included population and individual predicted versus observed concentration plots, as well as conditional weighted residuals (CWRES) versus population‐predicted values and time since last dose for cariprazine, DCAR, and DDCAR. The distributions of random effects and residuals were examined through histograms and quantile–quantile (QQ) plots; potential covariate‐parameter relationships through covariates against random effects plots. For prediction‐corrected VPCs, 500 simulated replicates of the dataset were generated using NONMEM (version 7.5.1, ICON plc, Dublin, Ireland). Observed data were then superimposed on simulated data and prediction‐corrected. Simulated median and 5th/95th percentiles, together with their 90% confidence intervals (CIs), were then used for comparison.

Pediatric Population Pharmacokinetic Model Simulations

Simulated cariprazine, DCAR, DDCAR, and total CAR concentration‐time profiles in pediatric and adult patients were created using the final pediatric and legacy adult popPK models, respectively. For adults, the existing patient dataset used in legacy adult model development was used. ¹⁵  For children (10‐12 years) and adolescents (13‐17 years), 1000 virtual patients were generated for each age group using the newly developed pediatric popPK model. The age of the pediatric population was sampled uniformly within the specified range for each age category. The resulting ages were then converted to months and subsequently mapped to the nearest corresponding age value in the Center for Disease Control (CDC) growth chart. ¹⁹  The body weight for a specific individual was determined by using the body weight distribution for a specific age and sex on the CDC growth chart and Equation (2):

$$X=M\times {\left(1+L\times S\times Z\right)}^{\wedge }\left(1/L\right)$$ (2)

where X is body weight, M is median body weight for a specific age and sex, S is the generalized coefficient of variation for a specific age and sex, L is the Box‐Cox transformed power, and Z is the z‐score. The M, L, and S parameters were obtained from CDC growth tables for each specific age and sex. ¹⁹ The z‐score was randomly generated from a distribution with a mean (±SD) of 0 (±1). Except for body weight, all other cariprazine, DCAR, and DDCAR model covariates followed the same sex (male/female) and race (White/Black/Asian/Japanese/other) proportions as the adult legacy population. ¹⁵

Both fixed and titrated dosing regimens were simulated in the three pediatric (children <40 kg, children ≥40 kg, adolescents) and one adult virtual groups as detailed in Table 1. The median and 5th/95th percentiles of predicted cariprazine, DCAR, DDCAR, and total CAR concentrations‐time profiles for each dosing regimen were graphically compared between adults and each of the three pediatric groups. Simulations were also conducted to evaluate cariprazine, DCAR, DDCAR and total CAR exposures (including maximum concentration [C _max], trough concentration [C _trough], area under the concentration‐time curve [AUC] over one dosing interval [AUC_tau; 1 day]) at steady state (C _max,ss, C _trough,ss, AUC_tau,ss) after 42 days of fixed cariprazine dosing. Model‐predicted cariprazine, DCAR, DDCAR, and total CAR exposures following comparable dosing were then compared between the pediatric and adult populations.

Table 1.

Simulated Cariprazine Dosing Regimens.

	Titration regimen	Full dose
Fixed dosing (steady‐state evaluations, daily dosing)
Children (10‐12 years old, <40 kg)	‐	1.5 mg, 3 mg, 4.5 mg
Children (10‐12 years old, ≥40 kg)	‐	1.5 mg, 3 mg, 4.5 mg
Adolescents (13‐17 years old)	‐	1.5 mg, 3 mg, 4.5 mg
Adults (≥18 years old)	‐	1.5 mg, 3 mg, 4.5 mg, 6 mg
Dosing with titration
Children (10‐12 years old, <40 kg)	0.5 mg on Days 1‐2	4.5 mg/day on Day 8 and beyond
	1.5 mg on Days 3‐4
	3.0 mg on Day 5‐7
Children (10‐12 years old, ≥40 kg)	0.5 mg on Days 1‐2	4.5 mg/day on Day 8and beyond
	1.5 mg on Days 3‐4
	3.0 mg on Days 5‐7
Adolescents (13‐17 years old)	0.5 mg on Days 1‐2	4.5 mg/day on Day 8 and beyond
	1.5 mg on Days 3‐4
	3.0 mg on Days 5‐7
Adults (≥18 years old)	None	1.5 mg/day for treatment duration
Adults (≥18 years old)	1.5 mg on Day 1	3.0 mg/day on Day 2 and beyond
Adults (≥18 years old)	1.5 mg on Days 1‐2	6.0 mg/day on Day 8 and beyond
	3.0 mg on Days 3‐4
	4.5 mg on Days 5‐7

Results

Patients

Data from 85 pediatric and 16 adult patients with SCZ or BPM with 4237 concentration records (cariprazine: n = 1570, DCAR: n = 1269, DDCAR: n = 1398) were included in the development of the pediatric popPK model for cariprazine, DCAR, and DDCAR. Briefly, 60% of patients were male, mean age was 16 years, and mean body weight was 59.6 kg (range: 26.9‐95.4 kg; Table 2).

Table 2.

Baseline Characteristics of Patients Included in the Development of Pediatric and Adult Population Pharmacokinetic Models of Cariprazine, DCAR, and DDCAR Concentrations.

	Pediatric model
	Study 1 9 (N = 49)	Study 2 a (N = 52)	All (N = 101)	Legacy adult model 15 (N = 2199)
Male, n (%)	23 (47)	38 (73)	61 (60)	1461 (66.4)
Age, years, mean±SD	13 ± 2	19 ± 8	16 ± 6	39.2 ± 10.8
Child (10‐12 years), n (%)	25 (51)	0	25 (25)	0
Adolescent (13‐17 years), n (%)	24 (49)	36 (69)	60 (59)	0
Adults (≥18 years), n (%)	0	16 (31)	16 (16)	2199 (100)
Body weight, kg, mean±SD	55.2 ± 15.7	63.6 ± 12.6	59.6 ± 14.7	78.9 ± 18.7
Range	26.9‐95.4	42.7‐90.6	26.9‐95.4	‐
<40 kg, n (%)	9 (18)	0	9 (9)	0
≥40 kg, n (%)	40 (82)	36 (69)	76 (75)	0
Adults, n (%)	0	16 (31)	16 (16)	2199 (100)
Race, n (%)
White	8 (16)	51 (98)	59 (58)	1003 (46)
Black	41 (84)	0	41 (41)	767 (35)
Asian	0	1 (2%)	1 (1)	351 (16)
Other	0	0	0	78 (4)
Cariprazine dose, n (%)				0.5‐21 mg b
1.5 mg	13 (27)	17 (33)	30 (30)	‐
3.0 mg	12 (24)	18 (35)	30 (30)	‐
4.5 mg	24 (49)	0	24 (24)	‐
6.0 mg	0	17 (33)	17 (17)	‐

Study 1 population: pediatric patients with schizophrenia (13‐17 years old) or bipolar I disorder (10‐12 years old); Study 2 population: pediatric (13‐17 years old) and adult (18‐40 years) patients with schizophrenia.

^a Unpublished data.

^b Cariprazine dose at time of pharmacokinetic sample collection. ¹⁵

Exploratory Pharmacokinetic Analysis

The dose‐normalized observed cariprazine, DCAR, and DDCAR concentration‐time since last dose (TSLD) profiles from the two pediatric studies were comparable. Therefore, data were pooled for the development of the pediatric popPK model (Figure S1).

Pediatric Population Pharmacokinetic Modeling

Cariprazine

Cariprazine plasma concentrations in patients with SCZ or BPM in two phase 1 pediatric studies were adequately described by the final pediatric popPK model. PK parameter estimates are summarized in Table 3. All PK parameters that were re‐estimated had a relative standard error of ≤73.4%. Shrinkage for cariprazine CL/F, V_c/F, and k _a were 5.34%, 18.8%, and 44.5%, respectively.

Table 3.

Cariprazine PK Parameter Estimates and Inter‐Individual Variability in Pediatric Patients With Schizophrenia or Bipolar I Disorder.

Cariprazine PK parameter	Population estimate	%RSE	95% CI
CL/F, L/h (fixed)	21.5	‐	‐
Vc/F, L (fixed)	266	‐	‐
Q3/F, L (fixed)	0.431	‐	‐
Vp1/F, L (fixed)	149	‐	‐
DUR/F, h (fixed)	2.57	‐	‐
k a, 1/h (fixed)	0.352	‐	‐
Q4/F, L/h (fixed)	100	‐	‐
Vp2/F, L (fixed)	501	‐	‐
Vc/F prop shift for 1st dose (fixed)	2.84	‐	‐
Q3/F prop shift for 1st dose (fixed)	39.4	‐	‐
Vp1/F prop. shift for 1st dose (fixed)	2.61	‐	‐
Power of CL/F for weight	0.0977	73.4	−0.0429‐0.238
Power of Vc/F for weight	0.998	28.6	0.438‐1.56
CL/F prop shift for race = Black (fixed)	−0.0907	‐	‐
CL/F prop shift for race = Asian (fixed)	−0.178	‐	‐
CL/F prop shift for race = Japanese (fixed)	−0.111	‐	‐
Proportional error full profile studies	0.107	2.59	0.102‐0.113
Proportional error (fixed)	0.154	‐	‐

Inter‐individual variability	Population estimate	%CV	%shrinkage
On CL/F	0.0608	25.0	5.34
On Vc/F	1.47	184	18.8
On k a	0.906	121	44.5

PK, pharmacokinetic; CL/F, apparent elimination clearance; V_c/F, apparent central volume of distribution; Q₃/F, apparent first distribution clearance; V_p1/F, apparent first peripheral volume of distribution; DUR/F, duration of zero‐order absorption; k _a, first‐order absorption rate constant; Q₄/F, apparent second distribution clearance; V_p2/F, apparent second peripheral volume of distribution; prop shift, proportional shift. %RSE was calculated as the standard error of the estimator divided by the absolute value of the mean of the estimator multiplied by 100. %CV was calculated as SQRT[exp(ω ²)−1]×100.

Goodness‐of‐fit plots showed a random distribution of observed versus individual or population model‐predicted cariprazine concentrations along the line of unity (Figure S2). Additionally, CWRES distributions showed no trends when plotted against time since last dose or population predicted cariprazine concentration (Figure S2). Lastly, QQ plots of CWRES and the interindividual random effects showed no model deficiencies (Figure S3).

The pediatric cariprazine popPK model adequately captured observed cariprazine concentration‐time course and variability in pediatric patients with SCZ or BPM, as demonstrated by prediction‐corrected VPCs over 24 h after the last dose and over the full concentration‐time curve (Figures S4 and S5). Further, the model adequately described cariprazine concentrations across different age and body weight ranges. The relationships between model covariates of interest (age, body weight) and the random effects IIV on cariprazine CL/F, V_c/F, and k _a are shown in Figure 1.

Figure 1.

IIV versus model covariates for cariprazine (k _a, V _c/F, CL/F), DCAR (V _c/F, CL/F), and DDCAR (V _c/F, CL/F). Solid lines represent locally estimated scatterplot smoothing (LOESS); circles represent individual data points. DCAR, desmethyl cariprazine; DDCAR, didesmethyl cariprazine. IIV, inter‐individual variation; CL/F, apparent oral clearance; V _c/F, apparent volume of distribution of the central compartment; k _a, absorption constant.

DCAR

DCAR concentrations in patients with SCZ or BPM from two phase 1 pediatric studies were adequately described by the final DCAR popPK model. All re‐estimated PK parameters had a relative standard error ≤174% and shrinkages for DCAR CL/F and V_c/F were 2.90% and 39.3%, respectively (Table 4).

Table 4.

DCAR Pharmacokinetic Parameter Estimates and Inter‐Individual Variability in Pediatric Patients With Schizophrenia or Bipolar 1 Disorder.

DCAR PK parameter	Population estimate	%RSE	95% CI
CL/F, L/h (fixed)	77.3	‐	‐
Vc/F, L (fixed)	128	‐	‐
Q3/F, L/h (fixed)	78.5	‐	‐
Vp/F, L (fixed)	347	‐	‐
Vc/F proportional shift for 1st dose (fixed)	1.27	‐	‐
Vp/F proportional shift for 1st dose (fixed)	0.535	‐	‐
Power of CL/F for weight (in kg)	0.0697	121	(−0.0963, 0.236)
Power of Vc/F for weight (in kg)	0.284	174	(−0.686, 1.25)
CL/F proportional shift for race = Black (fixed)	0.249	‐	‐
CL/F proportional shift for race = Asian (fixed)	−0.0861	‐	‐
CL/F proportional shift for race = Japanese (fixed)	−0.145	‐	‐
CL/F proportional shift for sex = female (fixed)	−0.160	‐	‐
Proportional error full profile studies	0.0688	3.08	(0.0647, 0.0730)
Proportional error (fixed)	0.141	‐	‐

Inter‐individual variability	Population estimate	%CV	%shrinkage
On CL/F	0.105	33.2	2.90
On Vc/F	1.56	194	39.3

The population pharmacokinetic (popPK) model for DCAR, a metabolite of cariprazine, included cariprazine popPK model parameters as fixed values. DCAR, desmethyl cariprazine; DDCAR, didesmethyl cariprazine; CL/F, apparent elimination clearance; V_c/F, apparent central volume of distribution; Q₃/F, apparent first distribution clearance; V_p1/F, apparent first peripheral volume of distribution; DUR/F, duration of zero‐order absorption process; k _a, absorption rate constant; %CV, percent coefficient of variation; DKTR/F, rate constant delaying DDCAR formation; IIV, inter‐individual variation. %RSE was calculated as the standard error of the estimator divided by the absolute value of the mean of the estimator multiplied by 100. %CV was calculated as SQRT[exp(ω ²)−1]×100.

The goodness‐of‐fit plots showed a random distribution of observed versus individual or population model‐predicted DCAR concentrations along the line of unity (Figure S2). The distributions of CWRES showed no trends when plotted against time since last dose or population predicted DCAR concentrations. The condition number of the final model was 2.73, indicating that all parameters could be simultaneously estimated. Additional diagnostic plots evaluating the final model (QQ plots of CWRES, inter‐individual random effects), showed no model deficiencies (data not shown).

Prediction‐corrected VPCs for DCAR concentrations in the first 24 h after the last dose and over the full concentration‐time curve demonstrated that the model adequately captured the observed DCAR concentration‐time course and variability (Figure S6). These also indicated that the model effectively captured accumulation of data over time. Prediction‐corrected VPCs stratified by age demonstrated that the model accurately described observed data across age and body weight ranges (data not shown). The effects of age and body weight on CL/F and V_c/F of DCAR were adequately described by the model (Figure 1 ).

DDCAR

DDCAR concentrations in pediatric patients with SCZ or BPM were adequately described by the final pediatric popPK model. All re‐estimated PK parameters had a relative standard error ≤54.7% and shrinkages for DDCAR CL/F and V_c/F were 2.09% and 6.07%, respectively. PK parameter estimates of the final model are provided in Table 5.

Table 5.

DDCAR Pharmacokinetic Parameter Estimates and Variability in Pediatric Patients With Schizophrenia or Bipolar 1 Disorder.

DDCAR pharmacokinetic parameter	Population estimate	%RSE	95% CI
CL/F, L/h (fixed)	9.24	‐	‐
Vc/F, L (fixed)	1310	‐	‐
Q3/F, L/h (fixed)	0.386	‐	‐
Vp/F, L (fixed)	258	‐	‐
DDKTR/F, 1/h (fixed)	0.0269	‐	‐
Power of CL/F for weight (in kg)	0.511	20.1	(0.310, 0.712)
CL/F proportional shift for race = Black (fixed)	0.547	‐	‐
CL/F proportional shift for race = Asian (fixed)	−0.194	‐	‐
CL/F proportional shift for race = Japanese (fixed)	−0.156	‐	‐
Power of Vc/F for weight (in kg)	0.244	54.7	(−0.0174, 0.506)
Power of Vc/F for age (fixed)	0	‐	‐
Vc/F proportional shift for race = Black (fixed)	0.676
Vc/F proportional shift for race = Asian (fixed)	−0.240
Vc/F proportional shift for race = Japanese (fixed)	0.0888
Proportional error full profile studies	0.0470	2.48	(0.0447, 0.0492)
Proportional error DCAR (fixed)	0.0875	‐	‐

Inter‐individual variability	Population estimate	%CV	%shrinkage
On CL/F	0.167	42.6	2.09
On Vc/F	0.295	58.6	6.07

The population pharmacokinetic (popPK) model for DDCAR, a metabolite of DCAR, included cariprazine and DCAR popPK model parameters as fixed values. DCAR, desmethyl cariprazine; DDCAR, didesmethyl cariprazine; CL/F, apparent elimination clearance; V_c/F, apparent central volume of distribution; Q₃/F, apparent first distribution clearance; V_p1/F, apparent first peripheral volume of distribution; DUR/F, duration of zero‐order absorption process; k _a, absorption rate constant; %RSE, percent residual error; %CV, percent coefficient of variation; DDKTR/F, rate constant delaying DDCAR formation; IIV, inter‐individual variability. %RSE calculated as the standard error of the estimator divided by the absolute value of the mean of the estimator multiplied by 100. %CV was calculated as SQRT[exp(ω ²)−1]×100.

Goodness‐of‐fit plots showed a random distribution of observed versus individual or population model‐predicted DDCAR concentrations along the line of unity (Figure S2). The distributions of CWRES showed no trends when plotted against time since last dose or population predicted DDCAR concentrations. The condition number of the final model was 14.2, indicating that all parameters could be simultaneously estimated. Further, diagnostic plots evaluating the final model (QQ plots of CWRES, interindividual random effects), showed no model deficiencies (data not shown).

Prediction‐corrected VPCs for DDCAR concentrations in the first 24 h after the last dose and over the full concentration‐time curve demonstrated that the model adequately captured the observed DDCAR concentration‐time course and variability of observed data (Figure S7). The model accurately described data across different age and weight ranges and effectively captured accumulation of data over time (data not shown). The effects of age and body weight on CL/F and V_c/F of DDCAR were adequately described by the model (Figure 1 ).

Total Cariprazine popPK Model Simulations and Pediatric Dosing Recommendations

Simulations using a fixed daily dose over 42 days were performed to predict steady‐state total CAR exposures (AUC_tau, C _trough and C _max) with daily cariprazine doses of 1.5, 3, 4.5, and 6 mg and are summarized in Figure 2 and Table S1 for each age group (adults, adolescents, children weighing <40 kg, children weighing ≥40 kg). In patients with SCZ, cariprazine administration of 1.5‐4.5 mg once daily (QD) in adolescents was predicted to result in a steady‐state total CAR AUC_tau, C _max, and C _trough within the range of adults administered 1.5‐6.0 mg QD cariprazine (efficacious dosing range; Figure 2). For BPM, cariprazine administration of 3.0‐4.5 mg QD in children and adolescents was predicted to result in a steady‐state total CAR AUC_tau, C _max, and C _trough within the range of adults administered 3.0‐6.0 mg QD (efficacious dosing range; Figure 2). Of note, steady‐state simulations yielded similar exposures across the three pediatric groups for each dose examined.

Figure 2.

Predicted total CAR AUC_tau, C _trough, and C _max at steady‐state in virtual adult (18+ years), adolescents (13‐17 years), and children (10‐12 years) following 42 days of fixed dosing. AUC_tau, area under the concentration‐time curve over one dosing cycle (1 day); total CAR, total cariprazine (cariprazine+desmethyl cariprazine+didesmethyl cariprazine); C _max, maximum concentration; C _trough, trough concentration; SS, steady‐state. Boxes show the median, interquartile range (IQR; 25‐75%) and whiskers represent 1.5 times the IQR.

Simulations of total CAR were also conducted during cariprazine titration using the pediatric and legacy adult ¹⁵  popPK models in virtual child, adolescent, and adult groups (Figure 3). During the initial titration phase, a single dose titration scheme (0.5 mg on Days 1 and 2, 1.5 mg on Days 3 and 4, 3.0 mg on Days 5 to 7, 4.5 mg on Day 8+) was applied for pediatric (10‐17 years old) simulations. Predicted total CAR concentrations for children and adolescents for these simulations allowed for a more gradual increase in total CAR exposure, remaining below that of adults administered the recommended titration scheme (1.5 mg on Day 1, 3.0 mg on Days 2, 4.5 mg Days 5‐7, and 6.0 mg on Day 8 and beyond; Figure 3 inset). Therefore, the examined pediatric dose escalation scheme is predicted to be safer for pediatric patients.

Figure 3.

Simulated total cariprazine (cariprazine+DCAR+DDCAR) concentration in adults (18+ years), adolescents (13‐17 years), and children (10‐12 years) following administration of various cariprazine dosing regimens with titration (Days 1‐8). Simulation duration was 42 days. Solid lines represent median simulated plasma concentrations, shaded regions represent 5‐95% of predicted concentrations, and numbers shown in titration period plots indicate daily cariprazine dose in mg (inset).

Discussion

The US FDA issued a general advice letter to sponsors (January 13, 2020) highlighting regulatory acceptance of adult to pediatric efficacy extrapolation of certain atypical antipsychotic agents for the treatment of SCZ (13‐17 years of age) and BPM (10‐17 years of age). Cariprazine has demonstrated long‐term safety in patients 10‐17 years of age (N = 303; NCT04578756). ²⁰ The current analysis extrapolated the efficacy in adult SCZ and BPM patients and supported approved pediatric cariprazine dosing (SCZ [13‐17 years]: 1.5‐4.5 mg daily) and BPM [10‐17 years]: 3 or 4.5 mg daily). ⁸ A similar efficacy extrapolation has been previously used to support brexpiprazole approval for the treatment of SCZ in adolescent patients. ²¹

The pediatric popPK model for cariprazine, DCAR, and DDCAR presented here was developed by updating an established adult popPK model ¹⁵ using observed concentration data from two phase 1 pediatric clinical trials. A previous preliminary pediatric PK model used to inform the dose selection for long‐term safety studies assumed a standard allometric scaling relationship between body weight and cariprazine, DCAR, and DDCAR PK parameters (CL/F and V_c/F; exponents of 0.75 for CL and 1 for V_c). That model suggested different dosing for children <40 kg and ≥40 kg (unpublished data). The current analysis updated the pediatric popPK model by re‐estimating body weight effect on both CL/F and V_c/F. For cariprazine, allowing the body weight exponent to be estimated led to optimized prediction of observed concentration data without biases in IIV trends over body weight and age in V_c/F or CL/F. As a result, steady‐state simulations at the same cariprazine dose yielded similar exposures between children 10‐12 years old who were <40 kg and ≥40 kg in weight. Therefore, dose adjustment based on body weight (<40 kg or ≥40 kg) for 10‐12 year old children is not supported by the current analysis. Similarly, estimating the weight effect on CL/F and V_c/F was also the preferred approach compared to assuming a fixed allometric relationship for describing brexpiprazole PK in adolescents with SCZ. ²² Overall, while standard allometric scaling can be used as an initial assumption in the absence of pediatric data, it does not necessarily reflect the variability in drug properties and elimination pathways across different drugs. ²³ Once pediatric data is available, the body weight effect on PK parameters should be determined based on the observed data.

The PK model‐based exposure‐matching described here for adult and pediatric patients with SCZ or BPM, along with the demonstrated long‐term safety of cariprazine use in pediatric patients, ²⁰ led to the regulatory approval of cariprazine for treating pediatric SCZ (13‐17 years) and BPM (10‐17 years). ⁸ Of note, a lower starting dose in children and adolescents (0.5 mg versus 1.5 mg in adults) was selected with a slower titration aimed to optimize tolerability in pediatric patients. This is in line with the common practice of starting at a lower dose as an added safety measure in pediatric patients initiating an antipsychotic agent.

These modeling analyses had several limitations. First, IIV on k _a over age and body weight in cariprazine was consistently above zero. However, this is not expected to impact exposure extrapolation between adults and pediatric patients because the effect of the bias is limited to C _max and has similar effect across age groups. Additionally, IIV on V_c/F was relatively large for cariprazine and DCAR, which was likely reflective of variability in dose administration times across studies during outpatient periods. Overall modeling and extrapolation results were likely not meaningfully affected as cariprazine and DCAR only account for a combined ∼36% of total cariprazine exposures at steady state. ¹⁵ In confirmation, model diagnostics indicated that clinical data were well‐captured for total cariprazine. Second, data from pediatric patients weighing <40 kg was sparse (N = 9 across both studies). Though the developed pediatric popPK model adequately predicted all active moiety exposures in this weight group, more granular weight analyses were not possible. Third, DCAR has a much shorter half‐life than cariprazine and DDCAR. As a result, fewer PK data points for DCAR were available, particularly in the lower pediatric weight range (DCAR is often undetectable after 336 h post dose). This may have led to the relatively high RSE that was observed on the DCAR weight effect estimate. However, because DCAR only accounts for <10% of total CAR exposure at steady state, ¹⁵ variability in DCAR concentrations would only minimally impact total CAR exposure across weight groups.

Conclusions

In conclusion, we employed a PK model‐based approach to extrapolate cariprazine efficacy from adult to pediatric patients through exposure‐matching. This analysis supported the approval of cariprazine for pediatric SCZ and BPM in the US without the need for pivotal efficacy trials in pediatric patients. Pediatric dosing recommendations were based on comparability to adult PK exposures. Doses of 1.5‐4.5 mg QD in adolescents (13‐17 years old) were predicted to yield exposures within the efficacious range of adults with SCZ administered 1.5‐6.0 mg QD. Doses of 3.0 or 4.5 mg QD in children (10‐12 years old) and adolescents were predicted to yield exposures within the efficacious range of adults with BPM administered 3.0‐6.0 mg QD. These dose recommendations were further supported by safety and tolerability evidence from a long‐term safety study in pediatric patients.

Author Contributions

Wrote manuscript: all authors. Designed research: Shams Ismaeil, Lucia Siovitz, Sven Stodtmann, and Hao Xiong. Performed research: Lucia Siovitz, Rahul K. Goyal, Doerthe Eckert, and Sven Stodtmann. Analyzed data: all authors; Interpreted data: all authors. All authors had full access to the data and gave final approval before submission. All authors agree to be accountable for the work.

Conflicts of Interest

All authors are employees of AbbVie Inc. and may hold stock and/or stock options in the company.

Funding Information

AbbVie funded this work and the clinical studies and participated in study design, research, analysis, data collection, interpretation of data, reviewing, and approval of the publication. All authors had access to relevant data and participated in the drafting, review, and approval of this publication. No honoraria or payments were made for authorship.

Supporting information

Supporting Information

Data Availability Statement

AbbVie is committed to responsible data sharing regarding the clinical trials we sponsor. This includes access to anonymized, individual, and trial‐level data (analysis data sets), as well as other information (e.g. protocols, clinical study reports, or analysis plans), as long as the trials are not part of an ongoing or planned regulatory submission. This includes requests for clinical trial data for unlicensed products and indications.

These clinical trial data can be requested by any qualified researchers who engage in rigorous, independent, scientific research, and will be provided following review and approval of a research proposal, Statistical Analysis Plan (SAP), and execution of a Data Sharing Agreement (DSA). Data requests can be submitted at any time after approval in the US and Europe and after acceptance of this manuscript for publication. The data will be accessible for 12 months, with possible extensions considered. For more information on the process or to submit a request, visit the following link: https://vivli.org/ourmember/abbvie/ then select “Home”.