Population pharmacokinetics and pharmacodynamics of canagliflozin in paediatric patients with type 2 diabetes mellitus

Abstract

Aims

Canagliflozin was recently approved in type 2 diabetes mellitus (T2DM) patients ≥10 and <18 years old as a single agent dosed once daily (QD) or in a fixed‐dose combination with metformin dosed twice daily (BID). We characterize canagliflozin pharmacokinetics in paediatric patients with T2DM, compare estimated individual and simulated exposure metrics for QD and BID doses in paediatric patients and adults, and investigate the relationship between canagliflozin plasma concentrations and change from baseline HbA1c with QD and BID doses.

Methods

Paediatric population pharmacokinetics and pharmacokinetics/pharmacodynamics models were based on the previously developed corresponding adult models, using Phase 1 and 3 study data, evaluating canagliflozin 100‐ or 300‐mg QD doses in paediatric patients with T2DM. Modelling encompassed exploratory analysis, external evaluation, model update and simulation of canagliflozin exposure metrics and HbA1c for QD and BID doses.

Results

Estimated and simulated steady‐state area under the concentration–time curve during 24 h (AUC_24h) and maximum drug concentration were comparable between paediatric patients and adults for canagliflozin 100‐ or 300‐mg QD. Simulated steady‐state AUC_24h was similar between QD and BID dosing for the same canagliflozin total daily dose in paediatric patients and adults with T2DM. QD to BID bridging was supported by simulated HbA1c and change from baseline in HbA1c in paediatric patients.

Conclusions

No canagliflozin dose adjustments are required alone or in combination with metformin to ensure similar plasma exposures between paediatric and adult patients with T2DM, as well as similar HbA1c lowering between QD and BID dosing in paediatric patients.

What is already known about this subject

• Canagliflozin significantly improved glycaemic control in paediatric patients with T2DM compared with placebo.

• Canagliflozin was well tolerated with a similar safety profile in paediatric patients and adults.

What this study adds

• Population pharmacokinetics and pharmacodynamics of canagliflozin can be extrapolated from adults to paediatric patients with T2DM.

• No dose adjustments are required to ensure similar plasma exposures between paediatric patients and adults.

• Daily and twice daily dosing (at the same total daily dose) of canagliflozin provides similar HbA1c lowering in paediatric patients.

1. INTRODUCTION

Type 2 diabetes mellitus (T2DM) is a chronic, progressive metabolic disorder that impacts patients across a broad age range. ¹ , ² The incidence of T2DM is increasing in children and adolescents and disproportionately impacts ethnic and racial minorities. ¹ , ² , ³ T2DM in paediatric patients is characterized by early onset and a higher prevalence of comorbidities, such as hypertension, dyslipidaemia and microalbuminuria at initial diagnosis, more rapid decline in β‐cell function and higher cumulative disease burden compared with adults. ⁴ Inadequate management of T2DM represents an extensive healthcare issue, with substantial economic burden, ⁵ and paediatric patients have an increased risk of diabetes‐related complications.

Canagliflozin is a sodium‐glucose co‐transporter 2 (SGLT2) receptor inhibitor, which lowers renal threshold for glucose excretion, thereby increasing urinary glucose excretion, approved worldwide as a single agent or in a fixed‐dose combination (FDC) with metformin for adults with T2DM as an adjunct to diet and exercise to improve glycaemic control, adding to metformin, insulin, glucagon‐like peptide‐1 agonists and other SGLT2 inhibitors in the T2DM armamentarium. ⁶ , ⁷ , ⁸ , ⁹ , ¹⁰ , ¹¹ Canagliflozin is initiated at 100 mg once daily (QD), which can be increased to 300‐mg QD in patients who require additional glycaemic control and have an estimated glomerular filtration rate (eGFR) of ≥60 mL/min/1.73 m². ⁶ In the FDC with metformin, canagliflozin is dosed as 50 mg twice daily (BID) or 150 mg BID, with similar titrations as for the single agent. The Phase 1 DIA1055 study (NCT02000700) assessed the pharmacokinetics (PK) and pharmacodynamics (PD) of canagliflozin in paediatric patients (≥10 and <18 years of age) with T2DM and showed that canagliflozin 100‐ and 300‐mg QD for 14 days had similar PK and PD characteristics to those observed in adult patients. ¹²

The Phase 3 DIA3018 study (NCT03170518) investigated the effectiveness of canagliflozin 100‐mg QD, with conditional uptitration to 300‐mg QD at Week 13 in patients with inadequate glycaemic control (glycated haemoglobin [HbA1c] ≥ 7%) and eGFR ≥60 mL/min/1.73 m², vs. placebo after 26 and 52 weeks of treatment (end of treatment), as well as overall safety and tolerability, in patients ≥10 to <18 years old. ¹³ HbA1c (expressed as a percentage) is a measure of the 3‐month average blood sugar level and therefore represents efficacy over this time period. The results from the Phase 3 study confirmed that canagliflozin significantly improved glycaemic control (change from baseline in HbA1c) in paediatric patients compared with placebo at Week 26 (least square mean difference [95% confidence interval] –0.76% [−1.25 to −0.27]; p = 0.002), with improvements in glycaemic control observed as early as at Week 6 and sustained until the end of the trial. ¹³ Canagliflozin was well tolerated and generally safe in paediatric patients, with a similar safety profile as observed in adults. ¹³ Based on these results, canagliflozin was approved by the United States Food and Drug Administration and the European Medicines Agency as an adjunct to diet and exercise to improve glycaemic control in paediatric patients aged ≥10 years with T2DM. ⁶ , ⁸

The population pharmacokinetics (popPK) of canagliflozin and PK relationship with HbA1c lowering has been described in healthy adult volunteers and adult patients with T2DM. ¹⁴ , ¹⁵ Although disease progression has been reported to be more rapid in paediatric patients than in adults, the basic pathophysiology of T2DM in paediatric patients is expected to be generally similar to adults. As paediatric and adult patients with T2DM are expected to share similar underlying physiology that influences canagliflozin clearance, particularly in paediatric patients aged ≥10 years, we hypothesised that extrapolation of response to canagliflozin from adult to paediatric patients would be appropriate. Here, we characterize the canagliflozin popPK and PK/PD effect on HbA1c lowering in paediatric patients with T2DM based on data from the Phase 1 and 3 studies and using the adult models as starting models. To support the FDC of canagliflozin with metformin in paediatric patients, we derive individual and simulated canagliflozin exposure metrics and simulate the change from baseline in HbA1c for QD and BID canagliflozin dosing.

2. METHODS

2.1. Study design and assessments

The study designs of the Phase 1 and 3 clinical studies have been described previously. ¹² , ¹³ Briefly, the Phase 1 study had an open‐label, sequential, multiple‐dose, multicentre design and evaluated the PK, PD and safety of canagliflozin 100‐ or 300‐mg QD for 14 days in children and adolescents aged ≥10 and <18 diagnosed with T2DM, defined as HbA1c of ≥6.1% to ≤10%, who were on a stable regimen of metformin (at a dose ≥100‐mg QD) for at least 8 weeks before screening. The Phase 3 study had a randomized, double‐blind, placebo‐controlled, parallel‐group, multicentre design and included patients with T2DM aged ≥10 and <18 years of age who had inadequate glycaemic control, defined as a HbA1c of ≥6.5% to ≤11.0%. Patients were randomized in a 1:1 ratio to canagliflozin 100‐mg QD or placebo. Patients who had an HbA1c of ≥7.0% and an eGFR of ≥60 mL/min/1.73 m² at Week 12 were re‐randomized at Week 13 in a 1:1 ratio to remain on canagliflozin 100‐mg QD (or matching placebo) or uptitrated to 300‐mg QD (or matching placebo). The primary endpoint was change in HbA1c from baseline to Week 26, and the total study duration was 52 weeks. The popPK model was developed on canagliflozin plasma concentrations in patients from both paediatric studies, while the PK/PD model was developed on HbA1c in patients from the Phase 3 study because HbA1c was not an endpoint in the Phase 1 study.

In the paediatric studies, venous blood samples were taken to determine canagliflozin plasma concentration. In the Phase 1 study on Day 14 (end of treatment), intensive PK samples were collected pre‐dose, and at 0.5, 1, 2, 4, 8, 12, 24, 48 and 72 h after dosing. In the Phase 3 study, PK samples to determine plasma trough concentrations were collected at Weeks 12, 26 and 52 (end of treatment) or at early withdrawal. In both studies, plasma samples were analysed to determine concentrations of canagliflozin using a validated, specific and sensitive liquid chromatography–tandem mass spectrometry method. The lower limit of quantification was 5 ng/mL. Blood samples for HbA1c analyses were collected at Week −3 (screening visit), Day 1 (baseline) and Weeks 6, 12, 20, 26, 34, 42 and 52 (end of treatment) or at early withdrawal, and at rescue visits. Full data handling is described in the Supporting information.

Both studies were conducted in accordance with the ethical principles that have their origin in the Declaration of Helsinki and that are consistent with Good Clinical Practices (GCP) and applicable regulatory requirements. The study protocols and amendments were reviewed by independent ethics committees or institutional review boards at each participating site, as reported in the primary publications. Ethics approval was obtained at all participating sites. Participants or their legally acceptable representatives provided written informed consent after being informed about the nature and purpose of the study, participation/termination conditions and risks and benefits of treatment. Known instances of nonconformance were documented and are not considered to have had an impact on the overall conclusions of this study.

2.2. popPK and PK/PD models

The paediatric popPK model was developed based on the previously developed adult popPK model. ¹⁴ Given that the underlying physiology that influences canagliflozin clearance is matured in children and adolescents ≥10 years of age, the previously developed adult popPK model was considered as a suitable starting model for paediatric patients in this age group. The adult popPK model consisted of an open two‐compartment disposition model with first‐order elimination and a sequential zero‐order input into the depot compartment followed by first‐order absorption from depot to central compartment, after a lag time (T _lag). This model was parametrized in terms of rate constants and distribution volumes. The adult popPK model includes a body weight effect on the apparent volume of distribution of the central compartment (V _c /F) with an estimated exponent of 0.583, and implies the same body weight effect on the apparent clearance (CL/F), because there is no body weight effect included on the elimination rate constant (k _e ) and CL/F=V _c /F*k _e . Body weight ranges overlapped between studies of paediatric patients (Phase 1: mean 107 kg [standard deviation 33] ¹² ; Phase 3: mean 81 kg [standard deviation 25] ¹³ ) and adults (mean 88 kg [standard deviation 19.6 kg] ¹⁴ ); therefore, no further body weight effects were included in the starting model for paediatric PK. A popPK model with fixed allometric exponents of 1.0 for V _c /F and −0.25 for k _e was considered in a sensitivity analysis. ¹⁶ Inter‐individual variability (IIV) was quantified on the adult popPK model parameters for V _c /F, k _e , distribution rate constant from the peripheral to the central compartment (k ₃₂ ), absorption rate constant (k _a ) and T _lag , assuming a lognormal distribution. An additive error model was used to quantify the residual unexplained variability on the log‐transformed plasma concentration scale, assuming different variances for rich and sparse PK data.

The adult PK/PD model was previously developed to characterize the exposure–response relationship of canagliflozin on HbA1c lowering in adult T2DM patients. ¹⁵ Briefly, the adult PK/PD model consisted of a turnover model for HbA1c with a placebo‐corrected effect of canagliflozin on HbA1c formation using an E _max model. IIV was quantified on the model parameters for baseline HbA1c and HbA1c‐lowering effect of placebo (Ef _p ), assuming a lognormal and normal distribution, respectively. Residual unexplained variability (RUV) was assumed to be additive on the log‐transformed HbA1c scale. Further details about the adult popPK and PK/PD models are provided in the Supporting information.

The modelling approach for the popPK and PK/PD analyses encompassed exploratory analysis, external evaluation, model update and simulation. The exploratory analysis for the popPK analysis assessed canagliflozin plasma concentration, baseline demographics and covariates from the adult popPK model included in the paediatric dataset, including a comparison with the adult dataset. The exploratory analysis for the PK/PD model was performed to assess the change from baseline in HbA1c concentrations, demographics and other relevant baseline characteristics included in the paediatric dataset.

For the external evaluation, prediction‐corrected visual predictive check (pcVPC), numerical predictive check, goodness‐of‐fit (GOF) plots and graphical analysis of empirical Bayes estimate (EBEs) were used to evaluate whether canagliflozin PK and HbA1c data were adequately described by the previously developed adult popPK and PK/PD models, respectively. If the external evaluation was unsuccessful, an update to the respective model was performed, considering updates to model structure, covariate model, IIV random effects model and/or parameter re‐estimation.

For the external evaluation of the adult PK/PD model, the final paediatric popPK model and corresponding individual popPK parameter estimates were used, while fixing the PD parameters to the estimates from the adult PK/PD model (sequential approach). For Ef _p , the estimate from the phase 3 DIA3006 study was used in the external evaluation as the study population and design were closest to the paediatric Phase 3 study analysed.

Using the individual parameter estimates obtained from the final paediatric popPK model, individual estimates of steady‐state area under the canagliflozin concentration–time curve during 24 h (AUC_24h) for paediatric patients with T2DM were calculated and compared with those obtained in adult patients with T2DM who received either 100‐ or 300‐mg QD of canagliflozin. Finally, steady‐state AUC_24h and maximum drug concentration (C_max) were simulated for canagliflozin administered as overencapsulated tablets at doses of 50‐ and 150‐mg BID (as in the FDC with metformin), and 100‐ and 300‐mg QD (as for the single agent) for 7 days in paediatric patients with T2DM and compared with those simulated in adult patients with T2DM based on the adult popPK model. PK profiles from 1000 female and 1000 male virtual T2DM patients were generated, accounting for IIV and RUV. For adults, covariate values of sex, age, body weight, body mass index (BMI) and eGFR were jointly sampled from the patients with T2DM in the studies that were used to develop the adult popPK model. For paediatric patients, the same covariates were considered if the external evaluation was successful or an updated set of covariates was considered in case the covariate model was updated. Sampled covariate values were kept fixed across all simulated dosing regimens.

Based on the final paediatric PK/PD model, HbA1c was simulated in virtual paediatric patients with T2DM for the same QD and BID canagliflozin dose regimens (extended to 26 weeks), sample sizes and sources of variability as in the popPK simulations described above. PK/PD model covariates were sampled jointly from the PK/PD analysis dataset, split by sex and kept the same for all simulated doses. PK as an input for the PK/PD relationship was simulated using the parameter estimates from the final paediatric popPK model, including IIV.

2.3. Statistical analysis and software

Nonlinear mixed‐effects modelling (NONMEM) software (v7.4, ICON plc) and the first‐order condition estimation approximation with INTERACTION (FOCE‐I) estimation option was used for both analyses. The NONMEM analysis was performed in a validated environment, based on good automated manufacturing practice and in accordance with 21 CFR Part 11 and GCP regulations. Modifications to create the modelling data file, exploratory analysis, post‐processing of NONMEM analysis results and diagnostic graphics were carried out in R v3.6.2 or higher (Comprehensive R Network, http://cran.r-project.org/). Simulations for popPK were performed in NONMEM, while PK/PD simulations were performed in mrgsolve v1.0.4 in R v3.6.2.

2.4. Nomenclature of targets and ligands

Key protein targets and ligands in this article are hyperlinked to corresponding entries online (https://www.guidetopharmacology.org) and are permanently archived in the Concise Guide to Pharmacology 2023/24. ¹⁷ , ¹⁸ , ¹⁹

3. RESULTS

3.1. PopPK analysis

For the popPK analysis, plasma concentrations below the quantification limit (BQL) were excluded from the analysis; the percentage of PK samples below the lower limit of quantification was 15.1% (full data handling in the Supporting information). There were no participants with missing covariates. Overall, there were 330 PK samples from 90 paediatric patients available for the analysis. The baseline demographics and covariates of paediatric patients with T2DM included in the popPK analysis are shown in Table 1. Baseline demographics and covariates in the two paediatric studies were comparable, although body weight (median 116 and 75.1 kg, respectively) and BMI (median 41.4 and 29.9 kg/m², respectively) tended to be higher for patients in the Phase 1 study compared with the Phase 3 study. Compared with the adults included in the previously developed popPK model, body weight and BMI were comparable, but eGFR was higher in the paediatric dataset (median 160 mL/min/1.73 m² [assessed using the Schwartz formula]) than in the adult dataset (87.0 mL/min/1.73 m² [assessed using the Modification of Diet in Renal Disease [MDRD] formula). The ratio of female to male patients differed slightly between the paediatric and adult populations.

TABLE 1.

Baseline demographics and covariates of paediatric patients and adults included in the popPK analyses.

Parameter	Phase 1 paediatrics (n = 17)	Phase 3 paediatrics (n = 73)	Total paediatrics (N = 90)	Diabetic adults b (n = 1219)	Total adults b (N = 1616)
Sex, n (%)
Male	5 (29.4)	24 (32.9)	29 (32.2)	647 (53.1)	769 (47.6)
Female	12 (70.6)	49 (67.1)	61 (67.8)	572 (46.9)	847 (52.4)
Race, n (%)
White, not Hispanic or Latino	2 (11.8)	13 (17.8)	15 (16.7)	676 (55.5)	942 (58.3)
Black, of African heritage or African American	12 (70.6)	6 (8.2)	18 (20.0)	52 (4.3)	102 (6.3)
White, Hispanic or Latino	0	20 (27.4)	20 (22.2)	154 (12.6)	203 (12.6)
Asian	0	30 (41.1)	30 (33.3)	226 (18.5)	245 (15.2)
Native Hawaiian or other Pacific Islander	0	0	0	1 (0.1)	2 (0.1)
American Indian or Alaskan Native	0	4 (5.5)	4 (4.4)	2 (0.2)	2 (0.1)
Other	3 (17.6)	0	3 (3.3)	108 (8.9)	120 (7.4)
Median age (range), years	15.0 (11.0–17.0)	15.0 (10.0–17.0)	15.0 (10.0–17.0)	57.0 (22.0–90.0)	55.0 (18.0–90.0)
Median body weight (range), kg	116 (48.2–170)	75.1 (44.8–161)	79.5 (44.8–170)	85.0 (42.5–167)	86.0 (42.5–167)
Median BMI (range), kg/m2	41.4 (17.9–54.4)	29.9 (19.0–50.4)	30.7 (17.9–54.4)	30.8 (18.4–54.7)	31.1 (18.0–54.7)
Median eGFR (range), mL/min/1.73 m2	142 (70.3–222)	165 (104–277)	160 (70.3–277)	87.0 (25.0–194)	87.0 (9.00–194)
Total daily dose, n (%)
25 mg	—	—	—	12 (1.2)	12 (0.7)
50 mg	—	—	—	69 (5.7)	164 (10.1)
100 mg	8 (47.1)	61 (83.6)	69 (76.7)	506 (41.5)	610 (37.4)
200 mg	—	—	—	69 (5.7)	148 (9.1)
300 mg	9 (52.9)	12 (16.4) a	21 (23.3)	480 (39.4)	596 (36.6)
400 mg	—	—	—	13 (1.1)	13 (0.8)
600 mg	—	—	—	70 (5.7)	70 (4.3)
800 mg	—	—	—	0	6 (0.4)
1200 mg	—	—	—	0	6 (0.4)
1600 mg	—	—	—	0	5 (0.3)

Abbreviations: BMI, body mass index; eGFR, estimated glomerular filtration rate.

^a After 12 weeks of 100 mg and re‐randomization, participants were uptitrated to 300 mg.

^b Adult participants included in the previously developed popPK model for canagliflozin. ¹⁴

Canagliflozin plasma concentration–time data in paediatric patients were comparable with those observed in adults (Figure S1); however, based on the external evaluation, the previously developed popPK model in adults did not appear to adequately capture the general trend and variability of PK data in paediatric patients, and a model update was therefore conducted. This was done by removing covariate relationships included in the adult model one by one, while keeping the other parameters fixed to those estimated in the adult model. The covariate for which removal resulted in the largest decrease in objective function value (OFV) was removed from the model, and this process was repeated until the decrease in OFV was less than 10.83 (i.e., p > 0.001 for a χ ² test with 1 degree of freedom).

The base model structure of the final paediatric popPK model was the same as the previously developed adult popPK model. The covariate relationships sex and body weight on V _c /F, total daily dose on k _e and overencapsulation on T _lag were retained, while age on V _c /F, eGFR on k _e and BMI on k _a and T _lag were removed in the final paediatric popPK model. All remaining parameters were fixed to the estimates from the adult model except for the RUV for rich and sparse PK data, which were re‐estimated based on the data from the paediatric studies (Table 2). Adequate precision was obtained for the estimated RUV as measured by a relative standard error below 15%. The shrinkage of the IIV random effects on V _c /F, k _e , k ₃₂, k _a and T _lag was high (>39%), which can be attributed to sparse sampling in the Phase 3 study, with mainly trough samples. The model appeared to adequately capture the central tendency and the variability of the paediatric data, as shown by the agreement between the observed and simulated 10th, 50th and 90th percentiles for the pcVPC (Figure 1). GOF plots confirmed the pcVPC results and are shown in Figure S2.

TABLE 2.

Parameter estimates of the final paediatric popPK model.

Parameter	Population estimate	RSE (%)	IIV (CV %)	Shrinkage (%)
V c /F (L)
Males	99.3 FIX		14.6 FIX	70.1
Females	82.6 FIX
Constants (h −1)
k e	0.145 FIX		23.1 FIX	39.1
k a	3.68 FIX		188 FIX	51.6
k 23	0.101 FIX
k 32	0.0856 FIX		36.2 FIX	65.1
T lag (h) a
Nonencapsulated tablets	0.147 FIX		93.6 FIX	65.7
Overencapsulated tablets	0.262 FIX
D 1 (h)	0.604 FIX
Effects of covariates
Body weight on V c /F b	0.583 FIX
Total daily dose on k e c	−0.0631 FIX
Residual error (SD)
Phase 1 paediatric	0.227	13.4
Phase 2 and 3 paediatric	1.14	9.12

Note: FIX = model parameters, including covariate and IIV random effects, were fixed to the estimates obtained from the adult popPK model. ¹⁴ CV% was calculated as $\left(\sqrt{\exp \left({\omega }^2\right)-1}\right)\times 100$.

Abbreviations: ${\omega }^2$, IIV random effect variance; CV, coefficient of variation; D1, duration of zero‐order input into gut compartment; IIV, inter‐individual variability; k ₂₃ , distribution rate constant from the central to the peripheral compartment; k ₃₂ , distribution rate constant from the peripheral to the central compartment; k _a , absorption rate constant; k _e elimination rate constant; RSE, relative standard error; SD, standard deviation; T _lag , lag time; Vc/F, apparent volume of distribution of the central compartment.

^a As only overencapsulated tablets were tested in paediatric patients, the effect of overencapsulation on T _lag was retained.

^b The body weight covariate on V _c /F was included as: ${\mathit{TV}}_{\frac{\mathit{Vc}}{F}}{\left(\mathit{WT}/75.8\right)}^{{\theta }_{\mathit{WT}}}$, where ${\mathit{TV}}_{\frac{\mathit{Vc}}{F}}$ is the typical value of V _c /F, WT is body weight and ${\theta }_{\mathit{WT}}$ is the scaling exponent.

^c The total daily dose covariate on k _e was included as: ${\mathit{TV}}_{k_e}{\left(\mathit{TDD}/100\right)}^{{\theta }_{\mathit{TDD}}}$, where ${\mathit{TV}}_{k_e}$ is the typical value of k _e , TDD is total daily dose and ${\theta }_{\mathit{TDD}}$ is the scaling exponent.

FIGURE 1.

pcVPC of the final paediatric popPK model for the Phase 1 (top) and Phase 3 (bottom) studies by dose. Prediction‐corrected concentrations (red dots) with median (solid lines) and 10th, 90th percentiles (dashed lines). Areas show the simulation‐based 95% pcVPC intervals in grey for the median and in blue for the 10th and 90th percentiles (N = 1000). pcVPC, prediction‐corrected visual predictive check.

The final paediatric popPK model did not reveal any trend regarding body weight: no relationship between the IIV random effect on k _e and body weight could be observed (Figure S3) and no trend could be revealed in the pcVPC and GOF plots stratified by body weight (Figures S4 and S5). Note that only six patients with body weight <95 kg were included in the Phase 1 study. Nevertheless, as a sensitivity analysis, a popPK model with fixed allometric exponents of 1.0 for V _c /F and −0.25 for k _e was fitted to the paediatric PK data. ¹⁶ Although both models entailed similar individual predictions, the final paediatric popPK model outperformed the model with fixed allometric exponents, which had larger OFV (increase of 20.7), an underprediction of C_max for 100 mg in the Phase 1 study (Figure S6) and a more left‐skewed distribution of individual EBEs for Vc/F. Therefore, in the remainder of this analysis, the final paediatric popPK model is considered.

The estimated individual AUC_24h for canagliflozin at 100‐ and 300‐mg QD at steady state in paediatrics were similar to adults with T2DM (Table 3). The simulated concentration–time profiles at steady state following 50‐mg BID, 100‐mg QD, 150‐mg BID and 300‐mg QD canagliflozin administration demonstrated that within the same regimen, there was good correspondence in exposure between paediatrics and adults with T2DM (Figure S7). Model‐based simulations show that AUC_24h at steady state, reflecting the total daily exposure, was similar between QD and BID dosing for the same canagliflozin total daily dose in both paediatrics and adults (Figure 2A, Table S2). C_max at steady state was higher following QD compared with BID dosing for the same total daily dose (Figure 2B, Table S2). For the same dose regimen, there was good overlap in simulated canagliflozin AUC_24h and C_max at steady state between paediatrics and adults (Figure 2A,B, Table S2). Finally, the exposure in females was slightly higher compared with males, although there was substantial overlap between these groups.

TABLE 3.

Summary statistics for the estimated individual AUC_24h (h·ng/mL) at steady state after oral administration of canagliflozin at 100‐ and 300‐mg QD, in paediatric and adult T2DM patients, stratified by sex.

Dose (mg) QD	Sex	N a	Mean (SD)	Median	10th percentile	90th percentile
Paediatric
100	Male	24	6574 (1774)	6353	4858	9072
	Female	54	8397 (2155)	8154	6143	10 355
300	Male	8	22 480 (7858)	19 013	15 501	32 336
	Female	10	29 930 (11 839)	25 502	21 681	50 845
Adult
100	Male	269	7183 (2146)	6684	4933	9975
	Female	237	8984 (2770)	8522	6048	12 425
300	Male	254	22 954 (6932)	21 374	15 768	32 555
	Female	226	28 391 (8111)	27 408	19 029	39 071

Abbreviations: AUC_24h, area under the concentration–time curve during 24 h; PK, pharmacokinetic; QD, once daily; SD, standard deviation; T2DM, type 2 diabetes mellitus.

^a All paediatric T2DM patients who contributed PK data on the specific dose regimen were considered, for example, patients in the Phase 3 300‐mg group were included for 100‐, 300‐mg QD or both doses depending on the PK samples available.

FIGURE 2.

Simulated (A) AUC_24h at steady state (B) and simulated C_max at steady state, stratified by sex, after administration of canagliflozin 50‐mg BID, 100‐mg QD, 150‐mg BID and 300‐mg QD in paediatric and adult patients with T2DM. AUC_24h, area under the concentration–time curve during 24 h; BID, twice daily; C_max maximum drug concentration; QD, once daily; T2DM, type 2 diabetes mellitus.

3.2. PK/PD analysis

For the PK/PD analysis, the original dataset was merged with the individual popPK parameter estimates from the final paediatric popPK model to generate a combined dataset. Overall, there were 702 HbA1c samples from 160 paediatric patients available for the analysis. Baseline demographics and covariates of paediatric patients with T2DM were reasonably well balanced across the canagliflozin and placebo groups for the PK/PD analysis (Table S1). In total, 46.2% of patients were on metformin, 28.7% on metformin and insulin, 13.8% on diet and exercise only and 11.2% on insulin. Baseline HbA1c tended to be lower in the canagliflozin group (median, 7.50%) compared with the placebo group (median, 8.00%).

Paediatric patients treated with canagliflozin had a more pronounced decline in HbA1c after the start of dosing compared with placebo (Figure S8). The external evaluation of the previously developed adult PK/PD model on HbA1c data indicated that a model update was required to adequately describe the HbA1c data for both treatment groups at a population level. This was done by re‐estimating model parameters one by one, while keeping the other parameters fixed to those estimated in the adult model. The model for which re‐estimation resulted in the largest decrease in OFV was selected and this process was repeated until the decrease in OFV was less than 10.83 (i.e., p > 0.001 for a χ ² test with 1 degree of freedom). Based on the graphical exploration, inclusion of potential covariates was tested in a stepwise manner.

The final paediatric PK/PD model had the same base model structure as the previously developed adult PK/PD model but included statistically significant covariate effects of background insulin and treatment group on baseline HbA1c, and of background insulin on Ef _p . In addition, the RUV was re‐estimated based on the paediatric HbA1c data. No differences were observed between paediatric and adults in terms of HbA1c half‐life, canagliflozin E _max , canagliflozin EC ₅₀ , that is, plasma concentration at which half‐maximal HbA1c‐lowering effect is reached, and IIV on baseline HbA1c and Ef _p (Table 4). The parameter estimates had acceptable standard errors and did not have correlations exceeding 0.90 in absolute value. The shrinkage of the individual random effects was below 21%, which was considered acceptable. The pcVPC results of the final PK/PD model for HbA1c and change from baseline HbA1c shows that the model adequately captured the central tendency and variability of the data, as reflected by the overall agreement between the 10th, 50th and 90th percentiles of the prediction‐corrected data and respective 95% pcVPC intervals obtained from the simulations (Figure 3). GOF plots confirmed the pcVPC results and are shown in Figure S9.

TABLE 4.

Parameter estimates of the final paediatric pharmacokinetic/pharmacodynamic model for HbA1c.

Parameter	Population estimate	SE	Shrinkage d (%)
Baseline HbA1c (%)
Insulin placebo	8.50	0.213
Insulin canagliflozin	7.82	0.270
No insulin placebo	7.91	0.190
No insulin canagliflozin	7.22	0.154
HbA1c half‐life t 1/2 (days) a	28.2 FIX c
Ef p insulin (%)	0.171	0.148
Ef p no insulin (%)	−0.238	0.107
Emax (%)	−0.738 FIX c
Log EC 50 (log ng/mL) b	4.12 FIX c
IIV of baseline HbA1c (variance)	0.011 FIX c		1.00E‐10
IIV of Ef p (variance)	0.369 FIX c		20.9
Residual error (SD)	0.0773	0.00522

Note: EC ₅₀ , canagliflozin plasma concentration at which half‐maximal HbA1c‐lowering effect is reached; E _fp , effect of placebo treatment on HbA1c at steady state for a typical participant (baseline HbA1c of 8.0%); IIV, inter‐individual variability; NONMEM, nonlinear mixed‐effects modelling; SD, standard deviation; SE, standard error.

^a HbA1c half‐life t _1/2  = log(2)/k _out .

^b Corresponds to an EC₅₀ estimate of 61.6 ng/mL.

^c FIX = model parameters were fixed to the estimates obtained from the adult PK/PD model. ¹⁵

^d Random effect shrinkage as reported in NONMEM.

FIGURE 3.

pcVPC of the final PK/PD model for HbA1c (A) and change from baseline HbA1c (B) in placebo and canagliflozin groups of the Phase 3 study. Lines are median (red solid line), 10th and 90th percentiles (red dashed lines) of prediction‐corrected concentrations (blue circles). Shaded intervals represent simulation‐based 95% pcVPC intervals in red for the median and in blue for the 10th and 90th percentiles (N = 1000). pcVPC, prediction‐corrected visual predictive check.

For total daily doses of canagliflozin 100 and 300 mg, there was a good overlap in simulated mean HbA1c (Figure 4A) and mean change from baseline HbA1c (Figure 4B) and their associated variability between QD and BID dosing regimens in paediatric patients with T2DM (Table S3), as was shown previously for adult patients. ¹⁵

FIGURE 4.

Simulated HbA1c (A) and simulated change from baseline HbA1c (B) after oral administration of canagliflozin at 50‐mg BID, 100‐mg QD, 150‐mg BID and 300‐mg QD in paediatric patients with T2DM. Dots represent mean (change from baseline) HbA1c at each time point, while error bars represent mean ± standard deviation. Figure shows total daily dose of 100 mg (left panel: 50‐mg BID, 100‐mg QD) and of 300 mg (right panel: 150‐mg BID, 300‐mg QD). BID, twice daily; QD, once daily.

4. DISCUSSION

The results of the current analyses support the extrapolation of canagliflozin popPK and PK/PD from adults to paediatric patients. It also supports that no canagliflozin dose adjustments are required as a single agent or as an FDC with metformin to ensure similar plasma exposures in paediatric patients aged ≥10 years compared with adult patients with T2DM. Finally, the results show that QD and BID canagliflozin dosing regimens provide similar HbA1c lowering in paediatric patients at the same total daily dose, supporting the use of canagliflozin‐metformin FDC in paediatric patients. ⁶

PopPK base model structure and model parameters were the same between paediatric and adults, except for the Phase 1 and Phase 2/3 RUV, which were re‐estimated based on the paediatric data. ¹⁴ Additionally, compared with the adult popPK model, covariate effects of age, eGFR and BMI were removed, while covariate effects of body weight, sex, overencapsulation and total daily dose were retained in the paediatric model. In the adult popPK analysis, these covariate effects were not deemed to be clinically relevant. ¹⁴ Although there was no indication that the relationship between body weight and popPK parameters in paediatric patients would be different from adults, it is noted in paediatric patients with body weights lower than those included in the clinical studies (i.e., <40 kg) that the extrapolation of results is made in the absence of observed data. Although a sensitivity analysis showed similar individual predictions between the paediatric popPK model and the popPK model with fixed allometric exponents, the latter model predicts higher exposures when extrapolating to low body weights. The paediatric patients included in the dataset were ≥10 to <18 years of age, and the age effect on Vc/F was estimated based on adult data (range, 18–90 years of age). This covariate relationship, which was mainly driven by elderly participants, did not appear to extrapolate to lower ages. Furthermore, paediatric patients had larger eGFR compared with the adults included in the adult popPK model, which may explain why the eGFR effect on k _e estimated on the adult data did not extrapolate to paediatric patients. It should be noted that eGFR for paediatric patients was estimated using the Schwartz formula, while the MDRD method was used for adults. ²⁰ , ²¹ Nevertheless, there was a high correlation between eGFR based on both methods for paediatric patients (data not shown). Although BMI was comparable between paediatric and adult participants included in the respective popPK model, an improved model fit could be obtained when removing the covariate effect of BMI on k _a and T _lag , which was mainly driven by a few patients from the Phase 1 study.

The estimated individual AUC_24h at steady state after oral administration of canagliflozin 100‐ and 300‐mg QD were similar in paediatric patients and adults. Additionally, simulated steady‐state AUC_24h was similar between QD and BID regimens at the same total daily dose of canagliflozin 100 or 300 mg in paediatric patients and steady‐state exposures were comparable between paediatric patients and adults. Consequently, no canagliflozin dose adjustments as a single agent or as an FDC with metformin are required to obtain similar plasma exposures in paediatric patients and adults.

In the paediatric PK/PD model, the basic model structure and parameters were the same between adult and paediatric patients, except for the placebo effect Ef _p , baseline HbA1c and RUV, which were re‐estimated. ¹⁵ In addition, compared with the adult PK/PD model, statistically significant covariate effects of background insulin and treatment group on baseline HbA1c and of background insulin on Ef _p were included in the paediatric PK/PD model. Paediatric patients with T2DM on insulin therapy (as monotherapy or combined with metformin) had a larger baseline HbA1c (typical value increase of ~0.6%) and a larger placebo‐induced change from baseline HbA1c compared with those without insulin therapy (typical values 0.171% vs. –0.238%). The estimate of −0.238% was consistent with the study‐specific estimates obtained for Ef _p in the adult PK/PD model (ranging from −0.483% to −0.137%), in which all patients were on metformin monotherapy without insulin. ¹⁵ No additional covariate effects were identified. Model‐based simulations evidenced that HbA1c and change from baseline HbA1c were similar between QD and BID regimens at the same total daily dose (100 or 300 mg) in paediatric patients. These results support the possibility to switch from canagliflozin QD and metformin BID as single agents to canagliflozin‐metformin as an FDC‐taken BID.

Limitations of these analyses include that high variability was observed in the sparse PK data from the Phase 3 study, which is expected given that sampling in late‐phase studies is often conducted less often than in early‐phase studies. This led to a re‐estimation of the RUV in the final paediatric popPK model. While the estimated Phase 1 residual error standard deviation only increased from 0.202 for adults to 0.227 for paediatric patients, the estimated Phase 2/3 residual error standard deviation increased more than two‐fold from 0.559 for adults to 1.14 for paediatric patients, which may be due to aberrations in dosing and/or dosing times vs. what was planned in the paediatric study. Further, inference regarding the canagliflozin exposure–response relationship for HbA1c lowering was limited as the PK/PD dataset included only 12 paediatric patients who were uptitrated to 300‐mg QD after 12 weeks of treatment with 100‐mg QD. Nevertheless, the final paediatric PK/PD model with drug effect parameters fixed to those from the adult PK/PD model adequately described the totality of the HbA1c data collected under different dose regimens in the Phase 3 study, and therefore, the data available did not provide evidence that the canagliflozin exposure–response relationship in paediatric patients was different than observed in adults. Finally, there were limited clinical data in children with a lower body weight (and absence in children <40 kg); therefore, caution is advised when up‐titrating canagliflozin to the 300‐mg dose in patients who weigh <50 kg.

Overall, the results support the same canagliflozin dosing (i.e., no dose adjustment) in paediatric patients compared with adult patients. Simulations showed that canagliflozin daily exposure (AUC_24h), and changes to HbA1c, were similar with canagliflozin QD and BID dosing at the same total daily dose.

AUTHOR CONRIBUTIONS

Iolanda Cirillo, Juan José Perez Ruixo, Nele Goeyvaerts and Oliver Ackaert contributed to the conceptualization of the study. Data were curated and formal analyses were conducted by Martine Neyens, Nele Goeyvaerts and Per Olsson Gisleskog. Methodology was developed by Iolanda Cirillo, Juan José Perez Ruixo, Nele Goeyvaerts, Oliver Ackaert and Per Olsson Gisleskog. Investigation was carried out by Iolanda Cirillo and Saberi Rana Ali. Software was developed by Martine Neyens, Nele Goeyvaerts and Per Olsson Gisleskog. Martine Neyens, Nele Goeyvaerts and Per Olsson Gisleskog contributed to data visualization. Juan José Perez Ruixo, Nele Goeyvaerts and Oliver Ackaert contributed to writing the original draft, and all authors contributed to the review and editing of the manuscript. The sponsor was involved in conceptualization, design, data collection, analysis, decision to publish and preparation of the manuscript.

CONFLICT OF INTEREST STATEMENT

N.G., M.N., J.J.P.R., I.C., S.R.A. and O.A. are employees of Johnson & Johnson and are potential stockholders of Johnson & Johnson. P.O.G. is an employee of POG Pharmacometrics.

Supporting information

Figure S1. Individual canagliflozin plasma concentration–time data in paediatric patients and adult patients and healthy volunteers up to 80 h post latest dose.

Figure S2. Goodness‐of‐fit plot of the final paediatric popPK model coloured by dose.

Figure S3. Relationship between the IIV random effects and body weight for the final paediatric popPK model.

Figure S4. pcVPC of the final paediatric popPK model for the phase 1 (top) and phase 3 (bottom) studies by body weight.

Figure S5. Goodness‐of‐fit plot of the final paediatric popPK model coloured by body weight.

Figure S6. pcVPC of the paediatric popPK model with fixed allometric exponents in the phase 1 (top) and phase 3 (bottom) studies by dose.

Figure S7. Simulated concentration–time profiles at steady state after administration of canagliflozin 50 mg BID, 100 mg QD, 150 mg BID and 300 mg QD in paediatric and adult patients with T2DM.

Figure S8. Individual change from baseline in concentration–time HbA1c for paediatric patients with T2DM, stratified by dose level.

Figure S9. Goodness‐of‐fit plot of the final paediatric PK/PD model for HbA1c.

Table S1. Baseline demographics and covariates for paediatric patients included in the PK/PD analysis.

Table S2. Simulated AUC_24h and C_max at steady state after oral administration of canagliflozin at 50 mg BID, 100 mg QD, 150 mg BID, and 300 mg QD in paediatric and adult T2DM patients, stratified by sex.

Table S3 Simulated HbA1c and change from baseline HbA1c at week 26 after oral administration of canagliflozin at 50 mg BID, 100 mg QD, 150 mg BID and 300 mg QD in paediatric patients with T2DM.

Goeyvaerts N, Gisleskog PO, Neyens M, et al. Population pharmacokinetics and pharmacodynamics of canagliflozin in paediatric patients with type 2 diabetes mellitus. Br J Clin Pharmacol. 2026;92(10):3622‐3633. doi: 10.1002/bcp.70644

DATA AVAILABILITY STATEMENT

The data sharing policy of Johnson & Johnson is available online (https://innovativemedicine.jnj.com/our‐innovation/clinical‐trials/transparency). As noted on this site, requests for access to the study data can be submitted through Yale Open Data Access (YODA) Project site (http://yoda.yale.edu).