Skip to main content
NCBI home page
Clinical and Translational Science logoLink to Clinical and Translational Science
. 2023 Apr 18;16(7):1197–1209. doi: 10.1111/cts.13523

Model‐informed approach to support pediatric dosing for the pan‐PI3K inhibitor copanlisib in children and adolescents with relapsed/refractory solid tumors

Peter N Morcos 1,✉, Jan Schlender 2, Rolf Burghaus 2, Jonathan Moss 3, Adam Lloyd 3, Barrett H Childs 1, Margaret E Macy 4, Joel M Reid 5, John Chung 1, Dirk Garmann 2
PMCID: PMC10339701  PMID: 37042099

Abstract

Copanlisib is an intravenously administered phosphatidylinositol 3‐kinase (PI3K) inhibitor which was investigated in pediatric patients with relapsed/refractory solid tumors. A model‐informed approach was undertaken to support and confirm an empirically selected starting dose of 28 mg/m2 for pediatric patients ≥1 year old, corresponding to 80% of the adult recommended dose adjusted for body surface area. An adult physiologically based pharmacokinetic (PBPK) model was initially established using copanlisib physicochemical and disposition properties and clinical pharmacokinetics (PK) data and was shown to adequately capture clinical PK across a range of copanlisib doses in adult cancer patients. The adult PBPK model was then extended to the pediatric population through incorporation of age‐dependent anatomical and physiological changes and used to simulate copanlisib exposures in pediatric cancer patient age groups. The pediatric PBPK model predicted that the copanlisib 28 mg/m2 dose would achieve similar copanlisib exposures across pediatric ages when compared with historical adult exposures following the approved copanlisib 60 mg dose administered on Days 1, 8, and 15 of a 28‐day cycle. Clinical PK were collected from a phase I study in pediatric patients with relapsed/refractory solid tumors (aged ≥4 years). An established adult population PK model was extended to incorporate an allometrically‐scaled effect of body surface area and confirmed that the copanlisib maximum tolerated dose of 28 mg/m2 was appropriate to achieve uniform copanlisib exposures across the investigated pediatric age range and consistent exposures to historical data in adult cancer patients. The model‐informed approach successfully supported and confirmed the copanlisib pediatric dose recommendation.

Study Highlights.

WHAT IS THE CURRENT KNOWLEDGE ON THE TOPIC?

Copanlisib monotherapy is approved for adult patients with relapsed follicular lymphoma and demonstrated superior progression‐free survival in combination with rituximab versus placebo in the adult phase III CHRONOS‐3 study.

WHAT QUESTION DID THIS STUDY ADDRESS?

We utilized a model‐informed approach to support and evaluate copanlisib pediatric dosing in children and adolescents.

WHAT DOES THIS STUDY ADD TO OUR KNOWLEDGE?

These analyses provide a quantitative assessment to propose and justify copanlisib 28 mg/m2 administered on Days 1, 8, and 15 of a 28‐day treatment cycle to pediatric patients aged 4 years and older.

HOW MIGHT THIS CHANGE CLINICAL PHARMACOLOGY OR TRANSLATIONAL SCIENCE?

The model‐informed approach provides a quantitative framework which leverages the strengths of physiologically based pharmacokinetic (PBPK) and population pharmacokinetic modeling to support and justify pediatric dosing of a small molecule anticancer agent.

INTRODUCTION

Copanlisib (BAY 80‐6946; Bayer Pharma AG, Berlin, Germany) is an intravenously administered small molecule inhibitor of phosphatidylinositol 3‐kinase (PI3K). In vitro, copanlisib shows preferential potent inhibition of p110α and p110δ isoforms (IC50 values of 0.5 and 0.7 nmol/L, respectively) relative to the p110β and p110γ isoforms (IC50 values of 3.7 and 6.4 nmol/L, respectively). 1 In PI3K mutant cancer cell lines and in mouse xenograft models, copanlisib showed significant antitumor activity with complete tumor regression seen at higher doses of copanlisib. 2 Nonclinical models demonstrated robust efficacy when copanlisib was administered either via continuous or intermittent dosing schedule with potential better tolerability of intermittent dosing by enabling recovery of normal tissue. 2 , 3 , 4

The clinical pharmacology properties of copanlisib have been comprehensively characterized in nonclinical and clinical studies. In the phase I first‐in‐human study (Study 12871), administration of copanlisib as a 1‐h intravenous infusion showed peak copanlisib concentrations which were reached by the end of infusion and thereafter declined in a multi‐exponential manner towards the end of the dosing interval. 5 Escalating doses administered on an intermittent dosing schedule showed approximately dose proportional pharmacokinetics (PK) up to and including the maximum tolerated dose (MTD) of 0.8 mg/kg administered on Days 1, 8, and 15 of a 28‐day cycle with no evidence of appreciable accumulation. 5 An approximate flat dose equivalent of 60 mg was supported based on evaluation of PK, safety, and activity. 6 Copanlisib disposition is characterized by metabolism primarily (>90%) via cytochrome P450 3A (CYP3A) and to a minor extent (<10%) by CYP1A1. 7 Copanlisib is a substrate of efflux transporters P‐glycoprotein (P‐gp) and breast cancer resistance protein (BCRP). 7 In humans, administration of 14C‐copanlisib revealed that parent copanlisib was the predominant circulating moiety representing 84% of total radioactivity. 8 Copanlisib was predominantly excreted via feces and partly via urine as unchanged drug and metabolites. 8 Copanlisib is eliminated with a half‐life of approximately 38 h. 5 A comprehensive population PK model described copanlisib PK via a three‐compartment model with first‐order elimination following intravenous administration. 9 Several covariates statistically influenced copanlisib PK variability including sex, patients from Japan, patients with some level of hepatic impairment, drug–drug interactions with strong modulators of CYP3A activity, and patients enrolled in the phase III study, CHRONOS‐3; however, individual covariates effects were generally limited in size. 9

The clinical efficacy and safety of copanlisib were investigated in two pivotal studies. In the phase II study, CHRONOS‐1, copanlisib monotherapy administered at 60 mg on Days 1, 8, and 15 of a 28‐day cycle demonstrated an objective response rate (ORR) of 59% and a manageable safety profile in adult patients with relapsed or refractory indolent B‐cell lymphoma who had previously received at least two therapies. 10 These results supported the approval of copanlisib in the United States, Taiwan, and Israel. 7 , 11 , 12 In the phase III study, CHRONOS‐3, copanlisib in combination with rituximab demonstrated superior progression‐free survival (PFS) over placebo plus rituximab in adult patients with relapsed indolent non‐Hodgkin lymphoma (iNHL) (hazard ratio [HR] 0.52; 95% confidence interval [CI] 0.39, 0.69). 13

Aberrations in the PI3K‐AKT signaling pathway have been implicated in high‐risk pediatric cancers with a reported frequency of ≥50% in neuroblastoma, 14 , 15 osteosarcoma, 16 Ewing sarcoma, 17 and rhabdomyosarcoma. 18 Copanlisib monotherapy was investigated in a dose escalation study in pediatric patients with relapsed/refractory solid tumors conducted in collaboration with the Children's Oncology Group (COG). 19 An empirically selected starting dose of copanlisib 28 mg/m2 was proposed for pediatric patients ≥1 year to represent 80% of the approved adult dose adjusted for body surface area (BSA; calculated as the adult copanlisib 60 mg dose ÷ an average adult BSA of 1.73 m2 × 80% = 28 mg/m2). This work describes the pharmacometrics analyses used to support and confirm copanlisib pediatric dosing (see Figure 1).

FIGURE 1.

FIGURE 1

Open in a new tab

Pharmacometrics analyses to support and confirm copanlisib pediatric dosing. ADME, absorption, distribution, metabolism, and excretion; AUC, area under the curve; PBPK, physiologically based pharmacokinetic; PK, pharmacokinetics; popPK, population pharmacokinetics.

METHODS

Adult PBPK model development

The PBPK modeling software Open Systems Pharmacology Suite (OSP Suite, www.open‐systems‐pharmacology.org, version 6.0) was used for the adult whole‐body PBPK model. 20 First, the model structure was built based on prior knowledge: the physicochemical properties of copanlisib, and in vitro information on metabolism, transport, and binding partners (Table 1). Second, physicochemical parameters for which literature data or in‐house information was available were set to the known values (Table 1). Similarly, the numerical values of the parameters describing the metabolism, transport, and binding processes were determined, either based on in vitro data or optimized in accordance with observed clinical datasets used in model development including clinical PK data for copanlisib 0.8 mg/kg from the first‐in‐human Study 12871 5 and mass balance information from Study 16353. 8

TABLE 1.

Input data values and model parameters for copanlisib adult physiologically based pharmacokinetic (PBPK) model development.

Parameter Copanlisib PBPK model
Model input value Source
Physicochemical
Molecular mass (g/mol) 480.5 Free base
Log P 2.1 a , 40 Estimated using parameter identification
pKa 8.64 Experimentally determined
7.41 Experimentally determined
2.11 Experimentally determined
Fraction unbound (%) 15.8 Experimentally determined
Aqueous solubility at pH 7 (mg/mL) 5 b Experimentally determined
Blood cell/plasma concentration ratio 1.71 Experimentally determined
Distribution
Partition coefficient model PK‐Sim® standard Willmann et al. 40 , 41
Cellular permeability model PK‐Sim® standard Willmann et al. 40 , 41
Binding partner
koff (1/min) 660 Estimated using parameter identification
Kd (nmol/L) 390.5 Estimated using parameter identification
Transport
P‐glycoprotein (P‐gp)
Km (μmol/L) 95.8 Estimated using parameter identification
Vmax (μmol/L/min) 87 Estimated using parameter identification
Metabolism
CYP3A4
Km (μmol/L) 51.8 Estimated using parameter identification
Vmax (μmol/L/min) 5.04 Estimated using parameter identification
Excretion
GFR fraction 1
Open in a new tab

Abbreviations: Glomerular filtration rate (GFR) fraction defines if drug is only filtrated (1), reabsorbed (<1), or additionally secreted (>1); Log P, octanol–water partition coefficient; pKa, negative decadic logarithm acid dissociation constant in log scale.

a

Identified using the distribution model by Willmann et al. 40 defined as the PK‐Sim standard.

b

For administration, the free base BAY 80‐6946 was formulated in a mannitol solution after addition of hydrochloric acid so that the hydrochloride salt was formed in the administration solution. The value was set sufficiently high to avoid solubility limitations.

PBPK model assumptions

Model building was carried out under consideration of model assumptions. Oxidative metabolism of copanlisib was predominantly catalyzed by CYP3A4, while CYP1A1 also contributed to a minor extent. 8 In the PBPK model, copanlisib is entirely metabolized by CYP3A4. Furthermore, copanlisib was characterized as a weak substrate of P‐gp and of BCRP in in vitro studies with different cell lines. 7 Both efflux transporters contribute to renal and biliary secretion located at the kidney proximal tubule and hepatocytes (canalicular). In the PBPK model, both processes were represented by implementation of P‐gp. In addition, a hypothetical binding partner of copanlisib is assumed and built into the PBPK model to better capture the observed variability in copanlisib PK around the end‐of‐infusion (see Results). It was chosen to spatially distribute the binding partner similar to the expression profile of PI3K. This is in line with the observed spatial distribution of copanlisib as revealed by whole‐body autoradiography in rats. 21

PBPK model evaluation

The developed model was evaluated through simulations. A virtual adult population of 1000 individuals in the age range of 20–80 years was created using the population simulation module of PK‐Sim®. 22 The population module creates virtual individuals within a given index age, body weight, and body height index range in a stochastic approach using age‐dependent distributions of anthropometric and physiological parameters. Variability of the reference concentration of the hypothetical binding partner, P‐gp, and CYP3A4 was considered as part of the respective ontogenies. Model qualification was determined based on goodness‐of‐fit plots and plots for residual distribution. Simulations were carried out for the virtual reference population using the intended dosing scenario as in the clinical study and statistical comparison of the distribution of individually observed data points around the simulated median in a virtual population. The clinical PK data used for the model building process (0.8 mg/kg from Study 12871) were first compared to model simulations. After refinement of the model with incorporation of a hypothetical binding partner, the PBPK base simulations were again compared to the datasets used for model development and independent (validation) datasets for copanlisib 0.1, 0.2, 0.4, 1.2 mg/kg from Study 12871 as well as clinical PK data from the phase II study, Study 16349A. 23

Pediatric PBPK model scaling

To establish the copanlisib pediatric PBPK model, information on the age dependencies of the relevant anthropometric measures (body height [BH], body weight [BW]) and physiological parameters (blood flows, organ volumes, hematocrit, and cardiac output) in children and adolescents were applied as incorporated into the system model in PK‐Sim. 24 The set of parameters that described the average adult individual served as the basis for the pediatric model (e.g., parameters for the active processes and the hypothetical binding protein). Reference concentrations and process parametrization were kept constant. The clearance capability was indirectly scaled along the pediatric age range by the age‐dependent changes in liver/kidney organ weight and hepatic/kidney blood flow. Additionally, the underlying process‐specific ontogenies accounting for activity levels and published previously were applied in the simulation scenarios. 25

Pediatric PBPK model application

The goal of the model application was to evaluate the empirically proposed body surface‐based dosing regimen(s) for copanlisib in a pediatric and adolescent population ranging from 1 year to 22 years of age in order to achieve a specific target exposure. A virtual population of individuals ranging in age from 6 months to 22 years was created for simulations with 2000 virtual individuals generated for five age‐bins applying a uniform age distribution for each age‐bin. Age groups for the purpose of the analysis were defined from ≥6 months to <1 year (infants), ≥1 to <2 years (toddlers), ≥2 to <6 years (pre‐school), ≥6 to <12 years (school), and ≥12 to <22 years (adolescents). The empirically proposed starting dose of 28 mg/m2 reflecting 80% of the adult dose corrected for BSA for children aged ≥1 year was simulated to predict respective exposures and subsequently compared with historical exposures in adults. Simulations are provided only for children aged ≥1 year as a lower empirically derived copanlisib dose (21 mg/m2) was proposed for younger children (aged <1 years) which was never investigated in the phase I study and thus any simulation results could not be confirmed with clinical data (see also Discussion). A sensitivity analysis was conducted to assess the impact of PBPK model parameters changes on copanlisib simulated area under the curve (AUC) for the different age groups within the virtual population, as previously done. 26 The sensitivity values are dimensionless quantities. For example, a sensitivity of −1.0 implies that a 10% increase of the parameters leads to a 10% decrease in AUC, and a sensitivity of +0.5 implies that a 10% increase of the parameters leads to a 5% increase in AUC. 26 Further details on the methodology are provided elsewhere. 26 Overall, sensitivities were calculated for 159 parameters, representing all primary (i.e., non‐derived) parameters. A forest plot was created to illustrate the impact of sensitivity analyses for the parameters.

Population PK modeling

The population PK (popPK) analysis (using NONMEM version 7.4; ICON Development Solutions, Hanover, MD, USA 27 ) included data collected from pediatric patients with histologically verified solid tumor or lymphoma malignancy at diagnosis for which there is no standard curative anticancer treatment or treatment is no longer effective enrolled in an open‐label, non‐randomized, single‐arm, stepwise phase I dose escalation study based on Rolling‐6 design. 19 Pediatric patients were enrolled in the first dose escalation cohort and received copanlisib doses of 28 mg/m2 and upon successful completion of the dose‐limiting toxicity (DLT) period, subsequent patients were enrolled into a second dose escalation cohort and received 35 mg/m2; both administered on Days 1, 8, and 15 of a 28‐day schedule. 19 Blood samples were collected for PK measurements at pre‐infusion (up to 30 min prior to start of infusion), 1–1.25 h, 1.5–3 h (patients ≥6 years only), and 22–24 h after start of the infusion at Cycle 1 on Days 1 and 15. Using the adapted adult population PK model (described later), these timepoints were predicted to enable accurate and precise estimation of copanlisib PK parameters, AUC from time = 0 to time = 168 h (AUC0–168) and maximum concentrations (Cmax), while limiting the extent of blood collection from the pediatric patients. Plasma samples were assayed for copanlisib concentrations using a validated liquid chromatography/mass spectroscopy (LC/MS) analytical method with a lower limit of quantification (LLOQ) of 0.5 ng/mL.

The popPK modeling approach utilized the comprehensive copanlisib adult popPK model which has been previously established based on data collected from phase I–III studies. 9 The adult popPK model was adapted through allometric scaling using BSA (centered around a typical adult BSA value of 1.88 m2) with an exponent of 1 on all rate (including clearance) and volume parameters. An exponent of 1 was initially considered based on the direct BSA‐based dosing approach investigated in the phase I study. Alternative exponents including commonly used allometric exponents (e.g., 0.75) were investigated in sensitivity analyses and were to be considered if the exponent of 1 was considered unsatisfactory in capturing the observed PK data from the pediatric patients. The adapted model was used without further modification or any parameter re‐estimation (i.e., typical values for the rate constants were fixed to the values in the adult model) to estimate individual exposure (AUC following three nominal copanlisib weekly doses [AUC0–168]) using the incoming pediatric PK. Goodness‐of‐fit plots and prediction‐corrected visual predictive checks (pcVPC) created by 1000 repeat simulations of the dataset were used to confirm that the adapted adult model can adequately describe the pediatric PK.

In addition, a population of 500 virtual pediatric patients with an age range consistent with the enrolled population in the phase I dose escalation receiving copanlisib 28 mg/m2 was simulated using a distribution of BSA sampled from the NHANES III database, 28 a survey containing demographic data of nearly 40,000 persons aged 2 months and older, and clearance and volume parameters sampled from the adapted adult popPK model. This approach assessed whether the adapted model could capture the distribution of exposure across all pediatric patients from the study and compared to historical adult exposures from the monotherapy pivotal study, CHRONOS‐1.

Ethics statement

The study protocol was approved by the institutional review boards of participating institutions and complied with the Declaration of Helsinki, current Good Clinical Practice guidelines, and local laws and regulations. Written informed consent for the use of PK data was obtained from all patients and participant parents/legal guardians.

RESULTS

PBPK modeling

A figure of the adult PBPK model simulation overlaid on observed concentration–time data from patients in the phase I study receiving copanlisib 0.8 mg/kg used in model development illustrates that the adult PBPK model generally captures the observed data with all clinically observed copanlisib plasma concentrations within the expected range defined by the minimum and maximum scenarios of the PBPK simulation (Figure 2a). Performance of the adult PBPK model was further evaluated based on data from the phase I study not used in model development and showed that across copanlisib 0.1–1.2 mg/kg the model generally captures the observed data, although notable high variability was seen during the infusion phase for lower doses (Figure 2b). Incorporation of an empiric hypothetical binding protein with a geometric mean and geometric standard deviation of the binding protein concentration of 149.5 and 2.6 μmol/L, respectively, and log‐normal distribution enabled descriptive capturing of the observed variability seen in the end‐of‐infusion concentrations between patients. Since the intention of the adult PBPK model was an adequate description of overall adult exposure, this approach was considered an appropriate fit‐for‐purpose approach. Overall, the adult PBPK model was determined to be suitable to extend to the pediatric population.

FIGURE 2.

FIGURE 2

Open in a new tab

Adult physiologically based pharmacokinetic (PBPK) model simulations and comparison to copanlisib pharmacokinetics (PK). (a) Overlay of adult PBPK model simulation and observed copanlisib PK used in model development from patients receiving copanlisib 0.8 mg/kg for the first dosing interval of 7 days (left) and the first 4 h (right). (b) Overlay of adult PBPK model simulation and observed copanlisib PK not used in model development from patients receiving copanlisib 0.1–1.2 mg/kg. Clinically observed (symbols) and simulated (lines, shaded areas) plasma concentration–time profile of copanlisib in adult cancer patients from the phase I study. The thick black line represents the median of the simulated plasma concentration in a virtual adult population with geometric mean concentration of hypothetical binding partner equal to 149.5 μmol/L and geometric standard deviation equal to 2.6. The gray shaded area represents the 5% (lower) and 95% (upper) of the simulated concentrations profiles. The dashed lines represent the minimum and maximum of the simulated plasma concentration profiles generated with a minimum and maximum reference concentration of copanlisib‐binding protein of 7.5961 μmol/L and 2415.3 μmol/L, respectively.

The pediatric PBPK model which incorporated age‐dependent absorption, distribution, metabolism, and excretion (ADME) processes was used to predict exposures in pediatric patients across age groups. Relative to the observed historical exposure from adult patients receiving copanlisib 0.8 mg/kg (body weight equivalent of 60 mg) in the first‐in‐human phase I study, 5 the pediatric geometric mean ratio based on predicted AUC0–168 following administration of copanlisib 28 mg/m2 was 86% for adolescents (age: ≥12 years), 88% for school‐children (age: 6–12 years), 95% for pre‐school children (age: 2–6 years), 103% for toddlers (age: 1–2 years), and 93% for overall pediatric patients aged ≥1 year. A sensitivity analysis revealed that most model parameters have only a minor or negligible influence on the system behavior and generally the sensitivities are similar across the age range (Figure 3). Those parameters that exert an influence do so in a comprehensible way reflecting their physiological role. Overall, the pediatric PBPK model was determined to be robust in providing predictions in the pediatric population and was used to support the starting dose of copanlisib 28 mg/m2 for patients aged ≥1 year in the pediatric phase I dose escalation study. 19

FIGURE 3.

FIGURE 3

Open in a new tab

Pediatric physiologically based pharmacokinetic (PBPK) model sensitivity analysis. Listing of the most sensitive parameters towards area under the curve (AUC) above the 90% cutoff in descending order. The x‐axis describes sensitivities and their positive or negative impact on AUC.

PopPK modeling

A total of 31 pediatric patients provided 191 copanlisib PK samples. Of the 31 patients, 24 received copanlisib 28 mg/m2 and 7 received copanlisib 35 mg/m2 on Days 1, 8, and 15 of a 28‐day schedule. A total of 30 PK samples collected prior to the first dose which were below the limit of quantification (BLQ) were omitted and 1 PK observation was omitted as it was kinetically implausible, leaving 160 PK observations in 30 patients total. For those remaining 30 patients, the baseline median (range) age was 13 (4–21) years and BSA was 1.37 (0.64–2.44) m2.

Goodness‐of‐fit plots illustrated that the adapted adult popPK model adequately captured the pediatric PK data with reasonable agreement between observed and predicted concentrations (Figure 4a). Conditional weighted residuals (CWRES) figures showed no evidence of significant bias and overall were acceptable with respect to concentration or time. The pcVPC also illustrated that copanlisib concentrations were generally well captured and model predictions were considered satisfactory overall given the relatively limited data and no re‐estimation of model parameters using the pediatric data (Figure 4b). Sensitivity analyses for the value of the allometric exponent supported the diagnostic plots and selection of an exponent of 1 as alternative exponents did not offer significant improvement in capturing the observed pediatric data.

FIGURE 4.

FIGURE 4

Open in a new tab

Adapted adult population pharmacokinetic (popPK) model performance in describing pediatric pharmacokinetics. (a) Goodness‐of‐fit plots. CWRES, conditional weighted residual; IPRED, individual prediction; PRED, population prediction; TAD, time after dose. (b) Prediction‐corrected visual predictive checks (pcVPC). Black circles: prediction‐corrected observations. Red horizontal lines: 50th percentiles of prediction‐corrected observations in bin. Yellow horizontal lines: 10th and 90th percentiles of prediction‐corrected observations in bin. Black dotted horizontal lines: 50th percentile of prediction‐corrected simulated values in bin. Gray shaded areas: range between 10th and 90th percentiles of prediction‐corrected simulated values in bin.

Overall, administration of copanlisib 28 mg/m2 demonstrated uniform exposure across the entire age range as evidenced by model‐predicted individual exposure estimates and based on population simulations of the virtual pediatric population (Figure 5). In addition, the copanlisib 28 mg/m2 achieved consistent exposures when compared with historical exposures in adults (Figure 5). Relative to adults in CHRONOS‐1, the geometric mean ratio for model‐predicted AUC0‐168 of all enrolled pediatric patients was 101%.

FIGURE 5.

FIGURE 5

Open in a new tab

Population pharmacokinetic (popPK) model simulations in pediatrics and comparison to historical adult data from the monotherapy pivotal study, CHRONOS‐1. Observed values: simulated AUC0–168 after three nominal doses of copanlisib at 28 mg/m2 using individual CL and V1 estimates from Study 19176. Simulated values: simulated AUC0–168 after three nominal doses of copanlisib at 28 mg/m2 of 500 virtual pediatric patients for each age (250 male, 250 female). Virtual patients had body surface area sampled from the NHANES III database and CL and V1 were derived by sampling random effects from the adjusted popPK model. AUC, area under the curve.

DISCUSSION

Copanlisib has demonstrated robust efficacy and acceptable safety in adult patients with relapsed B‐cell malignancies alone and in combination with rituximab based on results from two pivotal studies. 10 , 13 Given high‐risk pediatric cancers may be associated with activation of the PI3K signaling pathway, a phase I dose escalation study was initiated to investigate the safety, tolerability, antitumor activity, and PK of copanlisib in relapsed/refractory solid tumors. 19 While starting doses for copanlisib were proposed based on empirical considerations using BSA normalization and a safety adjustment to 80% of the total adult dose, a quantitative approach for assessment of pediatric dosing would provide a more systematic consideration of copanlisib pharmacology, pharmacokinetics, exposure–response relationships, and established knowledge of pediatric ontogeny and body size. Thus, the pharmacometrics analyses aimed to establish a quantitative framework to support and thereafter confirm copanlisib pediatric dosing. Given its specific mechanism of action as a targeted agent against the PI3K isoforms, the assumption for the pharmacometrics analyses was that copanlisib exposure–response relationships would be retained between adults and children and adolescents. Thus, the goal was to identify/confirm copanlisib doses that achieve uniform and consistent exposures to those achieved in adults with the approved dose, consistent with recommendations in the United States Food and Drug Administration (FDA) Guidance for Pediatric Studies. 29

The pharmacometrics analyses considered a stepwise model‐based strategy for copanlisib pediatric dosing using PBPK and popPK methodology (Figure 1). Initially, an adult PBPK model was established from bottom‐up by incorporating information on physicochemical properties, mass balance, and preclinical and clinical data for copanlisib. While a majority of the inputted data was taken directly from source experimentation, some parameters were optimized slightly to support incorporation into the PK‐Sim software and/or to best capture the observed clinical data that was used in model building or validation. Notably, a hypothetical binding protein with varying binding capability was introduced to capture the observed high end‐of‐infusion variability seen for copanlisib. This incorporation was required to adequately capture the observed copanlisib concentration–time profiles to support a suitable adult PBPK model and adequate extension to the pediatric population. The completed sensitivity analysis conducted with the pediatric PBPK model revealed that most model parameters have only a minor or negligible influence on the system behavior, and thus the developed PBPK model framework is robust against uncertainties in parameter values. The larger sensitivity across age groups for the lipophilicity impact on AUC compared with other parameters reflect age‐related changes in distribution organs (i.e., relative organ size with body weight) and copanlisib disposition properties (fast distribution phase). The lower sensitivity of CYP3A4 parameter impact on AUC reflects the rapid age‐related maturation of CYP3A. Results from the simulations supported the proposed empirically derived starting dose as appropriate to achieve copanlisib exposures within the range of historical adult exposures.

Evaluation of the investigated doses in the pediatric phase I study was conducted using the adapted adult popPK model incorporating allometrically scaled parameters according to BSA. Without any other adjustment, the adapted adult popPK model adequately captured the observed clinical PK data from the two evaluated doses, 28 and 35 mg/m2, in the pediatric phase I study. The analysis results support that following maturation of age‐dependent physiological processes involved in copanlisib disposition (predicted to be the case by the developed pediatric PBPK model for the age range investigated in the phase I study), copanlisib exposures scale to body size. BSA was utilized as the body size measure in the phase I study largely based on historical experience for use of anticancer agents by oncology clinicians. 30 , 31 The pcVPC demonstrating that the adapted adult popPK model well captures the observed clinical PK from the enrolled pediatric patients supports the use of BSA as a body size covariate and relevant variable for dosing. The popPK model confirmed that the MTD from the pediatric phase I study, 28 mg/m2, achieved uniform exposure (AUC) across the investigated age range (≥4 years) and demonstrated consistent exposure to the adult exposures from the CHRONOS‐1 monotherapy pivotal study. Given its mechanism of action as an inhibitor of the PI3K pathway which was assumed to be implicated in the tumorigenesis of the enrolled pediatric patients, the exposure–response relationships for copanlisib are assumed to be similar between adults and pediatric patients and thus the copanlisib 28 mg/m2 dose is expected to be tolerable and engage the target for the investigated pediatric population. These considerations are supported by the reported safety and pharmacodynamic (PD) data from the pediatric phase I dose escalation which confirmed this dose as the MTD and demonstrated significant PD changes similar to those reported in adult patients. 19 In addition, the popPK results corroborate and are in line with the simulations from the pediatric PBPK model, supporting the complementarity and consistency between these two model‐based approaches.

The application of model‐based approaches to inform pediatric dosing is becoming increasingly utilized in pediatric drug development. A review of pediatric clinical trial literature over a 10‐year period investigated the use and performance of PBPK models to support pediatric dosing. 32 Across 26 small molecule drugs with varying administration and disposition properties the overall prediction accuracy was nearly 80% across the pediatric age groups reviewed, with the exception of neonates, 32 illustrating the capability of the PBPK modeling approach to support pediatric dosing. A recent case example of application of PBPK modeling to support pediatric anticancer development was shown for selumetinib, a small molecule targeted agent, in which a developed pediatric PBPK model recovered observed data in pediatric patients and enabled prospective dosing regimen simulations for further planned pediatric trials. 33 Similar success of capturing pediatric data with PBPK modeling has been reported for other small molecule anticancer agents including sunitinib, 34 docetaxel, 35 and etoposide. 36 Likewise, popPK approaches are well established and have shown ability to characterize and predict exposures in pediatric patients by accounting for body size covariates for pediatric age ranges associated with maturation of physiological disposition processes. Indeed, popPK models with allometrically scaled body size covariates have successfully captured/predicted pediatric data for the anticancer agents ponatinib, 37 entrectinib, 38 and methotrexate. 39 Our example leverages the strengths of the complimentary modeling tools through incorporation of PBPK modeling accounting for maturation of copanlisib disposition properties to predict pediatric doses/exposures prior to study initiation and popPK modeling, which utilizes an established adult model to descriptively characterize and estimate exposures from emerging pediatric clinical PK data.

While the results from these analyses were used to support and confirm the copanlisib pediatric dose to achieve tolerable and target‐engaging exposures, there was a lack of sufficient evidence of monotherapy efficacy observed in the pediatric clinical study, 19 which potentially could have been due to the limited driving role of PI3K pathway activation in the cancer pathogenesis of the unselected enrolled population. 19 Ultimately, this precluded further enrolment of patients into the study, including additional patients at lower age ranges, which could have further evaluated the performance of the PBPK and popPK models to support appropriate copanlisib doses in those age ranges. Of note, a lower empirically derived BSA‐based starting dose, 21 mg/m2, was proposed for patients aged <1 year, which could not be investigated/confirmed due to lack of enrolment in this age group.

In conclusion, pharmacometrics analyses based on a stepwise application of PBPK and popPK modeling was used to support and subsequently confirm copanlisib dosing in pediatric patients aged 4 years and older. Incorporating knowledge on copanlisib drug properties and physiological changes with respect to age enables a quantitative and holistic approach to an informed dose selection. The complementary modeling approaches and availability of clinical data enabled assessment and comparison of model performance. The outcome of the clinical study and applied modeling approach support the copanlisib pediatric dose of 28 mg/m2 in pediatric patients aged 4 years and older. Notably, the established robust models could have alternatively been used in a complementary fashion to prospectively propose optimal pediatric dosing regimens. Additional clinical data are needed to confirm model predictions and dose selection for younger patients.

AUTHOR CONTRIBUTIONS

All authors wrote the manuscript. Barrett H. Childs, Margaret E. Macy, Joel M. Reid, and John Chung designed the research. Peter N. Morcos, Jan Schlender, Rolf Burghaus, Jonathan Moss, Adam Lloyd, and Dirk Garmann performed the research and analyzed the data.

FUNDING INFORMATION

This study was funded by Bayer AG.

CONFLICT OF INTEREST STATEMENT

PNM, JS, RB, BHC, JC, and DG are employees of Bayer HealthCare Pharmaceuticals, Inc. PNM, JS, RB, and DG have stock ownership of Bayer AG. JM and AL are employees of BAST Inc. Limited and receive consulting fees from Bayer AG. MEM has stock ownership in Johnson & Johnson, receives salary support and institutional support from Bayer HealthCare Pharmaceuticals, Inc., and consulting fees from Y‐mAbs Therapeutics. JMR is a consultant for Elucida Oncology, Inc.

Morcos PN, Schlender J, Burghaus R, et al. Model‐informed approach to support pediatric dosing for the pan‐PI3K inhibitor copanlisib in children and adolescents with relapsed/refractory solid tumors. Clin Transl Sci. 2023;16:1197‐1209. doi: 10.1111/cts.13523

Trial registration ID: NCT03458728.

REFERENCES

  • 1. Haike K, Stasik E, Soujon M, et al. Molecular mechanisms supporting inhibition of PI3K isoforms by copanlisib in blocking B‐cell signaling and tumor cell growth in diffuse large B‐cell lymphoma. Poster Presented at: 1st American Society of Hematology Meeting on Lymphoma Biology; August 10–13, 2014; Colorado Springs, CO, USA.
  • 2. Liu N, Rowley BR, Bull CO, et al. BAY 80‐6946 is a highly selective intravenous PI3K inhibitor with potent p110alpha and p110delta activities in tumor cell lines and xenograft models. Mol Cancer Ther. 2013;12:2319‐2330. [DOI] [PubMed] [Google Scholar]
  • 3. Paul J, Soujon M, Wengner AM, et al. Simultaneous inhibition of PI3Kdelta and PI3Kalpha induces ABC‐DLBCL regression by blocking BCR‐dependent and ‐independent activation of NF‐kappaB and AKT. Cancer Cell. 2017;31:64‐78. [DOI] [PubMed] [Google Scholar]
  • 4. Will M, Qin ACR, Toy W, et al. Rapid induction of apoptosis by PI3K inhibitors is dependent upon their transient inhibition of RAS‐ERK signaling. Cancer Discov. 2014;4:334‐347. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 5. Patnaik A, Appleman LJ, Tolcher AW, et al. First‐in‐human phase I study of copanlisib (BAY 80‐6946), an intravenous pan‐class I phosphatidylinositol 3‐kinase inhibitor, in patients with advanced solid tumors and non‐Hodgkin's lymphomas. Ann Oncol. 2016;27:1928‐1940. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 6. Reif S, Ahsman M, Jentsch G, et al. Use of a population pharmacokinetic approach and time‐to‐event analysis to support the clinical recommendation of a flat dosing of copanlisib in cancer patients. Clin Pharmacol Ther. 2016;99:S5‐S107.26803826 [Google Scholar]
  • 7. Bayer HealthCare Pharmaceuticals Inc . Aliqopa (copanlisib) [package insert]. U.S. Food and Drug Administration. 2022. https://www.accessdata.fda.gov/drugsatfda_docs/label/2022/209936s010lbl.pdf Revised February 2022. Accessed December 6, 2022 [Google Scholar]
  • 8. Gerisch M, Schwarz T, Lang D, et al. Pharmacokinetics of intravenous pan‐class I phosphatidylinositol 3‐kinase (PI3K) inhibitor [(14)C]copanlisib (BAY 80‐6946) in a mass balance study in healthy male volunteers. Cancer Chemother Pharmacol. 2017;80:535‐544. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 9. Moss J, Garmann D, Austin R, et al. Copanlisib population pharmacokinetics and exposure–response relationships in relapsed indolent non‐Hodgkin lymphoma: analysis from the 1‐year follow‐up of the phase III, randomized CHRONOS‐3 trial. Poster Presented at: Third AACR International Meeting, Advances in Malignant Lymphoma: Maximizing the Basic‐Translational Interface for Clinical Application; June 23–26, 2022; Boston, MA, USA.
  • 10. Dreyling M, Santoro A, Mollica L, et al. Phosphatidylinositol 3‐kinase inhibition by copanlisib in relapsed or refractory indolent lymphoma. J Clin Oncol. 2017;35:3898‐3905. [DOI] [PubMed] [Google Scholar]
  • 11. Taiwan Food and Drug Administration . ALIQOPA (copanlisib) Assessment Report. https://www.fda.gov.tw/tc/includes/GetFile.ashx?id=f637085445784753113. Accessed July 18, 2022
  • 12. The Israeli Drug Registry . ALIQOPA (copanlisib). https://israeldrugs.health.gov.il/#!/medDetails/165%2045%2035773%2000. Accessed July 18, 2022.
  • 13. Matasar MJ, Capra M, Özcan M, et al. Copanlisib plus rituximab versus placebo plus rituximab in patients with relapsed indolent non‐Hodgkin lymphoma (CHRONOS‐3): a double‐blind, randomised, placebo‐controlled, phase 3 trial. Lancet Oncol. 2021;22:678‐689. [DOI] [PubMed] [Google Scholar]
  • 14. Mosse YP, Laudenslager M, Longo L, et al. Identification of ALK as a major familial neuroblastoma predisposition gene. Nature. 2008;455:930‐935. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 15. Munoz J, Lázcoz P, Inda MM, et al. Homozygous deletion and expression of PTEN and DMBT1 in human primary neuroblastoma and cell lines. Int J Cancer. 2004;109:673‐679. [DOI] [PubMed] [Google Scholar]
  • 16. Freeman SS, Allen SW, Ganti R, et al. Copy number gains in EGFR and copy number losses in PTEN are common events in osteosarcoma tumors. Cancer. 2008;113:1453‐1461. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 17. Niemeyer BF, Parrish JK, Spoelstra NS, Joyal T, Richer JK, Jedlicka P. Variable expression of PIK3R3 and PTEN in Ewing sarcoma impacts oncogenic phenotypes. PLoS One. 2015;10:e0116895. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 18. Li HG, Wang Q, Li HM, et al. PAX3 and PAX3‐FKHR promote rhabdomyosarcoma cell survival through downregulation of PTEN. Cancer Lett. 2007;253:215‐223. [DOI] [PubMed] [Google Scholar]
  • 19. Macy ME, Cash T, Pinto N, et al. Phase I dose‐escalation study of the pan‐PI3K inhibitor copanlisib in children and adolescents with relapsed/refractory solid tumors. Eur J Cancer. 2022;174:S28‐S29. [Google Scholar]
  • 20. Lippert J, Burghaus R, Edginton A, et al. Open systems pharmacology community‐an open access, open source, open science approach to modeling and simulation in pharmaceutical sciences. CPT Pharmacometrics Syst Pharmacol. 2019;8:878‐882. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 21. Aliqopa FDA Multidisciplinary Review . https://www.accessdata.fda.gov/drugsatfda_docs/nda/2017/209936Orig1s000MultidisciplineR.pdf. Accessed December 6, 2022.
  • 22. Willmann S, Höhn K, Edginton A, et al. Development of a physiology‐based whole‐body population model for assessing the influence of individual variability on the pharmacokinetics of drugs. J Pharmacokinet Pharmacodyn. 2007;34:401‐431. [DOI] [PubMed] [Google Scholar]
  • 23. Dreyling M, Morschhauser F, Bouabdallah K, et al. Phase II study of copanlisib, a PI3K inhibitor, in relapsed or refractory, indolent or aggressive lymphoma. Ann Oncol. 2017;28:2169‐2178. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 24. Edginton AN, Schmitt W, Willmann S. Development and evaluation of a generic physiologically based pharmacokinetic model for children. Clin Pharmacokinet. 2006;45:1013‐1034. [DOI] [PubMed] [Google Scholar]
  • 25. Edginton AN, Schmitt W, Voith B, Willmann S. A mechanistic approach for the scaling of clearance in children. Clin Pharmacokinet. 2006;45:683‐704. [DOI] [PubMed] [Google Scholar]
  • 26. Schlender JF, Teutonico D, Coboeken K, et al. A physiologically‐based pharmacokinetic model to describe ciprofloxacin pharmacokinetics over the entire span of life. Clin Pharmacokinet. 2018;57:1613‐1634. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 27. Bauer RJ. NONMEM Users Guide: Introduction to NONMEM 7.4.0. Icon Development Solutions; 2017. [Google Scholar]
  • 28. National Center for Health Statistics . NHANES III (1988–1994). https://wwwn.cdc.gov/nchs/nhanes/nhanes3/default.aspx. Accessed December 6, 2022
  • 29. US Food and Drug Administration . General Clinical Pharmacology Considerations for Pediatric Studies for Drugs and Biological Products. https://www.fda.gov/media/90358/download. Accessed August 17, 2022
  • 30. Balis FM, Womer RB, Berg S, Winick N, Adamson PC, Fox E. Dosing anticancer drugs in infants: current approach and recommendations from the Children's Oncology Group's chemotherapy standardization task force. Pediatr Blood Cancer. 2017;64:e26636. [DOI] [PubMed] [Google Scholar]
  • 31. Smith M, Bernstein M, Bleyer WA, et al. Conduct of phase I trials in children with cancer. J Clin Oncol. 1998;16:966‐978. [DOI] [PubMed] [Google Scholar]
  • 32. Templeton IE, Jones NS, Musib L. Pediatric dose selection and utility of PBPK in determining dose. AAPS J. 2018;20:31. [DOI] [PubMed] [Google Scholar]
  • 33. Cohen‐Rabbie S, Zhou L, Vishwanathan K, et al. Physiologically based pharmacokinetic modeling for selumetinib to evaluate drug–drug interactions and pediatric dose regimens. J Clin Pharmacol. 2021;61:1493‐1504. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 34. Yu Y, DuBois SG, Wetmore C, Khosravan R. Physiologically based pharmacokinetic modeling and simulation of sunitinib in pediatrics. AAPS J. 2020;22:31. [DOI] [PubMed] [Google Scholar]
  • 35. Thai HT, Mazuir F, Cartot‐Cotton S, Veyrat‐Follet C. Optimizing pharmacokinetic bridging studies in paediatric oncology using physiologically‐based pharmacokinetic modelling: application to docetaxel. Br J Clin Pharmacol. 2015;80:534‐547. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 36. Kersting G, Willmann S, Würthwein G, Lippert J, Boos J, Hempel G. Physiologically based pharmacokinetic modelling of high‐ and low‐dose etoposide: from adults to children. Cancer Chemother Pharmacol. 2012;69:397‐405. [DOI] [PubMed] [Google Scholar]
  • 37. Hanley MJ, Diderichsen PM, Narasimhan N, Srivastava S, Gupta N, Venkatakrishnan K. Population pharmacokinetics of ponatinib in healthy adult volunteers and patients with hematologic malignancies and model‐informed dose selection for pediatric development. J Clin Pharmacol. 2022;62:555‐567. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 38. Gonzalez‐Sales M, Djebli N, Meneses‐Lorente G, et al. Population pharmacokinetic analysis of entrectinib in pediatric and adult patients with advanced/metastatic solid tumors: support of new drug application submission. Cancer Chemother Pharmacol. 2021;88:997‐1007. [DOI] [PubMed] [Google Scholar]
  • 39. Zhan M, Sun Y, Zhou F, et al. Population pharmacokinetics of methotrexate in paediatric patients with acute lymphoblastic leukaemia and malignant lymphoma. Xenobiotica. 2022;52:265‐273. [DOI] [PubMed] [Google Scholar]
  • 40. Willmann S, Lippert J, Sevestre M, et al. PK‐Sim®: a physiologically based pharmacokinetic ‘whole‐body’ model. Bios. 2003;1:121‐124. [Google Scholar]
  • 41. Willmann S, Lippert J, Schmitt W. From physicochemistry to absorption and distribution: predictive mechanistic modelling and computational tools. Expert Opin Drug Metab Toxicol. 2005;1:159‐168. [DOI] [PubMed] [Google Scholar]

Articles from Clinical and Translational Science are provided here courtesy of Wiley

RESOURCES