Extracorporeal membrane oxygenation (ECMO) is a cardiopulmonary bypass device that supports individuals with refractory heart and lung failure. In pediatrics, ECMO is commonly used to treat cardiac arrest, acute respiratory distress syndrome, and heart and lung transplantations. Specifically, the use of ECMO in pediatric cardiac arrest patients refractory to CPR is increasing due to better survival rates. However, one of the challenges in using ECMO in pediatric cardiac arrest is understanding and compensating for the effect of the extracorporeal circuit on drug pharmacokinetics (PK). Drug PK in ECMO patients can be altered due to the adsorption of drugs to ECMO components, physiological changes associated with ECMO (e.g., inflammation), and organ dysfunction due to the underlying critical illness. This change in drug PK may result in a difference in optimal dosing in this setting when compared with non‐ECMO patients.
The American Heart Association (AHA) guidelines for CPR and emergency cardiovascular care (ECC) recommend using amiodarone (5 mg/kg bolus dose, intravenous (i.v) or intraosseous) as a first‐line agent to treat ventricular arrhythmias in children. However, due to its high lipophilicity (LogP = 7.58), high protein binding (96%), and high percentage of the ionized fraction at physiological pH (87.3%), amiodarone has a high propensity for adsorption to the ECMO components leading to altered PK in ECMO patients. In an ex vivo ECMO circuit model, we previously showed that ~78% of the administered amiodarone dose was adsorbed onto the ECMO circuit components within 30 min of administration. Furthermore, Lescroart et al. reported a 32% decrease in the area under the curve (AUC) and 42% decrease in C max of amiodarone administered to pigs with potassium‐induced cardiac arrest and connected to ECMO. These data indicate that amiodarone concentration and PK are significantly altered in ECMO patients, and there is an unmet medical need to develop an optimal dose of amiodarone in pediatric cardiac arrest patients on ECMO.
Physiologically‐based pharmacokinetic modeling (PBPK) is a mathematical modeling technique that integrates drug properties (e.g., LogP, protein binding, clearance, etc.) and population information (e.g., organ size and blood flow) to predict the effect of various parameters (e.g., age, genetic variants, disease state etc.) on drug exposure. PBPK models are mechanistic in nature and structured to represent physiologically relevant spaces, with each virtual “organ” parameterized with differential equations describing the disposition of drugs within the compartment. Using data from ex vivo studies, an ECMO “organ” can be linked and parameterized to the PBPK model. This ECMO PBPK model can be used to predict the drug PK in ECMO patients and determine the optimal dosing in this vulnerable population. We previously used this approach to determine the optimal dosing of fluconazole in critically ill pediatric patients on ECMO.
In the current study, we developed a PBPK model of amiodarone in adults and extrapolated it to children. The pediatric model was evaluated with opportunistic clinical data. We then added an ECMO compartment to the amiodarone pediatric PBPK model and parameterized the model using the data from ex vivo studies. The final pediatric ECMO PBPK model was verified using amiodarone PK data in pediatric patients on ECMO from two different clinical studies. The verified pediatric ECMO PBPK model was used to predict the optimal dosing of amiodarone in children on ECMO across the pediatric age spectrum.
We followed the FDA guidance on PBPK model development and workflow in children to build our pediatric PBPK model. We first developed an adult PBPK model of amiodarone and verified the model prediction using amiodarone PK data from 5 adult PK studies. Model acceptance criteria were defined as: (1) Greater than 80% of observed concentrations captured within the 90% prediction interval of the model simulated concentrations; and (2) Average fold error (AFE) values of simulated concentrations within twofold of observed values.
Once the adult model met the acceptance criteria, we scaled the model to children. In pediatric model, we retained the physicochemical and drug‐specific absorption, distribution, metabolism, and elimination (ADME) parameters of amiodarone and replaced the anthropomorphic and physiological information with pediatric values using established age‐dependent algorithms in the model software (PK Sim®, Version 9.1, Open Systems Pharmacology Suite, Bayer Technology Services, Leverkusen, Germany). Pediatric model predictions were compared with observed data from a published pediatric amiodarone PK study and evaluated using the same acceptance criteria as for adults.
After confirming that the model acceptance criteria for the pediatric model were met, we added an ECMO compartment to the pediatric PBPK model to form the ECMO PBPK model. The ECMO compartment was parameterized using the adsorption data from our previously published ex vivo study. Model predictions in children on ECMO were evaluated using the data from Pharmacokinetics of Understudied Drugs Administered to Children per Standard of Care (POPS) trial and Primary Children's Hospital (PCH) opportunistic PK study.
PBPK modeling was performed in PK Sim® and MoBi® (Version 9.1 Build 2, Open Systems Pharmacology Bayer Technology Services, Leverkusen, Germany). We used R Statistical Software (v4.2.2; R Core Team 2022) for the calculation of amiodarone rate constants using the ex vivo data and GraphPad Prism® (Version 9, GraphPad LLC, CA, USA) for data visualization. Amiodarone PK parameters area under the curve (AUC 0‐24) and maximum plasma concentration (C max) in pediatric patients with or without ECMO, were calculated using noncompartmental method using PKanalix® (version 2021R1, Antony, France, Lixoft SAS, 2021. https://www.frankenthalerfoundation.org The area under the curve was calculated using the Linear Trapezoidal Linear method of PKanalix®. We used the software Plot Digitizer® (version 2.6.8) to extract the concentration versus time data from published PK papers used in model evaluation.
The clinical studies used for the adult PBPK model include two datasets with i.v administration: Cushing et al. – single 150 mg infused over 10 min, Ujhelyi et al. – 5 mg/kg infused over 15 min, and three datasets with oral administration: Meng et al. – 600 mg tablet, Teng et al. – 600 mg tablet, Andreasen et al. – 400 mg tablet. The datasets were selected based on the route of administration, study participant demographics, clinical characteristics, and bioanalytical method used (Table 1). The individual datafiles of each extracted study are provided as Data S1. The PBPK model was verified with oral datasets because our pediatric observed data included patients receiving amiodarone by both i.v and oral routes of administration.
We evaluated the amiodarone pediatric PBPK model using the clinical PK data obtained from the PK samples collected from the Pediatric Trials Network (PTN)‐sponsored Pharmacokinetics of Understudied Drugs Administered to Children per Standard of Care (POPS) trial. POPS (ClinicalTrials.gov: NCT01431326; protocol‐NICHD‐2011‐POP01) is a multicenter, prospective, PK, and safety study of understudied drugs administered to children (<21 years old) per standard of care. All dosing was at the discretion of the treating caregiver and samples specifically designated for PK analysis were collected opportunistically at the time of routine lab draws. The completely deidentified PTN PK analysis dataset was accessed through the Eunice Kennedy Shriver Data and Specimen Hub (DASH). The PTN PK analysis dataset was generated and formatted by the EMMES Corporation, MD, USA, the PTN Data Coordinating Center, and Duke Clinical Research Institute, NC, USA. Missing clinical data were imputed using the last value carried forward. For infants and children >120 days old, a gestational age of 40 weeks was imputed.
We evaluated the amiodarone pediatric ECMO PBPK model using two datasets: (1) clinical PK data from children on ECMO in the POPS study and (2) clinical PK data obtained from an opportunistic PK study enrolling children on ECMO (University of Utah, Institutional Review Board Number: 00138310) that was conducted at PCH, Salt Lake City, UT. This study enrolled children 0–18 years of age who were supported with ECMO and prescribed one of the study drugs of interest per standard of care. Dosing was at the discretion of the treating provider, and PK samples were collected at times of routine lab draws. For patients in both datasets, the ECMO circuit did not include hemofilter.
The plasma concentrations of amiodarone in the POPS study were quantified by the PTN central laboratory (OpAns LLC, Durham, NC) using a validated high‐performance liquid chromatography–tandem mass spectrometry (LC–MS/MS) assay. The plasma concentrations of amiodarone in the opportunistic study at PCH were quantified by ARUP Laboratories (ARUP, Salt Lake City, UT) using a validated LC–MS/MS assay. The linearity range was between 5–5000 ng/mL for PTN central laboratory assay and 300–6000 ng/mL for ARUP assay. The lower limit of quantification of 5 ng/mL for amiodarone for PTN central laboratory assay. The method validation parameters related to accuracy and precision were within the limits outlined by FDA guidance.
We used the standard whole‐body 15‐organ PBPK model implemented in PK Sim®. The physicochemical parameters and ADME data for amiodarone were collected from the literature (Table 2). For model predictions, we created a virtual adult population (N = 1000) using the demographic (sex, age, and weight) distribution reported in the adult PK studies of amiodarone selected from the development and verification datasets. We incorporated a 20% standard deviation based on literature to the fraction of the periportal zone to impart variability in the virtual pediatric populations. We also included the inter‐individual variability for fluid recirculation flow rates and plasma protein scale factor in the model based on literature.
For the i.v. amiodarone PBPK model, the cellular permeabilities for the barriers between interstitial and intracellular space for amiodarone were calculated from the physicochemical properties using the PK Sim® standard method. Amiodarone is mainly eliminated by hepatic metabolism, which was considered the only elimination mechanism. The total hepatic clearance value of amiodarone was obtained from the literature and added to the model. The total hepatic clearance value was partitioned a priori, with 50% of the total clearance attributed to CYP3A4 and 50% attributed to CYP2C8 based on literature‐reported enzyme activities. We compared the i.v amiodarone PBPK model predictions with the observed data reported in PK studies by Cushing et al, and Ujhelyi et al. We optimized the model parameters lipophilicity and unbound fraction of amiodarone using the parameter identification toolkit available in PK Sim®.
We developed the oral amiodarone PBPK model based on the i.v. PBPK model. The physicochemical and physiological parameters for the oral amiodarone PBPK model were retained from the i.v model. The oral amiodarone PBPK model performance was evaluated by comparing the model predictions with the observed data reported in PK studies by Meng et al, Teng et al, and Andreasen et al.
We evaluated the adult population PBPK model by calculating the AFE of observed amiodarone plasma concentration versus time data to the simulated data. We considered the model final if the mean ± SD of the simulated data captured >80% of the observed data within 90% prediction interval and the AFE values between observed and predicted plasma concentration were between 0.5 and 2.
We used the pre‐established age‐dependent algorithms in PK Sim® to generate the anatomical and physiological parameters, including body weight, height, organ weights and volumes, blood and lymph flows, cardiac output, total body water, lipid and protein concentrations.
We obtained the total hepatic clearance of amiodarone from the published adult clinical PK study. The total hepatic clearance was partitioned between CYP3A4 and CYP2C8 enzymes. We used the default setting of hepatic CYP3A4 and CYP2C8 ontogeny in PK Sim® for the pediatric population. Enzyme activity of CYP3A4 is, on average, 12% of the adult value at term, increases to 80% by the age of 1.3 years, and reaches adult activity by 5 years. Enzyme activity of CYP2C8 was observed in early infants with an approximately 8‐fold difference in CYP2C8 levels between individuals less than 35 days postnatal age and >35 days to 18 years.
The PK Sim® default values for gastric emptying time (GET) and small intestinal tran
