Developing Pediatric PBPK Models

The adopted modeling strategy is shown in Fig. 2. An adult model was first developed for the IV application, as this allows for the kinetics of drug disposition to be simulated in the absence of the complexities of the absorption process. Thus, the best set of input parameters, the suitable distribution model, and the most appropriate clearance that collectively gave the best visual description of the observed data used at this stage, were assigned. For the oral application, parameters values from the previous step were kept plus the values of additional parameters that control and influence drug absorption, such as intestinal permeability, GET, and SITT. In the previously mentioned steps (i.e., model building), only one fifth (n = 5) of the collected adult data set was used, whereas the remaining data (n = 22) were used later for a subsequent model verification. The adult model was slightly refined (logP and CL inputs) before the end model evaluation. The final adult model was then scaled down to children, taking into consideration the age dependencies of anatomical and physiological processes/parameters and the ontogeny of clearance pathways, which are already integrated into the modeling software, to predict pediatric sotalol exposure (see also, Clearance Scaling).Fig. 2

Schematic workflow of the developed PBPK models

The comparison of model results with observed data was based on simulations of virtual populations, where the main results of these simulations are concentration-time profiles. In adults, each virtual population consisted of 100 virtual subjects having the same age range, race, gender composition, and dosing as their respective real population. The resulting mean plasma concentrations were then compared with the mean observed concentrations for model evaluation. Population simulations performed with a higher number of virtual subjects (n = 1,000) did not produce any significant difference from the previous ones (using 100 replicates), and did not influence any differences seen between the results of both models. In children, a similar approach was used by performing a population simulation of 100 virtual children each with the same age, race, gender, and dosing information of a real child; however, the resulting median plasma concentrations were used along with the individual observed concentrations in the model evaluation. For all previous simulations, variability ranges for CL, GET, and SITT were assigned to account for the interindividual variability. These values were either set by the software, as in software 1—CL: mean value ± 30% CV; GET: mean value ± 38% CV; SITT: Weibull distribution around the mean value with α = 2.92 and β = 4.04; or were assigned manually based on a comprehensive literature search as in software 2: lognormal distribution with geometric standard deviation of 1.3 for CL, GET: uniform distribution of 0.2–1.9 h in adults (49 (link),50 (link)) and 0.2–2.1 h in children (51 (link)–54 (link)), and SITT: normal distribution with a mean value of 4 ± 1 h in both (49 (link),55 (link)).