Constructing a representative in-silico population for pediatric simulations: application to HIV positive African children.

AIMS Simulations are an essential tool for investigating scenarios in pharmacokinetics-pharmacodynamics. The models used during simulation often include the effect of highly correlated covariates such as weight, height, and sex and for children also age, which complicates the construction of an in-silico population. For this reason, a suitable and representative patient population is crucial for the simulations to produce meaningful results. For simulation in pediatric patients, international growth charts from the WHO and CDC provide a reference, but these may not always be representative for specific populations, such as malnourished children with HIV, or acutely unwell children. METHODS We present a workflow to construct a virtual pediatric patient population using WHO and CDC growth charts, suggest piecewise linear functions to adjust the median of the growth charts by sex and age, and suggest visual diagnostics to compare with the target population. We applied this workflow in a population of 1,206 HIV-positive African children, consisting of 19,742 observations with weight ranging from 3.8 to 79.7 kg, height from 55.5 to 180 cm, and an age between 0.40 and 18 years. RESULTS Before adjustment, the WHO and CDC charts produced weights and heights higher compared to the observed data. After applying our methodology, we could simulate weight, height, sex and age combinations in good agreement with the observed data. CONCLUSION The methodology presented here is flexible and may be applied to other scenarios where WHO and CDC growth standards might not be appropriate, in addition we provide R scripts and a large ready-to-use pediatric population.