Pediatric physiologically based pharmacokinetic (PBPK) models facilitate the prediction of PK parameters in children under specific exposure conditions. Pharmacokinetic outcomes are highly sensitive to fraction unbound in plasma (fup) as incorporated into PBPK models. Rarely is fup in children (fup_child) experimentally derived and prediction is based upon fup in adults (fup_adult) as well as a ratio of plasma protein concentrations between children and adults. The objectives were to (i) evaluate protein concentration vs. age profile derived from ontogeny models, (ii) assess predictive performances of fup ontogeny models, and (iii) determine overall uncertainty in fup_child prediction resulting from a combination of quantitative structure–property relationship (QSPR) model and ontogeny models. The plasma albumin and alpha-acid glycoprotein (AAG) concentration data for pediatrics and fup_child and fup_adult data were obtained from literature. The protein concentration vs. age profile derived from ontogeny models were compared to observed levels. Fup_child values were calculated according to ontogeny models using both observed and QSPR-predicted fup_adult as inputs and predictive performances of ontogeny models assessed by comparing predicted fup_child to observed values. Protein concentrations vs. age profiles derived from non-linear equations were more congruent with observed albumin levels than linear or step-wise models. When observed fup_adult values were used as input, the fup_child data were under-predicted with average fold error (AFE) amounts ranging 0.79–0.81 and 0.77–0.97 for albumin and AAG ontogeny models, respectively. When QSPR-predicted fup_adult values were used as input, AFE of fup_child ranged 1.2–1.35 and 0.98–1.2 for albumin and AAG models, respectively. The choice of ontogeny model with respect to prediction accuracy is more important for AAG, highly bound compounds and infants. For these compounds and scenarios, experimental determination of fup_child for inclusion into a pediatric PBPK model is necessary to have confidence in PBPK model outputs.

By Yejin Esther Yun & Andrea N. Edginton