Daniel Kaschek1, Benjamin Guiastrennec1, Henning Schmidt1, Rajendra Pratap Singh2
1IntiQuan GmbH, Basel, Switzerland; 2TEVA Pharmaceuticals, West Chester, PA, USA.
Objectives: Accurate determination of individual pharmacokinetic (PK) exposure parameters can be critical in the evaluation of a drug’s safety and efficacy profile. Non-linear mixed-effect modeling approaches have been used to leverage available data and better assess individual PK parameters in exposure-response trials. Unlike adult trials, the availability of PK sampling can be very sparse in pediatric exposure-response trials due to ethical constraints, and in turn can lead to unreliable estimates of PK exposure parameters. Here, we assess whether combining sparse pediatric PK profiles (2-3 samples per subject) with richly sampled adult data enables accurate estimation of individual PK exposure parameters in pediatric patients.
Methods: An adult population PK model developed on rich adult PK data was used to support the parameter estimation. The evaluation of individual PK exposure parameter predicted accuracy was done through a simulation-estimation approach. New pediatric PK samples were simulated for Nsubj patients with the original study design and new individual parameters, considering patient covariates and inter-individual variability. Parameters (population and individual) of the population PK model were re-estimated based on the combined adult and simulated pediatric dataset. Given the known individual parameter values, the precision and accuracy of individual parameter estimates was assessed. In particular, the derived PK exposure parameters, i.e., the maximal (Cmax,ss), minimal (Cmin,ss) and average (Cavg,ss) concentrations at steady state, were considered. Using an acceptability thresholds of log(1.25) as motivated by the definition of bioequivalence, success rates of parameter estimation were derived on both population and individual levels. The approach was repeated Ntrials times to obtain reliable statistics on the success rates.
Results: Distributions of estimated individual PK endpoints Cmax,ss and Cavg,ss were generally close to the reference distributions of known PK endpoints. The systematic shift between the distributions was below the acceptable threshold for all trials, corresponding to a 100% success rate. For Cmin,ss, the shift was considerably stronger (i.e., less than 15% of successful trials). Knowing the individual parameters, the deviance between estimated and known individual parameters was computed. For Cmax,ss and Cavg,ss about 75% of all subjects were within the acceptable range. For Cmin,ss still 50% of all subjects had acceptable individual estimates.
Conclusions: The analysis showed that individual pediatric PK endpoints Cmax,ss and Cavg,ss were generally well-characterized both on the population and the individual levels. The high accuracy and precision of estimated PK endpoints from just 2 observed samples was enabled by the combination of sparse pediatric data with rich adult data. The characterization of Cmin,ss was worse. Despite the rich adult data, 2 observed PK samples were not enough to yield satisfying estimates. The approach presented herein could be used to optimize the pediatric trial design, plan observed sampling time points, and determine the minimal number of samples to obtain satisfying individual PK endpoint estimates.
Reference: PAGE 29 (2021) Abstr 9813 [www.page-meeting.org/?abstract=9813]
Poster: Drug/Disease Modelling - Paediatrics