Population pharmacokinetics of treosulfan and development of a limited sampling strategy in children prior to hematopoietic stem cell transplantation

Abstract

Purpose: There is an increasing interest in use of treosulfan (TREO), a structural analogue of busulfan, as an agent in conditioning regimens prior to hematopoietic stem cell transplantation (HSCT), both in pediatric and adult populations. The aim of this study was to develop a population pharmacokinetic model and to establish limited sampling strategies (LSSs) enabling accurate estimation of exposure to this drug.

Methods: The study included 15 pediatric patients with malignant and non-malignant diseases, undergoing conditioning regimens prior to HSCT including TREO administered as a 1 h or 2 h infusion at daily doses of 10, 12, or 14 g/m². A population pharmacokinetic model was developed by means of non-linear mixed-effect modeling approach in Monolix® software. Multivariate regression analysis and Bayesian method were used to develop 2- and 3-point strategies for estimation of exposure to TREO.

Results: Pharmacokinetics of TREO was best described with a two-compartmental linear model with proportional residual error. Following sampling schedules allowed accurate estimation of exposure to TREO: 1 h and 6 h or 1 h, 2 h, and 6 h for a TREO dose 12 g/m² in a 1 h infusion, or at 2 h and 6 h or 2 h, 4 h, and 8 h for a TREO dose of 12 g/m² and 14 g/m² in a 2 h infusion.

Conclusions: A two-compartmental population pharmacokinetic model of TREO was developed and successfully used to establish 2- and 3-point LSSs for accurate and precise estimation of TREO AUC_0→∞.

Keywords: Area under curve; Hematopoietic stem cell transplantation; Infusions, intravenous; Population pharmacokinetics.

Fig. 1

Metabolic activation of treosulfan to its active mono- and diepoxide

Fig. 2

Spaghetti plot of treosulfan concentrations vs. time acquired for patients included in the study

Fig. 3

Goodness-of-fit plots for the final pharmacokinetic model. Panel A illustrates observed (OBS) vs. population predicted (PPRED) treosulfan concentrations and OBS vs. individual predicted (IPRED) treosulfan concentrations with an identity line and smooth. Panel B presents individual-weighted residuals (IWRES) vs. time and PPRED, population-weighted residuals (PWRES) vs. time and PPRED. Panel C presents normalized prediction distribution errors (NPDE) vs. time and PPRED with bold lines as 5th, median and 95th percentile of observed concentrations, light gray area as 50% interval of simulated data and dark gray areas as 95% intervals of simulated data

Fig. 4

Prediction-corrected visual predictive check (pcVPC) with dots as observed treosulfan concentrations, bold lines as 5th, median and 95th percentile of observed concentrations, light gray area as 50% interval of simulated data and dark gray areas as 95% intervals of simulated data