Eur J Clin Pharmacol DOI 10.1007/s00228-014-1692-z
PHARMACOKINETICS AND DISPOSITION
Population pharmacokinetics of intravenous busulfan in children: revised body weight-dependent NONMEM® model to optimize dosing Christian Diestelhorst & Joachim Boos & Jeannine S. McCune & Georg Hempel
Received: 12 November 2013 / Accepted: 24 April 2014 # Springer-Verlag Berlin Heidelberg 2014
Abstract Purpose We developed a new population pharmacokinetic (PopPK) model for intravenous (i.v.) busulfan in children to evaluate the optimal method to personalize its dosing without concentration-time data. Methods PopPK analyses were done with NONMEM® 7.2. First, a model from Trame et al. was evaluated using an external dataset consisting of 24 children. Second, a revised model was built in a separate dataset of 82 children. Model evaluation was performed by using a standardized visual predictive check (SVPC) procedure and a bootstrap analysis (internal evaluation) and by comparison to an external dataset (external validation). Results The final model included body surface area (BSA) as an exponential function on volume of distribution (V) and actual body weight (ABW) as an allometric function on clearance (CL). The dosing nomogram for every 6 h administration
Electronic supplementary material The online version of this article (doi:10.1007/s00228-014-1692-z) contains supplementary material, which is available to authorized users. C. Diestelhorst : G. Hempel (*) Department of Pharmaceutical and Medical Chemistry - Clinical Pharmacy, University of Münster, Corrensstr. 48, 48149 Münster, Germany e-mail:
[email protected] J. Boos Department of Paediatric Haematology and Oncology, University Children’s Hospital Münster, Albert-Schweitzer-Campus 1, 48149 Münster, Germany J. S. McCune Fred Hutchinson Cancer Research Center, University of Washington, Box 357630, 1959 NE Pacific St, Seattle, WA, USA
derived from the final model is: dose[mg]=target AUC[mg× h/L]×3.04L/h×(ABW/16.1)0.797. Compared to other dosing strategies, differences were observed for the very small and obese patients. Conclusions We revised our prior dosing nomogram after validation in a separate cohort of children. This dosing nomogram can be used to personalize i.v. busulfan doses without concentration-time data, but an additional prospective evaluation in the very small and obese children is needed. Keywords Busulfan . Pharmacokinetics . Cancer . Modeling
Introduction Busulfan is widely used for conditioning prior to hematopoietic stem cell transplantation (HSCT). The plasma exposure of busulfan, measured as area under the plasma concentration time curve (AUC) or average steady-state concentration (Css, calculated as AUC/dosing frequency), is associated with rejection, relapse, and toxicity in HSCT recipients [1, 2]. After intravenous (i.v.) busulfan administration, children typically experience greater pharmacokinetic (PK), and thus AUC, variability than that in adults [3, 4]. A minority of children (24.3 %) achieve their target busulfan AUC in recent clinical practice of i.v. busulfan dosing [5]. Simulations from population pharmacokinetic modeling indicated that a greater percentage of children would achieve the target AUC with the European Medicines Agency (EMA) dosing nomogram compared to the food and drug administration and the Children’s Oncology Group dosing nomograms [6, 7]. However, various small studies have shown variable success of the EMA dosing nomogram [8] to achieve the desired target exposure without concentration-time data [3, 4, 9, 10]. Therefore, there has been
Eur J Clin Pharmacol
considerable interest in developing population pharmacokinetic (PopPK) models to optimize i.v. busulfan dosing in children with the goal of improving clinical outcomes by more rapidly achieving the target AUC after the initial i.v. busulfan dose (i.e., dose 1). The aim of this analysis was to external validate the two models from our working group from Trame et al. [10] in order to investigate a possible bias and, building on this, to develop a revised PopPK model for busulfan with the aim to optimize dosing of i.v. busulfan in children. Subsequently, dosing based on the final PopPK model was compared with dosing strategies suggested by other groups in order to give an overview about differences of different dosing strategies.
Materials and methods Validation of two previously published models by Trame et al. [10] The Trame models (i.e., body surface area and allometric body weight model) were validated with an external dataset, which consisted of 24 children who received i.v. busulfan in a oncedaily schedule (q24hrs) at the University Medical Center Utrecht, The Netherlands. The corresponding analytical methods were previously described by Bartelink et al. [11]. All patient characteristics are described in Table 1. Population parameters were fixed based on final model results from Trame et al. [10]. Afterwards, each patient was simulated once, and the resulting predicted population plasma concentrations (PRED) were compared to the observed plasma concentration (OBS) by using the mean prediction error (MPE) calculated according to Eq. 1: X MPE ¼
ðPEÞ
N
ð1Þ
where N is the total number of patients in the dataset and PE is the prediction error according to Eq. 2:
PE½% ¼ 100
ðPRED−OBSÞ OBS
ð2Þ
To assess the possible bias caused by data derived from an investigation after administration of oral busulfan, we excluded the oral data from the development dataset, reestimated the population parameter with the remaining i.v. patients (n=40) of the development dataset, and repeated the simulation of the validation dataset.
Development of the revised NONMEM® model Dataset The model from Trame et al. was developed on a mix of oral and i.v. data with the minority data coming from the i.v. series (40/94 patients). It is well known that oral busulfan shows a large variability in drug absorption and bioavailability [12]. To evaluate the influence of the data after oral administration, we only used the i.v. data from the study of Trame et al. [10], and added 36 patients who received i.v. busulfan every 6 h (q6hrs) under the auspices of a protocol from the Fred Hutchinson Cancer Research Center (Seattle, WA, USA). In addition, another six children who received i.v. busulfan between 2008 and 2011 at the University Hospital Münster were collected through routine therapeutic drug monitoring (TDM). In total, this resulted in a development dataset consisting of 82 children, i.e., 46 from the Münster cohort and 36 from the Seattle cohort (Table 1). Model building Model building was done by using nonlinear mixed effect modeling with NONMEM® [version 7.2, ICON Development Solutions, Ellicott City, MD, USA], and the decision between each model building step was based on widely used PopPK modeling criteria [13]. More details about the model building steps are described in Supplementary text S1. Evaluation of the revised NONMEM® model Internal evaluation Besides evaluation steps during model development, the predictive performance of the final PopPK model was assessed by using a visual predictive check (VPC) procedure. To regard the described heterogeneity from the development dataset, a standardized visual predictive check (SVPC) was performed with 1,000 runs, and the resulting Pij percentiles of each observation in relation to its 1,000 simulated observations were plotted versus the time after the initial dose of busulfan [14]. Based on the findings from Wang et al. [14], the resulting Pij values (i.e., the percentiles of each observation in the dataset in relation to its 1,000 simulated observations derived from the final model) should be uniformly distributed between 0 and 1. Furthermore, a bootstrap analysis was carried out, also with 1,000 runs. The model was considered as unbiased, if the final parameter estimates from the development dataset fell within the 95 % confidence interval coming from the bootstrap analysis.
Eur J Clin Pharmacol Table 1 Patients’ characteristics (shown as mean (range))
Site Dataset Subjects (n) Samples (n) Age (years) Size (cm) Weight (kg) BMI (kg/m2) BSA (m2) Dosing frequency Infusion time (hours) Dose (mg/kg) Actual body weight (kg)
