Gudrun Würthwein (1), Claudia Lanvers-Kaminsky (1), Georg Hempel (2), Martin Schrappe (3), Mats O. Karlsson (4), Joachim Boos (1)
(1) University Hospital Muenster, Paediatric Haematology and Oncology, Albert-Schweitzer-Campus 1, Building A1, 48149 Muenster, Germany, (2) Department of Pharmaceutical and Medical Chemistry - Clinical Pharmacy, Corrensstraße 48, 48149 Muenster, Germany, (3) Department of Pediatrics, Christian-Albrechts-University Kiel and University Medical Center Schleswig-Holstein, Schwanenweg 20, 24105 Kiel, Germany, (4) Department of Pharmaceutical Biosciences, Uppsala University, P.O. Box 591, 75124 Uppsala, Sweden
Objectives:
The pharmacokinetics of the polyethylene glycol (PEG)-conjugated asparaginase Oncaspar® is characterized by an increase in elimination over time. An empirical transit compartment model was implemented to describe this time-dependency in pharmacokinetics (PK) [1].
Focus of our analyses were:
- to build up a covariate model
- to account for the significant differences in drug exposure between different parts of the protocol
- test the accuracy of the model (based on the PK model, individual drug exposure over the whole asparaginase therapy has to be predicted in order to correlate these data with clinical outcome and toxicity parameters).
Methods:
In paediatric acute lymphoblastic leukemia therapy (AIEOP-BFM ALL 2009; registered at www.clinicaltrials.gov as NCT0111744), two administrations of Oncaspar® (2500 U/m2 intravenously) in induction phase (14 days interval) and one single administration in reinduction were followed by weekly monitoring of asparaginase activity. For this analysis, the starting model was the previously published non-linear mixed-effects model [1]. Besides age and sex as potential covariates, the marked difference in drug exposure between induction and reinduction was a major focus during model development (median asparaginase activity: 1. administration in induction, day 7: 882 U/L, day 14: 534 U/L; reinduction: day 7: 1445 U/L, day 14: 748 U/L).
Predictivity of the model was tested for single observations as well as for the derived PK-parameters AUC (based on integration of the model) and time over 100, 250, 500 and 1000 U/L, resp.: for patients, where all 6 drug-monitoring samples were available, 2-4 samples were excluded and then the accuracy of the predicted metrics were calculated.
Results:
The previously published transit model included 14 compartments with 4 PK-parameters: V=volume of distribution, CLP=clearance from each compartment, CLE=additional clearance from the last compartment, Qtr=constant inter-compartmental clearance [1]. In a first step, the model could be simplified by replacing CLE with the transit-compartment clearance Qtr.
Inclusion of covariates on CLP (68% decrease at reinduction) and Qtr (60% increase at reinduction) were best to account for the differences in PK between induction and reinduction (dOFV= -1692). However, interpretation of these pronounced changes of PK-parameters on a pharmacological level seems to be difficult: CLP might be interpreted in terms of elimination by monocytes and macrophages; thus, lower elimination rates in reinduction seem to be plausible. However, the transit-clearance Qtr might reflect the rate of de-PEGylation of the molecule in vivo; here, changes observed in different protocol parts can only be postulated. Other pharmacologically more plausible attempts to model these differences resulted in less pronounced improvement of the model. Unsymmetrical distribution of ETAs for interoccasion variabilities on CLP and V as well as VPCs stratified on each Oncaspar® administration further indicated a slight underprediction of observed data for the 2nd induction administration. Thus, additional covariates on CLP and V for the 2nd induction administration were included in the model (dOFV= -214). Sex (dOFV= -22.1) and age (dOFV= -24.5) as additional covariates further improved the model.
The accuracy for AUC or time over predefined asparaginase activities were superior to the accuracy for individual observations: The percentage of individuals within a 10% error range from true parameters ranging from 60-100 % for AUC, 85-100 % for time over 100 U/L and 25-31 % for individual observations for the different scenarios tested (percentage of individuals within a 20 % error range: 87-100 % for AUC, 96-100 % for time over 100 U/L and 47-52 % for individual observations).
Conclusions:
Besides covariates for induction vs. reinduction, the PK model accounts for slight accumulation after subsequent Oncaspar® administrations. Test of accuracy demonstrates that derived PK-parameters AUC or time over predefined asparaginase activities are more promising parameters than predictions of individual asparaginase activities for further PK-PD correlations.
References:
[1]Würthwein G, Lanvers-Kaminsky C, Hempel G, Gastine S, Möricke A, Schrappe M, Karlsson MO, Boos J. Population Pharmacokinetics to Model the Time-Varying Clearance of the PEGylated Asparaginase Oncaspar® in Children with Acute Lymphoblastic Leukemia. Eur J Drug Metab Pharmacokinet 2017;42:955–63.
Reference: PAGE 27 (2018) Abstr 8614 [www.page-meeting.org/?abstract=8614]
Poster: Drug/Disease Modelling - Oncology