A pharmacometric framework for dose individualisation of sunitinib in GIST
M. Centanni, S.M. Krishnan, L.E. Friberg
Department of Pharmaceutical Biosciences, Uppsala University, Uppsala, Sweden.
Objectives: The rise of targeted cancer therapies has revolutionized the field of oncology, however, uncertainties remain regarding their best dosing regimen and individualization approaches with respect to adverse events and survival. For sunitinib, various strategies have been suggested, such as therapeutic drug monitoring (TDM) based on trough concentration measurements and toxicity-adjusted dosing (TAD) [1,2]. In addition, soluble biomarkers or, alternatively, neutrophil counts and blood pressure, have been related to overall survival (OS) in gastrointestinal stromal tumour (GIST) [2,3]. During the last years, TDM has gained momentum as a method to homogenize sunitinib exposure in clinical practice [4]. However, this approach does not account for the large inter-individual variability in the susceptibility of efficacy and safety endpoints [5]. Moreover, plasma concentration measurements for TDM may not be feasible in each country or treatment center due to practical and economical constraints. We explored an alternative, model-based, approach to increase OS in sunitinib-treated GIST patients, wherein dose-adjustments depend on pharmacodynamic biomarkers, such as adverse events (blood pressure and neutrophil counts), or soluble vascular endothelial growth factor receptor (sVEGFR)-3.
Methods: A previously developed pharmacodynamic framework describing the relations between sunitinib exposure, adverse events (hand-foot syndrome (HFS), fatigue, hypertension and neutrophil counts), sVEGFR-3 and OS [2,3] was further extended by a population pharmacokinetic (popPK) model [6]. The final framework of nine models was translated into mrgsolve [7] in order to evaluate dosing strategies by simulations. Intolerable toxicities were defined as ≥Grade 2 for HFS and thrombocytopenia and ≥Grade 3 for the remaining adverse events, as defined by the Common Terminology Criteria for Adverse Events v5.0 (CTCAE). Initial simulations were performed with fixed dosing regimens (4/2, 2/1 and continuous daily dosing [5]) and the best schedule, in terms of OS and adverse events, was selected as a base scenario. TDM, adverse event and sVEGFR-3-based dose adjustments were simulated according to an existing TDM schedule proposed by Lankheet et al. [4], with a discrete number of possible sunitinib doses (0 to 75 mg, by 12.5 mg increments). Finally, the accuracies of Bayesian maximum a-posteriori estimations were determined for various samplings schedules and the advantage of a model-based dosing algorithm for treatment optimization was explored.
Results: All models (including Markov and time-to-event models) were successfully implemented in mrgsolve. The continuous dosing schedule was found to give the best balance between AEs and OS, and therefore selected as a base scenario. AE-based dose adaptations increased median OS as compared to a fixed dose schedule (24.1 vs. 20.0 months; hazard ratio [HR] 0.90) and TDM-based dose adjustments (24.1 vs. 19.7 months; HR 0.81) without markedly raising the risk of intolerable toxicities. Similarly, sVEGFR-3-based dose adaptations increased median OS compared to fixed dosing (25.5 versus 21.7; HR 0.90) and TDM (25.5 versus 21.2 months; HR 0.77). Model-based predictions of blood pressure, sVEGFR-3 and neutrophils were accurate (80-125% of true value) for 28.5%, 64.6% and 73.5% of patients after three observations (routine sample at day 0, 15 and 29) and 35.1%, 76% and 85.6 % of patients after daily observations (day 0-29).
Conclusions: Biomarker changes were demonstrated to provide viable guidance for dose individualisation of sunitinib in GIST by increasing OS. AEs or sVEGFR-3 may therefore pose valuable alternatives to drug concentrations (TDM). Neutrophil-based dose adaptations may however be preferred over sVEGFR-3, as neutrophils are readily measured in clinical practice and will not necessitate additional hospital visits or expenses. To our knowledge, this is the first framework-based Bayesian decision support tool for dose-individualisation that includes drug concentrations, soluble biomarker, adverse effects and OS. An external validation with clinical data could help to further confirm our proposal of biomarker-based dose adjustments.
Acknowledgement:
This work was supported by the Swedish Cancer Society and the European University Consortium for Pharmaceutical Sciences
References:
[1] Sabanathan et al. Cancer Chemother Pharmacol (2017);80(2):385-393.
[2] Hansson et al. CPT PSP (2013);2(12):e85. [3] Hansson et al. CPT PSP (2013);2(11):e84.
[4] Lankheet et al. Br J Cancer (2014);110(10):2441-2449. [5] Khosravan et al. Clin Pharmacokinet (2016);55(10):1251-1269.
[6] Yu et al. Br J Clin Pharmacol (2015);79(5):809-819.
[7] Baron et al. (2017). mrgsolve: Simulate from ODE-Based Population PK/PD and Systems Pharmacology Models.
