Nadja Haas 1, Fenja Klima 2,3, Jonas Huhn 4, Juliane Staudinger 5, Georg Hempel 5, Oliver Scherf-Clavel 4, Charlotte Kloft 2,3, Ulrich Jaehde 1
1 Department of Clinical Pharmacy, Institute of Pharmaceutical Sciences, University of Bonn (Bonn, Germany), 2 Freie Universität Berlin, Institute of Pharmacy, Department of Clinical Pharmacy and Biochemistry (Berlin, Germany), 3 Graduate Research Training Program PharMetrX (Berlin/ Potsdam, Germany), 4 Department of Pharmacy, Faculty of Chemistry and Pharmacy, Ludwigs-Maximilians University of Munich (Munich, Germany), 5 Institute of Pharmaceutical and Medicinal Chemistry, Clinical Pharmacy, University of Muenster (Muenster, Germany)
Objective: Adverse drug events (ADEs) are a substantial burden for patients undergoing antitumor therapy and, in the worst case, may lead to dose reduction, treatment interruption, or discontinuation. In the assessment of ADEs, the patient perspective is gaining increasing importance, for example through the Patient-Reported Outcomes version of the Common Terminology Criteria for Adverse Events (PRO-CTCAE) developed by the National Cancer Institute [1]. It is particularly important to recognize ADEs early on in therapy with oral anticancer drugs, such as tyrosine kinase inhibitors, because patients manage their therapy independently and there are fewer doctor visits. Predicting patient-reported symptom burden may help to optimize therapeutic and supportive interventions in anticancer treatment in a more individualized and targeted manner. Minimal Continuous-time Markov models (mCTMMs), in combination with pharmacokinetic models, are particularly well suited to characterize the relationship between drug exposure and toxicity, because their simple model structure enables them to estimate precise parameters even with limited data. [2]. This project aimed to develop Markov models to describe the time course of selected patient-reported ADEs in patients with renal cell carcinoma receiving either axitinib or cabozantinib treatment within the ON-TARGET study [3].
Methods: Patients who received either axitinib at a dosage of 1–10 mg or cabozantinib at a dosage of 20–60 mg were included in the study. Plasma samples were taken at each visit and analyzed by validated LC-MS/MS assays. Moreover, patients were asked to complete a monthly diary documenting their daily drug intake and a PRO-CTCAE questionnaire covering 14 of the most common adverse events. Study centers documented additional information such as co-medication and comorbidity. A database was created in which the plasma samples collected were matched to the diaries and the composite scores for the PRO-CTCAEs were determined [4]. Clearance (CL) and volume of distribution were determined for each patient using post-hoc estimation, and the area under the concentration-time curve (AUC) was calculated for each visit. MCTMMs were developed for all 14 symptoms, e.g. hand-foot syndrome, and covariate influences were investigated. Age, sex, CL, and AUC were selected among others as potential covariates. The influence of these covariates was tested on the mCTMM parameter axis intercept α1 . Covariates were included in the model structure if they led to a significant reduction in the objective function value (OFV) (p-value ≤ 0.01).
Results: Between February 2021 and December 2024, 40 patients were included in the study. 392 questionnaires, 181 for axitinib and 211 cabozantinib, were submitted. The median time between two questionnaires was 28 days. For the symptom hand-foot syndrome, four patients receiving axitinib therapy and seven patients receiving cabozantinib therapy reported severity grades ≥ 2. In the mCTMM of hand-foot syndrome α1 was lower and the mean equilibration time higher for axitinib compared to cabozantinib . This shows that patients treated with axitinib were less likely to transition between severity grades and remained longer at the same grade. The likelihood of experiencing severity levels greater than 1 was also lower than with cabozantinib. In the covariate testing, CL and AUC were significant in the models when included separately. In both models, the inclusion of CL resulted in a lower OFV and higher parameter precision than including AUC. The inclusion of additional covariates such as age or sex did not lead to a significant reduction in OFV.
Conclusion: The development of mCTMMs in combination with PRO-CTCAE symptom scores was shown to be feasible for axitinib and cabozantinib. The hand-foot syndrome model was improved by the inclusion of covariates describing drug exposure, showing a the relationship between exposure levels and the development of hand-foot syndrome. This is a first step towards using Markov models to consider ADEs in precision dosing. Including biomarkers such as sVEGFR2 as covariates and evaluating alternative model structures, such as integrating a second MET or examining the effects of covariates on MET, could improve model performance and provide additional insights to enable more accurate patient-specific prediction of ADE progression.
References:
[1] Minasian LM et al. Patient Relat Outcome Meas (2022) 13, 249-258
[2] Schindler E et al. AAPS J (2017), 19, 1424-1435
[3] Mc Laughlin AM et al. Cancers (2021) 13, 6281
[4] Basch E et al. Clin Trials (2021) 18, 104-114
Reference: PAGE 34 (2026) Abstr 12293 [www.page-meeting.org/?abstract=12293]
Poster: Drug/Disease Modelling - Safety