Nadja Haas (1), Elli Husso (1), Eduard Schmulenson (1), Mathias Hoiczyk (2), Yon-Dschun Ko (3), Lothar Müller (4), Henning Schulze-Bergkamen (2), Ulrike Schwinger (5), Ulrich Jaehde (1)
(1) Department of Clinical Pharmacy, Institute of Pharmacy, University of Bonn, Germany, (2) Medical Clinic II, Marien Hospital Wesel gGmbH, Germany, (3) Department of Internal Medicine, Hematology and Oncology, Johanniter Krankenhaus Bonn, Germany, (4) Study Centrum Unter Ems, Practice for Oncology and Hematology, Leer, Germany, (5) Practice for Oncology, Medical care center at the Robert Bosch Krankenhaus, Stuttgart, Germany
Introduction: Adverse drug events (ADE) represent a great burden for patients in the treatment with antitumor drugs and can in the worst case be therapy-limiting. Regarding the evaluation of ADE, the patient’s perspective is becoming increasingly important, e.g. through the Patient-Reported Outcomes Version of the Common Terminology Criteria for Adverse Events (PRO-CTCAE) developed by the National Cancer Institute [1]. Predictions of patient-reported symptom burden may contribute to an individual and targeted optimization of therapeutic and supportive measures in anticancer treatment.
Objective: Development of Markov models to describe the time course of selected patient-reported ADE in patients treated with fluorouracil (5-FU).
Methods: Markov models, particularly minimal continuous-time Markov models (mCTMMs), continuous-time Markov models (CTMMs) and discrete-time Markov models (DTMMs) are well suited to describe the relationship between drug exposure, covariates, and toxicity in combination with pharmacokinetic models [2]. Schmulenson et al. developed an mCTMM to predict hand-foot syndrome in capecitabine-treated patients, revealing a correlation between the daily dose of capecitabine and the severity of hand-foot syndrome [3]. In this observational study (5-FU-PRO) colorectal cancer patients treated with 5-FU were recruited. The patients reported their symptom burden every seven days using a validated electronic questionnaire featuring selected PRO-CTCAE items [4]. Informations on treatment with 5-FU, comedications and concomitant diseases were documented by the study centers. The observation period lasted twelve weeks. mCTMMs were developed to predict the 16 measured PRO-CTCAE symptoms, e.g. diarrhea, pain, and nausea.
Results: Between November 2020 and December 2022, 42 patients were recruited for participation in the 5-FU-PRO study. For each of the 16 symptoms a mCTMM was developed. Covariates were identified in the selected model structure for seven PRO-CTCAE symptoms, such as diarrhea and pain. For the symptom diarrhea, inter-individual variability (IIV) of the logit baseline and two covariates were included in the final model: age and body-surface area (BSA). The probability of severe diarrhea decreased with increasing age and BSA. Both covariates had p-values <0.01 and the final model showed a reduction in objective function value (dOFV) of -110.912 with a p-value of 4.65 · 10-23 compared to the base model. Medication for anemia and medication for mental illness or sleep problems were identified as covariates for the symptom pain. The inclusion of the two covariates in the final model resulted in a dOFV of -131.849 with a p-value of 2.161 · 10-28 compared to the base model.
Conclusion: The development of mCTMMs in combination with PRO-CTCAE symptom scores was shown to be feasible. The models could be improved by identification and implementation of covariates. CTMMs and DTMMs are currently under development for each PRO-CTCAE symptom to test which of the three model types is best suited for PRO-CTCAE data sets.
[1] Minasian L et al. Patient Relat Outcome Mes, 13:249–258, 2022.
[2] Schindler E and Karlsson M. O. AAPS Journal, 19:1424–1435, 2017.
[3] Schmulenson E et al. Cancer Chemother Pharmacol, 86:435–444, 2020.
[4] Hagelstein V et al. Ann Oncol, 27:2294–2299, 2016.
Reference: PAGE 32 (2024) Abstr 11180 [www.page-meeting.org/?abstract=11180]
Poster: Drug/Disease Modelling - Safety