Burak Kürsad Günhan (1), Tianjing Ren (2), Yulia Vugmeyster (2), Sathej Gopalakrishnan (1), Wei Gao (2)
(1) Merck Healthcare KGaA, Darmstadt, Germany, (2) EMD Serono Research & Development Institute, Inc., Billerica, Massachusetts, USA
Introduction: Conventionally, single arm dose-expansion trials were widely used to select the therapeutic dose for oncology. Many gaps in the conventional practices have been discussed and Project Optimus was initiated by the FDA, highlighting the need of rational dose optimization, with the goal of reforming dose-optimization in oncology [1, 2, 3]. Collecting data from multiple dose levels, prior to the confirmatory trials, is critical to inform dose optimization strategy. As such, a randomized, parallel dose-response trial as well as backfill cohorts within the dose-escalation trial would provide invaluable data for characterizing dose/exposure-response relationship, and therefore inform optimal dose selection.
Use of exposure-response modelling better supports dose-selection (e.g. a more precise estimation of ED50) compared to dose-response analysis in different dose-finding trial settings, especially in the presence of high variabilities in dose-exposure relationship (i.e. large inter-individual variability in clearance) [4, 5]. Exposure-response modeling when impacted by confounding may result in false-positive exposure-response relationships [6, 7]. Dose-response analyses may not suffer from the same issue with certain trial designs (e.g. when patients are randomized into the dose levels in a randomized trial).
Objectives: We aim to investigate the potential benefits and drawbacks of exposure-response and dose-response analysis in the setting of a few typical oncology dose-optimization trials (i.e. 2 active dose levels and no placebo cohort), including the setting where dose-response analysis is not supported.
Methods: Trials were simulated following a dose-exposure-response modelling for different realistic dose-optimization oncology studies i) two non-placebo active dose levels and ii) two non-placebo dose levels with additional data from backfill cohorts from dose-escalation trial. In scenario i), dose-response is not supported and the dose selection often time relies on the pairwise comparison of the two dose levels. The dataset for the dose-exposure-response relationship was simulated as follows: The relationship between dose and exposure was characterized using a simple model where concentration was calculated by dividing the dosing rate by the clearance. The clearance is assumed to follow a log-normal distribution. Assuming binary responses, the exposure-response relationship was described by an Emax model with a logit link function in the data-generating process. Exposure-response model and dose-response models were compared for the estimation of different target doses including ED50 from the simulated data. Both models were implemented in fully Bayesian framework via Stan. As prior distributions, weakly informative priors were used [8].
Results: In the context of wide-range of doses, high inter-individual variability in CL (when true ED50 = 250 mg), exposure-response analysis results in more precise ED50 estimates (mean estimate 274.2 mg, CV: %46), compared to the dose-response analysis (mean estimate 277.7 mg, CV: %55.). The dose selected to be tested in the trial, relative to the location of ED50 or EC50, has an influence on the performance of the analysis methods. If the dose levels are well spaced around ED50 or EC50, the potency parameters are estimated more precisely (i.e. when true ED50 = 100 mg, a study design with dose levels of 100, 200, 300 result in the mean ED50 estimate of 103 mg (CV: %26), while a study design with dose levels of 25, 75, 100 resulted in the mean ED50 estimate of 111 mg (CV: %47)).
Conclusion: Our work highlights the advantages of exposure-response analysis for dose selection in oncology trials, particularly under restricted dose ranges and high variability in dose-exposure relationship. However, the risk of confounding necessitates a cautious and rational approach. We recommend designing the proper trials maximizing dose optimization information, leverage totality of evidence (e.g. biomarker data to inform EC50 or ED50) and apply relevant Model Informed Drug Development (MIDD) methods (e.g. dose-response and exposure-response analysis as complementary approaches) to enable rational oncology dose optimization.
References:
[1] U.S. Food and Drug Administration. Project Optimus: reforming the dose optimization and dose selection paradigm in oncology, https://www.fda.gov/about-fda/oncology-center-excellence/project-optimus (2022, accessed 05March2024).
[2] U.S. Food and Drug Administration. Optimizing the dosage of human prescription drugs and biological products for the treatment of oncologic diseases, https://www.fda.gov/regulatory-information/search-fda-guidance-documents/optimizing-dosage-human-prescription-drugs-and-biological-products-treatment-oncologic-diseases (2023, accessed 05March2024).
[3] Gao, W., Liu, J., Shtylla, B., Venkatakrishnan, K., Yin, D., Shah, M., Nicholas, T., & Cao, Y. (2023). Realizing the Promise of Project Optimus: Challenges and Emerging Opportunities for Dose Optimization in Oncology Drug Development. CPT: pharmacometrics & systems pharmacology, doi: 10.1002/psp4.13079.
[4] Hsu, C.-H. (2009), Evaluating potential benefits of dose–exposure–response modeling for dose finding. Pharmaceut. Statist., 8: 203-215. https://doi.org/10.1002/pst.392
[5] Berges, A., Chen, C. Dose finding by concentration-response versus dose-response: a simulation-based comparison. Eur J Clin Pharmacol 69, 1391–1399 (2013). https://doi.org/10.1007/s00228-013-1474-z
[6] Kawakatsu S, Bruno R, Kågedal M, et al. Confounding factors in exposure–response analyses and mitigation strategies for monoclonal antibodies in oncology. Br J Clin Pharmacol. 2021;87:2493–2501. https://doi.org/10.1111/bcp.14662
[7] Wiens MR, French JL, Rogers JA. Confounded exposure metrics. CPT Pharmacometrics Syst Pharmacol. 2023;00:1-5. doi:10.1002/psp4.13074
[8] Günhan BK, Meyvisch P, Friede, T (2022) Shrinkage estimation for dose-response modeling in phase II trials with multiple schedules, Stat Biopharm Res, 14:2, 249-261, doi: 10.1080/19466315.2020.1850519
Reference: PAGE 32 (2024) Abstr 11195 [www.page-meeting.org/?abstract=11195]
Poster: Methodology - Model Evaluation