Professor University of Texas MD Anderson Cancer Center Houston, Texas, United States
Disclosure(s):
Yisheng Li, PhD: No financial relationships to disclose
Objectives: To improve the efficiency of identifying the maximum tolerated dose (MTD) in phase I oncology trials, researchers have recently proposed dose-finding designs to incorporate pharmacokinetic (PK) and pharmacodynamic (PD) information, mostly in the form of summary statistics such as the area under the concentration-time curve (AUC) or maximum concentration (Cmax). Su et al. [1] proposed a semi-mechanistic dose-finding (SDF) design accounting for dynamic PK and latent PD biomarker information. However, the improvement of their design performance over the continuous reassessment method (CRM) [2] is small. It is also unclear how their trial design might perform for treatments with more complex PK mechanisms. We aim to extend the SDF design by 1) incorporating available PD data for toxicity, and 2) focusing on drugs whose actions are consistent with a potentially more complex PK mechanism than a 1-compartment IV bolus model.
Methods: Based on a motivating phase Ib/II trial of a bi-specific antibody T cell therapy in patients with non-Hodgkin’s lymphoma, we extend the SDF model framework to incorporate the observed data of IL-6, a PD biomarker for dose-limiting toxicity (DLT). We use Bayesian joint modeling of the PK, PD and DLT outcomes. We perform extensive simulation studies to evaluate the operating characteristics of the proposed design and compare their performance with existing phase I trial designs.
Results: Our extensive simulation studies show that on average the proposed design outperforms some common phase I trial designs, including the modified toxicity probability interval (mTPI) [3] and Bayesian optimal interval (BOIN) [4] designs, the continual reassessment method (CRM), as well as the SDF design assuming a latent toxicity PD biomarker (SDF-woPD) [1], in terms of the percentage of correct selection of maximum tolerated dose (MTD) and average number of patients allocated to MTD, under a variety of dose-toxicity scenarios. When the working PK model and the class of link function between the cumulative PD effect and DLT probability is correctly specified, the proposed design also yields better estimated dose-toxicity curves than CRM and SDF-woPD. Our sensitivity analysis results suggest that the design’s performance is reasonably robust to prior specification for the parameter in the link function, as well as misspecification of the PK model and class of the link function.
Conclusions: By incorporating PD biomarker data into the SDF model framework, the proposed design improves its performance upon the common designs, as well as the SDF design with a latent PD biomarker, for phase I dose-finding trials [5].
Citations: [1] Su X, Li Y, Müller P, Hsu C-W, Pan H, Do K-A. A semi-mechanistic dose-finding design in oncology using pharmacokinetic/pharmacodynamic modeling. Pharm Stat 21(6):1149-1166, 2022.
[2] O’Quigley J, Pepe M, Fisher L. Continual reassessment method: a practical design for phase 1 clinical trials in cancer. Biometrics. 1990;46(1):33-48.
[3] Ji Y, Liu P, Li Y, Bekele NB. A modified toxicity probability interval method for dose-finding trials. Clin Trials. 2010;7(6):653-663.
[4] Yuan Y, Hess KR, Hilsenbeck SG, GilbertMR. Bayesian optimal interval design: a simple and well-performing design for phase I oncology trials. Clin Cancer Res. 2016;22(17):4291-4301.
[5] Yang C, Li Y. An extended Bayesian semi-mechanistic dose-finding design in oncology using pharmacokinetic and pharmacodynamic information. Stat Med 43(4):689-705, 2024.
Keywords: cytokine release syndrome, dose finding, maximum tolerated dose
.jpg "Yisheng Li, PhD photo")