Xiao Su, Yisheng Li, Peter Mueller, Kim-Anh Do
Non-affiliated, The University of Texas MD Anderson Cancer Center, The University of Texas at Austin, The University of Texas MD Anderson Cancer Center
Objectives: While a number of phase I dose-finding designs in oncology exist, the commonly used ones are either algorithmic or empirical model-based. Other statistical designs that incorporate pharmacokinetic (PK) data mainly focus on summary PK information. We aim to: 1) propose an extended framework for modeling the dose-toxicity relationship, by incorporating dynamic PK and pharmacodynamic (PD) information; and 2) apply this modeling framework in the design of phase I trials.
Methods: We propose to jointly model the PK, latent PD, and dose-limiting toxicity (DLT) outcomes by using dynamic PK/PD modeling as well as modeling of the relationship between a latent cumulative pharmacologic effect and a binary DLT outcome. This modeling framework naturally incorporates the information on the impact of dose, schedule and method of administration (e.g., drug formulation and route of administration) on toxicity. The resulting design is an extension of the existing designs that make use of pre-specified summary PK information (such as the area under the concentration-time curve [AUC] or maximum serum concentration [Cmax]). We conduct extensive simulation studies to evaluate the performance of the proposed design and compare it with the existing designs. The performance of each design is summarized by the percentage of correct selection of the maximum tolerated dose (MTD), average number of patients allocated at the MTD, and average probability of patient experiencing DLT, in the simulations.
Results: Our simulation studies show, with moderate departure from the hypothesized mechanism of the drug action, that the performance of the proposed design on average improves upon those of the common designs, including the continual reassessment method (CRM), Bayesian optimal interval (BOIN) design, modified toxicity probability interval (mTPI) method, and a design called PKLOGIT that models the effect of the AUC on toxicity. In case of considerable departure from the underlying drug effect mechanism, the performance of the proposed design is shown to be comparable to that of the other designs. We illustrate the proposed design by applying it to the setting of a phase I trial of a $gamma$-secretase inhibitor in metastatic or locally advanced solid tumors. We also provide an R package to implement the proposed design.
Conclusions: The proposed design improves upon the existing designs. The proposed joint modeling framework may be considered promising for use in other settings of early phase oncology trial designs as well.
References:
[1] *Su X, Li Y, Müller P, Do K-A. A semi-mechanistic dose-finding design in oncology using pharmacokinetic/pharmacodynamic modeling. Pharmaceutical Statistics. Under revision.
[2] Ursino Moreno, Zohar Sarah, Lentz Frederike, et al. Dose-finding methods for Phase I clinical trials using pharmacokinetics in small populations. Biometrical Journal. 2017;59(4):804–825.
[3] R package at https://github.com/esuxiao/PKPDMTD.
Reference: PAGE 28 (2019) Abstr 8802 [www.page-meeting.org/?abstract=8802]
Poster: Methodology - Other topics