Sirin Yonucu, J.G. Coen van Hasselt, Mike Walker, Andrzej M Kierzek, Piet H. van der Graaf
Leiden University
Objectives:
Tumor cells can exploit host immune checkpoints to allow them to evade eradication by the immune system. For several immune checkpoint targets, therapeutic agents have been successfully developed, and for various additional targets, drugs are being actively developed. In this context, the checkpoint receptor PD-1 is an important drug target that inhibits T cell function after activation by its ligand. Five immune checkpoint inhibitors have currently been approved for blockade of PD-1 and its ligand PD-L1 [1]. The anti-tumor activity of these drugs is seen in a subset within a broad range of cancers, and the response is durable when it is achieved [2]. However, there are pressing questions about who will benefit from this treatment, what is the best dosing schedule and how should we combine them with other drugs to maximize the anti-tumor response. Research on these questions would also reveal the ways to convert non-responding patients to responders of immunotherapy. Preclinical models are crucial to evaluate novel drugs and drug targets. However, there are currently important discrepancies between preclinical and clinical results for immune checkpoint inhibitors, and several drug candidates have recently failed in clinical studies. An important reason for this could be the complexity of the underlying immune-pharmacological mechanisms that mediate the ultimately observed response.
Quantitative systems pharmacology (QSP) approaches can facilitate understanding of inter-species differences in response to immune checkpoint inhibitor therapy and may guide the design of optimal dose regimens for evaluation in clinical trials. In the current analysis, we aimed to develop a QSP model for tumor-immune system interactions for treatment with the anti-PD-L1 inhibitor pembrolizumab, based on a previously developed mathematical model in mice [3].
Methods:
The previously-developed mathematical model by Kosinsky et al. [3] consists of a system of ODEs for the immune system such as inactivated and activated T cells, PD-L1, dendritic cell maturation (DCm), systematic antigen presentation level (Agsys), immune suppressive cells accumulation (ISC) and immune activation rate (IAR) as systems variables to describe the tumor growth dynamics. We first considered the human pharmacokinetic (PK) of the drug using the study of Ahamadi et al. [4] and applied them to the model equations. Subsequently, we identified human parameters from clinical datasets in combination with principles of allometric scaling. For undetermined parameters, we used parameter sensitivity analysis to identify parameter sets that can reproduce clinical response as described in Chatterjee et al. [5].
Results:
The scaled human QSP model was able to describe similar treatment response distributions for different dosing regimens previously reported [5]. Sensitivity analysis of the model showed that the most sensitive parameters are tumor growth rate, T cell kill rate and sensitivity of dendritic cell maturation to tumor cell death rate. Variation in the distribution of drug response among patients could be reproduced by adding inter-patient variability to tumor growth rate. The initial tumor size in human simulations was much larger than the largest mouse tumors given in Kosinsky et al. [3]. Lumped parameters that determine ISC and IAR were re-evaluated since the magnitudes of these quantities depend on the tumor size. Reproduction of PD-L1 levels of patients at the start of the therapy is achieved by the addition of inter-patient variability to the parameter that stands for the sensitivity of PD-L1 expression up-regulation to activated T cell count. This parameter was fixed based on the fact that nearly half of the population had more than 50% of tumor cells with membranous PD-L1 staining [5].
Conclusions:
Simulation results showed the critical parameters that determine the tumor response to the drug in humans were tumor growth rate, T cell kill rate and sensitivity of dendritic cell maturation to tumor cell death rate. To achieve a similar inter-individual variation with the clinical outcomes, we needed to add inter-individual variability in tumor growth parameters in our model framework.
References:
[1] Ribas, Antoni, and Jedd D. Wolchok. “Cancer immunotherapy using checkpoint blockade.” Science 359.6382 (2018): 1350-1355.
[2] Jenkins, Russell W., David A. Barbie, and Keith T. Flaherty. “Mechanisms of resistance to immune checkpoint inhibitors.” British journal of cancer 118.1 (2018): 9.
[3] Kosinsky, Yuri, et al. “Radiation and PD-(L) 1 treatment combinations: immune response and dose optimization via a predictive systems model.” Journal for immunotherapy of cancer6.1 (2018): 17.
[4] Ahamadi, M., et al. “Model‐Based Characterization of the Pharmacokinetics of Pembrolizumab: A Humanized Anti–PD‐1 Monoclonal Antibody in Advanced Solid Tumors.” CPT: pharmacometrics & systems pharmacology 6.1 (2017): 49-57.
[5] Chatterjee, M., et al. “Systematic evaluation of pembrolizumab dosing in patients with advanced non-small-cell lung cancer.” Annals of Oncology 27.7 (2016): 1291-1298.
Reference: PAGE 28 (2019) Abstr 8967 [www.page-meeting.org/?abstract=8967]
Poster: Drug/Disease Modelling - Oncology