Mouse Clinical Trials Research Articles

Statistical classification of treatment responses in mouse clinical trials for stratified medicine in oncology drug discovery

Translational oncology research strives to explore a new aspect: identifying subgroups that exhibit treatment response even during pre-clinical phases. In this study, we focus on PDX models and their implementation in mouse clinical trials (MCT). Our primary objective was to identify subgroups with different treatment responses using Latent Class Mixed Model (LCMM).We used a public dataset and focused on one treatment, encorafenib, and two indications, melanoma and colorectal cancer, for which efficacy depends on a specific mutation BRAF V600E. One LCMM per indication was implemented to classify treatment responses at the PDX level, analyzing the growth kinetics of treated tumors and matched controls within the PDX models. A simulation study was carried out to explore the performance of LCMM in this context. For both applications, LCMM identified classes for which the higher the proportion of mutated BRAF V600E PDX models the greater the treatment effect, which is aligned with encorafenib use recommendations. The simulation study showed that LCMM could identify classes with large differences in treatment effects. LCMM is a suitable tool for MCT to explore treatment response subgroups of PDX. Once these subgroups are defined, characterization of their phenotypes/genotypes could be performed to explore treatment response predictors.

Abstract 3228: An online tool to balance baseline covariates for mouse clinical trials

Abstract INTRODUCTION: Randomization is a cornerstone of controlled studies such as mouse clinical trials (MCT). However, simple randomization schemes often lead to unbalanced group allocations, therefore reducing the statistical power of experimental studies. This issue becomes more relevant and critical in MCTs where it is usually necessary to allocate a large number of mice into multiple groups/arms, so that all groups have close tumor volumes (measured by average and variance) as well as other covariates such as body weight, while at the same time also using as few mice as possible. Possible allocations grow exponentially by mouse and group number, and it is computationally intractable to identify optimal allocation schemes; instead, heuristic methods are used to search for near-optimal solutions. Studylog® systems can balance tumor volume with at most one baseline covariate. Randmice, a newly developed tool by Stimunity, shows good performance in balancing bilateral tumors, but it cannot balance more than two baseline covariates. We have developed a fast, in-house, online tool to balance any number of baseline covariates and, in addition, users can assign different weights to each baseline covariate based on levels of importance. METHODS: We used a two-step method to balance designated baseline covariates for MCTs. First, we used a matched algorithm to find sub-matches from a population of mice, by grouping together mice with similar values of baseline covariates. Based on these sub-matches, we then used an optimization method to minimize the imbalance of baseline covariates guided by a weighted score function. Once the optimization was complete, a summary statistic table and a diagnostic plot for the top 10 group allocation solutions were displayed for picking the most suitable solution. RESULTS: We used our baseline covariate balancing tool for a benchmark set of mouse clinical trials with more than one baseline covariate to be balanced out, with the difference of key baseline covariates among groups well under control with close means. The tool completed the allocations in a few minutes by running on a single-CPU computer. CONCLUSIONS: We have developed a fast and powerful online tool to help preclinical researchers balance key baseline covariates in their studies, particularly in mouse clinical trials. Citation Format: Binchen Mao, Davy Ouyang, Henry Li, Sheng Guo. An online tool to balance baseline covariates for mouse clinical trials [abstract]. In: Proceedings of the Annual Meeting of the American Association for Cancer Research 2020; 2020 Apr 27-28 and Jun 22-24. Philadelphia (PA): AACR; Cancer Res 2020;80(16 Suppl):Abstract nr 3228.

In Vivo Pharmacology Models for Cancer Target Research.

Experimental animal tumor models have been broadly used to evaluate anticancer drugs in the preclinical setting. They have also been widely applied for drug target discovery and validation, which usually follows four experimental strategies: first, assess the roles of putative drug targets using in vivo tumorigenicity and tumor growth kinetics assays of transplanted tumors, engineered through gain-of-function (GOF) by overexpressing transgene or knock-in (KI) or loss-of-function by gene silencing using knockdown (KD) or knockout (KO) or mutation via mutagenesis procedures; second, similarly genetically engineered mouse models (GEMM), through either germline or somatic cell procedures, are used to test the roles of potential targets in spontaneous tumorigenicity assays; third, patient-derived xenografts (PDXs), which most closely resemble patient genetics and histopathology, are used in tumor inhibition assays for evaluating target-/pathway-specific inhibitors, including large and small molecules, thus assessing the drug target; and fourth, the targets can be assessed in population-based trials, mouse clinical trials (MCT), so that the validation can be generally meaningful as performed in human clinical trials. This chapter outlines the commonly used protocols in cancer drug target research: the first four sections describe four sets of different, specific pharmacology protocols used in the respective cancer modeling stages, with the last section summarizing the common protocols applicable to all four pharmacology modeling steps.

Abstract 4534: Theory and methodology for the design and analysis of PDX mouse clinical trials

Abstract Patient-derived xenograft (PDX) is a well-accepted experimental cancer model mimicking original patients in histo- & molecular pathology, as well as drug response. Mouse clinical trials (MCT) using PDX becomes increasingly prevalent in oncology research, yet the theory and methodology for trial design and analysis is underdeveloped. By analyzing tumor growths of 34081 PDX mice, we showed that majority of them can be satisfactorily modeled by exponential growth kinetics, based on which we derived a new drug response readout called the AUC ratio that is superior to the tumor growth inhibition (TGI) and its variants. We outlined a bootstrap algorithm to calculate its confidence interval, and showed that a 4:4 (4 mice in both treatment groups) or up design sharply increases measurement accuracy over the lesser design. Next, we introduce three approaches for drug evaluation and biomarker discovery in MCTs. In the end-point based analysis, we used a cohort of gastric PDXs to show a positive correlation between EGFR mRNA expression and cetuximab efficacy, in agreement with clinical trial results. We then used linear mixed models (LMMs) to describe MCTs as clustered longitudinal studies that capture the growth and drug response heterogeneities across PDXs and among mice within a PDX. Further, LMMs can separate prognostic and predictive biomarker effects, quantify response difference to a drug for multiple cancers, or efficacy difference of multiple drugs to a cohort of PDXs. Thirdly, we used additive frailty models to perform survival analysis on MCTs. We defined survival endpoints of PFS and OS in PDXs, and showed that hazard ratios can be more accurately estimated. We revealed an inherent connection between frailty and growth rate for PDXs. We performed computational simulations for the last two methods to show that statistical power improves tremendously with a 3:3 or up design. This work lays the foundation for rational design and analysis of PDX mouse clinical trials. Citation Format: Sheng Guo, Binchen Mao, Henry Q. Li. Theory and methodology for the design and analysis of PDX mouse clinical trials [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2017; 2017 Apr 1-5; Washington, DC. Philadelphia (PA): AACR; Cancer Res 2017;77(13 Suppl):Abstract nr 4534. doi:10.1158/1538-7445.AM2017-4534