Project summary Despite the recent development of new treatment modalities in oncology, the overall approval rate for cancer therapeutics is low: only 3% of drugs tested in a phase 1 setting are ultimately found superior to the standard of care in a phase 3 setting. Methods to more accurately understand drug activity in small patient populations, such as those in phase 1 clinical trials, could better estimate drug efficacy early in the drug development pipeline, help improve the success rate of clinical trials in oncology, and is one of the NCI’s 2020 “provocative questions.” Such methods will become critical as the number of novel drug monotherapies continues to grow and it becomes increasingly impractical to test all promising drug combinations. My proposal aims to develop statistical methods to make more precise estimates of combination drug efficacy from small amounts of clinical data, and to develop experimental methods to select combination therapies likely to be effective in human trials. Aim 1 will develop methods to make more precise estimates of drug efficacy from traditional early-phase (phase 1 and phase 2) clinical trials. Through the systematic analysis of 152 clinical trials for breast, colorectal, lung, and prostate cancer, I found that a single parametric form describes survival distributions across cancer types and therapies. I will test if application of this parametric form increases the precision of estimates for phase 3 drug efficacy from early-phase trials. I will make this dataset and methods publicly available to catalyze future progress in the analysis of clinical trials. Aim 2 will apply new statistical methodology, including that described in Aim 1, to analyze early-phase monotherapy data and to estimate the efficacy of drug combinations. We used this approach to analyze data from small numbers of patient-derived xenografts. We estimated the benefit expected for a novel combination under a mathematical "sum of benefits" model, in which monotherapies exert independent effects on tumor shrinkage, to identify a promising drug combination for T- cell lymphomas. I will use a similar approach to analyze early-phase human clinical trial monotherapy data in the setting of advanced solid tumors and model the expected survival benefit of drug combinations. Aim 3 will develop an experimental paradigm for selecting combination therapies likely to be successful in human clinical trials, and apply it to triple-negative breast cancer. Previous analysis of clinical trial data by our group demonstrates that in the setting of advanced solid malignancies, most successful drug combinations are made of effective single agents with nonoverlapping mechanisms of drug resistance, a principle described as independent action. Aim 3 will test combinations of drugs selected based on independent action across a heterogeneous panel of 18 triple-negative breast cancer cell lines, as well as those identified in Aim 2, to assess whether this design paradigm is likely to produce effective combinations. Overall, my proposal aims to precisely and accurately estimate the efficacy of therapies in oncology using small amounts of patient data and to identify promising drug candidates for use in combination therapies for triple-negative breast cancer.

Project narrative As the number of potential treatment options in oncology increases, along with the number of ways of combining existing therapies, it has become increasingly challenging to test all therapeutically promising drug combinations. The proposed research will develop statistical methods to precisely estimate the efficacy of different drug combinations using small amounts of patient data and will use preclinical data to select new promising combination therapies. My work will help guide the selection of new combination treatments in oncology across a variety of malignancies.