MIT study explores the key factors behind patient outcomes in clinical trials evaluating new treatments for non-small-cell lung cancer
Cambridge, MA, September 26, 2019—In a new study just published in JCO Clinical Cancer Informatics, researchers from the Massachusetts Institute of Technology applied machine-learning techniques to patient-level clinical trial data and simulations from a statistical model of tumor growth to predict response and survival among patients with non-small-cell lung cancer (NSCLC). In the era of precision medicine, therapeutic planning requires predictive models that provide patients and physicians with a better understanding of which treatment will work based on the specific characteristics of the individual. For patients with NSCLC, however, no such model has been widely implemented on such a large dataset until now.
Using data pooled from 17 clinical trials submitted to the FDA between 2007 and 2017 evaluating chemotherapy, targeted therapy, and immunotherapy in patients with NSCLC, the aim of the MIT researchers—who conducted their study in collaboration with Dr. Sean Khozin, director of the U.S. Food and Drug Administration’s Information Exchange and Data Transformation (INFORMED) program—was to determine the main drivers behind patient response and survival for each treatment modality. Reflecting recent advances in the treatment of NSCLC, their study is thought to be the largest of its kind that considers biomarker status and inhibitor therapy as candidate predictive variables.
The researchers found that biomarker status is the strongest predictor of overall response, progression-free survival, and overall survival in patients with NSCLC treated with immune checkpoint inhibitors and targeted therapies. However, they also found that single biomarkers have limited predictive value in patients treated with a programmed death-ligand 1 (PD-L1) immunotherapy. “Although genomic profiling is common in clinical trials these days, the data isn’t collected in a coordinated way to be able to examine it holistically,” noted Kien Wei Siah and Chi Heem Wong, both Ph.D. students in MIT’s Laboratory for Financial Engineering (LFE) and Computer Science and Artificial Intelligence Laboratory (CSAIL), who co-authored the study.
Being able to more accurately predict clinical outcomes in patients based on their individual characteristics is a major component in accelerating the entire drug development process and improving healthcare delivery. “Researchers and clinicians can make more informed decisions when designing clinical trials, all stakeholders can better measure, and thus manage, the risks associated with therapeutic development, and most importantly, patients with NSCLC can receive better and more effective care,” said Andrew W. Lo, senior author of the study and director of the LFE as well as a CSAIL Principal Investigator.
“Our study highlights the need to include more relevant data points—for example, composite multi-omic signatures, electronic medical records, and text-based information extracted from the biomedical literature—in clinical trial data collection pipelines that can serve as predictive variables,” Lo adds. “We hope to pursue these datasets in future research and welcome potential collaborators to engage with us at the LFE.”
About the MIT Laboratory for Financial Engineering
The MIT Laboratory for Financial Engineering (LFE) is a research center focused on the quantitative analysis of financial markets and institutions using mathematical, statistical, and computational models and methods. The goal of the LFE is to support and promote academic advances in financial engineering and computational finance that can be directly applied for the betterment of the world. To do that, LFE faculty, students, and staff engage with industry professionals, regulators, policymakers, and other stakeholders to develop and apply new financial technologies to practical and socially important settings.