Research spotlight: Using Data Analytics for Better Healthcare Delivery

Unlocking an analytics-fueled approach to chemotherapy combinations

Researcher: Dimitris Bertsimas

Dimitris Bertsimas and his co-researchers are pioneering new ways to apply state-of-the-art analytics and machine learning to the task of designing safe and effective chemotherapy regimens. They are focusing their efforts on gastric cancer, a cancer type for which there is currently no best-in-class chemotherapy treatment regimen. Having constructed a database of 414 clinical trials for advanced gastric cancer, Bertsimas and his colleagues have now trained statistical models with randomized and non-randomized clinical data to more accurately predict survival and toxicity outcomes from combinatory chemotherapy treatments—and to evaluate those predictions.

Equipped with these first-of-their-kind methodologies that can help predict outcomes (including identifying 10–20 percent of the trials with high toxicity or efficacy issues as well as high-promise clinical trials before they are run), providers will increasingly have access to data-driven methods for selecting combination chemotherapy regimens—a step forward with safer, more precise, and more effective weapons in the battle against cancer.

1. “MIT Professor Leverages Machine Learning to Find Promising Cancer Treatments.” MedTech Boston (2016).
2. “An Analytics Approach to Designing Combination Chemotherapy Regimens for Cancer.” Bertsimas, D., O’Hair, A., Relyea, S., & Silberholz, J. Management Science Vol. 62, No. 5 (2016): 1511–1531.