by John W. Mitchell
, Senior Correspondent | May 09, 2016
When his father was diagnosed with Stage 4 gastric cancer in 2007, Dimitris Bertsimas, co-director of the Operations Research Center at MIT, Sloan School of Management, did what any son would do. He took his father to the best cancer centers in the region for three opinions. That’s were the confusion began.
“I got three different opinions about what chemo treatments my father should receive,” Bertsimas told HCB News. “It was very confusing and I couldn’t figure out what to do.”
The three different treatment protocols each had different tradeoffs between survivability and toxicity. Bertsimas ended up applying his expertise in applied mathematics to do a few quick calculations on the back of an envelope to help him decide what was the best chemo treatment for his father. His father, who only had a life expectancy of eight months, lived “reasonably well” for two more years.
This experience led to a large-scale study just published in Management Science
. The research unleashed the power of big data to create an efficacy model to predict the best outcomes for Stage 2 and Stage 3 gastric and gastro-esophageal cancers. The study compiled a database of over 400 published clinical papers.
According to Bertsimas, the big data approach was able to identify the optimal chemo treatment options for gastric cancer patients to extend survivability for, on average, four months.
“Our main objective was [to] find a way to help patients and doctors to make individualized decisions about the most effective treatments,” he said. “The way treatment decisions are made now is very much based on intuition and on the luck of the draw.”
He said patients usually go to one doctor whose recommendation tends to be based on their own experience. That’s because a physician may only know about one or two treatment chemo protocols that may not be an ideal option for that patient compared to other proven treatments.
Bertsimas said that big data solutions could also assist pharmaceutical companies in clinical trials. The government could also benefit in its quest to find the best outcomes for the most affordable costs.
Bertsimas and his team are now working to compile and analyze breast and prostate cancer big databases. The gastric big database results and recommendations are being rolled out for clinical trials with patients at three hospitals, including Presbyterian New York and The Dana Farber Cancer Center. The goal will be to offer the same two-dimensional comparison of toxicity and survivability Bertsimas relied on to help his father.
“I feel the era of big data in medicine to develop predictive and prescriptive analysis is very promising,” he said. “It utilizes the totality of the data the entire medical community — really all of human knowledge — has created.”