IBM, which markets products based on its Watson supercomputer to improve cancer treatment, unveiled three studies showing Watson agreed with doctors on a treatment course for a range of cancers up to 96% of the time.
IBM’s Watson for Oncology is designed to help oncologists make treatment decisions for individual patients. It reads patient data and runs through a wealth of medical literature much more quickly than a team of experts would. It then provides recommendations to the clinician whether the standard treatment should be followed, considered or abandoned. The system is being trained at Memorial Sloan Kettering Cancer Center to support treatment decision-making in a variety of cancer types.
One study, conducted at Manipal Comprehensive Cancer Centre in India, found Watson agreed with experts on a molecular tumor board 96% of the time on treatments for lung cancer, 81% for colon cancer and 93% for rectal cancer. Meanwhile, a study at Bumrungrad International Hospital in Thailand posted an average concordance rate or 83% across colorectal, lung, breast and gastric cancers. The highest rate was 91% for lung and the lowest was 76% for breast cancer.
The third concordance study, conducted at Gachon University Gil Medical Centre in South Korea, had a mixed bag—Watson and the doctors agreed on treatment plans for 73% of colon cancer cases, but only on 49% of gastric cancer cases. This is because some drugs are used in the U.S.—where Watson was trained—but not in South Korea and vice versa, the investigators explained.
So the low rate shouldn’t be discouraging, according to Andrew Norden, M.D., deputy chief health officer for oncology and genomics, IBM Watson Health. It just means the tool should be customized for the way doctors treat cancer in different countries. And while the high concordance rates in other cancer types are “reassuring,” Norden said, “We think there’s real value in doing at least one concordance study in every country where Watson for Oncology is used.”
Watson for Oncology would be most useful in areas that lack deep subspecialty expertise, Norden said. There is a relatively small number of cancer centers in the U.S. and around the world where there are experts who focus on specific cancers, he said. Watson could bring this specialized expertise to community cancer centers, where most cancer patients are treated. It would also be useful in environments where the healthcare infrastructure is less well-resourced, he said.
IBM also presented a qualitative study, in which oncologists in Mexico found Watson for Oncology useful, particularly in clinics that do not have subspecialty expertise.
Finally, the company announced that its Watson for Clinical Trial Matching cut down the amount of time taken to screen patients for clinical trial eligibility by 78%. The study, conducted with Novartis and Highlands Oncology Group involved 2,620 lung and breast cancer patients. Watson analyzed patient records, doctors’ notes against clinical trial protocols from Novartis to automatically exclude patients who were ineligible for the trial.
Nine new hospitals and healthcare systems adopted Watson oncology tools, the company said in the statement, bringing the total number to 55.
"These studies demonstrate that Watson technologies are doing what we expect them to do: helping physicians augment their own experience and expertise to deliver evidence-based care," Norden said in a statement.