With oncology trials accounting for 25% of the roughly 1,000 clinical trials conducted at Mayo Clinic, Mayo oncologists have turned to IBM’s Watson supercomputer for a pilot program to match the medical center’s oncology patients with the appropriate clinical trials.
Since last fall, Mayo Clinic has been testing Watson’s cognitive computing capabilities in breast, lung and colorectal cancer patients by scanning their individual electronic health records (EHR) and other data to suggest the most appropriate clinical trial for each patient. The IBM computer also may prompt the clinician for additional information such as a particular test that may not be listed in the patient’s medical record. That information is updated by the computer as new data.
“We are primarily validating Watson’s accuracy in matching patients to clinical trials,” said Tufia Haddad, M.D., an oncologist at the Mayo Clinic Cancer Center. “As Watson’s ability to read unstructured medical text increases, so will Watson’s efficiency.”
Haddad explained that Watson is still in the early stages of development in providing a preliminary list of trials for which a patient may be eligible or clearly ineligible. Using natural language processing and powerful data analytics capabilities, Watson is expected to quickly sift through millions of pages of clinical trial and patient data to complete the process.
Currently, providers manually go through each clinical trial “checklist” for patient eligibility, bouncing back and forth throughout the EHR and studying protocols to see if a patient meets specific eligibility criteria— a process repeated for each clinical trial opportunity that can take upward of 30 minutes to review three or four protocols.
By contrast, the current version of Watson is providing preliminary matches for up to 20+ clinical trials in a matter of seconds, said Haddad, adding that Watson also can show how close an eligible patient is to a clinical trial match at this time.
“Patients need answers and Watson helps provide them quickly and more thoroughly,” Steven Alberts, M.D., a Mayo Clinic oncologist, said in a statement. “We are excited by Watson’s potential to efficiently provide clinical trials information at the point of care.”
Although this is the first time that Watson has matched patients with clinical trials, the IBM supercomputer is not new to cancer. Watson was “trained” with oncologists at New York-based Memorial Sloan-Kettering and at MD Anderson Cancer Center in Texas as early as 2012. That’s where its artificial intelligence platform was learning vocabularies and terms associated with cancer, including scanning oncology research literature along with data from cancer drugs.
Also contributing to the Mayo Clinic’s decision to implement Watson is how medical technology has impacted cancer, where the role of genetics and precision medicine have prompted more trials but fewer patients per trial than five to seven years ago.
“As cancer is increasingly being treated according to the presence or absence of genetic aberrations, we will need to screen larger populations to find small, select groups of eligible people who have these genetic aberrations,” said Haddad. “We are training Watson to read our pathology reports and genomic sequencing reports so that Watson can identify all potentially eligible patients in a tumor-specific practice. Right now it is a manual process to abstract this type of information from the electronic health record, and we do not have the resources to screen thousands of patients for these low-frequency tumor-genetic aberrations.”
As for which part of the EHR is most influential in finding the right oncology trial, Watson’s focus varies for each type of cancer. For example, in breast cancer, clinical trials are separated into studying disease that is operable (stage I–III) versus inoperable (stage IV or metastatic). Studies also take into account certain molecular subtypes of breast cancer, such as estrogen receptor (ER)-positive or HER2-positive, since the proposed therapy is tailored to these unique disease biologies.
Watson pulls out disease stage and ER and HER2 status of the tumor from the electronic health record. This information may reside in a clinical note or a pathology report, said Haddad. She noted that patient lab values, cardiac function and prior therapies are other key attributes enabling Watson to extract them from the EHR to increase the accuracy of the clinical trial matches.
“Ultimately, we believe Watson will help advance discoveries into promising new forms of care that clinicians can use to treat all patients,” Mike Rhodin, senior vice president, IBM Watson Group, said in a statement.
Email comments to Ronald at firstname.lastname@example.org. Follow @RonRCW
This article was reprinted from Volume 19, Issue 23, of CWWeekly, a leading clinical research industry newsletter providing expanded analysis on breaking news, study leads, trial results and more. Subscribe »