For most patients, a diagnosis of stage 4 non-small cell lung cancer comes with a dire prognosis. But for patients with specific mutations that cause the disease, there are potentially life-saving therapies.
The problem is that these mutations, known as ALK and EGFR, are not always identified in patients — meaning they never get the treatment.
A new study from the Fred Hutchinson Cancer Research Center in Seattle used machine learning to find these needle-in-a-haystack patients. The idea was to leverage cancer databases to see if patients were being tested for the mutations and receiving these personalized treatments.
“We know these patients should be treated up front with specific oral targeted therapies that are highly effective and less toxic than chemotherapy and recommended by all the guidelines,” Dr. Bernardo Goulart, lead author on the study, said in an announcement.
For Shelly Engfer-Triebenbach, who was diagnosed with stage 4 non-small cell lung cancer at age 40, finding out she had one of the mutations was “like winning the lung cancer lottery.”
“We wanted to know, in real practice, if this patient population was getting timely access to these oral drugs and what their outcomes were. We also wanted to know whether patients were getting tested,” Goulart said.
Goulart’s team collaborated with Emily Silgard, a natural language processing engineer, to design an algorithm that could filter through cancer patient databases to find people with the mutations. The National Cancer Institute, which operates a national database of cancer patients called SEER, funded the study.
The researchers pitted their algorithm against humans who checked data on two separate cancer registries, one in Washington state and the other in Kentucky. The algorithm performed remarkably well on the Washington registry, but it struggled to accurately identify patients in Kentucky.
While the study looked at past patient records, Goulart said he hopes the tool can be used for real-time screening. “The ultimate goal is to have a way to identify patients treated in real-world scenarios and clinics whose lung cancers carry those mutations,” he said.
Goulart hopes the same process will also be used to identify other mutations in patients with different forms of cancer who might also benefit from targeted therapies. “This study could absolutely serve as a prototype to look at molecular mutations in other tumors,” he said.