NEW YORK – Researchers have developed a new test to distinguish between patients with pancreatic cysts who need surgery or further surveillance and those who do not.
Pancreatic cysts are common and are found in about 4 percent of people in their 60s and in 8 percent of people over the age of 70. While some pancreatic cysts — most often ones that produce mucin — can develop into pancreatic cancer, not all do, which can make it difficult to determine the best treatment course for patients.
A Johns Hopkins Kimmel Cancer Center-led team has developed a new test, dubbed CompCyst, that it hopes to commercialize. It incorporates clinical features, imaging, and molecular data to classify pancreatic cysts into three groups: one with no potential to become cancerous, one with a small risk of doing so, and one with a high risk of doing so. This approach is better at determining the proper treatment approach than the current clinical standard alone, the researchers reported today in Science Translational Medicine, and could prevent patients from undergoing unneeded, risky surgeries.
"If these results translate into clinical practice, a large number of patients will be spared an unnecessary operation with associated mortality and lifelong morbidity," co-author Christopher Wolfgang, the director of surgical oncology at the Johns Hopkins Kimmel Cancer Center, said during a press briefing.
The researchers developed CompCyst by analyzing a cohort of 862 patients, all of whom had undergone surgical resection of their pancreatic cysts. That way, the cyst's true histopathology was known as well as what would have been, in retrospect, the best way to manage the cyst.
They fed a combination of molecular data — such as gene alterations and aneuploidy information — as well as clinical and imaging data from 436 patients into a machine-learning program to classify those patients into the three cyst categories. They then validated their approach in the remaining 426 patients.
This CompCyst approach outperformed the current clinical approach, the researchers reported, though they noted it was still not perfect. Overall, CompCyst had a 69 percent accuracy for classifying patients into those three groups, compared to 56 percent for the current standard of care.
For instance, based on the samples' pathology, 53 patients from the validation cohort had a benign, non-mucin-producing cyst that required neither surgery nor surveillance. But current clinical management only suggested that course of action for 10, or 19 percent, of the patients, while CompCyst suggested it for 32, or 60 percent, of the patients.
This, the researchers estimated, indicates that had CompCyst been used when the patients presented with their pancreatic cyst, surgery would have been avoided for more than half of the patients who didn't require it. According to Wolfgang, pancreatic cyst surgeries have a high complication rate — it can be as high as 50 percent, depending on the type of operation needed — and should be limited to those who would truly benefit from it.
Still, the researchers noted that neither CompCyst nor the current standard of care discharged any patient for whom surgery was indicated. In the press briefing, senior author Anne Marie Lennon, a professor of medicine at Hopkins, said that they chose to err on the side of caution.
"What we don't want to do is to discharge someone who could have either a cancer or a mucin-producing cyst," she said. "We specifically developed the test so that we would have an extremely high specificity for identifying non-mucin-producing cysts, which have no risk of developing cancer."
According to Bert Vogelstein, the co-director of the Ludwig Center at Hopkins' Kimmel Cancer Center, CompCyst will be available at Hopkins within the next six to 12 months there.
The researchers also plan to launch a prospective study of about 1,000 patients with the goal of seeking approval for the test from the US Food and Drug Administration. Vogelstein added Hopkins has licensed the technology to the cancer detection firm Thrive, which launched earlier this year. It is also commercializing Vogelstein's CancerSEEK tool, which relies on DNA alterations and blood protein measurements to detect early stage cancer.