To have surgery or not to have surgery? That is the big question for hundreds of thousands of Americans diagnosed each year with a pancreatic cyst.
The dilemma? While cysts are very common, most will not actually lead to cancer. But for doctors, determining which will from which won't is often a shot in the dark. So many patients undergo a highly invasive cyst-removal operation out of an abundance of caution, often unnecessarily.
Enter "CompCyst." New research suggests that this experimental, artificial intelligence-driven test could help slice the number of such unwarranted surgeries nearly in half.
"In those few patients who would go on to develop a cancer from their cyst, the benefit of these big operations in preventing cancer strongly outweigh the risks of surgery," noted study author Dr. Christopher Wolfgang, from the Johns Hopkins Kimmel Cancer Center in Baltimore. "In particular, because we know this cancer has approximately a 90% mortality once developed."
Speaking at a media conference, he cautioned that these types of surgeries "are among the largest abdominal operations performed." That means a high risk for complications, and post-surgery recovery time typically approaches three months.
"Thus, surgery must be reserved for only those who would truly benefit," said Wolfgang, co-director of Hopkins' Precision Medicine Center of Excellence for Pancreatic Cancer.
With that goal in mind, he and his colleagues analyzed cyst samples from more than 860 surgical patients at 16 different medical centers around the world.
Such fluid-filled pancreatic lesions are found in roughly 800,000 Americans every year, investigators noted, including about 4% of those in their 60s and 8% of those in their 70s. Only a small percentage, known as "mucin-producing cysts," have the potential to develop into cancer.
The samples were subjected to protein analyses, conventional image scanning and molecular tests, to identify significant genetic mutations and chromosomal changes.
Results were then fed into a special computer algorithm called "MOCA." Over time, MOCA "learned" to use that information to search for millions of telltale markers for cancer risk.
MOCA then predicted which cysts represented no risk for cancer, a small risk for cancer (requiring follow-up monitoring), or a high likelihood for cancer and a real need for surgery.
The result: The CompCyst screening method concluded that cyst removal surgery had been unnecessary for 45% of those patients who had gone under the knife.
At the media conference, study team member Dr. Marco Dal Molin pointed out that standard screening tools were only able to identify about 20% of those patients whose cysts posed no cancer risk.
By comparison, the algorithm correctly pinpointed 60% of no-surgery-needed cases, said Dal Molin, who is a postdoctoral research fellow at the Kimmel Cancer Center.
But he also stressed that the new computer test is not intended to replace standard screening tools. Instead, the hope is that CompCyst can be deployed alongside current imaging and genetic tests "to integrate all of these aspects together."
Still, the team acknowledged that more research is needed, including another study to try to predict risk among new patients who have not yet undergone surgery.
The findings were published July 17 in the journal Science Translational Medicine.
Further investigation will be critical to assessing the new screening method's "real-world generalizability," said Dr. Cesar Castro, an assistant professor of medicine at Harvard Medical School, and director of the Cancer Program at MGH Center for Systems Biology in Boston. He was not part of the study.
But given that "a doctor is only as good as his or her intel," Castro said the new screening test could prove to be an important advance in trying to get a better handle on true pancreatic cancer risk.
"Pancreatic cyst removal should not be confused with straightforward day surgeries," Castro said. "Fatalities and severe complications may arise. As a medical school instructor once told us: 'One should never mess with the pancreas.'"
Castro did point out that the new test is somewhat invasive itself, as physicians would need to collect pancreatic cyst fluid and draw blood to conduct the analysis.
But he added that when it comes to patient risk, CompCyst "certainly pales in comparison" to unnecessary, major surgery.
For more on pancreatic cysts, see Memorial Sloan Kettering Cancer Center.