Imaging Predicts Pancreatic Cyst Cancer Risk

Imaging features can help assess the risk of developing pancreatic cancer from pancreatic cysts, potentially sparing patients from years of surveillance and associated costs, according to research presented Monday.

“With the increase in scans that people are getting for unrelated medical reasons, we’re finding more pancreatic cysts, and we don’t have great answers on what we should do when we find them,” said study presenter Andrea Bejar, a fourth-year medical student at Northwestern University’s Feinberg School of Medicine in Chicago.  

Pancreatic cysts are fluid-filled sacs that form in the pancreas. Most are benign, but a small percentage will progress to cancer. Abdominal CT and MRI for other indications provide a means to opportunistically detect these cysts.

Bejar and colleagues undertook a study to identify imaging and patient risk factors for pancreatic cancer development from cysts in a large group of patients who had abdominal CT/MRI. 

They identified 3,604 incidental pancreatic cysts in 112,361 patients. Of the 3,604 patients with cysts, 138 developed pancreatic ductal adenocarcinoma (PDAC), the most common type of pancreatic cancer. Of those, 21 developed PDAC more than one year after cyst diagnosis. The mean time from cyst diagnosis to PDAC development was 6.24 years. 

A comparison of patients with pancreatic cysts who developed PDAC after more than one year of follow-up imaging with those who did not develop PDAC after more than 10 years of follow-up revealed that imaging features like larger cyst diameter, main pancreatic duct dilation and septations were key risk factors for developing cancer. 

According to Bejar, the results suggest that imaging surveillance of more than five years may be warranted for cysts with specific risk factors, thereby optimizing early PDAC detection while reducing unnecessary follow-up in low-risk patients. 

“These higher-risk features aren’t always present immediately, so surveilling and then understanding more about how these features develop and in which patients will be very important for understanding who should be followed and for how long,” Bejar said. 

AI To Aid with Surveillance Decisions

The findings represent the first part of a study whose ultimate goal is to use AI to predict cancer risk from earlier scans. 

“Say we have a scan from 10 years earlier and the patient ended up developing cancer,” Bejar posited. “We want to know if we can take that scan and use AI to predict whether the patient will develop cancer and if we can prevent the surveillance period and potentially invasive procedures.”

Currently, an optimal way to know whether a cyst will progress to cancer is lacking. Surveillance with MRI and endoscopic US (EUS) is recommended, but there are no clear recommendations on when that surveillance should be stopped. Some patients will get a scan every year for up to five years, while others get them every two years. 

“Hopefully, getting a better understanding of which cases we should be worried about and seeing if AI can help us make those decisions will help save a lot of patients from annual MRIs and EUS,” Bejar said.

Access the session, “Imaging Biomarkers for Pancreatic Adenocarcinoma Risk Stratification in Incidental Pancreatic Cysts: A Multi-Hospital Cohort Study,” (M3-SSGI04-2) on demand at RSNA.org/MeetingCentral.