Algorithm Could Be 'Major Leap Forward' for Pancreatic Cysts

Pam Harrison

July 18, 2019

A new algorithm has been developed that can reliably distinguish benign from potentially malignant pancreatic cysts, sparing patients whose cysts would never develop into cancer from highly invasive abdominal surgery, researchers are reporting.

"We currently do not have an accurate method that reliably tells us whether to immediately remove the cyst with surgery or closely observe the patient or  discharge [patients] with no further care," one of the investigators, Chris Wolfgang, MD, PhD, professor of surgery, Johns Hopkins Kimmel Cancer Center, Baltimore, Maryland, told a press briefing.

"Our study directly addresses these limitations and if these results translate into clinical practice, a large number of patients will be spared an unnecessary operation with its associated mortality and life-long morbidity," he added.

Indeed, the findings have "the potential to be a major leap forward in the management of pancreatic cysts," he emphasized.

The research was published online July 17 in Science Translational Medicine, by Simeon Springer, PhD, also from Johns Hopkins Kimmel Cancer Center, and colleagues.

The article is the first in the Mind the Gap series of in-depth stories that aim to highlight the clinical translational potential of the journal's studies.

CompCyst Outperformed Standard of Care

As speakers at the briefing noted, most pancreatic cysts never develop into cancer but so-called mucin-producing cysts can turn malignant, and current guidelines recommend patients with mucin-producing cysts undergo surveillance or surgery, depending on how advanced the cysts are.

The cyst classifier test, referred to as CompCyst, emerged from an amalgamation of known clinical features of pancreatic cysts, their imaging characteristics, and genetic and biochemical markers of benign and malignant cysts.

Researchers then used deep learning techniques to train the platform to stratify patients into one of three clinically relevant groups: those who require no further follow-up because their cysts are benign; those who require surveillance because of potentially malignant features of their cysts; and those who need to have their cysts resected because of high-grade dysplasia or overt cancer.

After training the test protocol with 436 patients whose cysts had been surgically resected, the team then validated the CompCyst algorithm by testing it in 426 other patients who had also undergone surgical resection.

Surgical pathology is currently the gold standard of pancreatic cyst classification.

Within the validation cohort, researchers then compared CompCyst-based recommendations for cyst management with current clinical guidelines as to what the optimal treatment strategy should have been.

Overall, CompCyst outperformed current clinical practice in all three patient groups, co-investigator Marco Dal Molin, PhD, Johns Hopkins Kimmel Cancer Center, explained.


For example, current clinical management correctly identified only 19% of patients who were deemed suitable for discharge because of the benign nature of their cysts.

This compared with 60% of patients in the same risk group who were correctly identified on the CompCyst test, as Dal Molin noted.

"If we look at patients who should be monitored — that is patients with mucin-producing cysts without high-grade dysplasia or invasive cancer — application of current clinical management correctly identified 34% of them whereas CompCyst correctly identified 49% of them," he added.

CompCyst Cuts Unnecessary Surgeries but Still IDs Necessary Ones  

This meant that relative to current clinical practice, "CompCyst would have decreased the number of operations done for benign, nonmucin producing cysts by 74%, and overall, the use of CompCyst would have avoided unnecessary surgery in 60% of patients included in this study," Dal Molin emphasized.

On the basis of known histopathology, surgery was indicated in the remainder of the cohort.

In this group of patients, the current standard of care correctly identified 89% of patients who required surgery for cyst resection compared with 91% of those who were identified on CompCyst testing.

Overall, CompCyst correctly classified 69% of the cohort into one of the three groups of potential follow-up care — surgery, surveillance, or discharge — compared with 56% for the current standard of care, the researchers report.  

They also stress that they developed the CompCyst algorithm to be able to classify patients as having either no, to low, to a high risk of developing pancreatic cancer with a high degree of specificity.

As senior investigator Ann Marie Lennon, MD, professor of medicine, Johns Hopkins Kimmel Cancer Center, explained: "What we don't want to do is discharge a patient who could have either a cancer or a [potentially malignant] mucin-producing cyst. The CompCyst test performed very well [in this regard] in that only 1% of patients who were discharged might have required surveillance and no patient [was discharged] who required surgical resection."

"As a clinician this is very important," she added, "because, with a 99% specificity rate, you can say to patients who have a cyst that it is safe not to follow them," she added.

Pancreatic Cysts Are Incidental Diagnoses

Approximately 80 million CT scans are done every year in the United States, out of which approximately 800,000 patients will be incidentally diagnosed with a pancreatic cyst.

"This poses several clinical problems," as Wolfgang explained.

Firstly, physicians need to follow hundreds of thousands of patients with expensive and sometimes invasive tests in order to identify the minority of patients with pancreatic cysts who will progress to cancer.

Follow-up strategies also inevitably subject patients to the risks of radiation and potential complications from procedures, he also observed.

"Moreover, not knowing the level of risk of progression for cancer is anxiety-provoking for both patients and physicians," Wolfgang said.

And on a practical level, following this number of patients over many years is a poor use of healthcare resources.

Decision to Operate Is Fraught With Anxiety

For surgeons such as himself, the decision whether to operate or not is often fraught, Wolfgang explained. On the one hand, if surgeons choose to operate, "We know we may be doing an operation that in hindsight was not necessary."

And the surgery is not lightweight, he stressed. The procedure required to remove pancreatic cysts is among one of the largest abdominal operations performed today, with a 1% to 3% mortality rate and complication rates as high as 50% that can have a life-long impact on quality of life.

"Current criteria are based on the philosophy that we would rather do an unnecessary operation than miss a cancer, so while we seldom miss a cancer, we accept a relatively high rate of nontherapeutic operations as a trade-off," he noted.

Hence, the study has the potential to revolutionize the management of pancreatic cysts, he concluded.

Investigators expect the CompCyst test will be available within 6 to 12 months within the Johns Hopkins medical system.

In the long-term, they also expect a prospective study will be done to further validate the algorithm.

These results will hopefully lead to the approval of the test, at which point the technology will be commercialized through Thrive, a company that has licensed the technology from Johns Hopkins University.

Wolfgang, Dal Molin, and Lennon have reported no relevant financial relationships. A number of co-investigators have reported receiving equity or royalties from companies and inventions licensed to Johns Hopkins University.

Sci Transl Med. Published online July 17, 2019. Abstract

For more from Medscape Oncology, join us on Twitter and Facebook.

Comments

3090D553-9492-4563-8681-AD288FA52ACE
Comments on Medscape are moderated and should be professional in tone and on topic. You must declare any conflicts of interest related to your comments and responses. Please see our Commenting Guide for further information. We reserve the right to remove posts at our sole discretion.
Post as:

processing....