AI-Powered Blood Test Detects Early Pancreatic Cancer in Study of 1,800 Patients

An investigational blood test that combines three biological markers with an artificial intelligence scoring system correctly identified early-stage pancreatic cancer in a large international study, according to findings published this week in Nature Medicine.

The test, called PANXEON, is the first liquid biopsy to pull together circulating microRNAs, exosomal microRNAs, and the protein CA19-9 into a single composite risk score. According to a Technology Networks report on the study, that score is calculated by an AI model that weights the three signals together rather than relying on any single marker in isolation.

The case for combining signals matters because the existing standard, CA19-9, has long been known to be an imprecise screener on its own. As reported by Medical News Today, the study enrolled nearly 1,800 people across the United States, Europe, and Asia, testing PANXEON in patients with different levels of pancreatic cancer risk rather than comparing cancer cases only against healthy volunteers. That design is more clinically realistic, since the real-world screening population is full of people who have inherited risk, family history, pancreatic cysts, or chronic pancreatitis.

The results, as reported by Healio, were that PANXEON correctly identified stage 1 and 2 cancers 87% of the time, with a false-positive rate of just 3% in low-risk groups. The false-positive rate climbed to 16% in high-risk groups, which is the population a screening program would most likely target. The test also showed sensitivity for high-grade dysplasia, a precancerous condition that could open a treatment window before invasive cancer develops.

That last point is the one oncologists will lean on most when they read this study. Pancreatic cancer kills roughly 52,740 Americans this year, according to Medical News Today, largely because most patients don't know they have it until it's past the point where surgery can help. The disease is nearly always asymptomatic in its early stages.

"Pancreatic cancer remains so deadly largely because we find it after the window for cure has begun to close," senior study author Ajay Goel, PhD, chair of City of Hope's department of molecular diagnostics and experimental therapeutics, said in a statement reported by Healio. Goel's lab led the research.

But the results need context before anyone starts expecting this test at their next physical. The study is prospective, not yet the kind of large, multi-site, blinded validation trial that would support a regulatory submission. Healio noted the investigators themselves wrote that additional large-scale prospective studies are warranted to confirm the test's utility. That's an honest accounting, and it's what the data actually support at this stage.

Patients most likely to see this test in a clinical setting first, if and when it clears validation, are those already in high-risk surveillance programs: people with a strong family history of pancreatic cancer, known genetic mutations like BRCA2 or PALB2, or imaging findings like worrisome pancreatic cysts. For that group, as Technology Networks reported, current monitoring relies on imaging that's expensive, invasive, and only moderately useful for catching cancer at its earliest stages. A reliable, repeatable blood draw could change that calculus in a meaningful way.

Separately, a $225,000 grant from the Pancreatic Cancer Action Network reported by the University of Miami's InventUM is funding a parallel study at Sylvester Comprehensive Cancer Center, also focused on blood-based ctDNA testing for metastatic pancreatic cancer and pairing molecular profiling with AI tools. That work uses samples from PanCAN's Precision Promise trial and will benchmark liquid biopsy results against tissue testing from the same patients.

For now, PANXEON remains investigational. The path from a promising 1,800-person study to a test a gastroenterologist can order runs through regulatory review, reimbursement decisions, and a clinical infrastructure that still doesn't have a clear framework for pancreatic cancer screening. None of that is a knock on the science. It's just what the road looks like.

Sources cited:
- Medical News Today (https://www.medicalnewstoday.com/articles/ai-enhanced-biomarker-blood-test-may-detect-pancreatic-cancer-earlier)
- Healio (https://www.healio.com/news/hematology-oncology/20260916/blood-test-may-detect-pancreatic-cancer-earlier-allow-for-meaningful-intervention)
- Technology Networks (https://www.technologynetworks.com/diagnostics/news/blood-test-shows-promise-for-detecting-pancreatic-cancer-at-its-earliest-stages-416609)
- InventUM (University of Miami) (https://news.med.miami.edu/pancan-grant-ctdna-blood-test-pancreatic-cancer/)