Pancreatic cancer kills most patients because it is almost never found early. A new Nature Medicine study just validated a blood test that detects it at stage 1 or 2 in nearly 9 out of 10 cases.
Of all the cancers that medicine has struggled to treat, pancreatic cancer stands apart in a specific way. The problem is not primarily that effective treatment does not exist. For patients whose tumors are caught before they spread, surgery combined with chemotherapy produces five-year survival rates above 80%. The problem is that catching pancreatic cancer before it spreads is extraordinarily rare in clinical practice.
The pancreas sits deep in the abdomen, behind the stomach, producing no symptoms that would alert a patient or clinician to early disease. By the time pain, jaundice, weight loss, or digestive changes appear, the cancer has almost always infiltrated surrounding tissue or spread to lymph nodes or distant organs. At that stage, surgery is no longer possible. Systemic chemotherapy can extend life by months. Cure is not the expected outcome.
The result is a five-year survival rate of approximately 12%, the lowest of any major cancer. More than 66,000 Americans will be diagnosed with pancreatic cancer this year. Roughly 57,000 will die from it.
The fundamental bottleneck is detection. For decades, researchers have searched for a blood-based test capable of identifying pancreatic cancer early enough to allow curative surgery. The tumor marker CA19-9 exists and is widely used, but it is an inadequate screening tool: it misses a substantial proportion of early tumors, and it is elevated in several benign conditions, generating false positives that lead to unnecessary anxiety and invasive procedures. MRI and endoscopic ultrasound are more accurate, but they are expensive, resource-intensive, and available only to high-risk individuals who have already been identified as candidates for surveillance.
A new study published in Nature Medicine has produced the most compelling validation yet of a blood-based early detection test for pancreatic cancer. The test is called PANXEON.
What PANXEON measures and how it works
PANXEON was developed by Ajay Goel and colleagues at City of Hope National Medical Center in collaboration with investigators across the United States, Japan, Italy, and South Korea. Its design combines three distinct biological signals into a single integrated score.
The first signal comes from circulating microRNAs, small noncoding RNA molecules that regulate gene expression and are secreted by cells into the bloodstream. Specific microRNA patterns are associated with pancreatic cancer, and prior research had identified signatures that distinguish cancer patients from healthy controls. PANXEON uses a 13-microRNA signature derived from both freely circulating microRNAs and microRNAs packaged inside exosomes, tiny membrane-bound vesicles that cells release as a form of cellular communication.
The second signal is CA19-9, the existing tumor marker. On its own, CA19-9 is insufficient for early detection. Combined with the microRNA signature, it adds complementary information that improves both sensitivity and specificity.
The third element is artificial intelligence. An AI-derived algorithm integrates the microRNA signature and CA19-9 levels into a single composite score called PANXEON, calibrated against a training dataset and then tested in prospectively collected samples from patients who had not contributed to the algorithm’s development.
The approach addresses a long-standing criticism of cancer biomarker research: that many tests perform impressively in laboratory conditions but collapse when tested in independent, prospective cohorts that reflect real clinical diversity. PANXEON was developed with that failure mode explicitly in mind.
What the international study found
The prospective study enrolled 1,785 individuals from 12 institutions across four countries: the United States, Japan, Italy, and South Korea. Participants included people with confirmed pancreatic ductal adenocarcinoma at various stages, people with high-grade dysplasia (a precancerous condition in the pancreas that can progress to invasive cancer), people with other gastrointestinal cancers, and healthy controls. After excluding samples with inadequate RNA quality, 1,757 plasma samples from 1,649 participants were analyzed across training, validation, independent testing, treatment-monitoring, and cross-reactivity cohorts.
The key performance metrics came from the independent testing cohort, which was prospectively collected and completely separate from the data used to develop or refine the algorithm.
In this cohort, the full PANXEON composite score achieved an area under the receiver operating characteristic curve of 91.7% for distinguishing pancreatic cancer and high-grade dysplasia from controls. This is a measure of overall discriminative performance, where 100% represents perfect discrimination and 50% represents chance.
For early-stage pancreatic cancer specifically, defined as stage I or II disease for which curative surgery is possible, PANXEON achieved a sensitivity of 86.8% with a specificity of 96.8% in average-risk controls. In practical terms, this means PANXEON correctly identified approximately 87 out of every 100 early-stage pancreatic cancers while falsely flagging fewer than 4 out of every 100 negative tests.
Compared with CA19-9 alone at the same specificity threshold, PANXEON substantially reduced both false-negative and false-positive classifications.
“Our hypothesis was that combining cell-free and exosomal microRNAs would give us a broader molecular view of what is happening in the pancreas than either compartment alone,” Goel said. “The results in an international, independent testing cohort validated that hypothesis at a scale we had not previously achieved.”
The precancerous detection finding
One of the most clinically significant results in the study concerns high-grade dysplasia, an advanced precancerous condition of the pancreas that represents the stage immediately before invasive cancer. High-grade dysplasia is often considered “stage 0” pancreatic cancer. It can be found in pancreatic cysts that are detected incidentally during imaging for other reasons, and the clinical question of which cysts require intervention and which can be monitored safely is one of the most difficult decisions in gastroenterology.
PANXEON detected high-grade dysplasia in 64.2% of cases. While this is lower than its sensitivity for invasive early-stage cancer, it represents a meaningful signal for a condition that is currently detected primarily through invasive procedures or incidental imaging findings. Identifying precancerous pancreatic disease through a blood test could allow earlier, less invasive intervention in a population that currently either goes undetected or undergoes unnecessary procedures based on imaging alone.
“This finding could help physicians better identify which pancreatic cysts require active monitoring or intervention before invasive cancer develops,” Goel said.
Why sensitivity and specificity both matter in this context
The balance between sensitivity and specificity is not an academic detail for pancreatic cancer screening. It has direct consequences for how a test can be used and who benefits from it.
High sensitivity means the test catches most cancers that are present, minimizing the number of patients whose disease is missed. High specificity means the test rarely raises a false alarm in people who do not have cancer, minimizing the number of unnecessary follow-up procedures, anxiety, and costs in unaffected individuals.
For a blood test to function practically as a screening tool, both must be high. A test with 90% sensitivity but 60% specificity would generate an enormous number of false positives across a screened population, overwhelming clinical capacity with unnecessary biopsies and imaging while potentially causing harm through the procedures themselves. Conversely, a test with 99% specificity but 40% sensitivity would miss more than half of early-stage cancers, providing false reassurance to many patients who need treatment.
PANXEON’s combination of 86.8% sensitivity and 96.8% specificity represents a performance profile that had not previously been achieved in a prospective, international study of pancreatic cancer detection. Whether that performance holds in a true population-level screening context, where the prevalence of pancreatic cancer is much lower than in a clinically recruited study cohort, is the next critical question.
“A false-positive blood test can lead to additional imaging, invasive procedures and considerable anxiety; a false-negative result can delay evaluation of a potentially curable cancer,” the research team wrote. “PANXEON reduced both types of errors compared with CA19-9 alone.”
What this test does not yet offer and what comes next
PANXEON is investigational. It is not currently approved by the FDA or any regulatory body, and it is not available in clinical practice. The study published in Nature Medicine establishes its performance in a prospective international cohort, which is a necessary but not sufficient condition for clinical use.
The critical next step is a prospective population-level screening trial, in which PANXEON would be administered to a large cohort of individuals with no known pancreatic cancer, and the results would be tracked to determine how many cancers it identifies, how many are false positives, and what happens to patients whose cancers are caught through screening versus standard clinical detection.
This type of trial takes years and requires thousands of participants to be followed across a period long enough for some of them to develop cancer. The performance metrics achieved in a clinically recruited cohort, where participants are generally referred because they have symptoms, a known risk factor, or a cyst under surveillance, cannot be directly extrapolated to population-level screening of asymptomatic people.
The study also does not establish that finding pancreatic cancer earlier through PANXEON will improve survival in a screened population, which is the ultimate clinical goal. Demonstrating that requires a randomized trial comparing screened to unscreened populations, which has not yet been conducted.
What the study establishes is the foundation for those trials. The performance profile achieved in 1,785 patients across four countries justifies moving PANXEON into prospective screening studies. Previous tests have not cleared that bar. PANXEON has.
Who benefits most if PANXEON reaches clinical practice
While population-level screening would be the ultimate goal, the most immediate beneficiaries of a validated early detection test would be people already known to be at elevated risk for pancreatic cancer, for whom no adequate blood-based test currently exists.
This includes people with first-degree relatives who have had pancreatic cancer, individuals carrying specific genetic mutations including BRCA1, BRCA2, PALB2, and ATM, people with hereditary pancreatitis, Lynch syndrome carriers, and people with new-onset diabetes after age 50, who have a somewhat elevated pancreatic cancer risk.
For these groups, current surveillance relies on MRI and endoscopic ultrasound, which require specialized facilities, carry procedural risks, and must be repeated at regular intervals for years or decades. A reliable blood test would provide a less invasive, more frequent, and more accessible monitoring option that could complement or, over time, partially replace imaging-based surveillance.
“A validated blood-based biomarker like PANXEON could transform how we approach pancreatic cancer surveillance in high-risk populations,” the research team noted, “by providing an accessible, repeatable, and minimally invasive complement to imaging.”
What the cross-reactivity finding means
A critical challenge for any cancer blood test is specificity not just against healthy controls but against other cancers. A test that detects pancreatic cancer but also triggers elevated scores in colorectal, gastric, or biliary cancers would be of limited value in clinical settings where multiple gastrointestinal cancers are in the differential diagnosis.
The PANXEON study specifically tested for cross-reactivity against other gastrointestinal malignancies including colorectal, gastric, and hepatocellular cancer. The test showed minimal cross-reactivity with these cancers, maintaining its discriminative performance when gastrointestinal cancer patients were included in the control group.
This cross-reactivity profile is clinically important because many patients who would be evaluated for pancreatic cancer are also at risk for or undergoing evaluation for other gastrointestinal malignancies. A test that cannot distinguish between them would generate ambiguous results in exactly the clinical scenarios where clarity matters most.
The study, “Liquid biopsy for early detection of pancreatic ductal adenocarcinoma”, was authored by Cuncong Xu, Minyi Gu, Alok K. Tewari, and Ajay Goel at City of Hope National Medical Center and collaborating institutions in the United States, Japan, Italy, and South Korea, and published September 16, 2026 in Nature Medicine.
Source: City of Hope National Medical Center. DOI: 10.1038/s41591-026-04625-x
