PANXEON blood test catches 87 percent of early pancreatic cancers

A blood test built at City of Hope detected 86.8 percent of stage I and II pancreatic ductal adenocarcinomas in a prospective study of 1,785 people across the United States, Europe and Asia. The paper appeared in Nature Medicine on 16 September. False-positive rates were 3.2 percent in low-risk controls and 15.6 percent in high-risk controls. The same assay flagged high-grade dysplasia, a precancerous state inside high-risk pancreatic cysts, 64.3 percent of the time.

The test is called PANXEON, for PANcreatic cancer eXosome Early detectiON. It combines a panel of circulating and exosomal microRNAs with the older protein marker CA19-9 and then uses a model to emit a single risk score. The microRNA list in the paper includes hsa-miR-142-3p, hsa-miR-30c-5p, hsa-miR-335-5p, hsa-miR-340-5p, hsa-miR-200b-3p, hsa-miR-1260b, hsa-miR-145-3p, hsa-miR-145-5p, hsa-miR-429 and hsa-miR-200a-3p. On its own the miRNA signature produced an area under the ROC curve of 88.6 percent in the testing cohort and 83.8 percent sensitivity for early-stage disease, with little cross-reaction against other gastrointestinal cancers.

Ajay Goel, chair of molecular diagnostics and experimental therapeutics at City of Hope and senior author of the study, put the clinical claim in a sentence. "A stage shift is not just a statistic. The earlier we find pancreatic cancer, the greater the chance that meaningful intervention is still possible." Five-year survival for the disease sits at about 14 percent, the worst among common solid tumours, because most cases are found after they have left the pancreas.

CA19-9 by itself misses too many early tumours and rises in benign disease. Imaging finds masses that are already large. What the field has lacked is a blood test that can sit in a surveillance clinic for people who already carry risk: a strong family history, a known mutation, chronic pancreatitis, or a cyst that radiologists do not trust. PANXEON is aimed at that room, not at the general population. Goel said as much. The test is a triage tool. A high score would send a patient to endoscopic ultrasound or MRI, not straight to an operating theatre.

A smaller group of 19 patients offered a second signal. Signature levels fell during neoadjuvant chemotherapy and after surgery, then rose before recurrence. That pattern, if it holds in a larger series, would give oncologists a blood trace to watch between scans.

The dysplasia result is the piece other coverage has underplayed. High-grade dysplasia inside a cyst is often treated as stage 0. Finding 64.3 percent of those lesions, while staying negative in 75.5 percent of cysts without high-grade dysplasia, is not good enough for a stand-alone decision. It is good enough to argue that the miRNA signal appears before invasion. The paper says the miRNA arm, not CA19-9, carried most of that early signal.

Limits sit in the same tables. Fifteen percent false positives in high-risk controls means a large surveillance clinic will generate extra scans. The study is observational. It is not a randomised trial that shows lives saved. The authors write that large prospective studies are still required. No regulator has approved the assay for sale.

What changed on 16 September is the size of the evidence, not the standard of care. A 1,785-person, four-country series is now in a major journal with named molecules, named error rates and a defined use case. The next number that matters is how the score behaves when the people drawing the blood are ordinary hospital labs, not the teams that designed the panel.