Nature Medicine September 16, 2026

Liquid biopsy for early detection of pancreatic ductal adenocarcinoma

Caiming Xu, Alessandro Mannucci, Haiyong Han, Ruben M. Munoz, Derek Cridebring, Sourat Darabi, Yuji Toiyama, Yoshinaga Okugawa, Gagandeep Singh, Mustafa Raoof, Ajay Goel, et al.

Bottom Line

An international prospective biomarker study developed and validated PANXEON, a liquid biopsy assay combining a 10-miRNA signature with CA 19-9, demonstrating high sensitivity for detecting early-stage pancreatic ductal adenocarcinoma and pre-cancerous high-grade dysplasia.

Key Findings

1. The standalone miRNA signature achieved an area under the receiver operating characteristic (AUC) curve of 88.6% in the testing cohort, with a sensitivity of 83.8% for early-stage PDAC.
2. When combined with carbohydrate antigen 19-9 (CA 19-9) levels, the composite PANXEON blood assay achieved a sensitivity of 86.8% for stage I-II PDAC.
3. The assay maintained low false-positive rates of 3.2% in low-risk controls and 15.6% in high-risk controls in the testing cohort, while demonstrating minimal cross-reactivity with other gastrointestinal cancers.
4. In a small sub-cohort of 19 individuals, the targeted miRNA signature levels successfully tracked disease burden, decreasing during neoadjuvant chemotherapy and after surgical resection, and increasing prior to disease recurrence.
5. PANXEON detected high-grade dysplasia in individuals with high-risk pancreatic cysts with a sensitivity of 64.3%, indicating strong potential for pre-cancer surveillance.

Study Design

Design
Prospective Observational
N/A
Sample
1,785
Patients
Duration
N/A
Median
Setting
Four countries
Population 1,785 individuals with and without pancreatic ductal adenocarcinoma, comprising patients with PDAC, low-risk controls, high-risk controls, and individuals with high-risk pancreatic cysts.
Intervention Evaluation of blood samples using PANXEON, an assay integrating a specific 10-miRNA signature with carbohydrate antigen 19-9 (CA 19-9).
Comparator Clinical and pathological confirmation of diagnosis (early-stage PDAC, high-grade dysplasia, or benign/healthy control status).
Outcome Diagnostic performance for early-stage (stage I-II) PDAC and high-grade dysplasia, quantified by sensitivity, false-positive rates, and area under the receiver operating characteristic curve (AUC).

Study Limitations

The subset evaluating longitudinal disease tracking (neoadjuvant chemotherapy, surgery, and recurrence) was very small (n=19), limiting robust conclusions about its utility for monitoring therapeutic response.
The 15.6% false-positive rate among high-risk controls could still lead to a substantial number of unnecessary invasive procedures or significant psychological distress if broadly implemented without careful clinical triaging.
While the assay shows strong diagnostic accuracy, further large-scale prospective clinical trials are necessary to validate its real-world impact on survival and clinical decision-making before adoption into standard screening protocols.

Clinical Significance

Pancreatic ductal adenocarcinoma is typically diagnosed at advanced, incurable stages due to the lack of early symptoms and reliable screening tests. PANXEON offers a non-invasive, highly sensitive (86.8%) liquid biopsy for stage I-II PDAC and precancerous lesions (64.3%), potentially enabling surgical resection and curative interventions before the disease metastasizes.

Historical Context

Historically, CA 19-9 has been the only established blood biomarker for pancreatic cancer, but it lacks the sensitivity and specificity needed for early-stage screening, often presenting as elevated in benign conditions like biliary obstruction. Over the past decade, efforts to create viable screening tools have shifted toward liquid biopsies, focusing on circulating tumor DNA (ctDNA) and exosomes. This study represents a significant leap by combining multi-omic markers (exosomal/circulating miRNAs) with traditional protein markers (CA 19-9) via machine learning to successfully identify early-stage disease.

Guided Discussion

High-yield insights from every perspective

Med Student
Medical Student

How does the biological role of microRNAs (miRNAs) make them suitable biomarkers for early-stage malignancies like PDAC compared to traditional circulating protein biomarkers?

Key Response

miRNAs are small non-coding RNAs that regulate gene expression post-transcriptionally. Because they are often packaged in exosomes or bound to Argonaute proteins, they are highly stable in circulation, resisting RNase degradation much better than cell-free DNA or some proteins. This stability, combined with their early dysregulation in the adenoma-carcinoma sequence (e.g., during PanIN progression), makes them excellent candidates for early-detection liquid biopsies.

Resident
Resident

If a high-risk patient (e.g., new-onset diabetes over age 50 or BRCA mutation carrier) tests positive on the PANXEON assay but has a completely negative pancreas protocol CT, what is the next best step in clinical management?

Key Response

A positive liquid biopsy with negative cross-sectional imaging creates a clinical dilemma. The next best step typically involves Endoscopic Ultrasound (EUS), which is highly sensitive for detecting sub-centimeter solid lesions or high-grade dysplasia that CT scans often miss. This highlights the assay's role as a screening adjunct that requires subsequent localizing and tissue-sampling modalities.

Fellow
Fellow

The PANXEON assay detects both early-stage PDAC and high-grade dysplasia. In the context of pancreatic cystic neoplasms such as IPMNs, how might the ability to specifically identify high-grade dysplasia alter our current endoscopic surveillance paradigms?

Key Response

Current IPMN surveillance relies on morphologic features (e.g., Fukuoka or AGA guidelines) which have only moderate accuracy for predicting high-grade dysplasia or invasive cancer. A highly sensitive miRNA/CA19-9 liquid biopsy could definitively risk-stratify patients with indeterminate cysts, potentially reducing unnecessary prophylactic surgeries for low-grade lesions while expediting EUS/FNA or surgical resection for those harboring high-grade dysplasia.

Attending
Attending

Given the historically low incidence of PDAC in the general population, how does Bayes' theorem and the positive predictive value (PPV) of the PANXEON assay dictate which patient populations we should actually screen in clinical practice?

Key Response

Even a highly specific test (e.g., 98-99%) will yield massive numbers of false positives if applied to the general population where PDAC prevalence is extremely low. Therefore, to achieve a clinically acceptable PPV and avoid a cascade of invasive, morbid diagnostic workups (like unnecessary pancreatectomies), PANXEON must be restricted to high-risk cohorts, such as those with familial pancreatic cancer syndromes, Peutz-Jeghers, or chronic pancreatitis.

Scholarly Review

Critical appraisal through the lens of expert reviewers and guideline development

PhD
PhD

How does integrating a 10-miRNA signature with CA 19-9 methodologically address the biological limitations of CA 19-9, specifically regarding false negatives in Lewis antigen-negative populations and false positives in benign biliary obstruction?

Key Response

Roughly 5-10% of the population lacks the Lewis antigen enzyme (FUT3) and cannot produce CA 19-9, while benign obstructive jaundice often falsely elevates CA 19-9. Integrating a multi-target miRNA panel utilizes orthogonal epigenetic pathways to rescue false negatives in Lewis-null patients and improves specificity by mathematically distinguishing true neoplasia-driven epigenetic signatures from simple inflammatory or cholestatic profiles.

Journal Editor
Journal Editor

In evaluating the prospective validation cohort for the PANXEON assay, what specific biases in the control group composition must a peer reviewer scrutinize to ensure the reported specificity is not artificially inflated?

Key Response

A classic flaw in biomarker studies is 'spectrum bias,' where advanced cancer patients are compared to young, perfectly healthy controls. A rigorous reviewer would demand that the control group closely matches the intended screening population, including older adults with comorbidities, chronic pancreatitis, and benign biliary disease, to ensure the assay actually discriminates against clinically relevant confounders rather than just identifying perfect health.

Guideline Committee
Guideline Committee

Current CAPS (International Cancer of the Pancreas Screening) consortium guidelines recommend MRI/EUS for high-risk individuals but do not endorse routine biomarker screening. What specific evidentiary threshold must the PANXEON assay meet to be incorporated into updated CAPS or NCCN guidelines as a first-line screening modality?

Key Response

To update CAPS or NCCN guidelines, the assay must demonstrate clinical utility beyond high sensitivity/specificity. Specifically, prospective trials must prove that integrating PANXEON into screening algorithms leads to definitive 'stage-shifting' (detecting stage I or high-grade dysplasia instead of stage III/IV) and ultimately improves overall survival, without causing unacceptable harm from false-positive-driven invasive procedures.

Clinical Landscape

Noteworthy Related Trials

2020

DETECT-A Trial

n = 10,006 · Science

Tested

CancerSEEK multi-cancer liquid biopsy test

Population

Women aged 65 to 75 without prior cancer history

Comparator

Standard of care screening alone

Endpoint

Feasibility and safety of multi-cancer early detection

Key result: The blood test safely doubled the number of cancers first detected by screening without inducing excessive futile diagnostic procedures.
2021

CCGA Study

n = 15,254 · Ann Oncol

Tested

Targeted methylation analysis of circulating cell-free DNA

Population

Participants with newly diagnosed untreated cancer and healthy volunteers

Comparator

Non-cancer control subjects

Endpoint

Sensitivity and specificity for cancer detection and tissue of origin

Key result: The liquid biopsy assay demonstrated 99.5 percent specificity and robust sensitivity across more than 50 cancer types, including those without standard screening options.
2022

PATHFINDER Trial

n = 6,662 · Lancet

Tested

Galleri multi-cancer early detection (MCED) test analyzing cfDNA methylation

Population

Adults aged 50 years and older

Comparator

Standard clinical evaluation

Endpoint

Diagnostic resolution time and positive predictive value

Key result: The MCED test achieved a positive predictive value of 43.1 percent and accurately predicted the cancer signal origin in 85.3 percent of true positive cases.

Tailored to your role

Want this tailored to you?

Add your specialty or training stage to get role-specific takeaways and more questions.

Personalize this analysis