Endoscopic ultrasound pushes beyond biopsy in diagnostics

Endoscopic ultrasound (EUS) is moving past its traditional role as a diagnostic imaging and sampling tool toward enabling early disease detection, molecular profiling, and personalized treatment, according to a comprehensive review published in Gastroenterology.

The review, led by Michael B. Wallace, MD, MPH, the Fred C. Andersen Professor of Medicine in the Division of Gastroenterology and Hepatology at Mayo Clinic, Jacksonville, Fla., describes how new technologies and workflows spanning molecular assays, advanced needle designs, and artificial intelligence (AI) are redefining the diagnostic potential of EUS across gastrointestinal and hepatobiliary diseases.

“In the past five years or so, there’s been a great deal of appropriate attention paid to therapeutic interventions of endoscopic ultrasound, particularly because we’ve seen such a dramatic shift in therapeutic options with the advent of lumen-apposing metal stents, and radiofrequency ablation in particular,” Dr. Wallace told GI & Hepatology News. “I think the advances in diagnostic endoscopic ultrasound have been underappreciated, a little bit overshadowed by those advances. So, I think that this is a very timely review.”

Highlights from the article, which Dr. Wallace assembled with coauthors Shounak Majumder, MD, of the Division of Gastroenterology and Hepatology at Mayo Clinic, Rochester, Minn.; Peter Storz, PhD, of the Division of Gastroenterology and Hepatology at Mayo Clinic, Jacksonville; and Jeanin E. van Hooft, MD, PhD, of the Department of Gastroenterology and Hepatology, Leiden University Medical Center, Leiden, the Netherlands, include the following:

Pancreatic and functional testing

In chronic pancreatitis (CP), EUS remains the preferred diagnostic modality when computed tomography or magnetic resonance imaging findings are inconclusive. Secretin-stimulated endoscopic pancreatic function testing (ePFT) can complement morphologic evaluation by measuring bicarbonate concentration in duodenal fluid to diagnose exocrine pancreatic insufficiency.

A prospective study cited in the review involving 145 patients with suspected early CP found diagnostic concordance between EUS and ePFT in 88 cases, allowing CP to be confidently confirmed or excluded. However, discordant findings limited certainty. The authors noted that added procedural time and interpretive expertise have restricted ePFT’s broader adoption.

Liquid biopsy for early pancreatic cancer detection

Secretin-stimulated pancreatic juice sampling via the duodenum — performed concurrently with EUS — now enables molecular “liquid biopsy” testing without cannulation of the pancreatic duct. In one prospective cohort of 88 pancreatic cancer cases and 134 controls cited in the review, a methylated DNA marker panel (C13orf18, FER1L4, and BMP3) combined with serum carbohydrate antigen 19-9 achieved an area under the receiver operating characteristic curve of 0.95, surpassing CA 19-9 alone (0.91; P = .0135). Sensitivity and specificity were 89% and 88%, respectively.

The review’s authors noted that these biomarkers, along with extracellular vesicle–derived microRNAs, are being investigated for cancer surveillance in high-risk individuals and cystic neoplasms.

Endohepatology: Pressure measurements and liver biopsy

According to the review, EUS-guided portal pressure gradient (PPG) measurement directly quantifies portal and hepatic vein pressures and strongly correlates with histologic fibrosis and clinical markers of portal hypertension. In multiple studies cited, the procedure demonstrated high technical success and low adverse event rates. AGA recently recognized EUS-PPG as a clinically available alternative to the transjugular approach.

EUS-guided liver biopsy also offers comparable yield, adequacy, and safety to percutaneous or transjugular routes, with the added advantage of sampling both hepatic lobes in the same session as PPG assessment. Dr. Wallace and his coauthors note that elastography integrated with EUS now permits noninvasive fibrosis staging, which can potentially reduce dependence on biopsy.

Enhanced imaging: Contrast and elastography

According to the review, contrast-enhanced (CE) EUS improves visualization of microvasculature and tissue perfusion using intravenous microbubbles, distinguishing hypovascular pancreatic adenocarcinoma from hypervascular neuroendocrine tumors and inflammatory lesions. CE-EUS is especially valuable for lesions ≤2 cm and for targeting fine-needle aspiration (FNA) when gray-scale imaging is inconclusive.

Meanwhile, EUS elastography provides real-time tissue stiffness mapping to help differentiate malignant from benign pancreatic masses and lymph nodes. Combined CE-EUS and elastography improve diagnostic accuracy to more than 95% for gastrointestinal stromal tumors versus leiomyomas, according to cited evidence.

Needle designs and acquisition techniques

Advances in fine-needle biopsy (FNB) technology have driven a shift away from FNA. According to data cited in the review, third-generation “end-cutting” needles, such as Franseen and fork-tip designs, now achieve diagnostic accuracy above 90% without increasing complication risk. Meta-analyses of up to 29 randomized controlled trials found 22-gauge fork-tip and Franseen needles provided the highest diagnostic accuracy, often requiring only two passes compared with four for FNA when rapid on-site cytopathology evaluation (ROSE) is unavailable.

In the realm of aspiration procedures cited in the review, wet-suction techniques, or using saline to fill the needle, produced the best tissue quality and lower blood contamination than dry suction, whereas no-suction or slow-pull techniques were preferred for hypervascular lesions. Macroscopic on-site evaluation (MOSE) and visual on-site evaluation (VOSE) emerged as practical alternatives to ROSE, allowing real-time specimen adequacy assessment without cytopathologist assistance.

Integrating AI

According to the article, AI-based models are being developed to enhance lesion characterization, automate reporting, and aid tissue interpretation. One deep-learning model analyzing EUS images achieved 94% sensitivity and 82% specificity for distinguishing malignant from benign pancreatic lesions. Another convolutional neural network differentiated autoimmune pancreatitis from pancreatic cancer with 90% sensitivity and 93% specificity.

AI also shows promise for standardizing EUS reports. A deep learning–driven automated report system trained on more than 235,000 images achieved 91% accuracy and higher completeness versus manual documentation (91% vs. 78%).

Cyst diagnostics and emerging tissue models

For pancreatic cysts, a meta-analysis cited in the review found that cyst fluid glucose levels <50 mg/dL yielded 90.8% sensitivity and 90.5% specificity for identifying mucinous cysts — performance comparable to or exceeding carcinoembryonic antigen. The combination of KRAS and GNAS mutation detection further improved molecular characterization.

Through-the-needle microforceps biopsy achieved diagnostic sensitivity and specificity of 80%, with a pooled adverse event rate of 7% across several meta-analyses. Needle-based confocal laser endomicroscopy provided 99% diagnostic accuracy for cyst typing in a recent meta-analysis, with low adverse event rates.

According to the review, EUS-acquired tissue now enables patient-derived tumor organoids, xenografts, and organotypic slice cultures that preserve the tumor microenvironment for testing individualized therapies. FNB samples have yielded successful organoid growth in 50–80% of cases, outperforming FNA specimens. 

 GI & Hepatology News invited first author Michael B. Wallace, MD, MPH, to elaborate on the article. Dr. Wallace is the Fred C. Andersen Professor of Medicine in the Division of Gastroenterology and Hepatology at Mayo Clinic, Jacksonville, Fla.,

How might this review influence clinical practice?

Dr. Wallace: I think we’re at a point now where screening for pancreatic cancer with EUS and associated imaging technologies like MRI have really come into the standard of care yet are underutilized. We now have very strong data from both the US and Europe cohorts in screening for pancreatic cancer that active screening with EUS and MRI is shifting the survival curve for pancreatic cancer through early detection. So, individuals who are at increased risk of pancreatic cancer, either because of family history or a known genetic variant, or most commonly a pancreatic cystic lesion, should undergo regular surveillance of their pancreas, EUS and MRI are the recommended technologies.

In liver disease, EUS has emerged as a true one-stop approach for patient care. It allows us to perform high-resolution endoscopic imaging, obtain liver biopsies, directly measure portal and hepatic venous pressures — often with greater accuracy than indirect measurements obtained via the internal jugular approach — and perform elastography in a single session. In addition, EUS enables both the diagnosis and treatment of varices.

Altogether, this comprehensive approach allows for thorough evaluation of the impact of liver disease on the portal venous system and varices. Multiple key studies have demonstrated that this method is highly accurate, safe, and more cost-effective compared with performing separate radiologic biopsies, pressure measurements, and endoscopic procedures.

What knowledge gaps remain in this field, and what research should be done next?

Dr. Wallace: I think still we have room to improve pancreatic cancer screening. Although we’ve made important advances, the ability to screen for solid cancers as opposed to cystic cancers is still limited because it’s very difficult to detect the lesions at stage one or especially the high-grade dysplasia stage.

This review highlights a highly promising area that warrants further investigation: biomarkers, particularly pancreatic juice–based biomarkers and circulating cell-free DNA assays. Blood-based biomarkers are gaining popularity, especially as commercial tests become more widely available; however, their diagnostic accuracy remains modest.

Pancreatic juice–based biomarkers may represent an important intermediate step. For patients already undergoing upper endoscopy, pancreatic juice can be collected after secretin stimulation without the need for ERCP, making this approach relatively convenient. Additionally, improving our ability to image pancreatic cysts and reliably distinguish high-risk lesions potentially without the need for biopsy — will become important, as many patients present with small, low-risk cysts that do not require invasive sampling.

Avoiding biopsy, which remains the most accurate diagnostic test but is invasive, would represent a significant advance. We are seeing rapid developments in artificial intelligence, both in EUS and in cross-sectional imaging modalities such as MRI. These innovations are likely to enable highly accurate classification of pancreatic cysts without the need for invasive tissue sampling.

Finally, once patients develop a full-blown adenocarcinoma of the pancreas, we currently really only have two options that are FDA-approved chemotherapy regimens, and it's very empiric as to what option we choose. Those two options would be combining the chemotherapy drugs gemcitabine and Abraxane or using the combination drug Folfirinox. We don't have a good way of predicting which of those two or other regimens patients will respond to, and so the ability to culture tumor cells and predict response to individual therapies, in addition to genetic factors in the tumor, I think will be an important advance.

Dr. Wallace disclosed that he has served as a consultant for Boston Scientific, ClearNote Health, Cosmo Pharmaceuticals, Endostart, Endiatix, Fujifilm, Medtronic, Surgical Automations, Ohelio Ltd, and Venn Bioscience; he has received research grants from Fujifilm, Boston Scientific, Olympus, Medtronic, Ninepoint Medical, and Cosmo/Aries Pharmaceuticals; he owns stock or stock options in Virgo Inc and Surgical Automation; and he has served as a consultant on behalf of Mayo Clinic: Boston Scientific, and Microtek.