EUS-FNA of Pancreatic Cysts: When to Aspirate, How to Read CEA 192, and When to Operate

Clinical Bottom Line

Decision	Do this	Do not do this
Aspirate?	Unclear cyst in a surgical candidate whose result will change management. Sample any mural nodule or solid component, not only the fluid.	Classic honeycomb serous cystadenoma with a central scar. Obvious pseudocyst after pancreatitis. Already an absolute surgical indication. Patient unfit for pancreatectomy.
Read CEA	Mucinous vs nonmucinous. Classic cutoff 192 ng/mL (Brugge [8] accuracy 79%). Very low CEA (<5) favors SCA or pseudocyst. Very high CEA (>800) is more specific for mucinous.	Treat CEA 192 as a guarantee. Use CEA to call cancer or to send to surgery.
Glucose	Low (<=50 mg/dL) favors mucinous. High glucose argues against mucinous. Cheap, small volume.	Use glucose to grade dysplasia. Pretend it is in ACG 2018 [2].
Amylase	Amylase <250 U/L virtually excludes a pseudocyst.	High amylase "confirms" a pseudocyst. Branch-duct IPMN is high-amylase too.
Molecular	KRAS/GNAS classifies mucinous/IPMN. TP53, SMAD4, CDKN2A, or PIK3CA flag advanced neoplasia. Use NGS when the result will change surveillance or surgery.	Treat KRAS alone as cancer.
Surgery	High-risk stigmata or multiple worrisome features plus multidisciplinary review. MD-IPMN in fit patients. SPN.	CEA elevation alone. Size alone without other features. Guidelines disagree on the size cut.

What do I do tonight?

1. If MRI already shows a classic honeycomb serous cystadenoma, an obvious pseudocyst, or high-risk stigmata that already send a fit patient to surgery, skip FNA.
2. If the patient is not a surgical candidate, skip FNA. The fluid will not change the plan.
3. If you aspirate, sample any mural nodule or solid component, not only the fluid.
4. Read CEA as mucinous vs nonmucinous (classic cutoff 192 ng/mL). Do not use CEA as a cancer test or a ticket to the operating room.
5. Low glucose (<=50 mg/dL) favors mucinous. High amylase does not prove a pseudocyst. Amylase <250 helps exclude one.
6. Operate on Kyoto [5]/Fukuoka high-risk stigmata (jaundice from a head cyst, enhancing mural nodule >=5 mm, MPD >=10 mm, suspicious cytology), not on CEA height.

What do the societies say?

Do not use cyst-fluid CEA as a cancer test. Do not treat high amylase as proof of a pseudocyst. Do not apply the ASGE 2024 FNB-over-FNA solid-mass recommendation to cyst fluid.

When should I skip EUS-FNA?

ACG rec 4 (strong): patients not medically fit for surgery should not undergo further evaluation of incidental cysts. Rec 5: asymptomatic pseudocyst or serous cystadenoma, no treatment. Rec 6: EUS-FNA when diagnosis is unclear and results will alter management. European 2018: do not FNA if imaging already made the diagnosis or there is already a clear indication for surgery. ESGE 2017: skip sampling of lesions <=10 mm with no high-risk stigmata. AGA 2015 is more conservative still: EUS-FNA only if at least two of size >=3 cm, dilated main duct, or solid component, and stop surveillance at 5 years if stable. Name the guideline you are following. Do not collapse AGA, ACG, European, and Kyoto into one clock.

ASGE 2024 FNB-over-FNA is for solid pancreatic masses. For a purely cystic lesion, fluid FNA remains the default. If there is a mural nodule or solid component, sample the solid part with solid-mass technique (ESGE 2017). Through-the-needle microforceps raise histologic yield and also raise pancreatitis and intracystic bleeding versus fluid FNA. Not default for every cyst.

Infection after cyst FNA is uncommon. Colan-Hernandez 2020 (ciprofloxacin vs saline): one FNA-related infection in 205 completed follow-ups. ESGE 2017 still suggested prophylaxis (weak). The RCT is the better 2026 citation that routine multi-day antibiotics are not required. Local practice may still give a single peri-procedural dose.

Does cyst-fluid CEA diagnose cancer?

Brugge 2004 (cooperative pancreatic cyst study, 112 resected): CEA at 192 ng/mL had AUC 0.79 and accuracy 79%, better than EUS morphology or cytology. That is the origin of the number. Gaddam [10] 2015 (surgical histology n=226): at 192 ng/mL, sensitivity 61%, specificity 77%; would miss 39% of mucinous cysts. Kwan [11]/Pitman 2024 (MGH, including the original Brugge cohort plus 20 more years, histology n=394): at 192, sensitivity 56%, specificity 78%. CEA remains useful and is less specific than 2004 reported.

Cizginer: mean CEA in malignant mucinous cysts 2,558 vs 4,700 ng/mL in benign mucinous cysts. No separation. Ngamruengphong's meta of CEA for malignancy: pooled sensitivity 63%, specificity 63%. Do not send to surgery on CEA height. van der Waaij: CEA <5 favors SCA or pseudocyst (specificity 95%); CEA >800 favors mucinous (specificity 98%) with low sensitivity.

Amylase <250 U/L virtually excludes a pseudocyst (van der Waaij: specificity 98%). High amylase is compatible with pseudocyst and with IPMN. The old stub's "high amylase confirms a pseudocyst" line is false.

What does cyst-fluid glucose add?

Carr (pathologically confirmed n=153): glucose <=50 mg/dL, sensitivity 92%, specificity 87%, accuracy 90%, versus CEA >192 accuracy 69%. Glucose, like CEA, did not diagnose malignant disease. Zikos: laboratory glucose <50, sensitivity 95%; glucometer works. McCarty and Faias metas: glucose pooled sensitivity about 90% vs CEA about 56-67%. Glucose metas are smaller than CEA metas. Say so. Pair glucose with CEA when you have fluid. Do not replace CEA overnight, and do not use either test to grade dysplasia.

When do I send cyst fluid for molecular testing?

Singhi 2018 NGS (surgical subset n=102): KRAS/GNAS 89% sensitivity and 100% specificity for a mucinous cyst. KRAS/GNAS plus TP53/PIK3CA/PTEN: 89% sensitivity and 100% specificity for advanced neoplasia in that subset, better than duct dilation, mural nodule, or malignant cytopathology. GNAS is IPMN-restricted and helps separate IPMN from MCN (Kyoto CQ5). Sanger-era KRAS/GNAS was less sensitive. PANDA is the early DNA study, not the current assay. CompCyst is a research classifier on resected cohorts, not a community lab test.

When do I operate?

Kyoto 2024 high-risk stigmata: obstructive jaundice with a head cyst; enhancing mural nodule >=5 mm or solid component; MPD >=10 mm; suspicious or positive cytology. Worrisome features now include growth >=2.5 mm/year and new or recently exacerbated diabetes. Fukuoka 2017 growth was >5 mm/2 years. European absolute indications align on jaundice, enhancing nodule >5 mm, and MPD >10 mm, but use 40 mm (not 30 mm) as a relative size cut and allow observation of some asymptomatic MCN <40 mm without a nodule. ACG sends jaundice, pancreatitis from the cyst, mural nodule, MPD >5 mm, mucinous cysts >=3 cm, and concerning cytology to EUS and a multidisciplinary group rather than straight to the OR. AGA wants solid component plus dilated duct, or concerning FNA.

Unselected neoplastic cyst cancer risk is about 0.24% per year (AGA technical review / ACG). Resected-series cancer rates (about 15% of resected cysts, higher in resected IPMN) are selected. Do not equate mucinous with malignant. Postoperative mortality is about 2% in high-volume pancreas centers vs higher in mixed SEER data. Refer resection to a high-volume center.

Surveillance clocks disagree. ACG: size-banded MRI, stop if not a surgical candidate. Kyoto: optional stop for <20 mm cysts with no morphologic change and no worrisome features after 5 years. European: lifelong follow-up of IPMN while fit. AGA: stop at 5 years if stable. Print the conflict. Book the next MRI only after you name the document.

Pitfalls

Pitfall	Better move
CEA >192 "guarantees" a mucinous cyst	Brugge accuracy was 79%. Gaddam and Kwan show worse sensitivity at the same cut.
High CEA means high cancer risk	Malignant and benign mucinous CEA means overlap (Cizginer).
High amylase confirms a pseudocyst	Low amylase excludes. High amylase does not confirm.
Citing ASGE solid-mass FNB data as cyst-fluid data	FNB for solid components. Fluid FNA for a pure cyst.
One algorithm for AGA, ACG, European, and Kyoto	Tabulate the conflict. Name the guideline on the report.
Using resected-IPMN cancer rates for an incidental 1.5 cm BD-IPMN	Use the unselected ~0.24%/year estimate for that sentence.

Selected references

1. Vege SS, et al. AGA guideline: asymptomatic neoplastic pancreatic cysts. Gastroenterology. 2015.
2. Elta GH, et al. ACG Clinical Guideline: pancreatic cysts. Am J Gastroenterol. 2018.
3. European Study Group. European evidence-based guidelines on pancreatic cystic neoplasms. Gut. 2018.
4. Tanaka M, et al. Fukuoka 2017 revisions for IPMN. Pancreatology. 2017.
5. Ohtsuka T, et al. Kyoto guidelines for IPMN. Pancreatology. 2024.
6. ASGE. Role of endoscopy in cystic pancreatic neoplasms. Gastrointest Endosc. 2016.
7. Dumonceau JM, et al. ESGE: indications for EUS-guided sampling. Endoscopy. 2017.
8. Brugge WR, et al. Cooperative pancreatic cyst study. Gastroenterology. 2004.
9. van der Waaij LA, et al. Pooled cyst fluid analysis. Gastrointest Endosc. 2005.
10. Gaddam S, et al. Suboptimal accuracy of CEA. Gastrointest Endosc. 2015.
11. Kwan MC, et al. 20-year institutional review of cyst-fluid CEA. Gut. 2024.
12. Cizginer S, et al. CEA does not distinguish benign from malignant mucinous cysts. Pancreas. 2011.
13. Carr RA, et al. Cyst fluid glucose. Surgery. 2018.
14. McCarty TR, et al. Meta-analysis of cyst fluid glucose. Gastrointest Endosc. 2021.
15. Singhi AD, et al. NGS of pancreatic cyst fluid. Gut. 2018.
16. Colan-Hernandez J, et al. Antibiotic prophylaxis RCT for cyst FNA. Gastroenterology. 2020.