In most institutions, pancreatic EUS-FNA aspirates are prepared for cytologic review by the direct smear method. Direct smears are usually performed in the endoscopy suite; as soon as the aspirated material is obtained, the material is expelled onto glass slides. Grossly visible solid particles are placed into formalin for histopathologic examination, while the remaining material is smeared on site. The smears are either placed directly into alcohol fixative for Papanicolaou or hematoxylin-eosin stain or air-dried for subsequent Diff-Quik stain. If smeared properly without crushing or air dry artifacts, Papanicolaou or hematoxylin-eosin stains result in smears with high nuclear detail, which is important for pathologic diagnosis of various pancreatic neoplasms. Direct feedback helps to improve cytology smears, including microscopic observation of the smear slide by the person who actually performs the smear. Unfortunately, cytology smear preparation and cytology interpretation are usually performed by different people. Romanowsky stains, such as DiffQuik, are useful for analyzing background material, such as mucin, and for examining cytoplasmic features. Cell blocks, which may be useful for ancillary immunohistochemical stains, may also be prepared from aspirates. Recently, liquid-based cytology (LBC; e.g. ThinPrep, SurePath) has also been used for pancreatic EUS-FNA cytology.

LBC has become a standard method for cytologic evaluation of some organs, including the uterine cervix and thyroid. LBC can eliminate unnecessary background inflammatory and blood cells to increase diagnostic accuracy. Recent studies reported that LBC showed similar results to conventional smear for diagnosis of pancreatic masses [3,4]. The use of LBC may be limited for mucinous neoplasms because mucin is more dilute and scant on LBC preparations.

Because EUS-FNA involves penetration of either the gastric or duodenal wall to access the pancreas, most EUS-FNA aspirates contain at least a small amount of normal gastrointestinal mucosa. This is a very important pitfall in the interpretation of pancreatic EUS-FNA specimens, as normal gastrointestinal contaminants may be mistaken for neoplastic epithelial cells. Gastric foveolar epithelium appears as large irregular folded or monolayered sheets of mucin-containing columnar cells (Fig. 1A). The columnar cells have basally oriented nuclei with a palisaded appearance, and the luminal border is frequently seen along one edge of the epithelial cell sheets. Duodenal epithelium is also characterized by sheets of columnar cells with interspersed goblet cells, resulting in a characteristic “starry-sky” appearance (Fig. 1B). However, the most common contaminants are red blood cells, which obscure the cytologic feature of EUS-FNA aspirates. Cell blocks are alternative tools to help overcome bloody smears; however, the efficiency of conventional cell block preparation is sometimes suboptimal. The tissue coagulum clot method reportedly increases cell block cellularity with endobronchial ultrasound-guided transbronchial fine-needle aspiration [5]. However, more validation studies are needed to apply this method to EUS-FNA for pancreatic lesions.

As cystic lesions comprise 2%–13% of pancreatic lesions [6], it is not uncommon for pathologists to encounter EUS-FNA specimens for pancreatic cystic lesions. While characteristic cytologic features may help in diagnosis when a mural or papillary nodule is present within the pancreatic cystic lesion, purely cystic lesions without mural or papillary nodules may yield paucicellular smears or only scattered macrophages. These cases would be interpreted as “inadequate” according to ordinary criteria for cytology diagnosis. However, it is difficult to obtain adequate cellularity from these cystic lesions despite repeated procedures. This is especially true for unilocular cystic lesions, such as pseudocysts, macrocystic serous or mucinous cystadenomas, and cystically dilated intraductal papillary mucinous neoplasms (IPMNs) without mural nodules; if the gross and microscopic findings of these cystic lesions are considered, it is logical that not enough cells or tissue may be obtained by the EUS-FNA procedure. In this situation, it may be more reasonable to interpret the findings as “benign cystic lesion” than “inadequate specimen.” Although a hidden malignancy may exist around the “benign cystic lesion”, this pattern is very rarely identified. In daily practice, benign cystic lesions, such as serous or mucinous cystadenomas, cystic IPMNs, and pseudocysts, are more common. Pathologists should correlate radiologic and EUS findings to avoid underevaluation of pathologic diagnosis for these cystic lesions. If imaging and EUS findings are concordant, it is not difficult to predict the biologic behavior of pancreatic cystic lesions. However, discrepancy between imaging and EUS interpretation for these cystic lesions is not uncommon. In this scenario, pathologists should raise the threshold for malignancy in pathologic diagnosis of EUS-FNA. Clinico-pathologic correlation is important for adequate pathologic interpretation of EUS-FNA from various pancreatic lesions.

Pancreatic ductal adenocarcinoma is the most common solid tumor of the pancreas and the most commonly encountered pancreatic neoplasm on EUS-FNA. Aspirated material usually demonstrates atypical cells with or without a necrotic background. The cellularity of the cytologic smear depends on many parameters, including tumor size, location, cellularity, type of aspiration needle, and endoscopist technique [7-9]. A pathologist is more likely to encounter a paucicellular smear than a highly cellular smear. If many atypical cells are found in a necrotic background (Fig. 2A), it is easy to conclusively diagnose a malignancy even without clinical information. However, pathologists may encounter paucicellular smears showing only a few atypical cells without a necrotic background. In such situations, it is difficult to differentiate reactive ductal cell atypia of chronic pancreatitis from malignancy. Pathologists should correlate with the clinical impression and radiologic findings of these pancreatic lesions. If the clinical finding impression is an overt malignancy, such as pancreatic tumor with vascular encasement or liver metastasis, the pathologic diagnosis of malignancy can be made with a very limited volume of atypical cells. Pancreatic ductal adenocarcinomas are often well-differentiated with very mild cytological atypia, and one should be careful not to mistake them for contaminated gastric foveolar epithelium. In such cases, the architecture of cell groups should be observed in detail, including cribriforming, nuclear stratification, and loss of regular honeycombing seen in normal and benign epithelia (“drunken honeycomb”). Well differentiated tumor cell nests usually exhibit more cellularity with overlapping and touching nuclei (Fig. 2B). The presence of scattered small clusters of atypical cells with necrotic cell debris in the background is also helpful, as pancreatic ductal adenocarcinomas frequently demonstrate necrotic backgrounds and scattered small atypical cell nests or clusters. However, this necrotic background should be discriminated from the fibrin clots that are frequently observed in bloody smears (Fig. 2C). Another EUS-FNA specimen that demonstrates diffuse necrosis is tuberculous lymphadenitis involving peripancreatic or retroperitoneal lymph nodes: if the EUS-FNA cytology/biopsy shows a diffusely necrotic background without apparently viable tumor cells (Fig. 2D), the pathologist should check the clinical impression and imaging findings. In many cases, the radiologic impression is lymphoma, especially if there are multiple enlarged lymph nodes around the pancreas or retroperitoneal region. Due to the high prevalence of tuberculosis in Korea, the possibility of tuberculous lymphadenitis should be considered, especially in EUS-FNAs from older patients. Special stains for acid-fast bacilli and molecular studies can be performed with biopsies or cell blocks to confirm the diagnosis of tuberculosis. Diffuse necrotic backgrounds are not limited to ductal carcinoma or tuberculosis. Necrotic cells may also be seen in EUS-FNA cytology/biopsy of solid pseudopapillary neoplasms (SPNs) and poorly differentiated neuroendocrine carcinomas. However, SPN and neuroendocrine carcinoma are usually accompanied by hypercellular components on cytologic smears, and these neoplasms are relatively easily differentiated on radiologic evaluation. The main differential diagnosis points are summarized in Table 1.

IPMN is one of the most common pancreatic cystic neoplasms, and EUS-FNA is frequently performed to evaluate the grade of cytologic atypia and presence of an associated invasive carcinoma. It is not uncommon to find mucinous content in cytologic smears from EUS-FNA. Although mucin may be found in EUS-FNA material from IPMN, it is not a specific finding. When a mucinous background is noted on cytology/biopsy slides (Fig. 3A), the pathologist should check the radiology findings. If the radiologic interpretation strongly suggests IPMN, the pathologist may suggest IPMN in the cytopathologic diagnosis of EUS-FNA. However, diagnosis of IPMN on EUS-FNA should be avoided if the radiologic and EUS findings do not suggest IPMN, as the radiologic findings of IPMN are relatively unique. In addition, it is relatively uncommon to observe mucin in cytology slides and/or biopsies from IPMN, especially in LBC. The appearance of aspirated mucin also depends on its viscosity. Main duct type IPMNs with intestinal phenotype usually show viscous mucin, similar to that seen in appendiceal mucinous neoplasms. Conversely, branch duct type IPMNs contain relatively less viscous gastric foveolar-type mucin, resulting in a more transparent mucinous background that is difficult to recognize on cytologic smears. The typical cytologic features of IPMNs include papillary clusters of columnar epithelial cells containing cytoplasmic mucin (Fig. 3B), and IPMNs may be graded according to the degree of nuclear atypia. IPMNs without intraductal or intracystic papillary lesions may not demonstrate typical papillary epithelial cell nests on EUS-FNA slides. Sometimes it is difficult to differentiate papillary cell nests from contaminated intestinal mucosa. If the IPMN is branch duct type with gastric foveolar type epithelium, it may be difficult to differentiate between mucinous epithelium of the IPMN and contaminated gastric mucosa. Pathologists should assess the radiologic findings of IPMN. If there are no mural or papillary nodules of IPMN on radiologic evaluation, the cellular nest of epithelial mucosa probably corresponds to contaminated gastric mucosa, not the epithelial component of IPMN. The presence of scattered goblet cells within monotonous-looking cell nests is a helpful clue that can differentiate normal intestinal mucosa from IPMNs (Fig. 1B). It is therefore important that pathologists assess the anatomic location of target lesions on clinical EUS-FNA reports; lesions in the pancreatic head and uncinate process are usually targeted via the duodenal wall, while those located in the body and tail are usually accessed via the gastric wall. If pathologists identify small intestinal mucosa-like epithelial cell nests in EUS-FNA from the pancreas body or tail with a clinical impression of main duct type IPMN, it likely corresponds to the epithelial component of IPMN because the EUS-FNA needle penetrated the gastric wall, not the duodenal mucosa. However, if radiologic findings indicate branch duct type IPMN, which is mostly composed of gastric phenotype epithelium, it may be difficult to discriminate between contaminated gastric mucosa and the epithelial component of gastric phenotype IPMN. The papillary nests within cystic lesions are relatively small in branch duct type IPMN. If a pathologist finds relatively large nests of intestinal mucosa, it may corresponds to contaminant intestinal mucosa rather than papillary nests of branch duct type IPMN.

Serous cystic neoplasms (SCNs) are not infrequently submitted for EUS-FNA evaluation. It is uncommon to observe the typical serous lining cells on cytology smears and/or biopsy, especially in aspirates from unilocular or macrocystic lesions. However, if the SCN is a multilocular cystic lesion, some serous epithelial cell clusters composed of cuboidal cells with clear cytoplasm may be seen in cytology and/or biopsy slides. If the pathologist can find scattered subepithelial capillaries in the biopsy slide, it is supportive of an SCN. In most cases, SCN is easily detected by radiologic evaluation, except for macrocystic SCN. Radiologic correlation is essential in pathologic diagnosis of SCN in EUS-FNA.

Mucinous cystic neoplasm (MCN) is a relatively uncommon cystic neoplasm of the pancreas. Mucinous epithelial lining and an underlying ovarian-type stroma are characteristic histologic findings. However, it may be difficult to observe well-preserved mucinous epithelium and ovarian-type stroma, even in resected specimens. It is even harder to identify mucinous epithelial cells and/or ovarian-type stroma in EUS-FNA cytology/biopsy. If some mucinous epithelial cells are seen, the pathologist should check the clinical findings and radiologic interpretation. The characteristic clinical setting is a middle-aged female with a pancreatic unilocular cystic neoplasm or multilocular septated cyst without ductal communication. Invasive carcinomas may arise from MCNs, and the biologic behavior of MCN-associated invasive carcinomas is aggressive even at an early stage. It may be difficult to detect early invasion by EUS-FNA, as these early invasive components do not create mass lesions on radiologic or EUS findings.

Aspirated material from pancreatic neuroendocrine neoplasms usually results in a hypercellular smear. Although some high-grade neuroendocrine carcinomas have an accompanying necrotic background, well-differentiated pancreatic neuroendocrine tumors (NETs) rarely show necrosis on cytology smears. Pancreatic NET is classified by a 3-tier scheme [10]. Grade 1 and 2 NETs are well-differentiated neuroendocrine tumors with mitotic rate of <2/2 mm² or Ki-67 index <3% and mitotic rate of 2–20/2 mm² or Ki-67 index 3%–20%, respectively. Well-differentiated grade 3 NETs show a mitotic rate of >20/2 mm² or Ki-67 index >20%. Tumor cells from well-differentiated NETs usually show mild cytologic atypia, sometimes with characteristic salt-and-pepper chromatin pattern, which is more frequently found on cytology (Fig. 4A). NET cells on biopsy slides are more discohesive epithelial cells with round to vesicular nuclei and plump cytoplasm (Fig. 4B). While NETs usually demonstrate scant cytoplasm, some NETs may have relatively abundant cytoplasm (Fig. 4C). In such cases, it may be difficult to distinguish NET from SPN. The cytologic findings of discohesive cells and plasmacytoid features suggest the diagnosis of SPN. Immunostaining for β-catenin is a useful and reliable differential diagnostic tool, as other neuroendocrine markers (e.g., CD56, synaptophysin) can be expressed in both tumors [11]. NETs usually express β-catenin with a membranous pattern whereas SPNs exhibit nuclear β-catenin staining. Pathologists should correlate with radiologic findings before performing immunohistochemical stains. NETs are usually well demarcated hyper-enhancing masses. If a pancreatic mass is partially solid and cystic on radiologic evaluation, it is more likely to be an SPN than a NET. Grading of NET on EUS-FNA biopsies is well correlated with that of surgically resected tumors [12,13]. However, considering the uneven distribution of Ki-67 hot spots in tissue sections, it is still possible that the Ki-67 labeling index of a biopsy does not represent the proliferation index of the whole tumor [14]. Aspirates from high-grade small cell neuroendocrine carcinomas demonstrate hyperchromatic small cell carcinomalike features with or without a necrotic background (Fig. 4D). Differential diagnosis of metastatic lung small cell from may be difficult because the cytologic findings are similar in both lesions. Ancillary immunostaining for thyroid transcription factor 1 (TTF-1) may be helpful, as metastatic small cell carcinoma from the lung usually expresses TTF-1 immunoreactivity. However, pathologists should also correlate with the clinical history and radiologic findings for systemic metastasis. If patients have a past medical history of lung cancer with multiple metastasis, it should be easily suspected regardless of ancillary test results.

The characteristic clinical setting of SPN is a middle age female with a pancreatic mass in a distal pancreas location, either the body or tail. Although exceptional cases may occur, pathologists should assess these two clinical findings when diagnosing SPN. Radiologic examination typically shows a solid and cystic tumor with or without necrotic contents. However, SPNs smaller than 2 cm may present as solid tumors and may be interpreted as NETs on radiologic evaluation. Cytologic smears of SPN are usually discohesive and hypercellular, and the cells demonstrate mild nuclear atypia and plasma cell-like features (Fig. 5A). The presence of myxoid pseudopapillae is a characteristic pathologic finding of SPN. However, it is relatively uncommon to see the typical myxoid pseudopapillary pattern of SPN in cytologic smears of EUS-FNA (Fig. 5B). Scattered pleomorphic or bizarre cells from SPN are reported to reflect degenerative change rather than high-grade malignancy, which may be overestimated on EUS-FNA pathologic diagnosis [15]. Sometimes, tumor cells of SPN show clear cell change that mimics renal cell carcinoma. Pathologists should correlate with the patient’s past medical history before confirming a diagnosis of metastatic renal cell carcinoma to the pancreas.

EUS-FNA is rarely performed to evaluate autoimmune disease that involves the pancreas or bile duct. If the clinical history or radiologic findings are characteristic, the pathologic findings of EUS-FNA biopsy may provide typical histologic findings of dense lymphoplasmacytic infiltration and/or storiform fibrosis. However, if the patient history or radiologic findings are vague, it may be difficult to differentiate conventional chronic pancreatitis from autoimmune pancreatitis or cholangitis. Especially when presenting as a local mass-forming lesion, it is very difficult to suspect autoimmune disease if EUS-FNA cytology/biopsy provides only surface mucosa [16]. It should be emphasized that the pathologic diagnosis of autoimmune disease should be made after excluding malignancy in the differential diagnosis. IgG4 immunostaining is essential in evaluation of IgG4-related autoimmune disease that involves the pancreas or bile duct. If IgG4 positivity is sufficient by immunostain study, IgG4-related autoimmune disease can be diagnosed. However, pathologists should be careful to exclude autoimmune disease with insufficient IgG4 positivity because increased IgG4-positive plasma cells can be found in other inflammatory conditions or even in some malignancies [17].

Some malignant tumors may metastasize to the pancreas; renal cell carcinoma is one of the most common metastatic tumors in the pancreas. However, it is uncommon for pathologists to see metastatic tumors on EUS-FNA cytology and/or biopsy because it is difficult to evaluate metastasis on pancreas EUS-FNA. It is challenging to consider metastatic renal cell carcinoma without pertinent clinical information. Metastatic renal cell carcinoma sometimes does not exhibit characteristic clear cell features on EUS-FNA cytology and/or biopsy specimens, instead showing monotonous cell nests with abundant capillaries (Fig. 6A). Metastatic small cell carcinoma from the lung is difficult to differentiate from primary poorly differentiated neuroendocrine carcinoma of the pancreas because the pathologic findings are very similar (Fig. 6B). Immunostain for TTF-1 and correlation with clinical history are essential for differential diagnosis of metastasis.

a few atypical cells that may be insufficient for a conclusive diagnosis of a malignancy. Therefore, it is often difficult for general surgical pathologists to confidently make a conclusive diagnosis on EUS-FNA specimens. In these cases, it may be better to adjust the threshold of malignancy according to the clinical and radiologic findings. For example, if a patient has multiple liver metastases and peritoneal seeding, the pathologist can suggest “malignancy” by lowering the threshold of malignancy, even with a few atypical cells on EUS-FNA. However, if the same cytology and/or biopsy is encountered in the setting of a clinical impression of “rule out autoimmune pancreatitis vs. hidden malignancy,” it may be better to provide a diagnosis of “atypical cells, uncertain malignant potential or cannot exclude malignancy.” Pathologists should raise their threshold of malignancy in such situations. Although this may be criticized as unprofessional, pathologists should be flexible to enable better communication with clinicians and improve patient management in diagnosis of EUS-FNA for pancreatic lesions. Clinical-pathologic correlation is essential in pathologic diagnosis of EUS-FNA from pancreatic lesions. If an EUS-FNA does not provide any atypical cells in a clinical setting of malignancy, pathologists should suggest the possibility of a non-representative sample in the pathology report. Suggested terminology is: “The possibility of non-representative sample remains. Re-aspiration is recommended to rule out malignancy.” Many clinicians think that a pathologic diagnosis of malignancy is difficult on EUS-FNA specimens. However, pathologic confirmation of a true negative is more difficult than confirmation of a true positive. Pathologists should be aware of inter-observer disagreement in radiologic interpretation of various pancreatic lesions. Due to the deep location of the pancreas, it is not feasible to obtain as many tumor cells as in cytology smears of the thyroid, breast, and uterine cervix. An open biopsy in the operation room would be required to obtain such quantities of tumor cells, which would be a burden to patients. In the era of molecular testing for precision medicine therapy, tumor volume is certainly an important issue. Given the small amounts of tumor cells obtained by EUS-FNA, more reliable sensitive molecular assays should be developed, such as circulating tumor DNA or mutant KRAS in circulating exosome [31,32]. Although there have been some advances in precision medicine strategies for pancreas cancer, there are currently few targets for precision medicine in pancreatic tumors [33,34]. Until now, the main goal of the EUS-FNA cytology/biopsy is confirmation of a malignancy or non-neoplastic disease to guide decisions for the next step in therapy and patient management.