Introduction

Pancreatic cysts are common in the general population. The reported incidence of cysts varies widely in the literature, ranging from 0.7% to 24.3%. The incidence estimated from cross-sectional imaging is lower than that estimated from autopsy studies. The incidence of truly asymptomatic cysts in the general population is estimated to be around 2.6% and increases with advancing age. In the last 2 decades the proportion of pancreatic resections performed for cystic lesions has grown substantially. As pancreatic cysts can be either benign or neoplastic, with variable malignant potential based on their subtype, there has been a substantial effort to establish an accurate preoperative diagnosis using minimally invasive techniques. Endoscopic ultrasonography has been a valuable tool in this effort, given its ability to provide not only high-resolution imaging of these cysts but also the ability to safely aspirate cyst fluid for detailed analysis.

Types of pancreatic cystic lesions

Pancreatic cysts can be classified as either neoplastic or nonneoplastic. Nonneoplastic cysts include pseudocysts, retention cysts, lymphoepithelial cysts, duplication cysts, and infectious cysts. Neoplastic pancreatic cysts comprise a heterogeneous group of lesions that are histopathologically different and demonstrate a diverse natural history. These cysts can be broadly classified as mucinous and nonmucinous cysts. Nonmucinous neoplastic cysts include serous cystadenoma (SCA) and solid pseudopapillary neoplasm (SPN). Mucinous cysts are premalignant lesions, and include mucinous cystic neoplasm (MCN) and intraductal papillary mucinous neoplasm (IPMN). IPMN can be further divided into a main-duct subtype and a branched-duct subtype, depending on the relationship with the main pancreatic duct. The main-duct subtype is considered to carry a higher risk of malignancy than the side-branch subtype. Morphologic variations of IPMNs have also been recognized, with 4 distinct subtypes distinguished based on the histomorphologic features of papillae and the immunohistochemical features of mucin glycoproteins. Recent consensus studies have labeled these IPMN types as gastric, intestinal, pancreaticobiliary, and oncocytic, and suggest that each type may demonstrate a distinct clinical course.

Types of pancreatic cystic lesions

Pancreatic cysts can be classified as either neoplastic or nonneoplastic. Nonneoplastic cysts include pseudocysts, retention cysts, lymphoepithelial cysts, duplication cysts, and infectious cysts. Neoplastic pancreatic cysts comprise a heterogeneous group of lesions that are histopathologically different and demonstrate a diverse natural history. These cysts can be broadly classified as mucinous and nonmucinous cysts. Nonmucinous neoplastic cysts include serous cystadenoma (SCA) and solid pseudopapillary neoplasm (SPN). Mucinous cysts are premalignant lesions, and include mucinous cystic neoplasm (MCN) and intraductal papillary mucinous neoplasm (IPMN). IPMN can be further divided into a main-duct subtype and a branched-duct subtype, depending on the relationship with the main pancreatic duct. The main-duct subtype is considered to carry a higher risk of malignancy than the side-branch subtype. Morphologic variations of IPMNs have also been recognized, with 4 distinct subtypes distinguished based on the histomorphologic features of papillae and the immunohistochemical features of mucin glycoproteins. Recent consensus studies have labeled these IPMN types as gastric, intestinal, pancreaticobiliary, and oncocytic, and suggest that each type may demonstrate a distinct clinical course.

Diagnostic modalities to evaluate pancreatic cysts

Differentiating among these cysts is challenging, and a variety of modalities including cross-sectional imaging, endoscopic imaging, cytology, and cyst fluid analysis have been found to be useful. The 2 noninvasive imaging modalities used most frequently to evaluate pancreatic cysts are computed tomography (CT) and magnetic resonance imaging (MRI). CT is a good test for cystic lesions of the pancreas because of its widespread availability and ability to detect cysts, and it is often the modality with which cystic lesions are initially suspected or diagnosed. Pancreas protocol CT scanning has become a preferred modality to evaluate the pancreas given its ease, relatively low expense, and accuracy with cyst detection. Pancreas protocol uses an intravenous contrast bolus timed for both arterial and venous phases, typically with water as the oral contrast, to minimize artifacts arising from denser contrast media. MRI has the advantage over CT of not involving ionizing radiation, and MR cholangiopancreatography (MRCP) has the added benefit of being able to examine the relationship of the lesion and the pancreatic duct in a noninvasive manner. T2-weighted images acquired on MRI (both MRCP and cross-sectional MRI) are excellent for serial follow-up of cyst size. Both CT and MRI, however, have limited ability in differentiating between mucinous and nonmucinous cystic lesions of the pancreas.

Endoscopic retrograde cholangiopancreatography (ERCP), although more invasive than MRCP, is very useful in defining the communication of the cyst with the main pancreatic duct and provides another method for tissue acquisition. However, diagnostic ERCP, with its associated risk of pancreatitis, has largely been replaced by MRCP and endoscopic ultrasonography (EUS).

EUS has many attributes that make the procedure an ideal tool for evaluation of cystic lesions in the pancreas. EUS is a low-risk procedure that produces high-resolution imaging of the pancreas (both parenchyma and ducts) and surrounding structures. More importantly, it affords the ability to sample cyst contents for analysis. The diagnostic accuracy of EUS morphology alone is widely variable. In a large prospective multicenter trial using morphologic criteria to distinguish mucinous cysts (macrocystic septations or adjacent mass) from nonmucinous cystic lesions (unilocular, honeycombed, or thickened wall), the sensitivity and specificity were low, at 56% and 45%, respectively, which resulted in a poor accuracy of 51%. Furthermore, a study by Ahmad and colleagues evaluated the degree of agreement among endosonographers for EUS diagnosis of neoplastic versus nonneoplastic pancreatic cystic lesions and the specific type of lesion. There was only fair agreement between endosonographers for the diagnosis of neoplastic versus nonneoplastic lesions (κ = 0.24). Agreement of individual types of lesions was moderately good for serous cystadenoma (κ = 0.46) but only fair for the remainder of lesions. As a result, the investigators concluded that there is little more than chance interobserver agreement among experienced endosonographers for the diagnosis of neoplastic versus nonneoplastic and specific type of pancreatic cystic lesions. The reason for this low diagnostic accuracy, despite excellent visualization by EUS, probably lies in the considerable overlap of morphologic features among the different kinds of cystic structures.

Indications for EUS-guided fine-needle aspiration

The imaging modalities discussed are very sensitive for detecting pancreatic cystic neoplasms. However, once a pancreatic cystic lesion is identified, the key clinical issue becomes diagnosis of the cyst type, at the least categorizing the cyst as benign, premalignant, or malignant, to guide subsequent management decisions. Imaging information alone is often not sufficient to accurately characterize the lesions. In cases of diagnostic uncertainty, EUS-guided fine-needle aspiration (EUS-FNA) can be valuable because of its ability to evaluate viscosity, cytology, chemistry, tumor markers, and molecular arrangement in the cyst fluid. As such, the American Society of Gastrointestinal Endoscopy support the use of EUS-FNA for pancreatic cyst diagnosis. However, before performing EUS-FNA, the question must be asked as to how management would be affected by the information obtained. For example, in the case of a macrocystic lesion in the head of the pancreas, given the operative morbidity with a resection of a pancreatic head cyst, confirmation that the cyst is mucinous before surgery is advisable. Alternatively, if the patient is symptomatic and is a good surgical candidate, resection without EUS-FNA may be the appropriate next step. Similarly, if a lesion is a benign-appearing unilocular cyst, or the lesion has characteristic morphology for SCA or demonstrates overt malignant features, the incremental value of FNA to confirm the clinical impression may be limited.

Cyst size is often the most important determinant of success for cyst aspiration and acquisition of adequate fluid for analysis. Walsh and colleagues conducted a study to determine whether cyst size or cyst location predicted success of cyst fluid collection and analysis. It was concluded that successful aspiration of cysts was independent of cyst location in the pancreas and that the larger the cyst, the larger number of diagnostic variables (cytology, carcinoembryonic antigen [CEA], amylase) were able to be obtained. The investigators stated that a minimum cyst size of 1.5 cm was needed to successfully result in at least 1 variable with an 84% success rate. The authors agree with this threshold of 1.5 cm, and endorse FNA of pancreatic cysts 1.5 cm or larger.

Viscosity

In an early study by Lewandrowski and colleagues, pancreatic cyst fluid was obtained by surgical exploration, and relative viscosity (RV) was measured using a viscometer. Water was used as a standard, and had an RV equal to 1. Cyst fluid from serous cystadenomas and pseudocysts had mean RV values of 1.24 and 1.25, respectively. By contrast, mean RV values of mucinous cysts were significantly higher, ranging from 1.2 to 30. In this study it was calculated that an RV of less than 1.63 seemed helpful for distinguishing mucinous from nonmucinous cystic lesions, with an 89% sensitivity and 100% specificity. Similar findings were found in a more recent study by Linder and colleagues, who concluded that viscosity greater than 1.6 predicted mucinous cystic adenoma/adenocarcinoma.

In a study by Leung and colleagues, a novel surrogate marker of cyst viscosity was used, described as the “string sign.” This test, which was performed by the endosonographer, was determined by placing a drop of fluid between the thumb and index finger and measuring the maximum length of stretch before disruption of the mucous string. Benign lesions in the study had a median string length of 0 mm compared with a significantly longer string length of 3.5 mm in potentially malignant/malignant cysts ( P <.0001). The investigators concluded that a long string sign was associated with premalignant or malignant lesions. The string sign is a simple manual technique that can quickly and easily be performed at the time of EUS-FNA and may complement existing tests. The authors use the string sign routinely, and believe that a positive string sign is highly predictive of a mucinous lesion.

Cytology

Pancreatic cystic neoplasms have epithelial cells that either are columnar and stain for mucin (in mucinous neoplasms) or are cuboidal and stain for glycogen (in SCA). These cell types are easily distinguishable by cytology and when positive results are yielded, cytology is one of the most accurate methods of cyst diagnosis. However, obtaining sufficient cells for diagnostic cytology is often difficult because of the limited volume and low cellularity of aspirated cyst fluid, with sensitivity generally less than 50%. In a large multicenter study from the United States, for diagnosis of a MCN based on cytology the sensitivity was 35%, specificity 83%, and accuracy 59%, which was similar to the overall accuracy of EUS morphology. By contrast, in a single-center study whereby all the cytopathologic examinations were performed by the same pathologist, FNA provided a correct diagnosis in 65 of 67 cases. In this study, 77% of the FNA contained enough material for definitive diagnosis. Sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) was 97%, 100%, 100%, and 95%, respectively. A pooled analysis including the aforementioned studies and 5 others that evaluated cytologic examination calculated sensitivities of diagnosing mucinous neoplasm, malignancy, and SCA at 45%, 48%, and 38%, respectively.

A recent study by Rogart and colleagues showed that EUS-FNA with cyst-wall puncture increased the cytologic yield by 37% compared with simple FNA with fluid analysis alone. The investigators in this study used a 22-gauge FNA needle and sent fluid for CEA and amylase if at least 1 mL was obtained for each. Any additional fluid beyond 2 mL was sent for cytology. Without removing the needle, FNA of the cyst wall was then performed by puncturing the far wall of the cyst and moving the needle back and forth through the wall to sample the wall epithelium. For 32 cysts in the study, despite inadequate fluid obtained for CEA or fluid cytology, cyst-wall puncture was performed and a positive diagnosis was still obtained in 15 of 32 cysts (47%). These results are very encouraging, and this technique warrants further study.

Chemistry

Amylase and lipase levels can be elevated in any cystic lesion having a communication with the pancreatic ductal system, such as pseudocysts, IPMN, and also some MCN. As a result, elevated amylase levels are often not helpful clinically in differentiating mucinous from nonmucinous cysts. A low amylase level, however, can be helpful because pseudocysts rarely demonstrate low amylase level in cyst fluid. In a pooled analysis by Van der Waaij and colleagues evaluating amylase concentration in a total of 155 cysts, an amylase level of less than 250 U/L indicated SCA, mucinous cystadenoma, or mucinous cystadenocarcinoma, with sensitivity of 44%, specificity of 98%, PPV 98%, NPV 53%, and accuracy 65%, thus virtually excluding pseudocyst.

A recent single-center study by Park and colleagues found that amylase may also be helpful in differentiating benign from malignant mucinous cysts. The study included 126 patients, and evaluated CEA and amylase in differentiating (1) mucinous from nonmucinous cystic lesions, (2) benign mucinous from malignant mucinous cystic lesions, and (3) pseudocysts from nonpseudocysts. The median amylase level for benign mucinous cysts was 5090 IU/L, compared with 60 IU/L for malignant mucinous cysts and 94 IU/L for malignant cystic neuroendocrine tumors ( P = .0008). The investigators speculated that malignant transformation may be associated with uncontrolled cellular growth that occludes any microscopic ductal connections. These results should be accepted with caution given the limitations of the study, including that this was a retrospective analysis in a single-center tertiary-care center with likely referral and selection bias. Also, use of amylase to differentiate benign and malignant mucinous neoplasms was not appreciated in the pooled analysis by Van der Waaij and colleagues.

Tumor markers

Since Hammel and colleagues first demonstrated that tumor markers were present in high concentrations in the cyst fluid from MCNs, multiple subsequent studies have evaluated the use of tumors markers to predict mucinous and malignant cystic pancreatic lesions. A large Unites States multicenter study conducted by Brugge and colleagues evaluated CEA, carbohydrate antigen (CA) 72-4, CA 125, CA 19-9, and CA 15-3. In this study CEA was found to be the most accurate marker in differentiating mucinous from nonmucinous cysts. The median cyst CEA concentration for mucinous cysts was 500 mg/mL and that for nonmucinous cysts 21 ng/mL. A cutoff value of 192 ng/mL provided the greatest accuracy (79%) for differentiating between mucinous and nonmucinous cysts, with moderate sensitivity (73%) and specificity (84%). Other investigators have found similar results. Khalid and colleagues reported on fluid CEA on 76 cysts, and the cutoff value of 192 ng/mL yielded a sensitivity of 64% and specificity of 83% for detecting mucinous cysts. Park and colleagues demonstrated in 104 patients that a CEA cutoff level of 200 ng/mL had a sensitivity of 60%, specificity of 93%, and diagnostic accuracy of 58% for mucinous cysts. In the pooled analysis by Van der Waaij and colleagues, a CEA level greater than 800 ng/mL showed a 79% accuracy for distinguishing mucinous adenoma/adenocarcinoma from a serous cystadenoma or pseudocyst, with 48% sensitivity and 98% specificity. Another large study by Snozek and colleagues analyzed tumor marker levels in 442 patients with pancreatic cysts, in which an optimal CEA cutoff of 30 ng/mL resulted in 79% sensitivity, 73% specificity, and 84% PPV for detection of a mucinous cyst. A low CEA level has also been shown to be useful in predicting serous cystadenoma and pseudocyst. In an early study, Hammel and colleagues showed that a CEA level below 4 ng/mL predicted an SCA with a sensitivity of 100% and specificity of 93%. In the pooled analysis by Van der Waaij and colleagues, a CEA level below 5 ng/mL predicted SCA or PC with a specificity of 95%, sensitivity of 50% and accuracy of 67%.

As one can appreciate, there is a wide variation in the literature regarding the optimal CEA level at which to predict a mucinous cystic lesion, which is largely due to the significant overlap in CEA levels between mucinous and nonmucinous cysts. Other factors may include laboratory variation (see Technical Considerations section), differences in study design, and differences in study population. Nonetheless, studies have consistently shown that increasing the CEA value raises specificity for a mucinous lesion but at the cost of falling sensitivity for detection. At the authors’ institution, for clinical decision making, a cutoff of greater than 192 ng/mL is used to differentiate mucinous from nonmucinous cysts.

Several other tumor markers have been evaluated for differentiating mucinous from nonmucinous cysts but have not performed as well as CEA. For example, in the pooled analysis by Van der Waaij and colleagues a cutoff CA 19-9 value of less than 37 U/mL predicted SCA or pseudocyst with a sensitivity of 19%, specificity 98%, and accuracy 46%. In the United States multicenter study already discussed, a CA 72-4 value greater than 7 ng/mL demonstrated a sensitivity of 80% and specificity of 61% for differentiating mucinous and nonmucinous lesions.

Molecular analysis of cyst fluid

Molecular testing of cyst fluid has increased over the past few years, largely due to the availability of a commercially available test (PathFinderTG; RedPath Integrated Pathology, Pittsburgh, PA, USA). Similar to pancreatic ductal adenocarcinoma, molecular alterations in neoplastic mucinous cysts have demonstrated multistep genetic changes involving KRAS mutation, p53, loss of p16, and SMAD4. Detection of these underlying molecular changes using cyst fluid DNA from exfoliated epithelial cells has become a major research focus. Early studies showed that increased levels of DNA, the presence of the KRAS mutation or 2 or more loci of allelic imbalance was each associated more significantly with a mucinous neoplasm, whereas a high amplitude of mutations and allelic loss was predictive of malignancy.

The largest multicenter prospective study (PANDA study) evaluating the molecular analysis of cyst fluid included 113 cysts with histologic follow-up based on surgical resection or cytologic findings from EUS-FNA. The goals of this study were to evaluate molecular characteristics of cyst fluid that differentiated mucinous from nonmucinous cysts as well as premalignant from malignant cysts. The investigators found that among the parameters of DNA level, presence of KRAS mutation, and 2 or more loci of allelic loss, presence of KRAS mutation showed the highest specificity for mucinous cysts at 96% but sensitivity was low, at 45%. It was therefore concluded that in the presence of the KRAS mutation, the cyst very likely represents a mucinous cyst. When the presence of KRAS mutation was combined with CEA, the sensitivity of CEA (cutoff level 148 ng/mL) was improved from 67% to 84% while specificity remained at 67%. In the absence of a KRAS mutation, the cyst fluid CEA remains significantly associated with mucinous cysts, although with a less optimal sensitivity (69%) and specificity (68%). For premalignant versus malignant cysts, the most accurate test diagnosing a malignant cyst was an allelic-loss amplitude of greater than 80%, achieving sensitivity of 70% and specificity of 85%.

Other studies have shown more mixed results with molecular analysis. Sawhney and colleagues conducted a retrospective study of 100 pancreatic cysts (with 17 mucinous neoplasms) that evaluated both CEA and molecular analysis in predicting a mucinous lesion. In this study, there was poor agreement between CEA and molecular analysis for the classification of mucinous pancreatic cysts (κ = 0.2). A CEA level higher than 192 ng/mL had a higher sensitivity of 82% for mucinous cysts, compared with 77% with molecular analysis. However, the combination of CEA and molecular analysis yielded a 100% sensitivity and specificity, indicating that these data may be complementary. The small sample size was the major limitation of this study. A recent retrospective study by Klochan, Dewitt and colleagues on 270 patients with pancreatic cystic lesions evaluated CEA, KRAS mutation, and DNA analysis for the diagnosis of mucinous cysts. CEA performed better than DNA analysis with sensitivity of 54.5% and specificity 80% compared with 45.5% and 80%, respectively. Using a combination of CEA, KRAS mutation, and DNA analysis, sensitivity increased to 75% but specificity decreased to 66.7%.

One of the major benefits of molecular analysis is that only 0.2 mL of cyst fluid is needed. Given that increasingly small cysts are being detected with fluid aspirate volumes often inadequate for CEA analysis or cytology, there may be a role for molecular analysis to provide information in these cases. In cysts with adequate aspirate volumes for CEA and cytology, the value-added benefit of molecular analysis may not be sufficient enough at this time to justify its high cost.