The incidence of intraductal papillary mucinous neoplasms (IPMNs) has increased significantly with the development and increasing use of cross-sectional non-invasive imaging, such as computed tomography (CT), magnetic resonance imaging (MRI), and magnetic resonance cholangiopancreatography (MRCP). These imaging modalities will reveal an incidental pancreatic lesion in up to 45% of patients [1
]. Many incidentally identified pancreatic cysts are IPMNs, a subset of which may be precursors of pancreatic adenocarcinoma (PDAC) [5
PDAC is a devastating diagnosis with a <10% 5-year overall survival rate for all comers, and unfortunately no method of screening for pancreatic cancer has been proven to be either effective or practical [8
]. Similar to colorectal and other cancers, the resection of a premalignant lesion such as IPMN is thought to improve survival. However, the unknown natural history of IPMN malignant degeneration into PDAC, as well as the significant morbidity of a pancreatic resection, has created controversy over which patients should undergo surgical resection. Therefore, the management for IPMNs remains controversial [9
The initial Sendai guidelines for IPMN management were published in 2006, updated to the Fukuoka guidelines published in 2012, and were updated again in 2017 [11
]. These guidelines (otherwise known as the International Consensus Guidelines or ICG) characterize IPMNs as main-duct IPMNs (MD-IPMN), branch-duct IPMNs (BD-IPMN), and mixed-type IPMNs based on imaging and histology. MD-IPMNs are defined as having a main duct dilation >5 mm without other identifiable reasons for obstruction. BD-IPMNs are cysts >5 mm that communicate with the pancreatic duct. The mixed type meet the criteria for both BD- and MD-IPMN.
These guidelines define three categories of all IPMNs; those considered low-risk, those with “worrisome features,” and those with “high-risk stigmata.” “Worrisome features” include a cyst size ≥3 cm, an enhancing mural nodule <5 mm, thickened enhanced cyst walls, a main pancreatic duct (MPD) size in the range 5–9 mm, an abrupt change in the MPD, lymphadenopathy, elevated CA 19-9, and a rapid cyst growth rate (more than 5 mm over 2 years). “High-risk stigmata” include obstructive jaundice with cystic lesions in the pancreatic head, an enhanced mural nodule ≥5 mm, and an MPD ≥10 mm.
The Fukuoka guidelines recommend the resection of BD- and MD-IPMNs with “high-risk stigmata” due to the increased risk of underlying malignancy. For BD-IPMNs, absolute indications for resection also include cytology positive for high grade dysplasia. IPMNs with “worrisome features” and endoscopic ultrasound (EUS) with negative findings should have imaging and EUS surveillance based on cyst size [13
]. In secondary analyses, the specificity of the Fukuoka guidelines reaches 94%–97% in some studies, with sensitivity only reaching 28%–62% [14
The American Gastroenterological Association (AGA) guidelines for pancreatic cysts also define “high-risk” lesion characteristics as a size ≥3 cm, a dilated MPD, and mural nodules [16
]. For high risk lesions, EUS with a fine needle aspiration (FNA) of cyst fluid is recommended for further analysis, which then informs the decision to observe or resect. These guidelines have demonstrated improved sensitivity (50%) and accuracy (73.7%) compared to the Fukuoka guidelines but are less specific (89.1%) for the management of pancreatic cysts [16
Although the specificity of these criteria are improving, there continues to be a significant percentage of patients who undergo pancreatic resection for noninvasive lesions; up to 77% may be benign on postoperative pathological examination [7
]. Additionally, the available data is based on retrospective studies of surgical patients, so the true percentage of invasive or high-risk lesions is even lower when including patients not selected for resection. Given the risk of surgical morbidity, this calls for further research into more objective tests to better identify patients with IPMN that are at the greatest risk of malignancy, namely IPMN with high-grade dysplasia (HGD) or patients associated with an invasive component. Furthermore, even in patients with HGD, the risk of progressing to an invasive malignancy has not been well defined. Current research has focused on tissue genetics, tumor markers, and cyst fluid analysis with promising results [13
]. Utilizing EUS–FNA to obtain cyst fluid samples for analysis in the pre-operative setting has demonstrated significant potential for diagnosis and the guidance of surgical management without incurring additional risks of peritoneal seeding [21
]. Here, we present an updated review on the most promising cyst fluid markers in order to summarize current strategies for risk stratification and to identify future directions for further investigation.
2. Single Marker Analyses
Cytology studies on EUS-obtained cyst fluid samples have historically been of great importance in the assessment of the malignancy risk of IPMN. However, the high sensitivity and specificity of cytologic studies have been impeded by multiple factors, including interobserver variability, low cellular yield, low volume yield, and even, when there is enough of a sample for cytologic studies, the percent of “indeterminate” or “nondiagnostic” results can reach greater than 40% [23
]. Efforts to improve the diagnostic yield of cyst fluid samples include additional techniques such as Moray micro forceps biopsy [25
]. While 0.4 mL is needed for DNA analysis, up to 1.0 mL is needed for carcinoembryonic antigen (CEA) analysis and more if seeking to perform cytology and send for sequencing studies [24
Given the limitations of cytology, the search for quantifiable cyst fluid biomarkers began with proteins and classically known tumor markers. Amylase in the cyst fluid would be consistent with ductal communication, and levels <250 U/L indicated lack of ductal communication with a specificity of 98% [28
]. Low amylase, however, does not rule out malignancy [30
]. CAE is an epithelial glycoprotein suggestive of mucin production [28
]. Elevated levels of CEA > 192 ng/mL in combination with amylase in cyst fluid have been shown to identify mucinous cysts with a sensitivity of 64%–78% and a specificity of 65%–89% (Table 1
]. However, these two markers do not discriminate between levels of dysplasia in IPMN and do not reliably distinguish between IPMN and mucinous cystic neoplasms (MCN); some MCNs also have elevated amylase [32
Other genetic mutations have also been shown to distinguish between mucinous and non-mucinous cysts and their subtypes, including serous cystadenoma (SCN) and solid-pseudopapillary neoplasm (SPN) (Table 2
). The guanine-nucleotide-binding protein-alpha stimulating (GNAS) and the Kirsten rat sarcoma viral oncogene homolog (KRAS), an oncogene and proto-oncogene, respectively, have also been identified as commonly mutated in pancreatic cyst fluid [26
]. GNAS in particular has been shown to distinguish IPMN from MCN. Wu et al. demonstrated 98% sensitivity and 100% specificity in diagnosing IPMN when testing cyst fluid [35
]. Other single markers, such as glucose and neutrophil gelatinase-associated lipocalin (NGAL), have also been investigated with no ability to stratify IPMN [36
]. Fortunately, there are now several single markers with the ability to both identify and grade IPMN dysplasia.
Mucins (MUC) are highly glycosylated proteins released into the cyst from the epithelium, and their subtypes vary with the histopathological IPMN subtypes: gastric, intestinal, pancreatobiliary, and oncocytic [39
]. Given that IPMN subtypes also correlate with specific degrees of dysplasia, Maker et al. investigated the ability of mucin to determine the degree of IPMN dysplasia [41
]. In this study, investigators demonstrated a significant increase in MUC2 and MUC4 among cysts with HGD and increased MUC2 in intestinal type IPMNs, which have a higher degree of malignant transformation [41
]. A more recent study demonstrated that mucins can carry abnormal glycoforms, finding that oligosaccharide linkages α
GlcNAc and β
GlcNAc associated with MUC5AC were associated with malignancy. The version of MUC5AC with α
GlcNAc was found primarily in IPMN cyst fluid compared to MCN cyst fluid, and the level of staining on final pathologic specimens increased with degree of dysplasia [42
]. However, the study did not show if cyst fluid levels also varied with degree of dysplasia, limiting its preoperative diagnostic utility.
Cytokine expression within pancreatic cyst fluid has also been shown to be prognostic. Higher concentrations of interleukins (IL-) 1b, 5, and 8 have been identified in cysts containing HGD or malignancy (p
< 0.0001) [43
]. By multivariate analysis, IL-1b was found to be an independent factor in predicting high-risk versus low-risk pancreatic cysts with a positive predictive value of 71% and a negative predictive value of 75%, as well as sensitivity and specificity reaching 79% and 95% [43
]. Cyst fluid IL-1b remains a prime target among the pool of cytokines that otherwise did not correlate or had very low expression levels.
Prostaglandin E2 (PGE2) levels have been previously shown to be elevated in pancreatic cancer tissue, prompting investigations into its utility in diagnosing premalignant pancreatic cysts. Schmidt et al. prospectively studied cyst fluid samples from 65 patients with pancreatic cystic neoplasms [44
]. Using enzyme-linked immunosorbent assays (ELISA), they quantified the concentration of PGE2 and found higher levels of PGE2 in IPMNs compared to MCNs (p
< 0.05) and demonstrated that PGE2 concentration correlated stepwise with the degree of dysplasia within an IPMN. It was noted that PGE2 concentrations were also higher amongst patients who had a PDAC not associated with the coexisting IPMN [44
]. Their results were subsequently replicated within a larger cohort of 100 patients with similar results. On multivariable analysis, PGE2 alone was significantly associated with HGD-IPMN dysplasia with a sensitivity of 63% and a specificity of 79% [45
2.4. Telomere Fusion Status
Telomere shortening and fusion have been identified in pancreatic malignant degeneration due to chromosomal instability. IPMNs with associated dysplasia have been shown to carry shortened chromosomal telomeres [46
]. Hata et al. were able to demonstrate telomere fusion in 0% of IPMNs with low-grade dysplasia (LGD) and increasing copy numbers with HGD-IPMN and IPMN with adenocarcinoma [47
]. In some patients, there were fusions detected within IPMNs after histological interpretation, but not initially in cyst fluid analysis. This is a limitation of using telomere fusion as a preoperative diagnostic tool because it depends on the shedding of DNA into the cyst fluid, which may be uncommon [47
MicroRNA (miRNA) profiling using Next Generation Sequencing (NGS) is a newer area of cancer research, with demonstrable aberrant miRNA expression in pancreatic cancer and pancreatic cysts [48
]. Wang et al. sought to investigate the so-called ”miRNome” of IPMN cyst fluid [50
]. Of the 15 miRNAs investigated, miR-216 was the most strongly associated with dysplasia, with a higher expression of miR-216 in HGD-IPMN and IPMNs with adenocarcinoma compared to LGD IPMN (p
= 0.011 and p
= 0.020). Although, there were no statistical differences between HGD and adenocarcinoma (p
= 0.540) [50
]. MicroRNA has, thus far, demonstrated significant potential in stratifying IPMNs.
2.6. CEP and mAb Das-1
The murine Das-1 monoclonal antibody (mAb) was created to react with a normal colon epithelial protein (CEP), based on the observation that these cell types are not normally present in gastric and pancreatic epithelium and are prone to developing invasive carcinoma when present [51
]. This unique immunoreactivity has been demonstrated on resected pancreas specimens with HGD-IPMN and PDAC, leading to the most recent study evaluating its applicability to preoperatively sampled cyst fluid [52
]. Das et al. investigated 169 patients with pancreatic cystic lesions across 4 institutions and found that non-mucinous and low-risk cysts displayed little reactivity, whereas HGD-IPMN and MCN lesions had significantly higher reactivity (p
< 0.001), with a sensitivity of 88.3% and a specificity of 92.7% when stratifying for HGD or invasive malignancy [52
]. Based on their internal comparative evaluation, Das-1 reactivity has significant potential in distinguishing HGD or malignancy.
5. Future Directions
The state of current biomarker analyses continues to progress towards higher diagnostic performance, with sensitivities and specificities outperforming published guidelines (Figure 2
), however, their current use remains as an adjunct to the decision-making process [68
]. Existing studies have been limited by (1) the decision to take patients to surgery based on other factors excluding the cyst fluid analysis, (2) inadequate cyst fluid samples for analysis, (3) limited follow-up for patients who do not undergo surgical resection, (4) assessment of the test accuracy retrospectively, although specimens may be collected prospectively [58
]. Given the improved performance of these cyst fluid analyses, should they be given greater weight in deciding how to proceed? Will there be a reduction in the rate of benign specimens without compromising the false negative rate (i.e., not resecting an invasive lesion)? Additionally, a subset of the studies is based on cyst fluid obtained postoperatively. There are many other limitations, and validation studies are needed for many of the above biomarkers. Thereafter, prospective clinical trials with longitudinal follow-up and outcome tracking are warranted.
Cyst fluid analyses may also have a role in surveillance. If a cyst is deemed to have a low enough risk to surveil safely rather than resect, current guidelines recommend radiographic surveillance. Clearly, there is a subset of invasive lesions that do not have characteristic radiographic findings or meet the guideline criteria [69
]. However, the uncertain natural history of IPMN malignant degeneration makes the timing to repeat cyst fluid sampling and analysis equally uncertain.
Fortunately, there are ongoing efforts to fund investigations to solve these unanswered questions, powered through the Pancreatic Cancer Action Network (PANCAN) and National Cancer Institute grants [70
]. Several grants have already been awarded to exploit and study biomarkers for the early detection of malignancy and precursor lesions. In addition, although there are several promising single and composite biomarker assays, there remain many new and untested markers [50
]. The critical role of biomarker analyses in the future of diagnosing and surveilling precursor pancreatic cysts is undeniable. The question of how it will be best utilized remains to be discovered.