Pathology of intraductal papillary mucinous neoplasms

Abstract

Background

Intraductal papillary mucinous neoplasms (IPMNs) represent a unique opportunity to treat and prevent a curable neoplasm before it has the chance to progress to incurable cancer. This prospect, however, has to be balanced with the real risk of over treating patients with lesions that would, in fact, never progress during the life of the patient.

Purpose

Informed clinical decisions in the treatment of IPMNs are first and foremost based on a deep understanding of the pathology of these lesions.

Conclusions

Here we review the pathology of IPMNs, with an emphasis on the clinical relevance of the important features that characterize these lesions.

Introduction

Our understanding of the pathology of intraductal papillary mucinous neoplasms (IPMNs) has evolved rapidly, and it is worth-while to review this history. It is hard to imagine, but in the 1970s, all cystic tumors of the pancreas were lumped together [1]. Entirely benign serous cystadenomas were grouped with precancerous mucin-producing neoplasms [1]. Then, in 1978, two groups, Compagno and Oertel at the Armed Forces Institute of Pathology and Hodgkinson and colleagues at the Mayo Clinic, separated mucin-producing neoplasms of the pancreas from serous cystadenomas [2–4]. Four years later, investigators from Japan, led by K. Ohhashi and colleagues, described what we recognize today as IPMNs, and in the years that followed, G. Zamboni and others clearly separated IPMNs from mucinous cystic neoplasms (MCNs) [5, 6]. It is now recognized that MCNs are defined by the presence of ovarian-type stroma, while IPMNs lack ovarian-type stroma and involve the pancreatic duct system [7–9]. The separation of IPMNs from the other cyst-forming neoplasms of the pancreas opened the doors to detailed studies of the gross, microscopic, and genetic features of these neoplasms. As a result, more recently, entities that were previously classified as variants of IPMN, such as intraductal oncocytic papillary neoplasms (IOPNs) and intraductal tubulopapillary neoplasms (ITPNs), are now recognized as distinctive neoplasms by the World Health Organization (WHO) classification scheme.

An integrated understanding of IPMNs, IOPNs, and ITPNs, one that incorporates gross, microscopic, and genetic features is now emerging. This new understanding has improved our ability to preoperatively classify cystic lesions of the pancreas and to prioritize lesions that are most likely to require surgical intervention.

Gross appearance

While the first IPMNs to be recognized were large lesions that diffusely involved the main pancreatic duct, over time, it became clear that some IPMNs predominantly involve the main pancreatic duct, while others predominantly involve branches of the main duct (Figs. 1, 2, and 3) [8–12]. Not surprisingly, IPMNs that involve the main pancreatic duct are designated “main-duct” IPMNs, those that are limited to branches off of the main duct are designated “branch-duct” IPMNs, and those that involve both the main and side-branches are designated “mixed” IPMNs.

Fig. 1.

a An example of IPMN arising in a branch-duct without a mural nodule. b Branch-duct IPMN with a mural nodule

Fig. 2.

Main-duct IPMN with an associated mural nodule. A hint of papillary projection can be appreciated (arrow)

Fig. 3.

Two distinct, clearly separated, branch-duct IPMNs

Studies correlating the gross type of IPMN with the microscopic degree of dysplasia have shown that main-duct IPMNs are significantly more likely to have high-grade dysplasia or an associated invasive carcinoma than are branch-duct IPMNs [13–15]. For example, a collaborative analysis of 1028 surgically resected IPMNs found that 71% of the main-duct IPMNs had “high-risk disease” (high-grade dysplasia or an associated invasive carcinoma) compared to only 29% of the branch-duct IPMNs [14]. Similarly, Fernandez-Del Castillo and colleagues reported on 145 surgically resected branch-duct IPMNs, and only 22% had high-grade dysplasia or an associated invasive carcinoma [13]. This has immediate clinical implications as main-duct and branch-duct IPMNs can be distinguished on imaging, providing a preoperative measure of risk that has proven critical in the clinical management of patients with an IPMN [10, 11, 16–18]. Most main-duct IPMNs are surgically resected because the risk of high-grade dysplasia or an associated invasive carcinoma is so high, while small branch-duct IPMNs without high-risk stigmata, such as associated symptoms, can be safely followed.

Mural nodules are grossly found in some IPMNs (Figs. 1b and 2) [12]. Some of these nodules are formed by aggregated bunches of papillae or by more complex solid masses of neoplastic cells. Mural nodules are more likely to harbor epithelium with high-grade dysplasia than are the flatter areas of an IPMN. These mural nodules can be detected on imaging, providing another preoperative measure of risk that has proven valuable in the clinical management of patients with an IPMN [19]. Several studies, including the collaborative study of 1028 patients mentioned earlier, have shown that the finding of a mural nodule on imaging is associated with high-risk disease [14, 20]. The finding of a mural nodule on imaging is therefore another indication in favor of surgery in the current management guidelines [11, 18, 21]. However, gross examination of surgically resected IPMNs has revealed that some mural nodules are composed of acellular globs of mucin or polyps of reactive/inflammatory cells secondary to erosion of the cyst wall. These harmless mural nodules can mimic a neoplastic mural nodule on imaging and can lead a surgeon to resect an IPMN that, in retrospect, did not need to be resected.

Gross examination of surgically resected IPMNs has also demonstrated that 20–40% are grossly multifocal (Fig. 3) [22]. For example, H. Matthaei and colleagues described 34 grossly multifocal IPMNs [22]. S. Fritz and colleagues reported that 18% of 287 patients who underwent surgical resection of an IPMN had multifocal disease [23]. The multifocality of IPMNs contrasts with mucinous cystic neoplasms (the neoplasms with ovarian stroma) which are almost always unifocal. As will be discussed later in this review, the multifocality of IPMNs has significant clinical implications [8, 9, 24].

Colloid carcinoma (CC), which is a rare variant of ductal adenocarcinoma of the pancreas, typically arises in an association with IPMN or MCN. Grossly, CC usually forms well-demarcated gelatinous stromal nodules (Fig. 4) [25].

Fig. 4.

Colloid carcinoma (arrow) arising in association with a main-duct IPMN, the probe is in the main duct

Microscopy

Microscopic examination of surgically resected IPMNs has revealed that the neoplastic epithelium can have a variety of directions of differentiation. These include intestinal, gastric-foveolar, and pancreatobiliary. However, individual IPMNs can show a mixture of directions of differentiations, and further research is needed to determine if pancreatobiliary and intestinal IPMNs are truly separate entities or if they represent neoplastic progression of a gastric-type IPMN [26, 27].

The neoplastic epithelium in IPMNs with intestinal differentiation typically forms long finger-like (villous) papillae, and the neoplastic cells have basophilic cytoplasm and enlarged oval and hyperchromatic nuclei (Fig. 5a) [12, 28]. Pseudostratification can be seen, and there is usually moderate to high-grade dysplasia. Basically, these IPMNs histologically resemble villous adenomas of the colon. Virtually all colloid carcinomas of the pancreas arise from an intestinal-type IPMN [25, 29].

Fig. 5.

a An IPMN with intestinal differentiation forming villous papillae. Note the basophilic cytoplasm and enlarged oval and hyperchromatic nuclei. b IPMN with gastric-foveolar differentiation. Note the flat epithelium with basally placed nuclei and abundant mucin cap resembling gastric epithelium. c IPMN with pancreatobiliary differentiation. The neoplastic cells are more cuboidal in contrast to IPMNs with other directions of differentiation. The cytoplasm is amphophilic and the nuclei often are enlarged

The neoplastic epithelium in IPMNs with gastric-foveolar differentiation often forms broader (thicker) papillae, and the cytoplasm of the neoplastic cells is eosinophilic, with basally placed nuclei (Fig. 5b) [12, 28]. The epithelium can also be flat, and typically has only low-grade dysplasia. Branch-duct IPMNs often have gastric-foveolar differentiation.

The neoplastic epithelium in IPMNs with pancreatobiliary differentiation forms thin, branching, complex papillae (Fig. 5c) [12, 28]. The neoplastic cells are not as columnar as in IPMNs with other directions of differentiation. The cytoplasm is amphophilic and the nuclei often are enlarged. These IPMNs typically have high-grade dysplasia, and if an associated invasive carcinoma is present, it is almost always a tubular (ductal) type of invasive adenocarcinoma.

The neoplastic epithelium in IOPNs, previously known as oncocytic subtype of IPMNs, forms thick, complex papillae with intracellular and intraepithelial lumina (Fig. 6) [12, 28]. The cytoplasm of the neoplastic cells is abundant and distinctly oncocytic (eosinophilic). The nuclei are round and typically contain prominent nucleoli. These IPMNs almost always have high-grade dysplasia. As will be discussed in the section on genetics, distinct translocations have been identified in these neoplasms with oncocytic differentiation [30]. They should therefore be clearly designated as “intraductal oncocytic papillary neoplasms,” and classified separately from the other IPMNs.

Fig. 6.

a Intraductal oncocytic papillary neoplasm (IOPN). Note the thick, complex papillae. b The neoplastic cells have abundant and distinctly oncocytic (eosinophilic) cytoplasm. The nuclei are round and typically contain prominent nucleoli

ITPNs are composed of small, tightly packed glands that form intraductal nodules (Fig. 7) [12, 28, 31–33]. The neoplastic cells are cuboidal and, on hematoxylin and eosin staining, typically do not contain significant mucin. These neoplasms can pose a significant diagnostic challenge for pathologists, as it can be difficult to distinguish noninvasive neoplastic cells growing in the branching duct system, from a nodule of neoplastic cells invading into the stroma. ITPNs are now classified separately from the other IPMNs.

Fig. 7.

Intraductal tubulopapillary neoplasm (ITPN) composed of small, tightly packed glands forming intraductal nodules. The neoplastic cells are cuboidal with no significant mucin. This particular case shows foci of comedo-type necrosis resembling intraductal carcinoma of the breast and it also has an associated invasive carcinoma with perineural invasion (arrows)

Histologic examination of IPMNs, IOPNs, and ITPNs has also revealed a spectrum of dysplasia. Although a three-tier grading system was employed originally, things have been simplified, and the degree of dysplasia in these neoplasms is currently assigned one of two grades [8, 9, 34, 35]. Low-grade neoplasms have only mild to moderate atypia, simple papillae, rare mitoses, and minimal to moderate pleomorphism [8, 9, 34, 35]. In general, the cells are well-oriented towards the lumina, and there is minimal stratification of the neoplastic cells. By contrast, high-grade neoplasms have marked atypia, complex papillae, marked pleomorphism, and nuclear stratification with loss of polarity [8, 9, 34, 35]. Mitoses can also be seen. If both low-grade and high-grade dysplasia are present, the neoplasms should be assigned the higher grade.

Although grade and direction of differentiation are not completely independent, essentially all studies have found that the grade of dysplasia is clinically more important than the direction of differentiation [36, 37]. High-grade neoplasms are more likely to have an associated invasive carcinoma, and, as discussed in greater detail below, high-grade dysplasia in a surgically resected intraductal neoplasm turns out to be a significant risk factor for recurrent disease in the remnant pancreas after surgery [36, 37]. Therefore clinical efforts are focused on identifying and treating high-grade neoplasms while avoiding the overtreatment of patients with a low-grade neoplasm [11, 14, 18, 21].

Histologic examination has also confirmed the radiologic and gross findings that IPMNs are often multifocal [22, 23]. This multifocality can be dramatic at the light microscopic level and includes not only multifocal IPMNs, but importantly, histologic examination has shown that the different locules that comprise a single IPMN can have different grades of dysplasia [34]. Furthermore, when multiple anatomically separated IPMNs are present, they too can have different grades of dysplasia. As discussed later, this heterogeneity has significant clinical implications.

Cytopathology

Fine-needle aspiration (FNA) plays an important role in the initial diagnosis of a cystic lesion suspected to be an IPMN on radiological imaging. The ability to bring the ultrasound transducer closer to the pancreas via endoscopic ultrasound (EUS) has revolutionized diagnostic sampling of even sub-centimeter cysts with good diagnostic yield. EUS-FNA has the ability to sample IPMNs throughout the entire length of the pancreas. A successful EUS-FNA diagnostic approach requires the presence of a well-trained cytotechnologist or cytopathologist for rapid on-site evaluation (ROSE) at the time of the procedure. ROSE ensures a higher diagnostic accuracy, reducing the number of unnecessary FNA passes and helps triage the aspirated fluid for ancillary testing such as chemistry and molecular analysis, as needed.

The cytopathologic evaluation of an IPMN includes a multimodal approach and is carefully done after correlation with radiological imaging findings as well as the chemical analysis of the aspirated cyst fluid for carcinoembryonic antigen (CEA) and amylase levels [38, 39]. Although it is generally recommended that each lab should establish its own cut-off values for CEA, a value of approximately 200 ng/ml or higher is considered strongly supportive of a neoplastic mucin-producing cyst. It is important to realize that cyst fluid CEA levels do not help differentiate between an IPMN and mucinous cystic neoplasm (MCN), as well as in distinguishing low-grade from high-grade IPMNs. A significantly high amylase level supports the diagnosis of a pseudocyst or IPMN, a common differential for a cystic lesion. Serous cystadenoma typically will show a low CEA and amylase levels. Molecular analysis of the aspirated cyst fluid is being increasingly performed in several practices and significantly enhances the diagnostic accuracy of an FNA interpretation [40, 41].

Cytopathologic evaluation not only establishes the diagnosis of an IPMN, but also further characterizes the grade of dysplasia as well as evaluates for the presence or absence of invasive component. In order to use a standardized terminology for reporting FNA of IPMN and other cystic lesions of the pancreas, the Papanicolaou Society of Cytopathology (PSC) has put forward its recommended guidelines [42]. IPMN is presently reported on FNA under the general category of “Neoplastic” with the added statement of “Neoplastic Mucinous Cyst.” Also included in the diagnosis is a discrete statement further elaborating on the “presence or absence of high-grade dysplasia or carcinoma.”

Cytomorphologic characteristics of the aspirated cyst form the cornerstone for accurate interpretation of IPMNs and the degree of the associated dysplasia [43]. The evaluation starts at the time of ROSE with the gross features of the aspirated fluid having a thick, viscous, glistening, and mucoid appearance. On microscopy, at low magnification, the majority of cases will display significant extracellular mucin. IPMNs often show thick inspissated mucin having a “colloid-like” appearance, also visible to the naked eye on the glass slide (Fig. 8). Most cases will contain mucinous glandular epithelium in the form of well-preserved intact, sheet-like tissue fragments (Fig. 9). A papillary architecture may or may not be evident in all cases but when present is considered quite specific (Fig. 10). The neoplastic cells appear disorganized with enlarged and crowded nuclei. Intracytoplasmic mucin is readily identifiable in most cells. Gastrointestinal tract material (particularly when EUS-FNA is performed through the gastric wall) may confound interpretation in heavily contaminated aspirates. Occasional cases of IPMN may not contain appreciable mucin, especially cysts with thin degenerated mucin.

Fig. 8.

IPMN, EUS-FNA; a-b. Occasional cases of IPMN display abundant, thick and inspissated mucin with a “colloid-like” appearance, which is considered diagnostic in the right clinicoradiologic setting (Papanicolaou and Diff-Quik stains)

Fig. 9.

IPMN, EUS-FNA; a-b Images showing mucinous glandular epithelium with a well-preserved and intact, sheet-like architecture. Intracytoplasmic mucin is readily identifiable (Papanicolaou stain)

Fig. 10.

IPMN, EUS-FNA; An IPMN depicting neoplastic mucinous epithelium with a classic papillary architecture (Papanicolaou stain)

Another critical aspect of cytomorphologic evaluation of IPMN aspirates is to assess for the presence or absence of high-grade dysplasia or carcinoma [44]. A cellular sample with well-preserved epithelial component is definitely needed to exclude high-grade dysplasia or carcinoma. For high-grade dysplasia, the most important cellular characteristics are cells with significantly increased nuclear to cytoplasmic (N/C) ratios, abnormal chromatic patterns, and background necrosis [45] (Fig. 11). Nucleolar prominence is unusual and can be seen even in the absence of a high-grade dysplasia.

Fig. 11.

IPMN, EUS-FNA; a-b Two cases of IPMN with cytomorphologic features consistent with at least high-grade dysplasia. The neoplastic cells have significantly high N/C ratios and show abnormal chromatin patterns. (Diff Quik and Papanicolaou stains)

Immunolabeling

Several studies have examined the immunolabeling profiles of IPMNs. Most IPMNs label with antibodies to pancytokeratin (AE1/AE3) and with antibodies to cytokeratins 7, 8, and 19 [8, 9]. The expression of the oncoproteins CEA and CA 19.9 is also common [8, 9]. IPMNs with intestinal differentiation label with antibodies to cytokeratin 20 [8, 9]. As shown in Table 1, the pattern of labeling for mucins depends on the direction of differentiation of the IPMN [46–48]. IOPNs label with antibodies to mitochondrial proteins, 111.3, CD117, and with hepatocyte-1 (HepPAR-1) antibodies [49, 50].

Table 1.

Patterns of immunolabeling in IPMNs, IOPNs and ITPNs

Tumor type	MUC1	MUC2	MUC5AC	MUC6	CDX2
Gastric-foveolar IPMN	−	−	+	±	−
Intestinal IPMN	−	+	+	±	+
Pancreatobiliary IPMN	+	−	+	+	−
IOPN	+	*	*	+	−
ITPN	+	−	−	+	−

^*Restricted to goblet cells. IOPN, intraductal oncocytic papillary neoplasm; IPMN, intraductal papillary mucinous neoplasm; ITPN, intraductal tubulopapillary neoplasm

While immunolabeling can be used to determine the direction of differentiation of an IPMN, this classification, as noted above, is not as important as the degree of dysplasia and is of minimal clinical importance. The recognition of which oncoproteins are expressed in IPMNs is, however, indirectly clinically important as high serum levels of CA19.9 in a patient with an IPMN correlate with increased risk of malignancy [51].

Somatic and germline mutations

The molecular revolution has changed our understanding of the genetic drivers of human neoplasms, and IPMNs are no exception. The exomes of IPMNs have been sequenced, and a number of driver genes have been identified (Table 2). Somatic activating mutations in the KRAS and GNAS genes are the most prevalent alterations, and a number of other genes, KLF4, PIK3CA, p16/CDKN2A, RNF43, SMAD4, TGFBR2, and TP53, can also be targeted [52–57]. Somatic BRAF mutations also occur, but are rare (6% in one series) [56]. Gene fusions, including ATP1B1-PRKACB and DNAJB1-PRKACA, are almost universally found in, and restricted to, IOPNs [30, 58]. Mutations in SMAD4, TP53, and TGFBR2 appear to be late events, occurring primarily in high-grade dysplasia and in the invasive components of invasive carcinomas arising in association with an IPMN. By contrast, KLF4 mutations are found more commonly in low-grade IPMNs than they are in high-grade IPMNs [59].

Table 2.

Genes commonly targeted in IPMNs

Gene	Chromosome	Function ofgene product
BRAF	7q	Serine/threonine kinase that functions in MAP kinase/ERK signaling pathway
GNAS	20q	Guanine nucleotide-binding protein functions in adenylyl cyclase activation
KLF4	9q	A Kruppel family transcription factor with a number of functions, including p53 mediated DNA damage repair
KRAS	12p	Member of the small GTPase family
PIK3CA	3q	This phosphoinositide-3-kinase participates in cellular signaling by phosphorylating downstream targets
P16/CDKN2A	9p	Cell cycle control
RNF43	17q	A RING-type E3 ubiquitin ligase that functions in Wnt signaling
SMAD4	18q	Transforming growth factor—beta signaling
TGFBR2	3p	Transforming growth factor—beta signaling
TP53	17p	Induces cell cycle arrest, apoptosis and other changes in response to cellular stresses

McCune Albright syndrome is caused by somatic mutations in GNAS that occur very early in development resulting in individuals who are mosaic for GNAS mutations [60]. In addition to polyostotic fibrous dysplasia and café-au-lait spots, many individuals with this syndrome develop IPMNs with GNAS mutations [61].

Germline genetic alterations have been reported in patients with IPMNs. N. Roberts and colleagues sequenced the germline of 315 patients with a surgically resected and pathologically confirmed IPMN, and 23 (7.3%) of the 315 patients had a deleterious germline variant associated with cancer risk [62]. These included deleterious variants in the ATM, PTCH1, and SUFU genes [62]. IPMNs have been reported in patients with the Peutz-Jeghers syndrome, and M. Goggins and colleagues have demonstrated biallelic inactivation of the STK11 gene in IPMNs resected from patients with this syndrome [63, 64].

The somatic genetic changes in IPMNs can be used as highly specific tools to study the pathology of these lesions. The analysis of the noninvasive and invasive components of IPMNs with an associated invasive cancer has demonstrated the same exact somatic mutation in each component, establishing, beyond a shadow of a doubt, that IPMNs are bona fide precursors to invasive cancer [59, 65]. The findings, however, are more complex than a simple progression. Genetic analyses have shown that some invasive cancers located adjacent to an IPMN, did not, in fact, arise from the IPMN [66]. This phenomenon, “neighbors but not always relatives,” is clinically important because it suggests that even if we were to predict the grade of an IPMN with perfect accuracy, we will not be able to rule out an independent cancer arising in the parenchyma adjacent to the IPMN [65, 66].

Matthaei and colleagues used somatic GNAS mutations as an IPMN marker, and examined small precursors in the pancreas that were in between PanINs and IPMNs in size [67]. They found that some of these lesions, histologically too small to be technically classified as an IPMN, harbored a GNAS mutation, suggesting that they were incipient IPMNs [67].

Genetic tools have also been used to examine the multifocality of IPMNs. L. Wood and colleagues have used the pattern of somatic mutations to show that anatomically distinct IPMNs in the same pancreas harbor different somatic alterations, confirming that IPMNs can be multifocal [68, 69]. When Wood and colleagues took the sequencing to the single cell level, they found even greater heterogeneity [70]. Some grossly unifocal IPMNs contained multiple distinct genetic clones, establishing significant genetic heterogeneity in IPMNs [70].

The neoplastic cells lining cysts in the pancreas shed their DNA into the cyst fluid. As a result, sequencing cyst fluid aspirated at the time of endoscopy can provide preoperative insight into cyst type [41, 71–74]. For example, A. Singhi and colleagues applied next generation sequencing to over 600 cyst fluid samples obtained by EUS-FNA and reported that KRAS/GNAS mutations were highly sensitive for mucin-producing cysts, including IPMNs [41, 75]. They further suggested that the combination of TP53/PIK3CA/PTEN mutations was indicative of high-grade dysplasia. Lennon and colleagues, in a study of 436 patients with a pancreatic cyst, reported that sequencing of cyst fluid can be highly sensitive and specific for cyst type, especially when combined with protein markers (VEGFA and CEA) and clinical findings [72]. These molecular approaches, particularly when combined with clinical features, are extremely exciting, but are imperfect when it comes to determining the degree of dysplasia. As discussed below in clinical implications, we believe that this, in large part, comes from the underlying heterogeneity of the cysts in IPMNs.

Advances in our understanding of the somatic genetic alterations in intraductal neoplasms have also helped improve the classification of these lesions. For example, for years, pathologists debated whether or not IOPNs should be considered a variant of an IPMN, or a separate entity. The discovery that virtually all IOPNs harbor fusion specific fusion genes (ATP1B1-PRKACB or DNAJB1-PRKACA) that are not found in other intraductal neoplasms, helped establish that IOPNs should be considered a completely separate entity from the other IPMNs [30, 58].

Methylation

The patterns of DNA methylation in IPMNs have also been described [76–79]. Targeted and whole genome analyses have revealed distinct patterns of methylation in IPMNs compared to normal ducts, and some genes, including BNIP3, CDO1, EBF3, NXPH1, PTCHD2, and SOX17, have been reported to be more commonly methylated in high-grade IPMNs than in low-grade IPMNs [79, 80]. Future studies may exploit whole genome methylation profiles to refine IPMN subtyping and classification, as is currently the practice for brain tumor classification and as is emerging for pancreatic neuroendocrine tumors [81]. Methylation profiles can be detected in cyst fluid, suggesting that the analysis of the patterns of methylation in EUS-FNA obtained cyst fluid could be used aid in the clinical classification of cyst type [76].

Microbiome

When we think of the microbiome, we tend to think of the tubular gastrointestinal tract. Recently, however, several reports have documented the presence of bacteria in some IPMNs [82–85]. It is unclear if these bacteria are harmless passengers or potential drivers of disease progression [86].

Clinical implications

The dramatic improvements in our understanding of the pathology and genetics of IPMNs, IOPNs, and ITPNs that has occurred over the past two decades have a number of significant clinical implications.

First and foremost, as discussed earlier, cyst fluid can be aspirated at the time of endoscopic ultrasound and the patterns of gene and protein alterations in the cyst fluid, when coupled with imaging and other patient characteristics, can be used to characterize cyst type with high accuracy [41, 71–74]. These analyses can also give a good, but imperfect, indication of the degree of dysplasia in a lesion. The challenge is, and we understand this because we understand the pathology, that the multiple locules that comprise IPMNs can be heterogeneous. Fluid aspirated from one locule provides incomplete information on the other locules in that IPMN. For example, an IPMN could be composed of five locules, four of which have high-grade dysplasia and one of which has low-grade dysplasia. If the endoscopist’s needle happens to sample the locule with low-grade dysplasia, even a perfect test of that fluid will miss the high-grade dysplasia in the adjacent locules. This is compounded by the fact that IPMNs are often multifocal, and not all foci can be practically sampled. Thus, the heterogeneity of IPMNs poses a significant hurdle to the preoperative clinical management of these patients.

The multifocality of IPMNs also has significant clinical implications for patients who undergo surgery for an IPMN [87–91]. For example, the Japanese Pancreas Society reported that 155 (15%) of 1074 patients who underwent surgical resection of an IPMN developed a postoperative recurrence [90]. The risk in the remnant pancreas appears greatest after the resection of main-duct IPMNs and IPMNs with high-grade dysplasia [90, 92]. This risk of metachronous disease is almost certainly a manifestation of the multifocality of IPMNs as it is not seen with mucinous cystic neoplasms, a neoplasm which is almost always a solitary lesion [24, 69]. These results suggest that patients should be carefully followed clinically after the surgical resection of an IPMN, especially if the IPMN is a main-duct IPMN or has high-grade dysplasia [87, 90]

An important clinical issue is the clinical relevance of IPMN lesions at surgical transection margins. The data are controversial, in part due to great variability of the definition “positive transection margin” [93]. It is well accepted that high-grade dysplasia or invasive carcinoma at the transection margin requires further surgical intervention, to prevent progression or recurrence [87]. Pflüger and colleagues reported that the size (>0.5 cm) of the dysplastic focus at the margin and the presence of high-grade dysplasia are associated with relapse of disease in patients with an IPMN [87]. However, Dhar and colleagues found that the majority of positive transection margins have only low-grade dysplasia and that these low-grade lesions are not associated with the development of recurrent disease in the remnant pancreas [94]. Furthermore, genetic analyses have suggested that a significant proportion of dysplastic foci at positive surgical margins are actually separate PanIN lesions unrelated to the primary IPMN [87].

Resection margins during intraoperative frozen section should be always regarded with caution. Tumor heterogeneity and de-epithelialization can mask the real histological picture of the lesion [93]. Good communication between pathologist and surgeon is therefore critical, and the clinical value of resecting additional margin should be carefully balanced with the potential harm caused by resecting pancreatic parenchyma [87, 93, 94].

On a positive note, the discovery of the genetic drivers of IPMNs has identified several potential therapeutic targets. Most IPMNs harbor KRAS mutations, and novel inhibitors of KRAS G12C mutations have been developed, and Zhou and colleagues reported on the use of bispecific antibodies to target neoplastic cells with KRAS mutations [95–97]. Similarly, RNF43 mutations are prevalent in IPMNs, and tumors with RNF43 loss of function mutations may be particularly susceptible to therapies that inhibit the Wnt pathway [98]. Targeted therapies such as these have the potential to be useful in not only treating IPMNs, but, perhaps more importantly, in treating the invasive cancers that arise from IPMNs.

Summary

Our understanding of IPMNs, IOPNs, and ITPNs has come far, but we still have far to go. Too many patients with lesions that mimic IPMNs and too many patients with low-grade IPMNs undergo surgery and are over treated. We need new approaches to more accurately predict the degree of dysplasia in an IPMN and the risk of progression. We need new approaches to account for heterogeneity and to define the risk of the entire pancreas, and we need to do a better job of predicting who will recur after surgery. We believe that improvements in our understanding of pancreatic pathology will form the basis for these advances.