Chapter 58. Cystic Neoplasms of the Pancreas

Cystic neoplasms of the pancreas and other cystic lesions, many of which cause “cyst-like” dilatations of the main or branch pancreatic ducts are collectively referred to as cystic lesions of the pancreas on cross-sectional imaging of the abdomen. The incidence of these cystic lesions increases with age. One autopsy study has demonstrated that up to a quarter of elderly individuals harbor cystic lesions of the pancreas at their demise.1 With the ever increasing use of computed tomography (CT) and magnetic resonance imaging (MRI), cystic lesions of the pancreas are being defined with progressively greater frequency, and an ever increasing number are asymptomatic at discovery.2,3 Some of these lesions will be malignant or have malignant potential at diagnosis, while others are clearly benign and may not warrant further surveillance. Resection of benign cystic pancreas lesions or those containing only carcinoma in situ leads to nearly universal survival, while surgery for invasive carcinoma associated with cystic neoplasms generally has a more favorable prognosis than the results for pancreatic ductal adenocarcinoma.4,5 Thus careful consideration must be given to the diagnosis and prognostic implications of these lesions.

An ideal diagnostic approach would allow for the resection of only those lesions with present or near-future risk of malignancy, while excluding from surgery those individuals with either benign lesions or a prohibitive operative risk, thus minimizing the potential occurrence of mortality and morbidity associated with the surgical treatment of these cystic lesions. Recent advancements in imaging by CT, MRI, and endoscopic ultrasonography (EUS), linked with refinements in the pathological understanding of cystic neoplasms of the pancreas have furthered this effort. History and clinical criteria, such as age, gender, presence of symptoms, location of the neoplasm within the pancreas, as well as, morphology by cross-sectional imaging and cyst fluid analysis by EUS with fine-needle aspiration (EUS-FNA), all may play a role in the diagnosis of pancreatic cystic neoplasms and assessment of the need for resection. A recent analysis using decision analysis with Markov modeling has indicated that for patients focused on overall survival, regardless of quality of life, surgery is optimal for branch duct lesions greater than 2 cm in size.6 For patients more focused on quality-adjusted survival, a 3-cm threshold is more appropriate for surgical intervention, except for the very elderly patients.

Pancreatic pseudocysts (or early postpancreatitis acute fluid collections) have been considered as the most common non-neoplastic cysts of the pancreas. Their diagnosis is aided by a history of acute or chronic pancreatitis.7 Congenital cysts are rare and include those associated with genetic diseases such as autosomal dominant polycystic disease,8 cystic fibrosis,9 and von Hippel-Lindau (VHL) disease.10,11 Lymphoepithelial cysts are rare benign lesions of the pancreas lined with squamous epithelium.12

Three cystic lesions make up 90% of the cystic neoplasms seen in the pancreas: serous cystic neoplasms (SCNs), mucinous cystic neoplasms (MCNs), and intraductal papillary mucinous neoplasms (IPMNs). SCNs rarely demonstrate a progression to malignancy. Unequivocal definition of an SCN may permit nonoperative management of these lesions, provided symptoms do not mandate resection. Mucin-producing lesions of the pancreas can be segregated into two types which may differ significantly in natural history. Restriction of the definition of MCNs, to include only those lesions with subendothelial ovarian-type stroma, has permitted an improved distinction between MCN and IPMN.13 Recent consensus guidelines developed by the International Association of Pancreatology14 may assist in the management of cystic neoplasms of the pancreas. The premalignant nature of MCN prompts resection in patients who are acceptable operative risks, while observation of some branch duct IPMNs, may be tenable with an eventual risk of malignancy less than the operative mortality of pancreatic resection.6 Finally, solid pseudopapillary neoplasms (which may have cystic components) are rare lesions occurring predominantly in young women, for which, resection of the primary tumor results in an excellent opportunity for cure.

The accurate pathological description of pancreatic cystic neoplasms has evolved significantly in the past two decades, influenced largely by an improved understanding of the malignant potential of MCNs, in comparison to the largely benign SCNs, and the emergence of an understanding of the pathogenesis and behavior of IPMNs. Current classification of these tumors follows the World Health Organization (WHO) International Classification of Tumors as published in 2000 (Table 58-1).15 While the diagnostic criteria and organizational schema for these tumors are likely to be adapted further in future editions, the current classification system provides a means to stratify these tumors in terms of prognosis and management. In this review, particular attention will be paid to the three most common lesions: SCNs, MCNs, and IPMNs.

SCNs, previously referred to either as serous cystadenomas, glycogen-rich adenomas, or microcystic adenomas, are almost always benign. Careful delineation of the radiological and clinical features that distinguish these lesions may support and facilitate nonoperative management (ie, observation) of these lesions.

Pathological Features

The majority of SCNs are polycystic or so-called microcystic adenomas, characterized by a well-circumscribed, soft mass which includes numerous small cysts filled with clear serous fluid arranged in a characteristic honeycomb-like pattern. Larger cysts may line the periphery of the lesion. The multiple small cystic loculations are well defined and often focus on a central stellate scar with or without calcifications. These features may be highly suggestive of an SCN when seen on CT or magnetic resonance imaging (MRI) (Fig. 58-1). A small number of SCNs (≤10%) are oligocystic adenomas, and present with one or more dominant cysts, rather than multiple conjoined microcysts. These unusual SCN lesions may be more difficult to distinguish radiographically from MCNs, IPMNs, pseudocysts, and other cystic lesions.

Beyond these gross distinctions, both microcystic and oligocystic adenomas are composed of a single layer of simple cuboidal epithelium with rounded nuclei and clear cytoplasm which is glycogen rich and stains periodic acid-Schiff-positive (Fig. 58-2). The cystic fluid is serous and typically has no mucin content, with a low carcinoembryonic antigen (CEA) level (<5 ng/mL), factors that may provide diagnostic information upon cyst aspiration. Cytology diagnostic for SCN is present in less than 50% of cases; however, when positive the sensitivity is high.

The malignant potential of SCN is so low that most experienced centers recommend management of these lesions as benign entities. Certainly the argument can be made that a clearly documented classic-appearing SCN need not be resected, unless symptomatic or enlarging. The incidence of serous cystadenocarcinoma is extremely low, as fewer than 25 cases have been definitively documented in the literature.16 In the largest single institution experience to date, Galanis and colleagues from Johns Hopkins reported on 158 patients with SCN, only one of whom had serous cystadenocarcinoma on presentation, while another patient, who had a “locally aggressive neoplasm,” returned 13 years postresection with metastatic disease.17 Thus, two cases of aggressive tumor were observed, out of 158 resected SCNs. Evidence of distant metastatic disease is typically necessary to confirm the rare diagnosis of serous cystadenocarcinoma, as both the primary and extrapancreatic disease may appear histologically indistinguishable from benign SCN.18 Vascular and perineural invasion, or local invasion of the stomach and duodenum, are not sufficient criteria for the diagnosis of malignancy of SCN.17,19

Clinical Presentation

SCNs occur predominately in women in the sixth decade of life, while men tend to present at a later age. Bassi and colleagues described 100 patients with SCN, 87 of whom were female, with a mean age at presentation of 52 years.20 The average age of the 13 male patients was 54 years. In another study from the Massachusetts General Hospital, 75% patients were women, and the female patients were significantly younger at presentation than were the men (60 vs 67 years, p = .018).21 In the Galanis study from Johns Hopkins a similar number of patients were women (75%), though no age difference was noted between the genders.17

In 25–75% of the patients with SCN, abdominal pain is the presenting complaint.17,20,21 Weight loss is seen in 14–22%,17,20 and fewer patients (10%)20 present with a mass or fullness. Symptoms typically associated with invasive disease, such as jaundice (6%) or pancreatitis, are uncommon.17 Nausea and vomiting related to compression of the upper gastrointestinal tract may occur in 7–10% of patients.20 Traditionally, SCNs have been described as having a predilection for the pancreatic body and tail, though Le Borgne and coworkers described a relatively even distribution throughout the gland in 170 lesions (38% head, 41% body, 20% tail).22 Large SCNs located in the head are surprisingly unlikely to cause biliary or duodenal obstruction, reflecting their slow pattern of growth, soft texture, and lack of invasive behavior. Rarely, extremely large tumors have been seen in elderly patients, with considerable symptoms of abdominal fullness, and occasionally gastroduodenal obstruction or jaundice.

One clinical condition that has been clearly associated with SCNs of the pancreas is the von Hippel–Lindau (VHL) syndrome. Simple pancreatic cysts or SCNs occur in 17–56% of patients with this heritable multisystem neoplastic syndrome.23 The VHL tumor suppressor gene is located on chromosome 3p25. Vortmeyer et al demonstrated deletion of 3p25 in 7 of 10 sporadic SCN cases studied, suggesting a role for the VHL gene in SCN tumorigenesis, even in the absence of the VHL syndrome.24

Diagnosis

As mentioned earlier, SCNs often have a characteristic imaging phenotype (see Figs. 58-1 and 58-3). Most are well-demarcated solitary multicystic masses composed of innumerable small cysts. Up to one-third have a central, calcified starburst scar.22,25 SCN may also present as oligocystic or unilocular cystic lesions, making differentiation from other cystic lesions of the pancreas difficult. Lee and colleagues reported on the preoperative diagnostic accuracy of CT in pathologically confirmed SCN.26 Radiological features led to a correct diagnosis in only 36% of unilocular SCNs, while honeycombed microcystic and multilocular macrocystic SCNs were appropriately defined in 81 and 88%, respectively (p = .005). Overall in their series, CT diagnosis was accurate in 71% of SCN lesions. In 164 patients with surgically verified pancreatic cystic lesions, 28 of whom had SCN, Shah et al suggested that the CT features predictive of the diagnosis of SCN are microcystic appearance (22/28, 78%), surface lobulations (25/28, 89%), and central scar (9/28, 32%).27 Stepwise logistic regression analysis showed that only a microcystic appearance was predictive for the CT diagnosis of SCN (p = .0001). MRI correctly predicted the pathological diagnosis of SCN with greater frequency than did CT in the study by Bassi and coworkers.20 CT allowed for the correct diagnosis in 54%, incorrect diagnosis in 34%, and was nondiagnostic in 12%. The results with MRI were 74, 26 and 0%, respectively.

The limitations of the radiological diagnosis of SCN may call for additional analysis, which is frequently sought by EUS-FNA with cyst fluid cytology and biochemical study. The risk of complications with EUS-FNA is relatively low, with significant bleed occurring in less than 1% of studies, though intracystic hemorrhage may be as high as 6%.28 The risk of cyst bacterial inoculation with subsequent infection is less than 1%29 and the rate of pancreatitis is 1–2%.30 Cyst fluid aspirates from SCN are frequently sparsely cellular and may be contaminated with columnar enterocytes and mucin from the scope and needle traversing the gastric or bowel mucosa, potentially clouding the diagnostic accuracy of cytology. Cytology alone was found to be diagnostic of SCN in only 7 of 21 cases studied by Huang and others from M. D. Anderson.31 Detection of intracytoplasmic glycogen was noted to enhance the diagnostic confidence for the diagnosis of SCN. A recent 10-year review of 317 patients evaluated by EUS for pancreatic cysts by Pausawasdi and colleagues revealed that those followed for asymptomatic incidental cysts less than 3 cm in size almost uniformly had stable disease over a relatively short mean follow-up of 28 months.32 While these results are confirmed by other EUS studies, others prefer serial imaging (typically MRI to avoid radiation exposure) over EUS surveillance.

Cyst fluid analysis is an additional adjunct (beyond cytology) to improve the diagnostic accuracy of EUS-FNA. Fluid from within an SCN is typically low in viscosity and amylase, due to a consistent lack of connection to the pancreatic ductal system.33 CEA levels less than 5 ng/mL have a sensitivity of 54–100%, and specificity of 77–86% in the differentiation of SCN from other pancreatic cystic lesions.34,35 The finding of a cyst fluid carbohydrate antigen (CA) 19-9 level less than 37 U/L and a CEA less than 5 ng/mL virtually excludes an MCN or IPMN.

Allen et al recently reported on the analysis of cyst fluid using a biomarker panel developed for pancreatic cancer.35 Assessment of protein expression within the cyst fluid led to an error in classification of lesions of 27%, when all three types of cystic neoplasms were evaluated (SCN, MCN, and IPMN). When limiting the analysis to separating SCN from IPMN, this method had an error rate of only 8%, compared with a 14% error rate for use of CEA levels alone. The greatest utility of protein expression analysis might be in the differentiation of cystic lesions of the head of the pancreas, as the vast majority of MCN occur in the body and tail of the pancreas. However, the cost of this method may not be justified by the relatively small improvement in diagnostic accuracy. When cross-sectional imaging is sufficiently compelling, EUS alone is often unnecessary, but EUS-FNA may further clarify the diagnosis, particularly with oligocystic and unilocular SCNs.

Treatment

Observation of patients with SCN may be appropriate in asymptomatic patients. When a secure diagnosis of SCN is made, modern series demonstrate that a growing number of SCN are being followed by serial imaging (Table 58-2). Typically a pathological diagnosis is not required. Bassi and colleagues followed 32 patients with the diagnosis of SCN for a median time of 69 months, without any observed development of malignancy or significant increase in diameter of the lesion.20 Rapid rate of growth of a lesion may heighten suspicion for the development of malignancy or increase the likelihood of developing symptoms. In a report from the Massachusetts General Hospital, Tseng and coworkers found a more rapid rate of growth in SCN greater than or equal to 4 cm in size at presentation compared with smaller tumors (1.98 cm/y vs 0.12 cm/y, p = .0002).21 Tumors less than 4 cm were less likely to be symptomatic than were those greater than or equal to 4 cm (22 vs 72%, p < .001). Resection was thus suggested by these authors, even for asymptomatic SCNs which were greater than or equal to 4 cm.

When the diagnosis of SCN is uncertain, pancreatic resection is most often performed according to oncological principles, as if the lesion was malignant or had malignant potential (Fig. 58-4). Standard procedures include distal pancreatectomy for lesions of the body or tail, or pancreaticoduodenectomy for right-sided lesions. This practice avoids performance of an inadequate cancer operation in cases in which a malignancy is found on final pathological analysis. However, if the diagnosis of SCN is confirmed either preoperatively or with intraoperative biopsy (which is not routinely performed), a less radical approach may be considered. Enucleation of SCNs has been shown to be technically feasible, although it can be associated with a significant risk of pancreatic fistula.36,37 A central pancreatectomy, with distal pancreatic reconstruction via pancreaticogastrostomy or Roux-en-Y pancreaticojejunostomy may be considered in selected patients with lesions of the pancreatic neck.38 Distal pancreatectomy with splenic preservation may also be considered, particularly for small lesions in the tail, where the splenic hilum is more easily dissected. Lesions in the head of the pancreas that are not amenable to enucleation are best treated with pylorus-preserving pancreaticoduodenectomy. In patients who have an otherwise normal pancreas, meticulous attention must be paid to the technique of pancreaticojejunostomy, since such patients have a significantly higher risk for developing a pancreatic fistula. There has been some enthusiasm recently for duodenum-preserving pancreatic head resection as well, although this procedure has not had widespread application.39 Patients with pathologically proven, completely resected SCN do not require serial imaging in follow-up. Recommendations for appropriate monitoring of unresected SCN vary, but serial imaging with either CT or MRI every 6 months for 2 years, and then annually thereafter seems reasonable.40

Progress in the diagnosis and management of pancreatic cystic neoplasms has been aided in large part by the recognition of distinct pathological features that distinguish MCNs from other cystic lesions.41,42 The distinction between MCN and SCN is critical, as the premalignant and malignant behavior of MCNs stand in stark contrast to the nearly universally benign SCNs. Many of the same diagnostic challenges that exist for SCN are true for MCN, but the management decisions may be quite different, due to the differing clinical phenotype of these lesions.

MCNs account for approximately 15–30% of cystic neoplasms of the pancreas, at least as recorded in recent series in which the distinction of these lesions from IPMN has been clarified.2,40,43 Clinical series published prior to the establishment of diagnostic criteria for IPMN in 1996 likely overestimated the relative prevalence of MCNs in comparison to other cystic lesions, since they included what are now categorized as IPMNs as various “mucinous tumors.”

Pathological Features

MCNs are typically spherical, thick-walled, septated or unilocular cysts with a tall columnar mucin-producing epithelium accompanied by a subendothelial ovarian-type stroma that appears as a dense layer of spindle cells with sparse cytoplasm and uniform, elongated nuclei (Figs. 58-5). This stroma regularly expresses progesterone receptors, and less frequently estrogen receptors, and over 60% stain for human chorionic gonadotropin.44 Both the WHO and the Armed Forces Institute of Pathology (AFIP) have defined the presence of this ovarian-like stroma as a requirement for the diagnosis of an MCN.41,42 The Sendai consensus guidelines have also required the presence of ovarian-type stroma as a necessary criterion for the diagnosis of MCN, so as to prevent the misclassification of IPMN as MCN.14 Given the similarity of the histology and immunohistochemistry between MCN and ovarian mucinous cystadenomas, MCNs have been postulated to arise from ovarian rests (or ovarian-like stem cells) within the pancreas.45

MCNs exhibit characteristics of an adenoma-carcinoma sequence. Dependent on the degree of atypia, they are classified as mucinous cystadenomas, borderline lesions, in situ lesions or invasive cystadenocarcinoma. Atypical changes within the lining epithelium may be patchy and sparse, with abrupt transitions to normal mucosa. Classification of MCN should be based upon the highest degree of atypia present, and the entire lesion should be examined pathologically.46,47 Invasive carcinomas arising within MCNs are usually tubular or ductal type, though some may be undifferentiated carcinoma with osteoclast-like giant cells,48 adenosquamous carcinoma,49 choriocarcinoma or even high-grade sarcomas.50 Colloid carcinomas are extremely rare in MCN, but they occur commonly in IPMN.47

Clinical Presentation

In light of the mandatory presence of ovarian-type stroma, not surprisingly, MCNs are now diagnosed almost entirely in women.44,45,51–56 This requirement, combined with the usual lack of communication with the pancreatic duct, defines a unique phenotype separate from IPMN. In a combined report from the University of Verona and the Massachusetts General Hospital, Crippa and colleagues reviewed their experience with 163 MCNs that met the WHO criteria for diagnosis.55 Of the 163 patients, 95% (155 patients) were perimenopausal females. Only 8 males were identified, and they were significantly older than the female patients (63 vs 44 years, p = .011). The location of MCN within the gland was almost entirely confined to the body and tail of the pancreas (97%), and only five lesions were found in the pancreatic head. In reviewing the literature regarding MCN, these researchers noted the importance of segregating studies according to whether or not the presence of ovarian-type stroma was required for inclusion of pathological specimens within collected reports. Goh et al reviewed those studies where the presence of ovarian-type stroma was a mandatory criterion for the diagnosis of MCN: 99.7% of the patients were women, the mean age at presentation was 47 (range, 18–95) years, and 95% of MCNs occurred to the left of the pancreatic neck.57 By comparison, when this criterion was previously not a prerequisite to diagnose MCN, patients were older, more often male and the lesions were located in the head with a frequency exceeding 30%.

Abdominal pain or discomfort is the most common presenting symptom, occurring in over 70% of patients.52–54 A history of acute pancreatitis may also be elicited in 9–13% of patients, although less commonly than in patients with IPMN.4,52,55 Patients with MCN with an associated invasive carcinoma present 11 years later than those with noninvasive neoplasms, likely representing the longer time required to progress to overt malignancy within these neoplasms.55

Diagnosis

Macroscopically, MCNs have some characteristic features that may be evident during imaging or operative evaluation. Classically, MCNs contain large septated cysts with thick irregular walls that may be well visualized on CT, MRI, or ultrasound evaluation. Papillary projections from the epithelium often extend into the cystic cavities and may be visible, particularly on high quality axial or endoscopic ultrasound imaging. In a minority of cases, the wall of the MCN may contain calcifications, a characteristic associated with a higher likelihood of malignancy.58 MCNs may also present as large unilocular cysts that may appear similar on cross-sectional imaging to long-standing pseudocysts (Fig. 58-6). Two distinguishing characteristics in this scenario are the lack of surrounding inflammatory changes beyond the wall of the neoplasm in MCN, and the absence of pancreatitis, two features common in evolving pseudocysts related to pancreatitis.7 Demonstration of ductal communication with the cyst by MRI or magnetic resonance cholangiopancreatography (MRCP) may distinguish pseudocyst or IPMN from MCN, though MCN can in rare instances exhibit connection with the pancreatic duct.57

Similar to SCN, determination of a treatment plan for MCN is predicated upon whether or not a given lesion is mucinous. Analysis of cyst fluid aspirated from MCNs typically show elevated levels of CEA and low amylase concentrations. The Cooperative Pancreatic Cyst Study demonstrated that a CEA value greater than 192 ng/mL achieved the greatest efficiency for differentiating mucinous from nonmucinous lesions.25 The accuracy of CEA (88/111, 79%) was greater than the accuracy of EUS morphology or cytology (p < .05). No combination of tests further improved diagnostic accuracy. A CEA level greater than 800 ng/mL has a specificity of 98% for predicting MCN, but a sensitivity of only 48%.59 Khalid and his coinvestigators tested the utility of DNA analysis of cyst fluid to diagnose mucinous and malignant cysts.60 The presence of a K-ras mutation was highly specific for a mucinous cyst (96%), but had a low sensitivity of 45%. A considerable selection bias was introduced by the study design which may have overestimated the ability of DNA analysis to define a mucinous cyst.61 Presence of a K-ras mutation in cyst fluid may provide additional information when CEA levels are not discriminative, particularly in lesions that appear to not have clear imaging patterns which allow separation of SCN versus MCN.

Treatment

In their pooled review of the literature, including 10 studies of MCN defined by ovarian-type stroma, Goh and coworkers noted that, in the 40 invasive carcinomas found in 344 patients, only one of the malignant MCNs was less than 4.5 cm at the time of resection.57 Crippa and colleagues noted that lesions containing either in situ or invasive carcinoma were larger (median size 80 vs 45 mm, p = .0001), and intracystic nodules or papillae were more frequently present (64.3 vs 4.4%, p = .0001), when compared with benign neoplasms.55 All lesions demonstrating cancer on pathology were either greater than 4 cm in diameter or contained nodules by preoperative imaging. Careful observation of asymptomatic lesions less than 3 cm in size, without the presence of nodules, appears to be a reasonable approach for MCN (see Table 58-2). However, a recent post-hoc analysis by Sawhney and associates has questioned whether size alone, based on the Sendai Consensus criteria,14 is a sufficient predictor of malignancy in pancreatic cysts.62 Their data indicated that the original consensus guidelines should be applied with caution, and that more accurate diagnoses might be generated by the combination of cyst size and main pancreatic duct dilation greater than 3 mm.

Biopsy of MCN should not be utilized to determine the presence of carcinoma, because the presence of invasion within a lesion may be patchy and a negative biopsy may be obtained based on sampling error. Due to the significant rate of malignancy and the risk of progression to malignancy associated with MCN, symptomatic neoplasms, lesions greater than 3 cm or those containing nodules or papillae should undergo resection. As with SCN, enucleation has been documented to be an effective strategy for resection in selected MCN cases.36,37,63 However, there is some risk of performing an inadequate oncologic resection for an MCN should it harbor an invasive component, while there is virtually no risk for SCN. Therefore enucleation should only be applied to highly selected cases of small, peripherally located MCNs with confirmation by extensive frozen-section analysis. Likewise, segmental pancreatic resections for lesions in the pancreatic neck and body (central pancreatectomy) or tail (spleen-preserving distal pancreatectomy) should be performed cautiously in selected patients without any indication of invasive disease. Larger tumors in older patients (ie, patients fitting the characteristics of MCN with an associated invasive cancer) should be treated with formal pancreatic resection to include specimen-associated lymph node harvest. Lesions in the pancreatic head are best treated with pancreaticoduodenectomy, while left-sided lesions are treated via distal pancreatectomy with en bloc splenectomy. Extended lymphadenectomy, which has not been shown to definitively improve locoregional control or survival in patients with pancreatic ductal adenocarcinoma, has no role in the treatment of patients with cystic neoplasms.64,65

The 5-year disease-specific survival for benign or noninvasive MCN is 100%, but falls to 50–60% for patients with invasive mucinous cystadenocarcinoma55 (Fig. 58-7). Failure to completely resect a noninvasive MCN may result in a later recurrence (persistence), and a missed opportunity for cure.

Adjuvant therapy for mucinous cystadenocarcinoma has been poorly investigated and has no proven benefit. A single case report describes the use of neoadjuvant chemoradiation and treatment monitoring by serum CEA level, but no prospective clinical trials have been performed.66 Some high-volume centers would likely offer adjuvant chemotherapy to patients with invasive cystadenocarcinoma, extrapolating from the experience with ductal adenocarcinoma.67 There are no data to support the utility of adjuvant radiotherapy. Follow-up with serial MR imaging every 6 months for 2 years and annually thereafter appears reasonable for patients with resected MCN with an associated invasive cancer.36 Patients with resected noninvasive MCNs receive no postoperative adjuvant therapy and are not typically followed with serial imaging.

In the past, these lesions were variably referred to as mucinous ductal ectasia, intraductal papillomatosis, intraductal adenoma or adenomatosis, intraductal mucin-secreting tumor, and intraductal papillary mucinous tumor. However, the earliest report of this “new” lesion is attributed to Ohashi and Maruyama and was published in the Japanese literature in 1982.68 This report described four malignant lesions associated with the main pancreatic duct and characterized the now well-described copious amounts of mucus that distend and emanate from the ductal system. The authors noted the comparatively better survival of these patients compared to those with classic invasive ductal adenocarcinoma of the pancreas. While many subsequent authors have helped to further characterize the subtleties of IPMN, these initial observations accurately depict typical cases.

In 1996, the WHO first formally recognized IPMN as a distinct entity, establishing criteria for the pathological diagnosis of these lesions.41 Characteristic features include a tall, columnar epithelium with marked mucin production, and cystic transformation of either the main pancreatic duct or one of its side branches (Fig. 58-8). More recent versions of these diagnostic criteria have allowed the stratification of noninvasive IPMNs based on their degree of dysplastic change, and the clear separation of noninvasive IPMN from IPMN with an associated invasive carcinoma.

Pathological Features

Histologically, IPMNs are characterized by intraductal proliferation of mucinous cells which form papillae. Secretion of mucin leads to dilatation of the pancreatic ducts (Fig. 58-9). Lesions may be localized, multicentric, or rarely involve the entire ductal system. The proliferation of mucinous cells may involve the main pancreatic duct (main duct type), or be confined to the branch ducts (branch duct type) or show a pattern spanning both areas in a combined type. Three different morphologic patterns of IPMN can be seen.69 Most branch duct type IPMN demonstrate papillae lined by tall columnar cells with basally oriented nuclei and abundant pale mucin. This pattern is also prevalent in the nonpapillary areas of main duct type IPMN and appears similar to the gastric mucosa. Scattered goblet cells are present and stain for MUC2. This pattern is called the gastric-foveolar type. Most main duct type IPMN closely resemble colonic villous adenomas and show molecular characteristics of intestinal differentiation, such that the cells express CDX2 and MUC2.70 These are classified as villous-intestinal type. Their papillae are also positive for MUC5AC. Cancers arising in these IPMN are typically colloid carcinomas (Fig. 58-10). Colloid carcinomas also express CDX2 and MUC2, but not MUC1. A small proportion of IPMN is more complex and lined by cuboidal cells which do not express MUC2 or CDX2. This form is referred to as the pancreatobiliary type. Invasive cancers associated with the pancreatobiliary morphology are usually tubular, with structure similar to ductal adenocarcinoma. The invasive component expresses MUC1, but not MUC2. While both the villous-intestinal and the pancreatobiliary types may be found alongside the gastric-foveolar type, it is however, uncommon to identify both the villous-intestinal and pancreatobiliary type of papillae in the same IPMN.69

IPMNs demonstrate a progressive precursor model of carcinogenesis similar to that seen in colon cancer.71 Tall mucin-producing columnar epithelial cells that remain well differentiated characterize IPMN adenoma (low-grade dysplasia). Little or no dysplasia is present in these lesions. IPMN borderline lesions (moderate grade dysplasia) are described as lesions with moderate epithelial dysplasia, characterized by moderate loss of polarity, changes in nuclear morphology, and pseudopapillary formation (Fig. 58-11). IPMNs with carcinoma-in situ have severe dysplastic changes. These lesions may be papillary or micropapillary, and severely dysplastic lesions may lose the ability to secrete mucin. IPMNs are pathologically similar to pancreatic intraepithelial neoplasia (PanIN). IPMNs, like PanIN, are intraductal lesions that may demonstrate a range of cellular atypia and malignant transformation. However, IPMNs may be distinguished based on their gross visibility and involvement of large ducts. PanIN should be considered a microscopic finding involving ducts less than 5 mm in diameter, while IPMNs are macroscopic findings.72 In addition, IPMNs often express the mucin MUC2, while PanINs usually express MUC1.

IPMNs appear also to have distinct molecular events contributing to the clinical and pathological behavior that further distinguish them from lesions in the PanIN–ductal adenocarcinoma sequence. Iacobuzio-Donahue and associates described the intact (normal) expression of the tumor-suppressor gene Dpc4 in the intraductal components of 79 IPMN.73 In contrast, Dpc4 inactivation has been shown to be relatively specific for pancreatic adenocarcinoma, and its persistence in both noninvasive and invasive IPMNs argues that these lesions may arise through a pathway that is distinct from the PanIN–ductal adenocarcinoma. IPMNs also appear to have a significantly lower rate of K-ras and p53 mutations, lesions that are common in ductal adenocarcinoma.74 Fritz and coworkers recently demonstrated that loss of chromosome 5q, 6q, and 11q was significantly higher in IPMN with high-grade dysplasia or invasion compared with ductal adenocarcinoma.75 These data and others suggest that IPMNs are unique pancreatic neoplasms, with a pathogenesis that is distinct from that of the PanIN–ductal adenocarcinoma sequence.

Clinical Presentation

The biologic behavior of IPMNs parallels their classification according to their distribution within the pancreatic ductal system. Main duct type and combined main duct and branch duct type lesions are more likely to present with symptoms, while strictly branch duct type IPMNs are more frequently detected as asymptomatic cystic neoplasms on cross-sectional imaging.76 Pancreatitis is seen more commonly in main duct type IPMN, possibly related to mucous plugging of the ampulla. In a combined experience of the Massachusetts General Hospital and the University of Verona reported by Salvia and colleagues, acute pancreatitis occurred in 23% of 140 patients with main duct variant IPMN.77

Both genders are affected by IPMNs, with a moderate male predominance in some series. Patients with IPMN tend to be older, with a mean age of 65 years, as compared with those having MCN who are predominantly perimenopausal. Similar to the situation with MCN, IPMN patients demonstrated to have a malignancy, a trend toward being older, again suggestive of an adenoma-to-carcinoma sequence of progression and the time necessary to undergo this transformation. Main duct type IPMNs are more likely to demonstrate the development of malignancy. Of note, malignant IPMNs are more likely to present with symptoms typically attributed to ductal adenocarcinoma, such as obstructive jaundice and weight loss.5,77

The development of symptoms more commonly attributed to ductal adenocarcinoma in patients with IPMN may herald the occurrence of carcinoma either synchronously or metachronously in the gland. Several studies have demonstrated the presence of an invasive ductal adenocarcinoma elsewhere in the pancreas, distinct from the location of the cystic neoplasm in up to 10% of IPMN patients.40,78,79 Ingkakul and colleagues recently showed that, in a multivariate analysis, worsening diabetes (odds ratio 15.73 [95% CI: 4.40–56.25]; p < .001), and an abnormal CA 19-9 (odds ratio 3.70 [95% CI: 1.19–11.48]; p = .024) are independent factors predictive of synchronous or metachronous separate ductal adenocarcinoma in patients with IPMN.79 In a report from Memorial Sloan-Kettering, Allen et al sought to characterize those cystic lesions which should be initially resected.40 Asymptomatic cystic neoplasms, less than 2.5 cm in size, without septations or a solid component were followed over time. Eventually, 28 of 369 patients, initially managed conservatively, were operated upon primarily for cyst growth. Malignancy was found in 11 of these 28 patients (38%): three were cystic neuroendocrine tumors, while 8 of 11 were ductal adenocarcinomas. Further review of these eight ductal adenocarcinomas demonstrated that the cancer arose adjacent to the initially discovered cyst. This study demonstrates the need to thoroughly investigate the entire pancreas when electing to observe a cystic neoplasm. In addition, the incidence of extrapancreatic malignancies appears to be higher in patients with IPMN.80,81 The development of colorectal adenomas and carcinomas, Barrett's mucosa and gastric carcinomas appear to be important entities seen in IPMN patients.

Diagnosis

CT scanning (with all of its advances up to multidetector imaging) has been the primary method for imaging the pancreas in the past. However, the use of MRI, particularly in combination with MRCP imaging has allowed for more thorough identification of IPMN. IPMNs characteristically appear as cystic masses resulting from dilatation of the main pancreatic duct or side branch ducts. Polypoid projections (mural nodules) into the cystic spaces may be present. Approximately half of IPMNs occur in the pancreatic head, though they may be present anywhere within the pancreas and can diffusely involve the entire gland. Currently MRCP is the modality of choice for defining mural nodules, demonstration of the communication of the cystic neoplasm with the pancreatic ductal system, and evaluating the extent of the pancreatic ductal dilatation.82 Use of MRCP has largely supplanted endoscopic retrograde cholangiopancreatography (ERCP) in the diagnosis of IPMN, since MRCP is noninvasive, does not require sedation, and does not carry the risks of pancreatitis and perforation which accompany ERCP.

With the increasing emphasis on managing asymptomatic branch duct type IPMN by observation, imaging studies have sought to define features associated with invasive carcinoma by correlating preoperative imaging with pathology obtained by resection. Several imaging features suggestive of the presence of malignancy have been demonstrated, including tumor size (cyst diameter ≥30, 40, or 50 mm), main duct type IPMN, main duct dilatation greater than or equal to 10 or 15 mm, patulous papilla, mural nodules (≥3, 5, or 10 mm in size), presence of biliary ductal dilatation greater than or equal to 15 mm, a solid mass, or occurrence of an area of abnormal attenuation in the surrounding pancreas.82–90 Importantly, a recent report from Verona, Italy by Salvia and colleagues notes that they followed 121 patients with multifocal branch duct IPMN (median diameter of the largest lesion being 1.7 cm) over a 40-month observation period.91 All of the 121 patients remained alive, without surgery, and all remained asymptomatic. Thus, there is clearly a role for conservatism in the management of patients with branch duct IPMNs and no additional worrisome features.

Endoscopic ultrasound (EUS) may provide additional information which may prompt resection or promote a more conservative approach. Ohno et al demonstrated that the finding of a papillary mural nodule or a nodule exhibiting an invasive component on EUS was predictive of malignancy with a sensitivity of 60%, specificity of 93%, and an accuracy of 76%.92 EUS-FNA may also be useful in reinforcing a decision not to resect a branch duct type IPMN, if it is otherwise without features predictive of malignancy. Marie and colleagues found that the combination of a CEA level less than 200 ng/mL and a CA 72.4 level greater than 40 U/mL retrieved from the cystic material of an IPMN together had a 96% negative predictive value for the diagnosis of malignancy.93

Treatment

The Japan Pancreas Society performed a multi-institutional, retrospective study of 1379 cases of IPMN drawn from 98 of their member programs. The clinicopathologic features of benign IPMN (see Table 58-2) (adenoma [low-grade dysplasia] and borderline lesions [moderate dysplasia]; n = 564) were strikingly different when compared with tumors containing frank adenocarcinoma (n = 445).85 Patients with adenocarcinoma were significantly older (67 vs 65 years, p = .0002) and more frequently symptomatic (49 vs 35%, p < .0001), as compared to the noncarcinoma group. Cancer occurred more commonly in either main duct type or combined-type tumors, as compared to branch duct type neoplasms (60, 65, and 30%, p < .001), respectively. The preoperative imaging of patients who were subsequently found to have adenocarcinoma on pathology, demonstrated a higher incidence (63 vs 28%) and size of mural nodules (12 vs 5 mm) when compared with those who had benign lesions (both p < .0001). Branch duct type tumors with cancer were larger (35 vs 28 mm, p < .0001) than those without cancer.

Based on the data generated in the earlier report, the International Association of Pancreatology convened a consensus conference in Sendai, Japan in 2004. The subsequent guidelines published in 2006, have become a new benchmark for the management of IPMN.14 These guidelines recommend the resection of all IPMN of a main duct type and mixed variants, those showing main pancreatic duct dilatation greater than or equal to 10 mm, as well as those with the presence of mural nodules, or a positive cytology, provided the patients are reasonable candidates for surgery with an acceptable life expectancy. All symptomatic IPMNs were deemed to warrant resection. These recommendations were predicated upon the risk of carcinoma in symptomatic or main duct type lesions. Branch duct IPMNs less than 30 mm in diameter, without evidence of mural nodules or main duct dilatation, were felt to be of low malignant potential and were candidates for careful observation. At follow-up examinations, appearance of symptoms, cyst expansion to greater than 30 mm, detection of positive cytology on FNA, development or identification of mural nodules or main pancreatic duct dilatation (≥6 mm) were deemed indications for resection.

Since the development of the Sendai guidelines, much of the subsequent literature has sought to examine the accuracy of the recommendations, particularly with regard to the observation of asymptomatic branch duct type IPMN. Pelaez-Luna and colleagues identified 147 patients with branch duct type IPMN, of whom 66 underwent resection at diagnosis and 81 were followed over time (of which 11 were resected during the follow-up period).94 Of the patients undergoing resection who demonstrated Sendai consensus guideline indications for surgical therapy, 9/61 (15%) had carcinoma on pathology, whereas none of the 16 patients without consensus indications for resection had malignancy (p = .1). A single guideline indication for resection taken as an indicator of carcinoma had a sensitivity, specificity, positive predictive value, and negative predictive value of 100, 23, 14, and 100%, respectively.

Several studies have suggested that the development of mural nodules is predictive of the risk of developing malignancy, while a progressive dilatation of duct size remains controversial. Schmidt and colleagues identified 103 patients with branch duct type IPMN.95 The mean size of the 20 malignant lesions was 2.0 ± 0.1 cm, while the mean size of the nonmalignant neoplasms was 2.2 ± 0.1 cm, suggesting that size alone is an insufficient indicator of malignancy. In multivariate analysis, only the presence of mural nodules and atypical cytopathology were predictive of the presence of carcinoma. Tanno et al prospectively followed 82 patients with flat lesions within branch duct type IPMN diagnosed by CT or MR and EUS.96 During a median follow-up of 59 months, 9/82 patients (11%) exhibited progressive dilatation of the cystic lesion. Six elected to continue regular screening, while three underwent resection; the IPMNs resected were staged as IPMN-adenoma in two and IPMN-borderline in one. Four patients (5%) developed mural nodules during a median follow-up of 105 months. All four of these individuals were resected, demonstrating IPMN-adenoma in three and carcinoma in situ in the fourth. Sixty-nine of the 82 patients (84%) showed no changes in their dilated branch duct lesions over a median follow-up of 57 months.

A recent study from Kyushu University attempted to determine whether cyst size is predictive of the malignant potential in flat branch duct type IPMN.97 One hundred seventy patients with branch duct type IPMNs without mural nodules were retrospectively identified from their previous 10-year experience. Seventy-three patients underwent resection of their IPMN: 26 patients had lesions less than 30 mm in size, while 47 patients had neoplasms greater than 30 mm in diameter. All of the noninvasive (n = 5) and invasive (n = 1) malignancies were seen in the IPMN of greater than or equal to 30 mm. In a similar report, Salvia and coworkers followed 89 patients with flat branch duct type IPMN less than or equal to 3.5 cm in size for a median time period of 32 months.98 Five patients (5.6%) exhibited an increase in diameter of the cystic lesion, none of which demonstrated carcinoma in the resection specimen pathologically. Clearly, longer follow-up will be needed to determine whether or not clear evidence of carcinoma develops in patients without Sendai consensus guidelines for surgical therapy. As might be anticipated, increasing knowledge and follow-up has raised questions about the universal accuracy of the consensus guidelines.

The majority of studies, particularly those following IPMNs conservatively in a prospective fashion, would suggest that the development of invasive carcinoma in flat branch duct IPMN less than 30 mm in size is unusual. The occurrence of high-risk stigmata (mural nodules, dilated main duct, or positive cytology) clearly have great predictive value for the ultimate finding of malignancy. EUS appears to be an important adjuvant to fully evaluate IPMN patients for the presence of mural nodules, as well as for aspiration of cytologic specimens. Some authorities insist that any lesion which is to be followed conservatively should be examined by EUS at regular intervals. We have tended to use MRI or MRCP for serial surveillance of small (<3 cm) branch duct IPMNs, as this is a noninvasive procedure (as compared to EUS) which avoids radiation exposure (as compared to CT).

Given the excellent survival following resection of IPMN free of an invasive component, every effort must be made to define lesions at risk for the development of carcinoma at the earliest point possible (Fig. 58-12). Schnelldorfer and coworkers have demonstrated that the survival after pancreatectomy of patients with IPMN with invasive adenocarcinoma is equivalent to that of a matched cohort of patients following resection of ductal adenocarcinoma (median survival, 32 vs 21 months; 5-year survival rate, 31 vs 24%; p = .26).99 Other studies have revealed that survival of patients without lymph node involvement and invasive IPMN is quite good, while patients with lymph node involvement and invasive IPMN have equivalent outcomes to patients with lymph node–positive pancreatic ductal adenocarcinoma.5 Despite the poor survival in patients with invasive disease, surgery remains the best opportunity for cure. Swartz et al have recently shown that adjuvant chemoradiotherapy confers a 57% decrease in the relative risk of mortality after pancreaticoduodenectomy for invasive IPMN after adjusting for major confounders.100 This effect was most significant in patients with lymph node metastases or positive surgical margins.

Use of the Sendai consensus guidelines means that the preponderance of resections for IPMN will be performed with at least a suspicion of the presence of carcinoma. Targeted pancreatectomies, either pancreaticoduodenectomy or distal pancreatectomy with en bloc splenectomy, have been advocated so as to adhere to oncologic principles of resection. Most centers have advocated the use of frozen-section examination of the pancreatic margin, with attempted clearance of microscopically malignant margins by re-resection and occasional conversion to total pancreatectomy when needed to achieve negative margins. Skip lesions clearly occur, such that a normal resection margin may not be indicative of a lack of neoplasia in the pancreatic remnant. A recent report by Nara et al from Tokyo analyzed 130 consecutive patients undergoing resection for IPMN with frozen-section analysis of the pancreatic margin101. While most initial frozen-section results showed no neoplasia at the margin, 29 patients had additional pancreas resected for “positive” frozen-section results (12 for low or moderate dysplasia, 10 for high-grade dysplasia, 1 for floating cancer cells, and 6 for invasive cancer). Most patients who recurred following re-resection had their recurrence at a distance from the pancreatic margin (peritoneum, liver, and lymph nodes), raising doubt about the true value of re-resection for margins determined to be positive at frozen section. The role of total pancreatectomy to achieve clearance of all dysplastic epithelium, even prophylactic total pancreatectomy, is controversial. Notably, of the 84 patients with noninvasive IPMN described by Sohn and colleagues,5 7 patients developed recurrent disease in the pancreatic remnant. Negative margins at resection do not eliminate the need for chronic surveillance of the pancreatic remnant, perhaps best done by annual MRI/MRCP.

Solid Pseudopapillary Neoplasm (Solid and Papillary Neoplasm)

These rare tumors are notable for several characteristic clinical and pathological features. The ratio of women to men is roughly 10:1, with lesions typically appearing in the second or third decade of life (mean age 22 years, range 2–85 years). Patients often present either with abdominal pain or a palpable abdominal mass. The lesions may be large, presenting in one review at a mean size of 6.1 cm (range 0.5–34.5 cm).102 On CT, these tumors often appear well circumscribed, with hypodense areas representing hemorrhage or necrosis (Fig. 58-13). Lesions tend to be evenly distributed throughout the pancreas. Most of these lesions harbor β-catenin mutations, express the beta subtype of estrogen receptors and stain for galectin-3.103 Though most of these tumors are benign, some may be considered low-grade malignancies, with local invasion into contiguous structures and occasional distant metastases (roughly 15–20% of cases). An aggressive surgical approach is warranted for both the primary and metastatic disease, as 5-year survival in completely resected patients exceeds 95%.104,105

Cystic Neuroendocrine Neoplasms

Neuroendocrine tumors showing partial or complete cystic components are uncommon. The Cooperative Pancreatic Cyst Study demonstrated only 5 of these lesions out of 341 cystic neoplasms. Immunohistochemical staining of cytological specimens obtained by EUS-FNA demonstrating endocrine markers confirms the diagnosis.106 In a retrospective review of 170 patients undergoing resection for a pancreatic neuroendocrine tumor at the Massachusetts General Hospital over a 30-year period, 29 cystic neuroendocrine tumors were identified.107 Ten (34%) of the cystic lesions were purely cystic, while 19 (66%) were partially cystic. Cystic neuroendocrine neoplasms were larger (49 vs 23.5 mm, p < .05), more likely to be symptomatic (73 vs 45%, p < .05), and most likely to be nonfunctional (80 vs 50%, p < .05), when compared with solid pancreatic neuroendocrine lesions. The propensity for metastases, invasion, and survival (87 vs 77% at 5 years, p = .38) in patients with cystic lesions was the same as in those with solid pancreatic endocrine neoplasms. Thus, these lesions have a favorable prognosis if completely resected and should be treated aggressively and with similar technique as for a solid neuroendocrine tumor in appropriate surgical candidates.

Cystic Acinar Cell Neoplasms

Acinar cell carcinoma of the pancreas is a rare neoplasm, however, several recent registry reviews and multi-institutional series have better defined this entity.108,109 This lesion has a 2:1 male predominance and although many individuals will present with advanced disease, stage-specific survival is statistically better than that seen in ductal pancreatic adenocarcinoma.108 Occasionally, acinar cell carcinoma may display an intraductal, papillary or papillocystic growth pattern and may appear to mimic IPMN with a cystic component. A significant proportion of these tumors (up to 40%) may demonstrate a concomitant endocrine neoplasm.109 Acinar cell neoplasms with intraductal growth patterns tend to present somewhat earlier than typical acinar cell carcinoma, secondary to the pancreatitis resulting from duct obstruction. Characteristic immunohistochemical staining for trypsin and chymotrypsin, as well as presence of eosinophilic granular cytoplasm in acinar cell carcinoma are helpful in establishing the correct diagnosis.109

Cystic Degeneration of Pancreatic Ductal Adenocarcinoma

While not truly a distinct lesion, it is important to realize that pancreatic ductal adenocarcinoma may present with cystic features. Thus all cystic lesions should at least be considered as potential pancreatic ductal adenocarcinomas until an alternative diagnosis is established. In a comparative review of symptomatic and incidental pancreatic cysts by Fernandez-del Castillo and colleagues, 9% of symptomatic lesions and 2% of incidental cysts proved to harbor pancreatic ductal adenocarcinoma.43 Adenocarcinomas that obstruct the pancreatic duct may be associated with retention cysts in up to 8% of patients.110