Moving beyond Morphology: New Insights into the Characterization and Management of Cystic Pancreatic Lesions 1

Note: This copy is for your personal non-commercial use only. To order presentation-ready copies for distribution to your colleagues or clients, contact us at www.rsna.org/rsnarights. Patrick C. Freeny, MD Michael D. Saunders, MD Moving beyond Morphology: New Insights into the Characterization and Management of Cystic Pancreatic Lesions 1 The frequency of detection of cystic pancreatic lesions with cross-sectional imaging, particularly with multidetector computed tomography, magnetic resonance (MR) imaging, and MR cholangiopancreatography, is increasing, and many of these cystic pancreatic lesions are being detected incidentally in asymptomatic patients. Because there is considerable overlap in the cross-sectional imaging findings of cystic pancreatic lesions, and because many of these lesions being detected are smaller than 3 cm in diameter and lack any specific cross-sectional imaging features, it has become difficult to make informed decisions about patient management when the precise diagnosis remains uncertain. This article presents the limitations of cross-sectional imaging in patients with cystic pancreatic lesions, details advances in knowledge of the genomic and epigenomic changes that lead to progression of carcinogenesis, outlines the current understanding of the natural history of mucinous cystic lesions, and includes the current use and future potential of novel tumor markers and molecular analysis to characterize cystic pancreatic lesions more precisely. The need to move beyond cross-sectional imaging morphology and toward the use of new techniques to diagnose these lesions accurately is emphasized. An algorithm that uses these techniques is proposed and will hopefully lead to improved patient management. REVIEWS AND COMMENTARY n STATE OF THE ART RSNA, 2014 Online supplemental material is available for this article. 1 From the Department of Radiology (P.C.F.) and Department of Medicine, Division of Gastroenterology (M.D.S.), University of Washington School of Medicine, 1959 NE Pacific St, Seattle, WA 98195. Received May 23, 2013; revision requested June 20; revision received October 29; accepted November 18; final version accepted December 9. Address correspondence to P.C.F. (e-mail: freeny@uw.edu). q RSNA, 2014 Radiology: Volume 272: Number 2 August 2014 n radiology.rsna.org 345

Multidetector computed tomography (CT), magnetic resonance (MR) imaging, MR cholan giopancreatography (MRCP), and endo scopic ultrasonography (US) are used to detect incidental cystic pancreatic lesions in asymptomatic patients with increasing frequency. Prevalence of incidental cystic pancreatic lesions at multidetector CT, MR imaging, and MRCP ranges from 0.47% to 44.7% (1 6). In addition, not only is Essentials nn Small cystic pancreatic lesions (,3 cm) often have no diagnostic imaging features, while there can be substantial overlap of findings between larger mucinous and nonmucinous cystic pancreatic lesions. nn Intraductal papillary mucinous neoplasms (IPMNs) are macroscopic precursors of invasive ductal carcinoma; the increasing frequency of detection of IPMNs when they are in the early stage of carcinogenesis presents a unique opportunity to perform resection before they become invasive ductal carcinomas. nn Progression of IPMNs from benign adenomas to invasive ductal carcinomas is associated with sequential genetic mutations (K-ras and GNAS), gene silencing, and chromosomal deletions that can be detected with molecular analysis of cyst fluid obtained with endoscopic US guided fine-needle aspiration (FNA); endoscopic US guided FNA also allows novel cyst fluid tumor markers and cytologic evaluation to be combined with molecular analysis for more precise diagnosis. nn An algorithmic approach that uses these new diagnostic techniques can improve patient management by providing early and more precise diagnosis of cystic pancreatic lesions so patients can be advised to undergo surgery or be enrolled in surveillance programs. the prevalence increasing, but the size of the detectable lesions is decreasing, and with this, the ability to characterize these smaller lesions at the time of initial detection is becoming more difficult (7,8). Many of these incidentally detected cystic pancreatic lesions represent mucinous cystic lesions that have malignant potential, particularly intraductal papillary mucinous neoplasms (IPMNs) and mucinous cystic neoplasms (MCNs). The increasing prevalence and decreasing size, coupled with limitations in characterizing these smaller lesions with cross-sectional imaging modalities, can result in many lesions going undiagnosed and untreated until they become invasive carcinomas. This is discussed in detail in the sections that deal with the natural history of cystic pancreatic neoplasms and in Appendix E1 (online). This article will present the limitations of cross-sectional imaging for evaluation of patients with cystic pancreatic lesions; detail the advances in knowledge of the genetics, progression of carcinogenesis, and natural history of mucinous cystic lesions; and outline the current use and future potential of tumor markers and molecular analysis to characterize these lesions more precisely. It will emphasize that because of the limitations of cross-sectional imaging, it is necessary to move beyond morphology and toward the use of cyst fluid analysis for precise characterization. A new algorithm is proposed that uses these techniques and will hopefully lead to improved patient management. Classification of Cystic Pancreatic Lesions Cystic pancreatic lesions represent a pathologic spectrum that ranges from simple benign cysts (epithelial and nonneoplastic mucinous cysts) and inflammatory cysts (pseudocysts) to cystic neoplasms. The neoplastic category represents a spectrum that ranges from benign serous lesions (serous cystadenoma) to malignant or potentially malignant mucinous neoplasms, including IPMNs and MCNs. Pseudopapillary epithelial neoplasms, cystic neuroendocrine tumors, and cystic pancreatic carcinomas will not be discussed in this article. IPMNs are grossly visible intraductal epithelial neoplasms composed of mucin (MUC)-producing columnar cells that show papillary proliferation and cyst formation (9 11). The intracystic epithelium may demonstrate a spectrum of dysplastic changes, from benign adenoma to invasive carcinoma, and the entire spectrum can be found within a single cystic lesion (9). IPMNs must be intraductal that is, they must connect to and/or arise from within the pancreatic duct. They are subclassified into those arising from the main pancreatic duct (mainduct IPMN), those arising from the side-branch ducts (branch-duct IPMN), and mixed IPMN, involving both side branch ducts and main duct (mixed-duct IPMN) (12). MCNs do not arise within the pancreatic ducts; they have an ovarian-type stroma and do not communicate with the pancreatic duct unless there is a malignant fistula (13). Detection and Diagnosis of Cystic Pancreatic Lesions The frequency of incidentally detected cystic pancreatic lesions in asymptomatic patients is increasing and is currently estimated to represent more than 60% of all detected cystic lesions in the pancreas (6 8,14). Gaujoux et al reported an almost 8% linear increase in prevalence between 1995 and 2010, along with a simultaneous de- Published online 10.1148/radiol.14131126 Radiology 2014; 272:345 363 Content code: Abbreviations: CA = carbohydrate antigen CEA = carcinoembryonic antigen FNA = fine-needle aspiration IPMN = intraductal papillary mucinous neoplasm MCN = mucinous cystic neoplasm MRCP = MR cholangiopancreatography Conflicts of interest are listed at the end of this article. 346 radiology.rsna.org n Radiology: Volume 272: Number 2 August 2014

Figure 1 Figure 1: Axial CT images demonstrate small, uncharacterized cystic pancreatic lesions. (a) Pseudocyst (arrow). (b) MCN (arrow). (c) Branchduct IPMN (arrow). (d) Serous cystadenoma (arrow). (Figure 1d courtesy of Byung I. Choi, Seoul, South Korea.) crease in cyst size from a mean of 2.4 cm between the years 1995 and 2005 to a mean of 1.6 cm between 2005 and 2010 (14). Cystic pancreatic lesions initially detected with multidetector CT, MR imaging, or MRCP and subsequently evaluated with endoscopic US and/or biopsy may have characteristic features that enable a specific diagnosis to be assigned (15 26) (see Appendix E1 [online]). In several series, a correct preoperative diagnosis rate of cystic pancreatic lesions varied from 22% to 86% (27 30). These series are discussed in detail in Appendix E1 (online). In many cases, however particularly incidentally detected cysts that are smaller than 3 cm in diameter specific imaging features may be absent in 13% 88%, or features of nonneoplastic or benign cysts, such as serous cystadenomas, may overlap with those of malignant or potentially malignant lesions, such as IPMNs or MCNs (Fig 1) (24,28,31 34). This is highlighted in Table 1, where 23% of resected cystic pancreatic lesions were benign, and 61% of these were serous cystadenomas. In addition, placing patients who have a cystic lesion without a specific diagnosis in an imaging surveillance program can delay treatment of a malignant or potentially malignant lesion. In the series of Gaujoux et al that included 719 patients in a surveillance program, 47 (6.5%) underwent subsequent resection for increasing cyst size and/or development of a solid tumor nodule within the cyst (14). Five (11%) of the resected lesions were serous cystadenomas, 12 (26%) were malignant, and eight (17%) had invasive tumors. A review of 11 surveillance series showed lesion progression in 16.6% of cases (Table 2). In those patients, the percentage of malignant and/or premalignant cystic pancreatic lesions resected during surveillance increased from 18.7% (Table 1) to 83.2% (282 of 339) (Table 2), and of those, 16.7% (47 of 282) had invasive carcinoma. Thus, the main limitations of crosssectional imaging are the substantial overlap of findings among cystic pancreatic lesions, and, most importantly, the increasing number of incidentally detected cystic pancreatic lesions in asymp- Radiology: Volume 272: Number 2 August 2014 n radiology.rsna.org 347

Table 1 Benign and Malignant Cystic Pancreatic Lesions Resected Study and Year of Publication No. of Lesions Resected* No. of Malignant Lesions Resected after Detection No. of Benign Lesions Resected No. of Benign Lesions Resected after Detection No. of Benign Lesions Resected during Surveillance No. of Serous Cystadenomas Resected # Chung et al (91), 2013 247 (28) 61/204 (29.9) 43/247 (17.4) 33/204 (16.2) 10/43 (23) 31/43 (72) Correa-Gallego 159 (100) 25/136 (18.4) 47/159 (29.6) 39/136 (28.7) 8/23 (35) 26/47 (55) et al (28), 2010 Fernandez-del Castillo 212 (37) 66/212 (31.1) 62/212 (29.2) 62/212 (29.2) NS 23/62 (37) et al (169), 2003 Ferrone et al (7), 2009 256 (71) 20/101 (19.8) 59/256 (23.0) 19/101 (18.8) 12/13 (92) 34/59 (58) Goh et al (92), 2008 176 (40) 55/176 (31.2) 70/176 (39.8) 70/176 (39.8) NS 47/70 (67) Grobmyer et al (178), 2009 78 (48) 13/78 (17) 28/78 (36) 28/78 (36) NS 15/28 (54) Lee et al (41), 2008 166 (27) 31/166 (18.7) 38/166 (22.9) 38/166 (22.9) NS 38/38 (100) Lee et al (179), 2007 92(100) 11/72 (15) 33/92 (36) 24/72 (33) 9/20 (45) 25/33 (76) Morris-Stiff et al (8), 2013 68 (100) 10/68 (15) 20/68 (29) NA NS 15/20 (75) Salvia et al (29), 2012 476 (41) 23/476 (4.8) 43/476 (9) NA NA 13/43 (30) Spinelli et al (5), 2004 49 (NA) 10/49 (20) 15/49 (31) NA NS 14/15 (93) Total 1979 325/1738 (18.7) 458/1979 (23.1) 313/1145 (27.3) 39/99 (39.4) 281/458 (61.4) Note. NA 5 data not available, NS 5 no surveillance. Numbers in parentheses are percentages. * Numbers in parentheses are the percentage of all patients who were asymptomatic with incidentally detected cystic pancreatic lesions. Data are cystic pancreatic lesions with malignancy (carcinoma in situ or invasive carcinoma) resected immediately after initial detection. Data are total number of benign cystic pancreatic lesions resected. Data are number of cystic pancreatic lesions, with benign lesions resected immediately after initial detection. Data are number of benign cystic pancreatic lesions resected during surveillance after initial detection. # Data are number of serous cystadenomas resected as a percentage of total benign (25.6%) to cystic pancreatic lesions. tomatic patients that are smaller than 3 cm and lack any specific morphologic features (Fig 1). In addition, there may be considerable variability in interobserver agreement and radiologists recommendations (35 37). Thus, to precisely classify many incidentally detected pancreatic cysts and to manage the lesions in an optimal manner, additional techniques are needed that do not depend on morphology. Current Guidelines for Evaluation of Patients with Cystic Pancreatic Lesions There are multiple published algorithms for both surgical and nonsurgical management of cystic pancreatic lesions, including the algorithm recommended by the American College of Radiology Incidental Findings Committee (38). Most algorithms generally follow the guidelines defined by the International Association of Pancreatology consensus group, known as the Sendai criteria, to decide between im- mediate surgical resection and imaging follow-up of cystic pancreatic lesions. The Sendai criteria were initially defined in 2006 and were revised in 2012 (12,39). The 2012 criteria for immediate resection of cystic pancreatic lesions were defined as high-risk stigmata, indicating a likely malignant mucinous cystic neoplasm (IPMN and MCN), with a main pancreatic duct diameter of at least 10 mm (a feature of main-duct IPMN), obstructive jaundice associated with a cyst in the pancreatic head, and a solid contrast material enhancing nodular lesion within the cyst. Patients with worrisome features (cyst 3 cm, thick or contrast-enhancing cyst walls, main pancreatic duct diameter of 5 9 mm, nonenhancing mural nodule, and abrupt change in main-duct caliber, with upstream parenchymal atrophy) are advised to undergo endoscopic US guided fineneedle aspiration (FNA) with cytologic evaluation. If cytologic findings are positive or suspicious for malignancy, or if a mural nodule or evidence of main-duct involvement is seen, patients are referred for resection. Finally, in the absence of indications for resection, follow-up imaging is recommended on the basis of cyst size. Virtually all algorithms include a recommendation for follow-up imaging with multidetector CT, MR imaging, MRCP, or endoscopic US at intervals ranging from 3 to 24 months or longer. During surveillance, patients are observed for growth, stability, or regression of cyst size and for development of morphologic criteria that indicate the presence of a malignant neoplasm that is, high-risk stigmata or worrisome features. The Sendai guidelines also include endoscopic US guided FNA and cytologic evaluation for patients with cysts at least 2 cm in diameter (12). Patients are then advised to undergo either surgical resection or additional follow-up imaging. 348 radiology.rsna.org n Radiology: Volume 272: Number 2 August 2014

Table 2 Premalignant and Malignant Cystic Mucinous Neoplasms (IPMNs and MCNs) Resected during Surveillance Study No. of Patients in Surveillance No. of Neoplasms That Progressed during Surveillance* No. of Progressed Neoplasms That Were Resected No. of Malignant and/or Premalignant Neoplasms That Were Resected No. of Neoplasms with Invasive Carcinoma Chung et al (91) 793 43 (5.4) 43 31 (72) 6 (19) Correa-Galleo et al (28)* 179 23 (12.8) 23 4 (17) 2 (50) Ferrone et al (7) 159 13 (8.2) 13 10 (77) 1 (10) Kang et al (161) 201 35 (17.4) 35 35 (100) 5 (14) LaFemina et al (159) 170 97 (57.0) 97 97 (100) 18 (18) Lee et al (179) 182 20 (11.0) 20 11 (55) 2 (18) Maguchi et al (160) 349 62 (17.8) 22 22 (100) 1 (4) Sawai et al (156) 103 29 (28.2) 11 11 (100) 1 (9) Spinelli et al (5) # 79 49 (62) 49 35 (71) 10 (28) Tanno et al (158) 82 13 (15.9) 7 7 (100) 0 Woo et al (180) 124 19 (15.3) 19 19 (100) 1 (5) Total 2421 403/2421 (16.6) 339** 282 47 Note. Numbers in parentheses are percentages. * Progression was defined as increased cyst size, appearance of mural nodules and/or intracystic mass, increasing main pancreatic or common bile duct diameter, and/or development of clinical symptoms. All cysts were smaller than 3 cm, with no specific features (n 5 596). Presumed clinical diagnosis was branch-duct IPMN (127 cases). Clinical diagnosis was IPMN, mostly branch-duct type. All cysts were detected incidentally. # Presumed clinical diagnosis not stated. ** No. of cystic mucinous neoplasms (403 neoplasms) that progressed and were subsequently resected, 339 of 403 (84.1%). No. of cystic mucinous neoplasms (403 neoplasms) in the total patients in surveillance, 339 of 2421 (14.0%). No. of cystic mucinous neoplasms with invasive carcinoma (47 neoplasms) in the total group of patients resected for progression (339 patients), 282 of 339 (83.2%). No. of cystic mucinous neoplasms with invasive carcinoma in patients with premalignant or malignant MCNs (282 neoplasms), 47 of 339 (13.9%); 47 of 282 neoplasms (16.7%). Most algorithms are based on the following assumptions that constitute a reasonable and practical approach: (a) asymptomatic patients with cystic pancreatic lesions that are either uncharacterized (ie, no demonstrable communication with the main pancreatic duct, mural nodules, septa, or thick walls) or are stable in size and smaller than 3 cm in diameter most likely have a benign lesion, such as a branch-duct IPMN, and (b) by following them with imaging until symptoms develop, a more specific diagnosis can be assigned, or new morphologic findings may develop that are indicative of malignancy (16,40 42). The use of a cyst size larger than 3 cm as a significant determinant of malignancy or a cyst size smaller than 3 cm as an indication of low risk of malignancy is found in most all of the algorithms and features prominently in the algorithm recommended by the American College of Radiology Incidental Findings Committee (38). However, even cyst size combined with morphologic abnormalities is often not a reliable indication of cyst type, and future development of symptoms alone may be a more reliable indicator for the need to resect cystic pancreatic lesions (5,28,43,44). The incidence of malignancy in cysts smaller than 3 cm, even in asymptomatic patients, may be as high as 25% (41,43,45). The presence of malignancy is increased in cysts that have positive Sendai criteria, but even in the absence of these criteria, Fritz reported that 17 of 69 patients (25%) with branch-duct IPMNs and no Sendai criteria (ie, no high-risk stigmata or worrisome features) had invasive carcinoma or carcinoma in situ at the time of resection (43). While these algorithms have been useful in the past, new insights into the molecular biology, histopathologic study, and natural history of cystic pancreatic lesions are now calling into question the adequacy of these guidelines and the assumptions on which they are based (5,8,28,43,45,46). New Insights into Cystic Mucinous Neoplasms IPMN: A Marker and Precursor of Pancreatic Ductal Adenocarcinoma Pancreatic ductal adenocarcinoma represents the culmination of the progressive process of carcinogenesis, beginning with a microscopic pancreatic intraepithelial neoplasm, or PanIN, and progressing in a stepwise fashion to pancreatic intraepithelial neoplasm level 3, or PanIN-3, lesions, and finally to invasive carcinoma, where the 5-year survival rate is less than 5%. Because these initial steps are microscopic, early detection has essentially been impossible. It is now accepted that IPMNs of the pancreas are macroscopic precur- Radiology: Volume 272: Number 2 August 2014 n radiology.rsna.org 349

Table 3 Serum and Cyst Fluid MUC Expression in IPMN Subtypes IPMN Cell Type Duct Origin Mucin Expression (serum and cyst fluid) Risk of Malignant Progression Invasive Carcinoma Cell Type Intestinal dark cell Main duct (20) MUC2*, MUC4*, MUC5AC High Colloid carcinoma Pancreatobiliary compact cell Main duct (7) MUC1, MUC4, MUC5AC High Ductal adenocarcinoma Oncocytic Main duct (approximately 3) MUC1, MUC5AC High Oncocytic adenocarcinoma Gastric clear cell Branch ducts (70) MUC5, MUC6, MUC5AC (serum levels are low) Low Ductal adenocarcinoma Note. Numbers in parentheses are percentages. * Cyst fluid levels were typically much higher than those of nonintestinal cysts; mucins MUC2 and MUC4 showed increased expression in IPMNs with high-grade dysplasia or invasive carcinoma (Maker et al [61]). Serum levels were elevated significantly in high-risk IPMNs. Represents 79% of high-risk (high-grade dysplasia and/or carcinoma) IPMNs; levels were significantly higher in patients with high-risk IPMNs (Maker et al [61]). Represents 91% of low-risk (low- to moderate-grade dysplasia) IPMNs (Maker et al [61]). Figure 2 Figure 2: Diagram shows carcinogenetic progression of four subtypes of IPMN. All subtypes progress via the classic adenoma-carcinoma sequence. Percentages are the approximate percentage of each subtype that progresses from a benign adenoma to carcinoma. CIS 5 carcinoma in situ. sors (ie, identifiable with imaging studies) to invasive ductal carcinoma (1). They follow a pathway from benign adenoma to invasive carcinoma in a fashion similar to colon carcinoma the adenoma-carcinoma sequence. This is an important insight, since cystic mucinous neoplasms are macroscopic markers for invasive pancreatic cancer that are now being recognized, not only in symptomatic patients but incidentally in asymptomatic patients and, thus, frequently prior to the invasive stage (47). This presents a unique opportunity to resect these lesions early, when they have only low- or moderategrade dysplasia, and, most importantly, prior to the development of invasive carcinoma. Resection of an IMPN or MCN prior to developing high-grade dysplasia is essentially a resection for cure with a virtually 100% survival rate. There are, however, two caveats. First, in a small number of cases, IPMNs are found to be associated with a separate and independent pancreatic ductal adenocarcinoma (48). Tanno and coworkers found that 5.4% of patients with branch-duct IPMNs had either synchronous (55.5%) or metachronous (44.5%) pancreatic ductal carcinomas (49). In another series of 765 patients with IPMNs, 31 (4%) had concomitant, separate pancreatic carcinomas (50). Second, some IPMNs are multifocal, particularly those arising in the main pancreatic duct. These patients may have both synchronous and metachronous IPMNs and have a poorer survival rate, though generally much better that that of pancreatic carcinoma (49,51,52). Subtypes of IPMN There are four phenotypes of IPMNs that are distinguished by the type of papillary epithelial lining cells: intestinal, pancreatobiliary, and oncocytic types arise from the main pancreatic duct epithelium; and the gastric type arises from the pancreatic branch duct epithelium (9) (Table 3, Fig 2). Yonezawa et al (53) and Fukushima et al (54) have also shown that IPMNs are composed of three additional cell types clear cells, dark cells, and compact cells each of which express different mucins (Table 3). While IPMNs progress to different carcinoma cell types, they each progress through the same adenoma-carcinoma carcinogenetic sequence (Fig 2, Table 3). The intestinal subtype represents about 20% of IPMNs and is the most common type that arises from main pancreatic duct epithelium (main-duct IPMN) (9). It progresses to invasive colloid carcinoma in 30% 50% of cases and has a more favorable prognosis than that of pancreatic ductal carcinoma, with a 5-year survival rate of about 57% (55 59). At the time of resection, the frequency of malignancy (carcinoma in situ, invasive carcinoma) in main-duct IPMNs ranges between 36% and 100% (mean, 62%), 350 radiology.rsna.org n Radiology: Volume 272: Number 2 August 2014

and 70% of those show invasive carcinoma (12). The pancreatobiliary subtype represents about 7% of IPMNs. It progresses to conventional tubular and/or ductal adenocarcinoma in more than 50% of cases (9). Those with high-grade dysplasia or invasive forms have a poor prognosis similar to that of conventional pancreatic ductal carcinoma (55). The oncocytic subtype is the least frequent type (approximately 3% of main-duct IPMNs), and there is minimal information on this neoplasm (9). It may progress to an oncocytic invasive carcinoma. The gastric subtype is the most common, composing about 70% of IPMNs, and arises from pancreatic branch duct epithelium, or PanIN-1 (9). While it is less aggressive than the other subtypes, it is associated with pancreatic intraepithelial neoplasm level 1 lesions and can thus progress to tubular and/or ductal adenocarcinoma (9). At the time of resection of branchduct IPMNs, malignancy ranges from 6% 51%, and 68% of those show invasive carcinoma (12,60). Proteomic profiling of pancreatic cyst fluid and serum is a developing technique that may aid in differentiating mucinous from nonmucinous cysts and benign from malignant mucinous neoplasms (61 64). Allen et al (65) identified differential cyst fluid protein expressions in serous cystadenoma and IPMNs with an accuracy rate of 92%, and Corcos and coworkers (66) have suggested that changes in pancreatic cyst fluid mucus proteomes with characteristic peaks may be potential biomarkers for malignancy. Mucins are glycoproteins that have a role in pancreatic carcinogenesis and can be assessed in pancreatic cyst fluid by using an enzyme-linked immunosorbent assay (61,67 69). Different apomucins (MUC1, 2, 4, and 5AC) are expressed by the four histopathologic subtypes of IPMNs and can be used to differentiate these subtypes and to identify those with high malignant potential that are more likely to develop invasive tubular-type carcinoma, with an increased likelihood of metastatic Table 4 Pancreatic Cyst Fluid Analysis Test Mucinous vs Nonmucinous Benign vs Malignant Amylase Include/exclude pseudocyst Not useful CEA Highly useful Questionable* CA19 9 Accepted use (MCN vs IPMN) Not useful Mucin stain Highly useful Not useful CA19 9 and MUC5AC Accepted use Not useful CA19 9 and CEACAM6 Accepted use Not useful CEA and CA72 4 Investigational Potential CEA and CA19 9 Investigational Potential Mucins (MUC1, 2, 4, 5AC) Accepted use, investigational Used to differentiate subtypes of IPMN, determine low- versus high-risk cysts Molecular analysis (K-ras and LOH) Highly useful Possible, potentially with other genes Inflammatory mediated proteins Investigational Branch-duct IPMN versus main-duct IPMN (potential with TGF-b1, IL1b) Telomerase Investigational Potential MicroRNA Investigational Potential Cytologic evaluation Highly useful Note. For investigational tests, see Appendix E3 (online). * Levels above 2500 ng/ml are highly indicative of a malignant lesion. See Table 3. See Haab et al (67), 2010. See Xu et al (93), 2011. dis ease and thus a poor prognosis (53,59,61,70,71) (Table 3). The intestinal subtype produces exten sive amounts of cystic mucins MUC2 and MUC5AC (Table 3). The pancreatobiliary subtype produces only small amounts of MUC1 and MUC5AC, while those with high-grade dysplasia or invasive carcinoma produce MUC4 and MUC5AC, the latter of which may also be found in high concentrations in the serum of these patients. The gastric subtype expresses MUC5 and MUC6 but does not express MUC1 or MUC2 (Table 3). Hara and coworkers were able to classify IPMNs into one of the four subtypes by using pancreatic juice cytologic findings obtained during en doscopic retrograde cholangiopancre atography and by using mucin stains for MUC1, 2, and 5AC in 89% of cases with a sensitivity, specificity, and overall accuracy for subtyping of 86%, 100%, and 94%, respectively (72). In another report, it was suggested that it may be possible to differentiate MCNs from IPMNs on the basis of identifying distinct mucin, cytokeratin, and CD10 expressions (73). While these reports are encouraging, much additional research must be done to confirm these initial findings. Cyst Fluid Markers for Diagnosis of Cystic Pancreatic Lesions Initial evaluation of patients with cystic pancreatic lesions should focus on differentiating serous from mucinous cystic lesions. There are rapidly expanding series on the evaluation of a variety of markers for differentiating mucinous from nonmucinous cysts and benign from malignant cystic mucinous neoplasms (Table 4). These markers are found in the fluid of cystic pancreatic lesions obtained with endoscopic US guided FNA, in pancreatic juice obtained during endoscopic retrograde cholangiopancreatography and also in serum. Amylase. Cyst fluid amylase concentrations are most helpful for excluding the diagnosis of pseudocyst, where Radiology: Volume 272: Number 2 August 2014 n radiology.rsna.org 351

levels are usually above 1000 ng/ml (16.7 mkat/l) (74). If levels less than 250 ng/ ml (4.175 mkat/l) are found, a pseudocyst is virtually excluded. However, amylase levels are not helpful for differentiating other nonmucinous cystic pancreatic lesions from mucinous lesions or for differentiating MCNs from IPMNs (75,76). Tumor markers (carcinoembryonic antigen [CEA], carbohydrate antigen [CA] 19 9, CA72 4, CEACAM6). Glycoproteins are found in the epithelium of mucinous pancreatic cysts, and concentrations can be used to differentiate mucinous from nonmucinous cysts (Table 4). The cyst fluid CEA has been shown to be a highly accurate test for differentiating nonmucinous from mucinous cystic pancreatic lesions, and the accepted cutoff value for diagnosis of a mucinous cyst is 192 ng/ml (192 mg/l) (76 82). In a meta-analysis of 450 patients with cystic pancreatic lesions uncharacterized at multidetector CT or MR imaging and with nondiagnostic cytologic findings at endoscopic US guided FNA, a CEA of less than 5 ng/ ml (5 mg/l) indicated a benign serous cystadenoma; levels between 2 and 200 ng/ml (2 and 200 mg/l) indicated an inflam matory pseudocyst; levels higher than 200 ng/ml (200 mg/l) indicated a mucinous neoplasm (MCN or IPMN); and levels above 1000 ng/ml (1000 mg/l) indicated a likely malignant mucinous neoplasm (74,83). Buscaglia et al reported that cyst fluid CEA levels of at least 3594 ng/ml (3594 mg/l) are highly indicative of malignancy (84). However, while most investigators believe that CEA is not accurate for the differentiation of benign from malignant cystic mucinous neoplasms, cysts with levels in excess of 2500 ng/ml (2500 mg/l) are thought to be likely malignant (80,83,85). Identifying the presence of mucin may add to the accuracy of cyst fluid tumor markers (Table 4). By combining CEA and the presence of mucin (cyst fluid periodic acid Schiff positive stain), the sensitivity, specificity, and positive and negative predictive values for differentiating mucinous and nonmucinous lesions were 96%, 34%, 25%, and 97%, respectively, while for CEA alone, they were 93%, 43%, 51%, and 91%, and for mucin alone, they were 80%, 40%, 61%, and 63% (81). Thus, a cystic pancreatic lesion with low CEA and without mucin was highly unlikely to be a mucinous lesion. While cyst fluid CEA levels have not been shown be of value for differentiating benign from malignant mucinous lesions, new work suggests that CEA combined with other markers may produce better results (77,78,85,86). By using a CEA level of more than 200 ng/ ml and a CA72 4 level of more than 40 U/mL, Maire and coworkers found that the sensitivity, specificity, and positive and negative predictive values for diagnosis of a malignant IPMN were 100%, 53%, 36%, and 100%, respectively (87). They believe that the 100% negative predictive value of combined cyst fluid levels of CEA and CA72 4 may be of value in deciding when not to resect some mucinous cysts (87). None of their patients with CEA levels up to 200 ng/ml (200 mg/l) and CA72 4 levels up to 40 U/mL had a malignant IPMN. Talar-Wojnarowska et al analyzed CEA and CA19 9 in cystic pancreatic lesions and also found high negative predictive values of 95.8% and 89.3%, respectively (88). Shin et al showed that combining CEA and CA19 9 yielded a sensitivity and specificity of 88.9% and 87.8%, respectively (89). Thus, in patients with indeterminant CEA levels, adding CA19 9 and CA72 4 values may be helpful for deciding to defer resection, particularly in high-risk patients. Serum Tumor Markers for Diagnosis of Cystic Pancreatic Lesions Recent series have shown that elevated serum levels of CEA and CA-19 9 are likely independent predictors of a malignant IPMN (89 94). In the study of Fritz et al, preoperative elevation of serum CEA and/or CA-19 9 occurred in 80% of patients with invasive IPMNs and in only 18% of patients with noninvasive neoplasm (90). In a subsequent series from the same institution, of 11 patients with invasive carcinomas found in resected branch-duct IPMNs smaller than 3 cm with no Sendai high-risk stigmata, eight (73%) had elevated serum CEA and/or CA19 9 levels (43). The authors concluded that elevated levels of serum tumor markers could be helpful in deciding between surgical and nonsurgical management and recommended strong consideration of resection in all patients with Sendai-negative branch-duct IPMNs if serum tumor markers are elevated. Molecular Analysis of Pancreatic Cyst Fluid The molecular genomic and epigenomic basis of the progression of IPMNs from benign to malignant and invasive carcinomas has been the subject of intensive research over the past 2 decades (95) (Tables 5, 6). Progression to invasive carcinoma has been associated with sequential development of genetic mutations, gene silencing, and chromosomal deletions. The genetic changes and expressions of a variety of markers include K-ras (codon 12) and GNAS mutations; MUC1, 2, and 5AC; human telomerase reverse transcriptase; Cox2; Shh; and loss of heterozygosity in 9p21 (p16) and 17p13 (p53) (96 99). Loss of heterozygosity (allelic imbalance) is associated with inactivation of tumor suppression genes and K-ras mutations and has been identified as an early event in the progression of pancreatic epithelial changes from lowgrade dysplasia to malignant carcinoma (100,101). While the number of K-ras mutations in IPMNs has been shown to increase during carcinogenesis, the frequency of mutations varies between 31% and 86% (9,102). However, Wada noted that K-ras mutations were frequently seen without a relationship to histologic grade, while loss of heterozygosity of the p16 gene was observed more frequently as the degree of histologic atypia increased (103). In addition, loss of heterozygosity of the p53 gene was seen only in invasive carcinoma, and the frequency of each genetic alteration in invasive carcinoma was the same as that in invasive pancreatic ductal adenocarcinoma. Wada concluded that alterations and accumulations of the p16 and p53 oncosuppressor genes are crucial events during the car- 352 radiology.rsna.org n Radiology: Volume 272: Number 2 August 2014

Table 5 Differentiation of Mucinous from Nonmucinous Cysts by Using Molecular Analysis and CEA Test Sensitivity (%) Specificity (%) Accuracy (%) CEA. 192 ng/ml Sawhney et al (111), 2009 82 100 84 Khalid et al (106), 2009 64 83 Molecular analysis Sawhney et al (111), 2009 76.5 100 79 Khalid et al (106), 2009 19 100 K-ras mutation Sawhney et al (111), 2009 11 100 21 Khalid et al (106), 2009 45 96* Allelic imbalance (loss of heterozygosity) Sawhney et al (111), 2009 70 100 73 Khalid et al (106), 2009 67 66 DNA concentration 40 ng/ml Sawhney et al (111), 2009 29 100 37 Khalid et al (106), 2009 46 68 CEA with molecular analysis Sawhney et al (111), 2009 100 100 100 Khalid et al (106), 2009 84 67 * Absence of K-ras mutation indicates a nonmucinous cyst, with a specificity of 96%. K-ras followed by allelic loss indicates a mucinous cyst, with a specificity of 100%. Table 6 Differentiation of Premalignant from Malignant Mucinous Cysts by Using Molecular Analysis, CEA, and Cytologic Evaluation Test Sensitivity (%) Specificity (%) Accuracy (%) CEA. 192 ng/ml 78 53 62 K-ras mutation 53 71 K-ras mutation followed by allelic loss Khalid (104), 2005 91 93 Khalid (106), 2009 37 96 Allelic imbalance (loss of heterozygosity) 92 36 Allelic loss amplitude Amplitude. 80% 70 85 Amplitude. 82% 90 67 DNA concentration 40 ng/ml 75 75 Cytologic evaluation 37 96 75 CEA and K-ras 94.1* Cytologic evaluation and molecular analysis 100 Note. Premalignant is defined as adenoma to moderate-grade dysplasia (or borderline). * Talar-Wojnarowska et al (88), 2012. All malignant cysts with false-negative cytologic findings had at least one positive DNA analysis variable associated with malignancy. (PathFinderTG; RedPath Integrated Pathology, Pittsburgh, Pa). Khalid et al were the first to report the use of molecular analysis of pancreatic cyst fluid to show that malignant IPMNs acquired a K-ras mutation first, followed by allelic loss, while benign or premalignant IPMNs acquired the K-ras mutation as the first step in DNA damage and without subsequent allelic loss (104,105). In IPMN cyst fluid obtained preoperatively with endoscopic US guided FNA, the presence of allelic loss after the K-ras mutation had a sensitivity and specificity for diagnosis of malignant IPMNs of 91% and 93%, respectively (Table 5). On the basis of the assumption that malignant IPMNs have uncontrolled cell growth and a constant release of high-quality DNA, it was determined that an elevated concentration of DNA, determined by measuring optical density and the cyclic threshold value, also correlated with malignant IPMNs (104,105). The Pancreatic Cyst Fluid DNA Ana lysis, or PANDA, study of 113 patients confirmed the usefulness of these molecular changes to indicate malignancy (106) (Table 6). However, while the sequence of K-ras mutation followed by allelic loss amplitude showed a decrease in sensitivity to 37%, it had a 96% specificity for the differentiation of benign from malignant cystic lesions. The presence of high-amplitude mutations with a cutoff value of 82% increased the sensitivity to 90%, but the specificity decreased to 67%, with a cutoff value of more than 80%; sensitivity and specificity were 70% and 85%, respectively (106). Mizuno et al indicated that a single-clonal convergence of K-ras mutation is associated with IPMN malignant progression and is a useful parameter for deciding which tumors to resect (107). In the future, a combination of markers, such as TP53, CDKN2A, and SMAD4, or identification of DNA copy number alterations may increase the specificity (108 110). The usefulness of molecular analysis to separate benign mucinous cysts from malignant cysts is still open to question, and large, prospective series are needed (98). However, multiple reports have confirmed the ability of molecucinogenesis and malignant progression of IPMNs. Molecular analysis is currently being used in some centers primarily to differentiate mucinous from nonmucinous cystic pancreatic lesions and secondarily to help differentiate benign from malignant mucinous neoplasms (Tables 5, 6). There is one commercially available laboratory that performs molecular analysis Radiology: Volume 272: Number 2 August 2014 n radiology.rsna.org 353

Table 7 Novel Biomarkers for Specific Evaluation of Cystic Pancreatic Lesions Biomarker Potential Use Studies Inflammatory mediator proteins (cytokines-interleukins, chemokines) Telomerase (reverse transcriptase directing synthesis of telomeric repeats [GGT-TAG] onto chromosomal ends) MicroRNA (class of 18 24 nucleotides RNA molecules frequently misexpressed in pancreatic carcinoma) Differentiate inflammatory pancreatic cyst (pseudocyst) from branch-duct IPMN Separate benign or borderline lesion from malignant IPMN Differentiate invasive from noninvasive or low- from high-grade dysplasia IPMNs; differentiate mucinous from nonmucinous cysts Lee et al (122), 2012; Lee et al (123), 2012; Maker et al (124), 2011 Yeh et al (125), 1999; Hiyama et al (126), 1997; Hashimoto et al (127), 2008; Inoue et al (128), 2001; Pearson et al (129), 2000; Nakashima et al (130), 2009 Ryu et al (131), 2010; Ryu et al (133), 2011; Habbe et al (132), 2011; Matthaei et al (134), 2012 lar analysis to help distinguish between mucinous and nonmucinous pancreatic cysts (75,99,111 115) (Table 4). Shen et al (113) found that molecular anal ysis allowed differentiation of mucinous from nonmucinous cysts with a sensitivity, specificity, and positive predictive value of 86%, 93%, and 95%, respectively, and Talar-Wojnarowska et al reported a sensitivity and specificity of 64.7% and 97.4%, respectively (116). Toll and colleagues used combined cyst fluid analysis of CEA, cytologic evaluation, and molecular analysis and found that molecular analysis added significant diagnostic value in the evaluation of patients with small (,3 cm) cystic pancreatic lesions, especially when cytologic findings were inadequate or CEA levels were not elevated (115). New Investigations in Molecular Analysis of Neoplastic Pancreatic Cysts Additional genetic mutations, other than K-ras, have been identified in neoplastic pancreatic cysts by using whole-exome gene sequencing (115 121). Detailed information is available in the online module (Appendix E2 [online]). Biomarkers Novel biomarkers that may be the future keys to finally enabling accurate discrimination of benign from malignant IPMNs and predicting which IPMNs are most likely to progress to invasive carcinoma are being investigated at several centers (122 134) (Table 7). Additional details are available in the online module (Appendix E3 [online]). Cytologic Evaluation Cytologic evaluation plays a prominent role in identifying malignancy in cystic pancreatic lesions and is used in the Sendai guidelines (12,61,135 147). Most investigators believe that endoscopic US guided FNA with cytologic evaluation and cyst fluid analysis should be performed in all cystic pancreatic lesions that are not obviously characterized with cross-sectional imaging (30,140 144). In particular, small (,3 cm) branch-duct IPMNs, which are often believed to be benign and nonaggressive, have been shown to be malignant in up to 25% of resected cases (43,148). In the series of Ono et al, in 31 patients with branch-duct IPMNs smaller than 3 cm and no specific morphologic imaging criteria, endoscopic US guided FNA cytologic findings with either high-grade atypia or positive results for diagnosis of a malignant IPMN had a sensitivity of 80%, specificity of 85%, and accuracy of 84% (146). In the absence of both highgrade atypia and positive cytologic findings, the negative predictive value was higher than 95%. The problem with aspirating cystic lesions smaller than 3 cm is the small volume of fluid and thus the lack of an adequate quantity of cells. This limitation has been addressed with two endoscopic US guided biopsy techniques: targeted cyst wall puncture and Trucut biopsy (149 151). In the series of Hong et al, in 30% of cysts with nondiagnostic CEA levels (,192 ng/ml) and in 67% of cysts where fluid volume was insufficient to determine CEA, diagnosis of a mucinous neoplasm could be made by means of cytologic findings derived from targeted cyst wall puncture (149). In the initial series of Levy et al, Trucut biopsy of the cyst wall resulted in a correct diagnosis in 70% of patients with previously nondiagnostic endoscopic US guided FNA cytologic findings (150). Natural History of Pancreatic Mucinous Cystic Neoplasms An increasing body of literature has shed new insight into the natural history of mucinous pancreatic cystic neoplasms and has led to a variety of different and controversial approaches to subsequent management. It is generally accepted that MCNs, main-duct IPMNs, and mixed-duct IPMNs should be resected because of their high malignant potential and because many already harbor a malignant or invasive component at the time of initial detection (12). However, management of branch-duct IPMNs, which represent an increasingly large pro portion of cystic pancreatic lesions, remains controversial because of the varied natural histories. Ferrone et al noted that while the prevalence of incidentally detected cysts in asymptomatic patients increased from 1997 to 2002, the prevalence of in situ or invasive carcinoma decreased from 40% to 26% in symptomatic patients and from 17% to 7% in asymptomatic patients (7). Branch-duct IPMNs slowly enlarge in 10% 15% of cases if followed up for more than 3 years (44). It has been estimated that it takes about 5 years for an IPMN to progress from benign ade- 354 radiology.rsna.org n Radiology: Volume 272: Number 2 August 2014

noma or low-grade dysplasia to highgrade dysplasia and/or invasive carcinoma (152 154). The risk of progression to malignancy for branch-duct IPMNs has been estimated to be 9% at 2 years and 15% at 5 years (155). The risk is accelerated for main-duct IPMNs, with 2- and 5-year risks of 58% and 63%, respectively. By using Kaplan-Meier curves, Traverso et al indicated that the natural history observed in 210 patients shows the progression to be as short as 2 years in 50% of cases, particularly with main-duct and/or mixed-duct IPMN, and was even shorter if the patient was symptomatic (44). At both 2 years and 5 years, 54% and 66% of symptomatic patients, respectively, with main-duct IPMNs had malignant neoplasms, while asymptomatic patients had both 2- and 5-year risks of malignancy of only 40%. For symptomatic patients with branchduct IPMNs, 2- and 5-year risks of malignancy were 30% and 37%, respectively, while none of the asymptomatic patients had malignant branch-duct IPMNs. While the presence of symptoms is a good predictor of malignancy, lack of symptoms does not exclude malignancy. In other series, the 5- and 10-year actuarial risk of development of invasive carcinoma in branch-duct IPMNs is estimated at 2.4% and 20%, respectively, while the 5-year risk for progression to high-grade dysplasia in main-duct and/ or mixed-duct IPMN is about 63% (155 158). Despite this estimated high risk of carcinogenic progression, there is wide variability in which IPMNs progress to invasive carcinoma and how long it actually takes. This becomes important when making a decision to resect or to clinically follow up uncharacterized cystic pancreatic lesions or presumed benign branch-duct IPMNs. The overall incidence of malignancy and invasive carcinoma in IPMNs at the time of initial resection varies between those limited to branch-duct IPMNs and those with main-duct involvement. Main-duct IPMNs are malignant 60% 92% of the time (mean, 70%) and are invasive in two-thirds of these cases, while mixed-duct IPMNs are malignant in 58% and invasive in 45%. Branch-duct IPMNs are malignant in 6% 46% and have invasive carcinoma in up to 42% (12,52,60). However, these figures would vary if the subtypes of IPMNs were analyzed separately (Fig 2). A number of surveillance series of patients with mostly branch-duct IPMNs have been reported, detailing the development of clinical and morphologic findings that suggest possible malignant progression and thus indicate the need for resection (8,28,156,158 160) (Table 2). Progression is usually defined as increasing cyst size ( 30 mm or a rate higher than 2 mm per year), development of a mural nodule or solid enhancing mass component within the cyst, increasing main pancreatic duct diameter ( 10 mm), and/or the appearance of symptoms, usually described as pancreatitis-like, likely caused by the cyst (12,44,161). Despite the use of these criteria, 27.3% of cystic pancreatic lesions resected immediately after detection were found to be benign at histopathologic examination, and 13.9% of cystic pancreatic lesions were resected because of progression during a surveillance program and were invasive carcinomas (Tables 1, 2). Interestingly, and likely reflecting in part the overlap of cross-sectional imaging findings between serous cystadenomas and mucinous neoplasms, 61.4% of resected benign cystic pancreatic lesions were serous cystadenomas (Fig 1, Table 1). Multiple investigators report varied results in patients known or suspected to have branch-duct IPMNs in surveillance programs. In the collected series, progression varied from 5.4% to 62% (Table 2). There are several possible explanations for the varied reported rates of neoplastic progression of branch-duct IPMNs. While it is known that the gastric type of branch-duct IPMN is less aggressive and progresses to invasive carcinoma less frequently than the intestinal and pancreatobiliary types, none of the investigators detailing progression to malignant neoplasms attempted to separate these subtypes (Table 2). Second, since the patients without progression did not undergo surgery, no final histopathologic diagnosis was obtained to know exactly what types of cystic pancreatic lesions were actually being followed up. Finally, there is some evidence to suggest that oncogene-induced senescence may occur early in carcinogenesis of IPMNs and thus prevent progression to malignancy (162). Miyasaka defined this as a permanent cell cycle arrest induced by telomere shortening that can be caused by an accumulation of cell doublings, DNA damage due to irradiation, oxidative stress, or activation of oncogenes (162). The mean number of patients with cystic pancreatic lesions that progressed during surveillance (range, 3 130 months) was 16.6%, and of those undergoing subsequent resection, 83.2% had either malignant or potentially malignant neoplasms (ie, mucinous neoplasms), and 16.7% of these were invasive carcinomas (Table 2). While there is no good correlation between the presence or development of malignant cysts and initial cyst size, there is a correlation of cyst histopathologic findings with the degree and rate of cyst growth. In the series of Kang et al, 201 patients with cysts smaller than 30 mm and no high-risk Sendai stigmata were followed up with cross-sectional imaging for 3.2 109.5 months (161). Thirty-five (17.4%) showed progression and underwent resection. All patients had branch-duct IPMNs, and eight were malignant (five invasive carcinoma, three noninvasive). The malignant cysts grew by a higher percentage than the nonmalignant cysts (69.8% vs 19.4%) and at a higher rate (4.1 mm per year vs 1.0 mm per year). There may also be a correlation between initial cyst size and subsequent growth potential, but conflicting results were reported in two series. In the patients of Kang et al with branch-duct IPMNs up to 3 cm that were followed up for a median of 27.9 months, smaller cysts (,20 mm) showed a higher mean increase in cyst diameter compared with larger cysts ( 2 cm and 3 cm) and also grew at a faster rate (161). However, Das et al followed up 166 patients that had incidentally detected cysts (117 MCNs and 49 IPMNs; median duration of 32 months) with an initial mean cyst diameter of 2 cm (86% were up to 3 cm, and 14% were larger than 3 cm) (40). The diagnosis of Radiology: Volume 272: Number 2 August 2014 n radiology.rsna.org 355