`15 (2005) 117 – 142
`
`Pancreatic neoplasms
`
`Michael J. Levy, MDa,*, Maurits J. Wiersema, MDb
`
`aDevelopmental Endoscopy Unit, Division of Gastroenterology and Hepatology, Mayo Clinic,
`200 First Street Southwest, Rochester, MN 55905, USA
`bGI Consultants, 1900 Carew Street, Suite 1, Fort Wayne, IN 46805, USA
`
`Endoscopic ultrasound (EUS) is a highly sensitive imaging modality for
`identifying pancreatic neoplasms, with detection rates over 90% (Table 1) [1–8].
`In most studies, EUS has been superior to transabdominal ultrasound (TUS), CT,
`endoscopic retrograde cholangiopancreatography (ERCP), and angiography in
`the detection of pancreatic tumors [1,3,9,10]. Rosch et al demonstrated greater
`sensitivity (99%) and specificity (100%) for detecting pancreatic tumors than
`TUS (sensitivity 67%, specificity 40%) and CT scan (sensitivity 77%, specificity
`53%) [1]. Recent studies comparing EUS to dual-phase helical CT, MRI, and
`positron emission tomography (PET) have found EUS to have a greater
`sensitivity for identifying pancreatic neoplasms [2,4,8,11]. In another study of
`34 patients with an elevated contrast angiography (CA) 19-9 and normal pancreas
`according to TUS and CT scan, EUS was 94% accurate in detecting a pancreatic
`or biliary neoplasm, with a positive and negative predictive value of 92% and
`100% respectively [12]. The advantage of EUS is even greater for recognizing
`tumors less than 2 to 3 cm in diameter [1,4,9,13,14]. Yasuda et al found that EUS
`had a detection rate of 100%, ERCP 57%, TUS 29%, CT 29%, and angiography
`14% for pancreatic tumors less than 2 cm [9]. Similarly, in a study by Rosch et al,
`the diagnostic sensitivity of EUS for detecting tumors smaller than 3 cm was
`100%, compared with 57% for TUS and 68% for CT [15].
`
`Neuroendocrine pancreatic tumors
`
`Neuroendocrine pancreatic tumors (NPTs) are rare, with an incidence of less
`than 1 tumor per 100,000 people [16]. Gastrinoma,
`insulinoma, and non-
`
`* Corresponding author.
`E-mail address: levy.michael@mayo.edu (M.J. Levy).
`
`1052-5157/05/$ – see front matter D 2004 Elsevier Inc. All rights reserved.
`doi:10.1016/j.giec.2004.07.014
`
`giendo.theclinics.com
`
`ARGENTUM
`Exhibit 1019
`
`000001
`
`
`
`118
`
`levy & wiersema
`
`Table 1
`Endoscopic ultrasound detection rates of pancreatic tumors
`
`Author/Year/Reference
`
`Sensitivity
`
`Specificity
`
`Positive
`predictive value
`
`Negative
`predictive value Accuracy
`
`Rosch, 1991 [1]
`Snady, 1992 [3]
`Yasuda, 1993 [7]
`Muller, 1994 [4]
`Baron, 1997 [5]
`Legmann, 1998 [2]
`Akahoshi, 1998 [6]
`
`99%
`85%
`–
`94%
`95%
`100%
`89%
`
`100%
`80%
`
`–
`100%
`88%
`93%
`97%
`
`100%
`89%
`–
`–
`95%
`–
`94%
`
`97%
`73%
`–
`–
`88%
`–
`93%
`
`76%
`83%
`100%
`96%
`
`–
`–
`
`94%
`
`functioning tumor are most common, with glucagonoma, somatostatinoma, and
`VIPoma less often reported. Preoperative determination of the location and extent
`is necessary to enable surgeons to plan the optimal surgical approach. Resection
`offers the only chance for cure and should be undertaken whenever possible
`because of the malignant potential of these tumors. Preoperative localization is
`also important because of the difficulty in identifying these tumors during
`surgery, which is the case in up to 20% of insulinomas, and as many as 50% of
`gastrinomas [16]. The approach to tumor localization is similar for all tumor
`types. Various imaging modalities are available for preoperative identification of
`NPTs. They include TUS, CT, selective abdominal angiography, selective venous
`sampling, radiolabeled octreotide (somatostatin–analog) receptor scintigraphy
`(SRS), intraoperative ultrasound, and most recently EUS.
`Endoscopic ultrasound studies report a localization rate of approximately 77%
`to 93% for insulinomas [16–23]. In these same studies CT, was able to locate the
`tumor in only 0% to 20% of patients, and somatostatin receptor scintigraphy
`(SRS) was able to locate the tumor in only 12% to 14% of patients. Insulinomas
`have a low density of somatostatin receptors, and as a result they often go
`undetected by SRS. The high detection rate of EUS for insulinomas likely is
`explained by the fact that 99% of insulinomas are confined to the pancreas
`[21,24,25]. Approximately 75% to 100% of pancreatic gastrinomas are identified
`by EUS [16–18,21,22,26] versus 0% to 67% of duodenal gastrinomas [16,18,22].
`EUS is comparable to SRS for detecting pancreatic gastrinomas, and both tests
`are clearly superior to CT. Even so, both techniques may miss a significant
`proportion of duodenal gastrinomas [16,18,22,27], which is important, given that
`30% to 45% of gastrinomas are located in parapancreatic locations, most
`commonly the duodenal wall or lymph nodes [24]. Despite focused examination
`of the duodenal wall by EUS, gastrinomas in this location commonly are missed
`by EUS unless previously identified endoscopically [21]. Therefore, at the time
`of EUS, the authors initially perform a careful forward- and side-viewing exam of
`the duodenal wall.
`The addition of fine needle aspiration (FNA) further increases the diagnostic
`accuracy for NPTs, with overall accuracy of EUS–FNA reported to be 75% to
`80% [26,28], which is superior to TUS, CT, or surgical biopsies [29–31]. In
`addition, EUS also may identify multi-focal tumors not seen by other imaging
`
`000002
`
`
`
`pancreatic neoplasms
`
`119
`
`modalities [28,31]. In a multi-center trial involving 37 patients with a suspected
`NPT undetected by TUS and CT, the sensitivity and specificity of EUS for tumor
`localization were 82% and 95%, respectively [1]. These tumors had a mean
`diameter of 1.4 cm (range 0.5 to 2.5 cm) and consisted of 31 insulinomas,
`7 gastrinomas, and 1 glucagonoma. In this same study, only 27% of tumors were
`identified by angiography. All patients underwent surgical resection, with 36 of
`37 considered cured based on clinical and laboratory parameters.
`The EUS appearance of NPTs is similar regardless of the type of tumor. They
`typically appear as round, well-delineated, homogenous, echo-poor lesions, with
`a surrounding hyper-echoic rim (Fig. 1). Cystic or calcified tumors, echo-rich
`lesions, an echo-poor border, or echo-texture, however, are similar to surrounding
`pancreatic parenchyma [1,32,33]. The EUS technique for localizing these tumors
`is identical to that for ductal adenocarcinoma, except that a more deliberate exam
`may be needed to find these small lesions. The parapancreatic region also should
`be examined carefully, not only to search for malignant lymph nodes but also to
`look for primary tumors [13,34]. Parapancreatic tumors may be attached by a
`pedicle or completely separate from the pancreas, and they are more difficult to
`locate than intrapancreatic tumors [21]. As with other tumors, infiltration into
`adjacent organs and vessels should be evaluated. EUS–FNA helps differentiate
`benign parapancreatic lymph nodes from a primary NPT, a distinction that can be
`difficult, especially for insulinomas [17,19–21,28,35–37]. EUS appearance also
`may predict the malignant potential of NPTs, which can be otherwise difficult to
`discern in the absence of extensive local invasion or distant metastasis [37,38].
`The presence of a hypo-echoic lesion with anechoic regions, an irregular central
`echogenic area, or pancreatic duct obstruction is indicative of malignant
`transformation [38]. The echogenic areas correspond with hemorrhage, necrosis,
`or hyaline degeneration, each of which suggests a malignant tumor [38].
`Once identified, it is important to accurately describe the location of tumor(s)
`to facilitate surgical resection. The authors recommend describing the location
`relative to pancreatic and peripancreatic structures. In a step further, Gress et al
`
`Fig. 1. Neuroendocrine pancreatic tumor. Gastrinoma identified in a patient with long-standing
`uncontrolled acid reflux symptoms and diarrhea.
`
`000003
`
`
`
`120
`
`levy & wiersema
`
`reported their experience in one patient employing EUS-guided fine needle
`tattooing [39]. After identifying a 1.9 0.5 cm insulinoma, they injected 4 mL of
`sterile India ink into the lesion and continued to inject as the needle was with-
`drawn from the pancreas. The ink and tumor were identified readily at surgery
`performed the same day. This is a method that may facilitate operative
`localization of NPTs and in particular assist when enucleation or laparoscopic
`resection is planned for small tumors. For most patients, however, NPT marking
`is likely to be of no benefit, and the authors discourage doing so outside of a
`research protocol. In addition to the risks inherent to pancreatic EUS–FNA,
`injection of India ink may induce peritonitis, phlegmonous gastritis, and luminal
`and periluminal abscess formation, ulceration, and necrosis [40–43].
`The cost-effectiveness of EUS for the preoperative localization of pancreatic
`endocrine tumors was demonstrated recently. Bansal et al compared the cost of
`performing tumor localization with and without EUS as part of the protocol, and
`found that the use of EUS significantly reduced the cost of preoperative staging
`($2620 versus $4846) [32]. Savings resulted from the reduced need for angio-
`graphy and venous sampling procedures and because of the reduction in surgical
`and anesthesia times. The cost per tumor located was $3144 when EUS was used
`versus $5628 when EUS was not employed.
`Endoscopic ultrasound is an accurate technique for detecting NPTs. EUS is
`being used increasingly to search for sporadic NPTs and in patients with multiple
`endocrine neoplasia (type 1) because of its the ability to identify small, previously
`undetected tumors [44] (Fig. 2). Although some favor its use only when non-
`invasive studies detect no metastases and no primary tumor is seen, the authors
`suggest performing EUS in all patients in whom surgery is planned. They favor
`this approach even when a lesion already has been identified to allow detection of
`unsuspected multi-focal or metastatic disease and clarify the relationship of the
`tumor to the main pancreatic duct. The added information obtained by EUS–FNA
`
`Fig. 2. Neuroendocrine pancreatic tumor. One of many small nonfunctioning neuroendocrine tumors
`identified by endoscopic ultrasound in a patient with multiple endocrine neoplasia (MEN) I syndrome.
`CT and somatostatin receptor scintigraphy failed to identify any of the lesions.
`
`000004
`
`
`
`pancreatic neoplasms
`
`121
`
`allows cytologic confirmation of the diagnosis with reduced false-positive
`imaging results and also allows the surgeon to plan the optimal strategy (eg,
`tumor enucleation versus pancreatic resection). Further study is needed, however,
`to determine the role, utility, and safety of EUS–FNA when noninvasive studies
`already have localized a tumor.
`
`Cystic pancreatic tumors
`
`Widespread use of high-resolution cross-sectional imaging has led to increased
`detection of cystic lesions of the pancreas, which may be benign, malignant, or
`result from an inflammatory process. Pseudocysts are most common (80% to
`90%), while cystic pancreatic tumors (CPTs) account for 10% to 20% of cystic
`pancreatic lesions and 1% of primary pancreatic neoplasms [45,46]. The
`differential also includes congenital cysts, acquired cysts, and extrapancreatic
`cysts. In addition, solid pancreatic tumors may undergo necrosis and cystic
`degeneration and be mistaken for a CPT [45]. Although accounting for a minority
`of lesions, CPTs are an important subgroup to identify given their often
`distinctive presentation, diverse pathological features, and usually indolent
`biological behavior. CPTs are classified broadly according to their malignant
`potential, which impacts prognosis and therapy. Mucinous lesions (mucinous
`cystic neoplasms and intraductal papillary mucinous tumors) are premalignant or
`malignant tumors, and surgical resection generally is recommended in operative
`candidates [47–49]. Nonmucinous lesions include serous cystadenomas that have
`a very low malignant potential, and pseudocysts, which are always benign, and
`generally only resected when causing symptoms or complications [47–49].
`Management and outcome of patients with CPTs critically depends on early
`tumor detection, distinction from pseudocysts, and accurate determination of
`tumor type. The appropriate use of clinical, imaging, laboratory, and pathology
`information is essential in this regard (Table 2). Detection is important even
`after malignancy has developed, because certain malignant CPTs have a bet-
`ter prognosis than ductal adenocarcinoma and a relatively high cure rate fol-
`lowing resection.
`
`Table 2
`Analysis of aspirated cystic pancreatic tumor fluid — general characteristics
`
`Viscosity
`
`Amylase
`
`CA 19-9
`
`CA 15-3
`
`CA 72-4
`
`CEA
`
`Cytology
`
`SCA
`MCA
`MCAC
`IPMN
`Pseudocyst
`
`Low
`High
`High
`High
`Low
`
`Variable
`Variable
`Variable
`High
`High
`
`Variable
`Variable
`Variable
`Variable
`Variable
`
`Low
`High
`High
`Variable
`Low
`
`Low
`High
`High
`Variable
`Low
`
`Glycogen
`Low
`Mucinous
`High
`Mucinous
`High
`Variable Mucinous
`Low
`Histiocytes
`
`intraductal
`Abbreviations: CA, carbohydrate antigen; CEA, carcinoembryonic antigen; IPMN,
`papillary mucinous neoplasia; MCA, mucinous cystadenoma; MCAC, mucinous cystadenocarcinoma;
`SCA, serous cystadenoma.
`
`000005
`
`
`
`122
`
`levy & wiersema
`
`Serous cystadenoma
`
`Serous cystadenomas (SCAs) usually appear as focal, well-demarcated
`lesions, containing multiple (at least six), small (less than 1 to 2 cm) fluid-
`filled microcysts (Fig. 3) [50–52]. Although some report that most (50% to 70%)
`are located in the pancreatic body or tail [53], others have found them more
`commonly in the head or neck region (63%) [47,54]. The individual cysts are
`interspersed within dense fibrous septations, producing a honeycomb appearance
`[55,56]. Central fibrosis or calcification may be seen, particularly in large lesions
`[34,57]. The resulting sunburst calcification, although pathognomonic, is present
`in only about 10% of patients [54,58–60]. A less common macrocystic variant
`contains larger (greater than 2 cm) cysts [8,51,61]. A solid variant contains
`numerous tiny cysts, each 1 to 2 mm, and appears as a homogenous hypoechoic
`mass that can be mistaken for a ductal carcinoma. Endoscopic retrograde
`pancreatography (ERP) infrequently demonstrates ductal distortion because of a
`mass effect [34], and rarely communication with the pancreatic duct [62,63].
`Angiography, although seldom performed, reveals the hyper-vascular nature of
`most SCAs. The presence of intracystic mucin or floating debris, pancreatic duct
`dilatation, echogenic ductal wall thickening, and focal cyst wall nodularity or
`thickening are distinctly unusual and raise the possibility of a mucinous tumor
`[51,57,59,64–66]. Cyst fluid usually has low viscosity and tumor marker levels.
`Cytologic analysis is diagnostic in only 50% of aspirates [67], with the presence
`of bland cuboidal glycogen staining cells establishing the diagnosis [67–69].
`Aspiration of SCAs may be technically challenging because of the small size of
`individual microcysts that limits the volume of fluid aspirated, thereby dimin-
`ishing the diagnostic accuracy. The vascularity of SCAs may cause bleeding
`during FNA and impair cyst fluid analysis.
`
`Fig. 3. Serous cystadenoma.
`
`000006
`
`
`
`pancreatic neoplasms
`
`123
`
`Mucinous cystadenoma/mucinous cystadenocarcinoma
`
`Mucinous cystic neoplasia (MCNs) are usually (66% to 75% of the time)
`located in the pancreatic body or tail and contain a smooth, glistening outer
`surface [34,49,70–73]. They typically are comprised of several fluid-filled
`cavities (each greater than 1 to 2 cm) separated by thin septations (Fig. 4) [55,64].
`The wall lining is thin and may contain peripheral eccentric calcifications that
`although pathognomonic, are only found in 15% of patients [49,57,74,75]. ERP
`is usually normal but may identify pancreatic duct strictures, obstruction, and
`displacement caused by a mass effect primarily resulting from malignant
`transformation [45]. Pancreatic duct communication seldom is seen, because
`the origin of MCNs is within the peripheral ductal system [47,62]. Although
`seldom obtained, angiography demonstrates the hyper-vascularity of most
`MCNs. These tumors may grow as large as 36 cm, with greater size correlating
`with malignancy [71]. Other evidence of malignancy includes cyst wall
`irregularity and thickening, intracystic solid regions, or an adjacent solid mass
`[64,72,76].
`As opposed to SCAs, the larger size of the individual cystic components
`simplifies FNA and facilitates complete drainage. Aspiration, however, may be
`impaired by the presence of viscous mucous. Prolonged aspiration or use of a
`larger caliber needle (19 Gauge) usually allows procuring of a fluid sample. The
`presence of mucin or elevated tumor marker (eg, carcinoembryonic antigen
`[CEA])
`levels strongly suggests a mucinous tumor [68,77–80]. Mucinous
`cuboidal or columnar epithelial cells are found in approximately 50% of cases
`and are diagnostic of a mucinous lesion but also may be seen with intraductal
`papillary mucinous neoplasia (IPMN) [78]. The results of FNA, however, can
`distinguish these lesions from SCAs and pseudocysts. Additionally,
`the
`interpreting pathologist must consider contamination from gastric or duodenal
`columnar epithelial cells. Individual tumors commonly contain a spectrum of
`histology ranging from regions of adenomatous change, to invasive carcinoma,
`with intervening denuded epithelium. The often sporadic distribution of
`
`Fig. 4. Mucinous cystadenoma.
`
`000007
`
`
`
`124
`
`levy & wiersema
`
`Fig. 5. Intraductal papillary mucinous neoplasia. Endoscopic visualization widely patent (gaping or
`fish mouth) papilla extruding mucous.
`
`dysplastic findings prohibits high diagnostic accuracy by biopsy alone [34,73].
`The sensitivity of FNA for diagnosing mucinous cystadenocarcinoma (MCAC)
`is 67%, in large part because of the focal distribution of malignancy [69,81].
`Surgical resection may be necessary to distinguish the specific type of CPT and to
`establish the presence of malignancy [71].
`
`Intraductal papillary mucinous neoplasia
`
`Endoscopic inspection of the papilla may reveal a widely patent (gaping or
`fish-mouth) papilla extruding mucous (Fig. 5) [82]. IPMN can be divided into
`predominantly main duct or side branch disease, with EUS demonstrating a
`diffusely dilated main duct or one or several dilated side branches (Figs. 6, 7)
`
`Fig. 6. Intraductal papillary mucinous neoplasia. Dilated main pancreatic duct in a patient with
`main duct disease.
`
`000008
`
`
`
`pancreatic neoplasms
`
`125
`
`Fig. 7. Intraductal papillary mucinous neoplasia (IPMN). Dilated uncinate branch in a patient with side
`branch IPMN.
`
`[51,64,83]. Although communication with the pancreatic duct is a feature of side
`branch IPMN and helps exclude MCN, the absence of communication does not
`exclude IPMN, because mucous can block the flow of contrast into the abnormal
`side branch. This often can be overcome by relatively forceful contrast injection,
`but risks inducing pancreatitis. Mucous or a mural tumor nodule (papillary
`projection) may cause filling defects. Patients rarely present with a predominantly
`solid mass that may be mistaken for a primary ductal carcinoma. Conversely,
`patients also may present with a cystic mass that may be misdiagnosed as a SCA
`or MCN (Fig. 8)
`[84,85]. Although IPMN can be mistaken for chronic
`pancreatitis, the finding of normal pancreatic parenchyma and mucous emanating
`from the papilla suggests IPMN. The latter finding is present in only 25% to 50%
`of patients with IPMN. Distinction from chronic pancreatitis may be difficult, as
`parenchymal changes can develop in IPMN as a result of ductal obstruction from
`intraductal tumor growth or inspissated mucous. Cytologic analysis of aspirated
`
`Fig. 8. Intraductal papillary mucinous neoplasia. Patient presenting with a predominant cys-
`tic component.
`
`000009
`
`
`
`126
`
`levy & wiersema
`
`Fig. 9. Intraductal ultrasound catheter exiting a standard side viewing duodenoscope.
`
`duct or cyst fluid demonstrates findings similar to mucinous cystadenomas,
`including the presence of columnar epithelial cells. Malignancy may be suggested
`by the finding of a focal hypoechoic mass, mural nodules, or a large unilocular
`cystic component [86]. Invasive carcinoma is suggested by: rupture of the main
`pancreatic duct wall with intrapancreatic spread of tumor, tumor invasion of the
`duodenum or common bile duct, malignant-appearing lymphadenopathy, and
`extrapancreatic spread or vascular invasion [87]. Intraductal ultrasound (IDUS)
`(Fig. 9) and pancreatoscopy are newer techniques that assist in the evaluation
`[88,89]. IDUS catheters are small-caliber (approximately 2 mm) miniprobes that
`are passed through standard duodenoscopes into the pancreatic duct [87,89].
`These probes operate at higher frequencies (12 to 30 MHz) than standard EUS,
`which improves image resolution (0.07 to 0.18 mm), but limits the depth of
`image penetration [90]. IDUS and pancreatography can distinguish main duct
`from side branch IPMN, identify papillary projections (Fig. 10) to assess the risk
`of malignancy, and determine the longitudinal extent of tumor spread and
`parenchymal invasion [89].
`
`Fig. 10. Papillary projections seen during pancreatoscopy in a patient with intraductal papillary
`mucinous neoplasia.
`
`000010
`
`
`
`pancreatic neoplasms
`
`127
`
`Endoscopic ultrasound for cystic pancreatic tumors
`
`The authors perform EUS in all patients with a suspected CPT to help exclude
`a pseudocyst, determine the specific type of CPT, and assess the risk for
`malignancy [33,64–66,91–93]. Doing so requires evaluation of the cyst wall
`(thickness, focal
`irregularity, mass, or papillary projections) and intracystic
`structures (septations, echo-dense mucous, debris). EUS examination of the
`pancreatic ductal anatomy, parenchyma, or the finding of previously undetected
`cystic or solid mass lesions can provide additional diagnostic information. The
`authors perform EUS even when resection is planned to assess for malignancy
`and locoregional or distant disease that would preclude surgical intervention.
`EUS features that correlate with malignancy include the presence of focal cyst
`wall thickening or irregularity, septal thickening, an adjacent solid mass, and the
`presence of collateral vessels. As with other imaging modalities, EUS alone
`cannot distinguish the tumor type accurately or identify malignancy, particularly
`when only few criteria are assessed [93–96]. In a study that solely considered the
`presence of an associated mass component as a sign of malignancy, the sensitivity
`and accuracy of EUS were only 65% and 75% for identifying premalignant or
`malignant cysts [96]. Others have shown equally poor results, with a sensitivity
`and specificity of 52% and 58%, respectively, when such a narrow spectrum of
`features was assessed [93]. In a prospective study involving 52 patients
`undergoing resection with tissue confirmation, however, EUS accurately
`categorized 92% of tumors [33]. Similarly, another group found that the presence
`or absence of at least two of three features (pancreatic parenchymal changes,
`septa, and mural nodules) offered a sensitivity and specificity of 94% and 85%,
`respectively. The limited depth of imaging with EUS reduces the diagnostic
`accuracy of large cystic lesions (greater than 6 cm) [33]. TUS, CT, and MRI may
`be particularly useful in this subgroup of patients with larger lesions.
`
`Cytology and cystic fluid analysis
`
`Although a recent study reported sensitivity of cytology greater than 95% [97],
`most centers describe a sensitivity ranging from 27% to 64% [93,94,98]. In
`contrast, the specificity of cytology approaches 100% in all studies [93,94,97,98].
`Although aspirated fluid from CPTs may contain denuded epithelium even in the
`presence of malignancy, biopsies are often falsely negative as a result of sampling
`error. When cytology is negative, complete surgical resection is required to
`exclude or establish presence of malignancy [71]. Addition of cyst fluid marker
`levels, amylase, and mucin stain to cytology alone can increase the diagnostic
`accuracy to 80% to 90% [94,99].
`Assessment of cyst fluid for tumor markers (CEA, CA 19-9, CA 15-3, and CA
`72-4) may improve diagnostic accuracy. The CEA level appears to be the most
`useful to discriminate nonmucinous (benign) from mucinous (premalignant or
`malignant) lesions. Studies vary as to the threshold value that offers ideal
`
`000011
`
`
`
`128
`
`levy & wiersema
`
`sensitivity and specificity for discriminating lesions. Lower values of CEA are
`thought to arise in pseudocysts and SCAs, while higher values are more common
`with mucinous tumors that can behave more aggressively with malignant
`transformation. A CEA level below 5 ng/mL offers a sensitivity of 57% to 100%
`and specificity of 77% to 86% [67,97,100]. Others, using a cut-off of greater than
`50 ng/mL, found sensitivity for CEA to be 90% for identifying premalignant or
`malignant lesions [96], versus a sensitivity and specificity of only 28% and 25%,
`respectively in a more recent study [93]. In one report, a CEA value greater than
`400 ng/mL provided 100% specificity in distinguishing MCNs from pseudocysts
`[101], compared with another study with a sensitivity and specificity of only 13%
`and 75%, respectively [97]. Although the CEA level from pseudocyst fluid tends
`to be very low, elevated levels are common in infected pseudocysts [47,92,102].
`Limited data suggest that the CA 15-3 level is useful in differentiating benign
`from malignant pancreatic mucinous cysts with an upper cutoff value of 30 U/mL
`reported to distinguish MCAs from MCACs with 100% sensitivity and 100%
`specificity [103]. In another report, CA 72-4 was more useful than CEA or CA
`15-3 for distinguishing MCNs, demonstrating a sensitivity and specificity of
`87.5% and 94%, respectively [104]. Similarly, a CA 72-4 level greater 40 U/mL
`has demonstrated 63% sensitivity and 98% specificity for distinguishing MCNs
`from SCAs and pseudocysts [101]. A CA 19-9 cut-off level between 50,000 and
`90,000 U/mL may distinguish malignant cysts [97,100]. CA 19-9 levels greater
`than 50,000 U/mL provide a sensitivity of 15% to 75% and a specificity of 81%
`to 90% for distinguishing mucinous from nonmucinous lesions [97,100]. The CA
`19-9 level, however, commonly rises secondary to inflammatory conditions and
`when biliary obstruction is present, thereby limiting the diagnostic utility [47,
`92,102].
`The amylase concentration helps narrow the differential, because high levels
`typically are found only in fluid from cysts that communicates with the pancreatic
`duct (pseudocysts and IPMN) [102]. An amylase level greater than 5000 U/L
`provides a sensitivity and specificity of 61% and 58%, respectively, for dif-
`ferentiating pseudocysts from other CPTs [97].
`Although EUS–FNA appears safe, the utility of morphologic assessment and
`cyst fluid analysis remains uncertain. Although the sensitivity of EUS–FNA for
`identifying malignancy may be limited, this finding alters therapy for patients in
`whom surgery is not intended but rather surveillance and periodic imaging are
`planned. Negative or benign findings do not necessarily exclude malignancy, and
`in these patients, surveillance imaging is suggested. The role of tumor markers is
`controversial, as is the threshold value that discriminates the lesion type with
`greatest accuracy. Another limitation is the tendency for sampling error when
`processing fluid from multi-locular cysts, whose fluid composition can vary
`within the lesion [79]. Of all tumor markers, the CEA level appears to have the
`most diagnostic value. The authors consider use of tumor markers to be largely
`investigational, however, and caution the role they should play on influencing
`care. The combination of cyst fluid marker analysis and cytologic examination
`may prove to be the most accurate diagnostic approach. When limited fluid is
`
`000012
`
`
`
`pancreatic neoplasms
`
`129
`
`available for analysis, the authors request serial evaluation for cytology with
`mucin stain, CEA, and amylase. In the authors’ practice, determination of other
`tumor marker levels and biochemical studies are requested only for investiga-
`tional purposes.
`Performing FNA largely depends on a physician’s approach to the manage-
`ment of CPTs. FNA ideally is reserved for situations when the results are
`expected to influence patient care, as for patients in whom the need for surgical
`intervention is debated because of diagnostic uncertainty, advanced age, or
`marginal health status. In general, the authors do not recommend FNA for
`classically benign-appearing lesions for which no intervention is intended or for
`resectable malignant appearing lesions for which surgery already is planned.
`Although a negative result does not exclude malignant or potentially malignant
`disease, it may support the decision for surveillance and periodic imaging. The
`finding of malignant cytology, a positive mucin stain, or elevated cyst fluid CEA,
`however, may support resection.
`
`Pancreatic adenocarcinoma
`
`The incidence of pancreatic adenocarcinoma is increasing, with an estimated
`28,000 new cases in the United States this year [105]. Although it is the 10th
`most common malignancy, it is the fourth leading cause of cancer-related mor-
`tality and the second most common cause of cancer deaths for all GI-related car-
`cinomas [106]. Most patients with pancreatic cancer present late in their course
`and have either locally extensive or metastatic disease with a median survival of
`only 4 to 6 months [107,108]. At the time of diagnosis, only 10% to 20% of
`patients are candidates for curative resection [109,110]. The late presentation,
`aggressive nature, and lack of effective therapies all contribute to the poor
`prognosis. Accurate staging of pancreatic adenocarcinoma is important
`to
`identify the subset of patients who have potentially resectable localized cancers.
`Although early detection is crucial to improve prognosis, the determination of
`resectability is important to help avoid unnecessary surgical intervention.
`Staging as defined by the TNM classification (Table 3) depends on char-
`acteristics of the primary tumor, namely tumor size and infiltration into major
`vessels, (T stage), regional lymph node involvement (N stage), and the presence
`or absence of distant metastasis (M stage). EUS can evaluate all necessary
`structures to allow locoregional staging of pancreatic adenocarcinomas and at
`times also detects hepatic metastases. Gress et al evaluated the use of EUS to
`stage 151 patients with pancreatic cancer. In the 81 patients undergoing surgical
`resection, the accuracy of EUS for T stage, N stage, and vascular invasion was
`85%, 72%, and 93%, respectively [11]. Similarly, Tio et al demonstrated the
`overall accuracy of EUS for T and N staging at 84% [111]. Although distant
`metastasis must be evaluated by other means, such as CT or laparoscopy, local
`resectability is predicted accurately in 75% to 90% of patients by EUS [2,3,112].
`The overall accuracy of EUS for predicting lymph node invasion (N stage) is
`
`000013
`
`
`
`130
`
`levy & wiersema
`
`Table 3
`American Joint Committee on Cancer staging of pancreatic adenocarcinoma
`
`Primary tumor (T):
`T1
`T2
`T3
`
`T4
`
`Regional lymph nodes (N):
`N0
`N1
`
`Distant metastases (M):
`M0
`M1
`
`Stage grouping:
`Stage
`
`I
`
`II
`III
`
`IVA
`IVB
`
`Tumor limited to pancreas, size b 2 cm in greatest dimension
`Tumor limited to pancreas, size N 2 cm in greatest dimension
`Tumor infiltration into duodenum, bile duct, papilla, peripancreatic
`tissue (retroperitoneal and mesenteric fat, mesocolon, greater/lesser
`sac, and peritoneum) or major venous structures (portal vein,
`superior mesenteric vein)
`Tumor infiltration (extension) into stomach, spleen, colon, or
`major arterial structures (superior mesenteric artery, celiac trunk,
`hepatic artery, but not splenic vessels)
`
`No regional lymph node metastasis
`Regional lymph node metastasis (pN1a = single regional node,
`pN1b = multiple regional nodes)
`
`No distant metastasis
`Distant metastasis
`
`T
`
`1
`2
`3
`1
`2
`3
`4
`Any
`
`N
`
`0
`0
`0
`1
`1
`1
`Any
`Any
`
`M
`
`0
`0
`0
`0
`0
`0
`0
`1
`
`lower than for primary tumor staging (T stage). In three studies [4,113,114], the
`accuracy of EUS for T stage was 82% to 91%, and for N stage, accuracy was
`64% to 73%.
`Early EUS reports demonstrated superior accuracy for preoperative staging
`of pancreatic cancer (85% to 100%) compared with dynamic CT (64% to 66%)
`and TUS (61% to 64%) [9,115,116]. Gress et al [11] reported on 81 patients
`with pancreatic adenocarcinoma who were evaluated preoperatively by dynamic
`CT and EUS. The results achieved with EUS were superior to dynamic CT
`for T staging (85% versus 30%, P b 0.0001) and N staging (72% versus 55%,
`P b 0.0001), and for vascular
`invasion (93% for EUS and 60% for CT,
`P b 0.0001).
`The use of rapid-scanning helical CT permits multiple scans to be obtained
`through the abdomen during different phases of contrast enhancement. This
`technology allows