pancreatic cancer staging

Because of its late clinical presentation, pancreatic cancer remains the fourth most common cause of cancer death. Surgery is the only option for cure, and advancements in surgical technique and neoadjuvant therapy have the potential to increase the number of patients who could undergo surgery for potential cure. Accurate diagnosis and staging of pancreatic ductal carcinoma are therefore of great importance. This paper reviews the state of the art for such diagnostic modalities as computed tomography, magnetic resonance imaging, endoscopic ultrasound, positron emission tomography, and laparoscopic surgery and laparoscopic ultrasound for diagnosis and staging of pancreatic cancer. Currently, accurate diagnosis and staging probably require the use of a combination of techniques, including multiphase helical or multidetector computed tomography and/or dynamically enhanced magnetic resonance imaging with endoscopic ultrasound with fine-needle aspiration. The role of positron emission tomography still needs to be determined. The role of laparoscopic surgery and laparoscopic ultrasound may be limited in those institutions with state-of-the-art imaging techniques.

In the United States, 27,000 patients are diagnosed yearly with pancreatic cancer. Most cases (75%–90%) are due to ductal adenocarcinoma. Because of its very poor prognosis, pancreatic cancer is the fourth most common cause of cancer death, most commonly afflicting patients 65–80 years of age. The poor prognosis of pancreatic ductal adenocarcinoma is due to its late clinical presentation and the inability to treat the disease effectively at that stage. Untreated, the median survival is 3–6 months. The only option for cure is surgery (pancreaticoduodenectomy, otherwise known as the Whipple procedure). At presentation, only 40% of patients are potential surgical candidates. The remainder have liver metastases, peritoneal carcinomatosis, or both. Currently, 5-year survival after the Whipple procedure is 18%–21%. The prognosis is worse for those with tumors larger than 2 cm, vascular invasion, or lymph node metastases.

Advancements have been made in therapy. Improvements in surgical techniques now allow for reconstruction of the portal or mesenteric veins in those with limited involvement of these structures by tumor. The use of neoadjuvant chemotherapy and radiotherapy has been shown to downstage tumors, increasing the percentage of patients who can undergo surgical resection. Neoadjuvant therapy has also been shown to decrease the incidence of local recurrence and may increase survival over those who have surgery alone or surgery and adjuvant therapy. The use of imaging studies is therefore important in aiding early diagnosis, identifying patients who are surgical candidates, and guiding treatment planning.
IMAGING STRATEGIES

The objectives for imaging are to provide accurate diagnoses for pancreatic masses, to accurately stage pancreatic cancer for treatment planning, to identify complications of treatment, and to evaluate treatment progress.

Currently, only a very small group of patients out of all those who develop pancreatic cancer has been identified to be at increased risk for developing pancreatic cancer: those with hereditary pancreatitis or pancreatic cancer. No simple, inexpensive screening technique is available. Clinical presentations include such signs and symptoms as painless obstructive jaundice, abdominal or back pain, weight loss, and, uncommonly, palpable abdominal masses. Unfortunately, clinical presentations can be nonspecific.

Initial imaging is usually performed to identify the cause of abdominal pain, weight loss, and/or signs of biliary obstruction. Transabdominal ultrasound is often used for the screening of patients with signs of common bile duct obstruction. Ultrasound can screen for the presence of gallstones, gallbladder wall thickening, pericholecystic fluid, tenderness directly attributable to the gallbladder (sonographic Murphy's sign), and distention of the common bile duct. It can help identify signs of acute or chronic cholecystitis and can often identify the level of obstruction of the common bile duct. However, evaluation for pathology within the pancreas is often limited because of obscuring overlying bowel gas and skill of the operator. Although transabdominal ultrasound can identify a large mass within the pancreas, it is limited in its sensitivity for small lesions and is very limited for determining vascular invasion. If choledocholithiasis is excluded as a diagnostic possibility, either clinically or by transabdominal ultrasound, screening computed tomography (CT) of the abdomen is often then utilized. In those patients in whom CT indicates the presence of a probable pancreatic mass that is potentially resectable, the goal of imaging is then to determine the extent of the tumor and its relationship to vascular structures to guide treatment planning. This is typically performed with multiphase, thin-section, helical CT or with the more recently developed technique of multidetector helical CT. Dynamic imaging with gadolinium-enhanced abdominal magnetic resonance imaging (MRI) can be used as an alternative and is particularly useful in patients with a history of allergy to iodinated contrast agents or those with compromised renal function in whom the use of iodinated contrast is also contraindicated.

Endoscopic ultrasound (EUS) has developed as a technique that is complementary to CT for diagnosis of pancreatic ductal adenocarcinoma. The technique is very sensitive for detecting lesions (90%–95%) and, when combined with fine-needle aspiration (FNA), is also very specific for the diagnosis of pancreatic cancer. It reportedly can detect lesions as small as 5 mm in diameter, and it has been shown to be superior to CT for the detection of small lesions (<2 cm). However, its role for staging is limited because of its narrow field of view, and prior reports have shown mixed results in regard to vascular invasion and malignant adenopathy. EUS is also unable to fully evaluate the liver for metastases. However, when combined with helical CT or multidetector CT, EUS with FNA provides an excellent means to characterize pancreatic lesions as benign or malignant and can provide additional information to aid accurate staging for planning treatment.

Endoscopic retrograde cholangiopancreatography (ERCP) can be used to treat benign or malignant strictures of the biliary or pancreatic ductal systems and can provide diagnostic information on disease involvement of these structures. Brushings can be obtained from regions of suspected ductal tumor involvement. However, because it cannot provide cross-sectional information, its role in staging is limited. A study that evaluated the combined use of ERCP and EUS showed an increase in sensitivity to 92% but no significant change in accuracy (89%).

The role of positron emission tomography (PET) in the routine work-up of pancreatic cancer has yet to be determined. PET has been suggested as a means to characterize pancreatic masses. However, PET has been reported to have a sensitivity of 68%–98% and a specificity of 53%–95% for the identification of pancreatic cancer. The range in reported sensitivities and specificities may be attributable to the use of different techniques, study populations, and criteria for differentiating benign from malignant disease. An increased rate of false-negative results has been described for patients who are hyperglycemic, and false-positive results occur in those with acute pancreatitis. PET has also been suggested as a means to detect liver metastases at an earlier stage than CT and MRI. However, false-positive findings for liver metastases have been described. Currently, PET cannot be utilized to exclude the presence of pancreatic malignancy and probably cannot be used alone to document the presence of liver metastases but may offer greater sensitivity than CT for liver metastases. A new technologic development, which combines a PET scanner with a multidetector computed tomographic scanner, would potentially provide information on abnormal glucose metabolism and precise anatomic information.

Despite advances in sensitivity and specificity, imaging techniques are limited in their ability to detect small peritoneal metastases. Laparoscopic surgical evaluation has been suggested as a means to increase the accuracy of staging prior to patients undergoing pancreaticoduodenectomy. Studies have shown increased sensitivity for peritoneal metastases over CT. A study at the Massachusetts General Hospital using laparoscopy and peritoneal cytology showed that approximately 30% of patients without metastases evidenced on CT showed occult metastatic disease on laparoscopy. However, other studies, using high-quality preoperative imaging, have shown that perhaps as few as 5% of patients would benefit from the additional information obtained from laparoscopic surgery. Laparoscopic ultrasound has been suggested as a means to enhance detection of vascular involvement, metastatic liver and peritoneal disease, and adenopathy. One study showed that specificity and accuracy for staging increased (88% and 89%, respectively) when laparoscopic ultrasound was utilized over laparoscopy alone (50% and 65%, respectively).
IMAGING TECHNIQUES

EUS is a technique in which a high-frequency ultrasonic transducer (usually 7.5–12 MHz) is introduced via a side-viewing endoscope to provide cross-sectional imaging information of the gastrointestinal tract wall and adjacent structures, such as the pancreas. As noted earlier, it reportedly can detect more reliably than CT lesions that are less than 2 cm in diameter. FNA of abnormalities identified by EUS can then be performed, often in a manner safer than by percutaneous CT or transabdominal ultrasound—guided biopsy. The use of FNA greatly increases the specificity of EUS, reportedly to 88%–100%. However, as with any biopsy technique, it is not always possible to attain adequate specimens. The success of the technique is operator dependent. Studies evaluating the role of EUS in assessing vascular involvement have shown mixed results. Various criteria have been offered for identifying involvement of a vessel by tumor, and sensitivities for vascular involvement have been reportedly as low as 17%–43%, depending on the criteria that were used and the vessel in question. The ability of EUS alone to identify lymph nodes infiltrated by tumor is limited, but it may improve with the use of FNA, although reports in the literature are mixed. There have been suggestions that EUS may be useful as a delivery mechanism for immune-based, viral-based, or chemical-based therapeutic agents. EUS has also been proposed for administering local nerve blocks for palliation.
Computed Tomography Techniques

If an initial transabdominal ultrasound examination fails to show evidence for choledocholithiasis, or if the patient's signs and symptoms do not suggest that choledocholithiasis is a likely cause, then CT is commonly used for further evaluation. Imaging by CT requires intravenous contrast enhancement with iodinated contrast to differentiate between normal solid organ parenchyma and most pathology. An appropriate injection rate is vital; the more rapid the injection rate, the greater the contrast between the lesion and the adjacent normal pancreatic parenchyma. Studies that have evaluated pancreatic and hepatic enhancement using different rates of intravenous contrast enhancement have indicated that peak enhancement of the pancreas occurs before peak liver enhancement. This difference in the timing of peak enhancement occurs because the pancreas derives its blood supply from the celiac axis and the superior mesenteric artery, whereas the liver's supply is from the hepatic artery and the portal vein. The portal vein, which receives blood from the splanchnic circulation, provides 75% of the liver's blood supply. The advent of helical CT has allowed for the acquisition of images of the abdomen in a single breath hold, which allows images to be obtained during a single phase of contrast enhancement. The best contrast between normal pancreatic parenchyma and pancreatic ductal adenocarcinoma is obtained 40 seconds after injection of intravenous contrast.

Current CT protocols are designed to optimize conspicuity of pancreatic lesions, peripancreatic vessels, and liver lesions to provide optimum information for diagnosis and staging. Pancreatic ductal adenocarcinoma is a relatively hypovascular tumor and is best detected during the pancreatic parenchymal phase, which occurs approximately 35–50 seconds after the beginning of injection of intravenous contrast. Liver metastases, which are relatively hypovascular compared with normal liver parenchyma, are best imaged during the portal venous phase of liver enhancement, which occurs approximately 60–70 seconds after the beginning of injection of intravenous contrast. Current CT protocols utilize multiphase scanning with either helical or newer multidetector computed tomographic scanners. A typical protocol on a helical computed tomographic scanner would utilize a slice thickness of 3–5 mm and a pitch of 1.5 to 2 to evaluate the entire pancreas during the pancreatic parenchymal phase. With a multidetector scanner, a slice thickness of 1.25–2.5 mm could be obtained of the pancreas and the liver in the same time period using a pitch of 3 to 6 to cover the liver and pancreas. With an injection rate of 5 mL/sec, the delay before beginning imaging during the pancreatic parenchymal phase would be 35–50 seconds. At an injection rate of 3 mL/sec, the delay would be 40–70 seconds. Both the liver and the pancreas are then scanned during the portal venous phase (60–70 seconds) to optimally visualize liver metastases.

The use of sophisticated computer-based postprocessing techniques has provided new perspectives for interpreting axially acquired computed tomographic imaging data. The use of axially acquired computed tomographic imaging data acquired during the portal venous phase combined with thick-slab minimum intensity projections (computed tomographic cholangiopancreatography) has been described as improving visualization of pancreatic and biliary anatomy to a level that approaches that of ERCP. This technique could enhance visualization of pathology that obstructs the biliary or pancreatic ductal systems and help define its relationship to those structures. The use of maximum intensity projection and shaded-surface techniques has been shown to enhance visualization of arterial and venous anatomy relevant to staging pancreatic cancer. Such postprocessing techniques require thin slice axial source images, necessitating the use of either multidetector or helical CT. Images as thin as 1.25 mm can be acquired of the abdomen by multidetector CT in a single breath hold and can be used to create images reformatted in coronally oriented planes to enhance visualization of tumor encasement of vessels.
MRI Technique

The primary benefits of MRI are inherently superior contrast between soft tissues, even before the administration of intravenous contrast, and the ability to visualize pathology in multiple planes without the use of ionizing radiation or iodinated contrast agents. MRI therefore can be performed in patients with a history of allergy to iodinated contrast or poor renal function. However, MRI is contraindicated in a subset of patients, including those with implanted electronic devices or indwelling metal that would be susceptible to movement in a magnetic field.

Most MRI imaging utilizes high field strengths (1.5 T or higher). Imaging can be obtained with a body coil, but optimal imaging is obtained with a phased-array surface coil. The most commonly utilized sequences are T1-weighted spin-echo with or without fat suppression, T1-weighted fast spoiled gradient breath hold with or without fat suppression, T2-weighted fast spin-echo with fat suppression, and dynamically enhanced (usually with gadopentetate dimeglumine) TI-weighted spoiled gradient breath hold with or without fat saturation. The dynamically enhanced techniques have been shown to increase lesion conspicuity. It has been reported that fat-suppressed TI-weighted images may also contribute to lesion detection. Slice thickness is not as crucial as it is in contrast-enhanced CT because of MRI's better soft tissue contrast before contrast administration; slice thicknesses of 5–6 mm are suitable for evaluation of the pancreas. An overview of the literature indicates that state-of-the-art MRI systems may improve lesion detection and/or staging over those obtained through CT.

Traditionally, contrast-enhanced imaging has been performed with gadopentate dimeglumine. Recent studies have looked at the role of mangafodipir trisodium, which was developed as a hepatocellular-specific contrast agent, for the evaluation of primary pancreatic tumors, with mixed results However, this agent may be useful for identifying liver metastases.

The use of T2-weighted imaging with prolonged echo times has been utilized for MR cholangiopancreatography (MRCP). This technique, used to acquire images in the axial and/or coronal planes, can be used noninvasively to detect the presence and the level of pancreatic and biliary obstruction, with a sensitivity and specificity similar to those of ERCP. Thin-section images can be used to evaluate for intraductal abnormalities, such as calculi and tumor, and when combined with advanced postprocessing techniques, can depict ductal structures in various projections. Thick-section images can be used to provide an overview of ductal anatomy and can be rapidly obtained in multiple projections, but they can obscure pathology within ductal systems. New breath hold “0.5NEX” techniques are able to rapidly acquire many images with T2 characteristics in a single breath hold, eliminating artifacts caused by respiratory movement or bowel motion. When combined with additional sequences, such as those providing dynamically enhanced TI-weighted images, MRCP offers the advantages of both ERCP and cross-sectional imaging for identifying causes of ductal obstruction.
PET Technique

PET traditionally utilizes the agent fluoro-deoxy-D-glucose (FDG) labeled with F. The radiolabeled tracer is administered intravenously, and imaging usually occurs between 40 minutes and 3 hours after injection, depending on an institution's protocol. Plasma glucose levels should be less than 130 mg/dL to avoid false-negative results. Patients should not have an active inflammatory process (i.e., acute pancreatitis), which can cause false-positive results. Imaging is usually deferred for patients with hyperglycemia or for those who may have an active inflammatory process.

The field of imaging can include the abdomen or the whole body, requiring between 15–90 minutes of imaging time, depending on the field of imaging that is being acquired. Attenuation correction is performed of the emission scan. Standardized uptake values (SUVs) can be determined to provide a semiquantitative evaluation of lesions. Because PET can acquire data with a volumetric technique, with near isotropic resolution, images can be reconstructed in multiple planes.

Because FDG-PET evaluates glucose metabolism, it cannot be used to provide anatomic information regarding local extension of tumor (i.e., vascular invasion). This means that PET cannot be used as a primary means to stage pancreatic cancer. However, it has been suggested as a means to differentiate focal changes of chronic pancreatitis from pancreatic cancer, but the presence of false-positive and false-negative results means that correlation with other studies and biopsy is still warranted. To overcome these limitations, manufacturers are working on developing PET units that are combined with multidetector computed tomographic scanners. The techniques for optimally utilizing such hybrid units are currently being developed.
FINDINGS FOR PRIMARY TUMOR CT Findings

Nearly 60% of pancreatic cancers are identified in the pancreatic head, 15% in the body, and 5% in the tail, with 20% of pancreatic cancers diffusely involving the pancreas. Pancreatic cancer usually has the appearance of a hypodense or, less commonly an isodense, mass on various phases of intravenous contrast-enhanced CT.

The imaging characteristics of regions of tumor seen on pancreatic parenchymal and portal venous phases of enhancement correlate with findings on histopathology. Regions that are hypodense to normal pancreatic parenchyma on early-and late-phase images have been shown on histopathologic examination to be necrotic or to contain mucin. Regions that are initially hypodense that become isodense or hyperdense on late phase images correlate with regions of dense fibrosis. Regions that are isodense on both early- and late-phase images show increased tumor cellularity. Correlation with histopathology indicates that size on CT correlates well with that determined by histopathology.

Approximately 40% of tumors that appear hypodense on the pancreatic parenchymal phase of imaging become isodense to the normal pancreas on delayed phases of imaging. Approximately 80% of small tumors appear isodense on the delayed phase. In small tumors, which may not distort the contour of the pancreas, visualizing this subtle difference in density may be the only means to identify the tumor. Overall, ductal adenocarcinoma has been reported to have a hypodense appearance on CT in 69%–96% of cases. This emphasizes the importance of imaging during both the pancreatic parenchymal phase and the portal venous phase to obtain information for diagnosis and staging.

A review of the helical CT literature on sensitivity for the detection of pancreatic cancer shows a detection rate of approximately 89% (78%–100%) for lesions greater than 2 cm in diameter and a detection rate of approximately 71% (40%–100%) for lesions less than 2 cm in diameter. The detection rates for the newer technology of multidetector CT, which can acquire thinner sections during a single breath hold, are unknown. Pitfalls for detection include small tumors located in the uncinate process, which may not be associated with common bile duct or pancreatic duct dilatation, and chronic pancreatitis, which can mimic the appearance of a pancreatic cancer.
MRI Findings

The findings on MRI, when intravenous gadolinium contrast agents are given, mirror those seen for contrast-enhanced helical/multidetector CT. Pancreatic adenocarcinoma is best identified on pregadolinium fat-suppressed TI-weighted images and immediate post-gadolinium, dynamically obtained, spoiled-gradient, TI-weighted images.In both instances, ductal adenocarcinoma has the appearance of a mass with lower signal intensity than that of the surrounding normal pancreatic parenchyma. On later phases, the tumor can become isointense in signal intensity to the surrounding normal pancreatic parenchyma. As with CT, the timing of imaging is crucial in order to optimize lesion detection. Tumor can have a variable appearance on T2-weighted images, usually appearing hyperintense or isointense to surrounding normal pancreatic parenchyma.

Dynamically obtained, intravenous gadolinium—enhanced MRI has been reported to have a detection rate of approximately 90%. However, as with CT, the detection rate falls for tumors less than 2 cm in diameter. Nishiharu et al, in a study of 57 patients that compared MR and CT, reported that MRI had an overall accuracy of 79%–81% and a sensitivity of 80%–95% for the detection of pancreatic cancer; however, both CT and MRI missed some cancers that were less than 1.5 cm in diameter.

The findings for ductal adenocarcinoma on MRCP are similar to those seen on ERCP. Tumor can encase and obstruct the common bile duct, the pancreatic duct, or both. The identification of dilatation of both pancreatic and biliary ducts is very suggestive of a pancreatic malignancy. However, a normal caliber pancreatic duct is present in approximately 20% of patients with pancreatic cancer.[ 103] The detection of pancreatic duct side-branch ectasia can raise the consideration for chronic pancreatitis.[ 103, 104] MRCP can also be used to identify the level of obstruction and to help exclude calculi as a cause of biliary obstruction. Thick-slab imaging or advanced postprocessing techniques can provide anatomic information that is useful for guiding stent placement.
EUS Findings

In a prospective study of 115 patients, the most commonly encountered features for pancreatic cancer were a inhomogeneous (85%) solid mass (88%), with irregular borders (78%) that appeared hypoechoic (98%) to normal pancreatic parenchyma. As noted earlier, EUS has a high sensitivity for detection of pancreatic cancer (90%–95%). However, in the same study of 115 patients, specificity, derived from morphologic features identified by EUS, was only 53%. Entities that were mistaken as pancreatic cancer in this study included neuroendocrine tumors, metastases to the pancreas, and focal changes from chronic pancreatitis. For this reason, FNA is commonly employed. As noted earlier, when EUS is combined with FNA, specificity increases to 88%–100%. Findings for laparoscopic ultrasound and intraoperative ultrasound would be similar to those for EUS, and therefore, the entities to consider in the differential diagnosis would also be similar.
PET Findings

Pancreatic cancer, like most cancers, has the appearance of a region of intense, usually well-defined, radiotracer activity because of its uptake of glucose. However, regions of inflammatory activity, such as may be seen in acute pancreatitis, can mimic the appearance of tumor, leading one author to suggest that all patients have obtained acute-phase protein values before PET to determine whether an inflammatory process is present.False-negative results have been reported for highly differentiated tumors, for tumors less than 1 cm, and in patients with blood glucose levels >130 mg/dL. Attempts have been made to use SUVs to discriminate between pancreatic cancer and pancreatitis. One study of 106 patients obtained SUV measurements 45 minutes after injection and found that pancreatic cancer had a value of 6.4 +/-3.6, whereas chronic pancreatitis had a value of 3.6 +/- 1.7. Another study of 47 patients with suspected pancreatic cancer found that 17 of 22 patients with benign disease showed a decrease in SUV at 2 hours after injection, whereas in 22 of 27 patients with malignant lesions, SUV increased at the same time point. When a retention index was combined with the SUV value obtained at 1 hour after injection, accuracy increased to 91.5% in discriminating benign from malignant disease. Other investigators have suggested other techniques and cutoff values with different sensitivities and specificities.

COMPARISON OF IMAGING MODALITIES

Literature comparing the imaging techniques of CT, MRI, EUS, PET, and laparoscopic surgery with/without laparoscopic ultrasound is very limited because comparisons are usually not obtained with state-of-the-art techniques for all of the modalities under review, partly because of the rapid changes in technology. Early reports that did not use state-of-the art techniques for CT and MRI found EUS to be more sensitive for lesion detection. However, other recent reports have found EUS and CT to be comparable. Some studies have suggested that high-quality MRI may be superior to CT in detecting lesions. However, another comparison study found CT to be more sensitive than MR in detecting peripancreatic tumor extension and vascular invasion. Other factors that limit comparison include the experience of the participating physicians (e.g., gastroenterologists, radiologists, surgeons), consistency among the equipment utilized, interpretation criteria, and patient-selection criteria.

An overall assessment of the literature appears to suggest that EUS may have a slight advantage over CT and MRI in detecting tumors smaller than 2 cm and better specificity when EUS is combined with FNA. CT and MRI probably are, for most institutions, the appropriate test for the initial diagnostic work-up and staging of pancreatic cancer, given their ability to identify and accurately size the primary pancreatic tumor, to identify liver metastases, and to document the extent of vascular invasion. The role of PET has yet to be determined.
DIFFERENTIAL DIAGNOSIS

The differential diagnosis for a hypodense mass on contrast-enhanced CT, or for a hypointense mass on gadolinium-enhanced MRI, includes chronic pancreatitis, unusual islet cell tumors, and lymphoma. The differential diagnosis for a hypoechoic mass on EUS includes neuroendocrine tumors, metastases, and chronic pancreatitis. Most of these conditions do not present with the single finding of a hypodense mass, and associated findings and the clinical history must be considered for an appropriate differential diagnosis to be formed. Conversely, if a patient has a history of painless jaundice and contrast-enhanced CT shows no mass or an isodense mass, then such entities as chronic pancreatitis, bile duct tumors, and ampullary tumors should be considered as more likely, but pancreatic cancer cannot be excluded. Further evaluation with EUS/ERCP with brushings or FNA should be considered and the findings correlated with those of cross-sectional imaging.
Chronic Pancreatitis

The most commonly encountered diagnostic difficulty in the work-up by imaging of a pancreatic mass is the differentiation of pancreatic carcinoma from chronic pancreatitis. Typical findings for chronic pancreatitis, seen in 60%–70% of patients, include diffuse dilatation of the pancreatic duct, atrophy of the pancreatic parenchyma, and extensive calcifications. However, in the 30% of cases in which chronic pancreatitis can present as a focal mass, the differentiation from pancreatic cancer is much more difficult. Although a focal hypodense pancreatic mass is likely to be a ductal adenocarcinoma, it can occasionally be caused by chronic pancreatitis. Uncommonly, an isodense mass can be due to ductal adenocarcinoma. Such problematic cases usually require further work-up, which may include percutaneous or EUS-guided FNA biopsy.
Islet Cell Tumor

Islet cell tumors are histologically vascular tumors with solid sheets of tumor cells that lack any significant fibrous component. The diagnosis is usually suspected clinically if the tumors are functional. Functional tumors usually have the appearance of small, hyperdense, enhancing masses during the early pancreatic parenchymal phase of imaging on contrast-enhanced CT. Similarly, they appear as focal regions of hyperintense signal on the early phase of gadolinium-enhanced, dynamically obtained, TI-weighted sequences on MRI.

However, when these tumors become large, as in the case of nonfunctioning islet cell tumors, regions of cystic degeneration can form that show low density on CT and low signal intensity on TI-weighted images on MRI. The peripheral, viable regions may remain hypervascular, and intense peripheral enhancement can help to distinguish these tumors from ductal adenocarcinoma. Although calcifications occur in only 50%–60% of islet cell tumors, they are rare in ductal adenocarcinoma and can serve as another means of differentiation. Biopsy should be considered in any cases in which diagnosis is problematic.
STAGING OF PANCREATIC DUCTAL ADENOCARCINOMA

Nearly 60%–70% of patients with pancreatic adenocarcinoma have advanced metastatic disease on presentation; therefore, only 30%–40% may be surgical candidates.

Both CT and MRI are very accurate when findings suggest that the tumor stage is advanced (e.g., vascular invasion, peritoneal and/or liver metastases). However, conventional CT (prehelical) and MRI (predynamic enhancement) had been criticized as being insufficiently sensitive for predicting resectability. The more recently developed techniques of multiphase helical CT and dynamic MRI, aided by their improved ability to determine the relationship between tumor and adjacent vessels, has been shown to more accurately predict resectability.

As described previously, how PET is performed, how its data are analyzed, and its role in staging pancreatic cancer are undergoing study at many institutions. A study of 159 patients, 89 with malignant pancreatic lesions and 70 with benign lesions, showed a sensitivity of 70% and a specificity of 95% for liver metastases, and a sensitivity of 49% and a specificity of 63% for lymph node staging. PET detected peritoneal metastases in 25% of cases. Unfortunately, PET is unable to provide such anatomic-based information as vascular involvement. Some studies have instead looked at SUVs as a means of providing prognostic information for pancreatic cancer patients. Their findings suggest that a low tumor SUV is predictive of increased survival.

EUS is limited, by its field of view, to evaluating structures immediately adjacent to the proximal gastrointestinal tract. This modality cannot evaluate liver or peritoneal metastases or tumor extension to the transverse mesocolon, and it can only evaluate nodal groups within its field of view. Its accuracy for nodal involvement increases when it is combined with FNA. Sensitivity and specificity for assessment of vascular involvement by tumor varies with the criteria that are utilized to judge involvement. Although EUS may be useful for evaluating the portal and splenic veins for invasion by tumor, it may be insensitive for evaluating the superior mesenteric vein.
VASCULAR INVOLVEMENT CT Criteria for Vascular Involvement

Current techniques for helical and multidetector CT can predict resectability of tumor in 80%–85% of cases. Criteria (Table 2) have been developed to indicate the probability of vascular involvement on the basis of the relationship between the tumor and adjacent vessels. In our experience, using these criteria, tumor was resectable without venous resection in 95% of patients when either a fat plane (type A) or normal parenchyma (type B) separated tumor from adjacent vessels. When tumor was inseparable from a vessel but the point of contact formed a convexity against a vessel (type C), tumor involvement could not be reliably predicted. When tumor partially encircled a vessel (type D), forming a concavity against a vessel, such as a vein, venous resection was usually necessary to remove the tumor. When tumor completely encircled a vessel (type E) or occluded a vessel, it was not possible to resect tumor with a negative margin. Using these criteria, helical CT has a sensitivity of 76%–88%, a specificity of 63%–100%, and an accuracy of 62%–92% for predicting vascular invasion. Accuracy appeared to be greater when helical, dual-phase (pancreatic parenchymal and portal venous) imaging techniques were used with higher contrast injection rates (3–4 mL/sec).
MRI Criteria for Vascular Involvement

A study by Sironi et al evaluated tumor-vessel contiguity and vessel encasement by pancreatic cancer by MRI. Tumor-vessel contiguity was identified when perivascular fat planes were infiltrated by tumor that extended to within 5 mm of a vessel and encompassed less than 180 degrees of its circumference. Encasement was identified when perivascular fat planes were obliterated by a cuff of tumor that encompassed 180 degrees or more of a vessel's circumference. Findings were confirmed by surgical examination or evaluation of pathology specimens. Accuracy for encasement was as high as 91% on precontrast and 94% on postcontrast spin-echo TI-weighted images. Corresponding accuracies for pre- and postcontrast multiple planar gradient-recalled images were much lower (74% and 76%, respectively). Other studies have shown lower accuracies of 58%–86% for visualization of vascular invasion by MRI.

EUS Criteria for Vascular Involvement

The sensitivity and specificity of EUS for vascular invasion depend on the criteria used for judging vascular invasion. Findings suggestive of vascular involvement include an “irregular venous wall,” presence of tumor in the lumen of a vessel, complete obliteration of a vessel, and/or presence of collateral vessels. In a study of videotapes of 75 patients with pancreatic cancer, in which the portal vein and its confluence and the superior mesenteric vein were evaluated, the sensitivity for venous invasion was 43% and the specificity was 91% when the presence of tumor in the lumen of a vessel, the complete obliteration of a vessel, and the presence of collateral vessels were used as criteria. When “irregular tumor-vessel” relationship was added, sensitivity rose to 62%, but specificity fell to 79%. Another study found the following accuracies for various signs for tumor involvement: irregular venous wall (87%), loss of interface (78%), and proximity of mass to vessel (73%). As noted earlier, although irregular venous wall was the most accurate sign in this study, the sign's low sensitivity rate (47%) was because of its inability to detect superior mesenteric vein invasion (sensitivity of 17%).Other studies have also shown similar rates of accuracy.

LIVER METASTASES

Liver metastases are unfortunately a common finding for patients presenting with pancreatic cancer, especially when cancers arise from the pancreatic body or tail. CT has been shown to have a sensitivity of approximately 75%–80% for liver metastases. A study of 58 patients that compared dual-phase helical CT, rapid-acquisition MRI (with dynamic gadolinium-enhanced, MRCP, and echoplanar sequences), angiography, and ERCP reported an accuracy for detection of liver metastases of 93.5% and 87% for MR1 and CT, respectively. However, sensitivity decreases for both modalities when the size of metastases decreases. CT and MRI have been reported to have sensitivities as low as 26%–42% in earlier studies for small metastases (<2 cm) to the surface of the liver.

Laparoscopic surgery, with laparoscopic ultrasound, may improve the ability to detect small liver metastases and peritoneal implants. However, studies in which high-quality preoperative imaging were utilized have shown that perhaps as few as 5% of patients would derive additional benefit from laparoscopic surgery.

The role of PET in the diagnosis and staging of liver melastases is still being determined. As noted earlier in the study by Diederichs et al, PET was reported to have a sensitivity of 70% and a specificity of 95% for liver metastases. As with CT and MRI, sensitivity for liver metastases decreases as lesion size decreases,with one study reporting a detection rate of 97% for lesions > 1 cm and 43% for lesions =1 cm. False-positive results have been described in patients with severe intrahepatic cholestasis and have been reported in patients with intrahepatic abscesses, and basilar pulmonary metastases.

EUS, because of its limited field of view, can evaluate only a small portion of the liver but can be used for FNA if a lesion is visible by EUS.

The current literature and our experience at our institution indicate that accurate staging probably requires the use of a combination of modalities, such as multiphase CT and/or dynamic MR and EUS. CT and MRI, when properly performed, can provide information on tumor size, vascular involvement, local extension of tumor, and presence of liver metastases. EUS with FNA can increase sensitivity for detecting small lesions and can provide high specificity. Using a combination of these techniques, detection rates of 90%–95%, resectability prediction rates of 85%–90%, and accuracy of 90% for detection of vascular involvement are probably possible. The role of developing technologies, such as PET, has yet to be determined. Advancements are needed to detect smaller lesions (<1 cm), to better predict resectability, and to improve detection of hepatic, peritoneal, and lymph node metastases.

Last updated Jan 2/07