EUS is the most sensitive nonoperative imaging test for the detection of benign or malignant pancreatic lesions (Table 30.2) [1–8]. Some studies included benign pancreatic disease and ampullary tumors [1, 2, 5], which may have led to a bias of tumor detection in favor of EUS. In studies that compared EUS and CT, the sensitivity of EUS for mass detection was superior to CT [1–8]. EUS is clearly superior to conventional CT [1–3, 6] and transabdominal ultrasound (US) [1–3, 5]. A few comparative studies between EUS and multidetector-row CT (MDCT) for pancreatic tumors have demonstrated the superiority of EUS for tumor detection compared to 4-row CT. Agarwal et al [7] reported an EUS sensitivity of 100 % for the diagnosis of cancer compared to 86 % for MDCT. Similarly, DeWitt et al [8] reported that the sensitivity of EUS (98 %) was statistically superior to MDCT (86 %) in a cohort of 80 patients with pancreatic cancer. There are relatively sparse comparative data between EUS and MRI for tumor detection with at least one study showing superiority of EUS [4]. Future studies comparing EUS and 3.0 or higher Tesla MRI are needed to define the roles of each for the diagnosis of pancreatic masses.

EUS is particularly useful for the identification of small tumors that have been undetected by other imaging modalities [1, 4, 7, 8]. For tumors ≤ 30 mm in diameter, EUS was found to have a sensitivity of 93 % compared to 53 % for CT and 67 % for MRI [4]. With thinner slice imaging and precisely timed contrast administration coupled with multiplanar reconstruction (often referred to as pancreas protocol), CT may now be able to identify small pancreatic masses that previously were undetected by conventional or even single detector dual-phase imaging [8]. We recommend that EUS be performed in all patients with obstructive jaundice or dual pancreatic and bile duct dilations in whom CT or MRI do not definitively identify a pancreatic lesion, both to detect any tumor and to exclude non-neoplastic diseases.

EUS may fail to identify true pancreatic masses in patients with chronic pancreatitis , a diffusely infiltrating carcinoma, a prominent ventral/dorsal split, or a recent episode (< 4 weeks) of acute pancreatitis. In a study of 80 patients with clinical suspicion of pancreatic cancer and a normal EUS, Catanzaro et al [9] found that no patient with a normal pancreatic EUS developed cancer during a follow-up period of 24 months. Therefore, a normal pancreas by EUS examination essentially rules out pancreatic cancer although follow-up EUS or other studies should be undertaken in the setting of chronic pancreatitis due to impaired visualization. It is also important to remember that acoustic shadowing caused by an indwelling biliary or pancreatic stent may also impede visualization of a small pancreatic mass.

Due to the ability of EUS to provide high-resolution images, there has been increasing interest in using this technique to screen asymptomatic high-risk cohorts for early cancer detection. Canto et al [10] evaluated an EUS-based screening approach in a prospective cohort of 38 asymptomatic individuals with Peutz–Jeghers syndrome or from kindreds with at least two affected relatives with pancreatic cancer . Five benign and one malignant pancreatic lesions were found by EUS. The diagnostic yield for detecting clinically significant pancreatic neoplasms was 5.3 % (2 of 38). Another study found that EUS is superior to MRI among high-risk asymptomatic patients, and may disclose adenocarcinoma and branch duct IPMN during first-time screening in individuals with family history of pancreas cancer or other familial cancer syndromes [11]. A recent consensus statement by the International Cancer of the Pancreas Screening Consortium recommended screening with EUS and/or MRI for the following groups: first-degree relatives (FDRs) of patients with pancreatic cancer from a familial pancreatic cancer kindred with at least two affected FDR, Peutz–Jeghers syndrome, p16, or BRCA2 mutations, and hereditary nonpolyposis colorectal cancer (HNPCC) mutation carriers with ≥ 1 affected FDR [12]. However, the optimal screening modality, interval, need for FNA, and screening abnormalities of sufficient concern for surgery remain unknown and further studies are required to answer these questions.

Autoimmune pancreatitis (AIP) may mimic pancreatic adenocarcinoma and accurate preoperative detection may avoid unnecessary surgery. The EUS morphology of AIP may include diffuse pancreatic enlargement, a focal mass, focal hypoechoic areas, bile duct wall thickening, or peripancreatic lymphadenopathy (Fig. 30.1) [13]. EUS-FNA may demonstrate a nonspecific plasmacytic predominant chronic inflammatory infiltrate but this finding has variable sensitivity and poor specificity. Diagnosis may be confirmed by EUS-guided core biopsies with staining for IgG4 plasma cells [14, 15].

Imaging-based technologies such as contrast-enhanced ultrasonography (CE-EUS) may be able to differentiate pancreatic adenocarcinoma from pancreatic neuroendocrine tumor (pNET) and inflammatory pseudotumors, which can all present as a hypoechoic mass, whereas ductal adenocarcinomas typically demonstrate hypoenhancement, pNET and inflammatory pseudotumors are hyperenhancing or isoenhancing. A recent meta-analysis of 12 studies involving 1139 patients undergoing CE-EUS reported a pooled sensitivity, specificity, and area under the curve receiver operator characteristic (ROC) of 94 %, 89 %, and 0.9732, respectively for pancreatic adenocarcinoma [16]. EUS elastography is another emerging technique based on the different stiffness of benign and malignant tissue. In a meta-analysis of 13 studies involving 1044 patients, the pooled sensitivity, specificity, and ROC was 95 %, 67 %, and 0.90 for elastography differentiating benign from malignant pancreatic masses [17]. However, several limitations to the routine use of these image-based techniques exist and include costs, the lack of both agent availability and expertise with the technique, and need for improved accuracy especially with EUS elastography.

Staging of pancreatic malignancy is done according to the American Joint Committee for Cancer (AJCC) Staging TNM classification, which describes the tumor extension (T), lymph node (N), and distant metastases (M) of tumors, respectively (Table 30.3). Reported accuracies of T staging by EUS range from 62 to 94 % (Table 30.4) [2–4, 6, 8, 18–23]. This wide variation may be due to improved detection of distant metastasis or vascular invasion by MDCT, resulting in less operative management for suspected locally advanced or metastatic disease. The exclusion of such patients may have resulted in the decreased T staging accuracy of some recent studies compared to earlier ones. For the last decade, some tertiary referral centers will attempt to achieve negative surgical margins by surgical resection with or without reconstruction of the portal and/or superior mesenteric vein in patients with venous invasion without thrombosis or occlusion. To better reflect this surgical trend, the 2010 staging criteria in the AJCC Manual—7th edition distinguishes potentially resectable (T3) from unresectable (T4) tumors [24]. Currently, only vascular invasion of the celiac or superior mesenteric arteries is classified as T4 cancer.

Despite the changes in T-staging criteria, nodal (N) metastases have uniformly been classified as absent (N0) or present (N1) across all AJCC editions, including the latest 7th edition. The accuracy of EUS for N-staging of pancreatic tumors ranges from 50 to 86 % [2–4, 6, 8, 19–21]. Various criteria have been proposed for endosonographic features of metastatic lymph nodes including size greater than 1 cm, hypoechoic echogenicity, distinct margins, and round shape. When all four features are present within a lymph node, there is an 80–100 % chance of malignant invasion [25]. However, sensitivity of EUS for malignant lymphadenopathy is often lower, presumably for two reasons. First, most metastatic lymph nodes do not have all four endosonographic features described above. Second, peri-tumoral inflammation and large tumor size may obscure visualization of adenopathy. The specificity of EUS alone for the diagnosis of metastatic adenopathy in pancreatic cancer is 26–100 % [3, 4, 19, 21], although most report specificities above 70 %. It is presumed that the addition of EUS-FNA of suspicious lymph nodes may increase specificity, although little data support this. For tumors involving the head of the pancreas, malignant lymph nodes are removed en bloc with the surgical specimen and accurate detection of these lymph nodes is not essential. However, since preoperative identification and EUS-FNA of celiac nodes may preclude surgery, meticulous survey of this region is critical during staging of all pancreatic tumors. Mediastinal lymph node metastases occur in a minority of patients, and thus, a brief survey of this region may be helpful during staging of pancreatic lesions.

Although early studies found EUS to be superior to conventional CT for tumor [3, 4] and nodal [2–4] staging of pancreatic cancer , most recent studies have found that the two are equivalent for both tumor [19, 21] and nodal staging [8, 19, 21]. Similarly, early experience reporting on the superiority of EUS over MRI [3, 4] has been replaced by more recent data that have found no difference [21, 23, 26]. Clearly, the initial advantage demonstrated by EUS over other imaging modalities for the staging of pancreatic tumors has narrowed considerably. Future studies that compare EUS to MDCT and higher Tesla MRI are needed to confirm these findings and further define the role of EUS for the locoregional staging of pancreatic tumors.

For the detection of non-nodal metastatic cancer, CT and MRI are superior to EUS due to both anatomic limitations of normal upper gastrointestinal anatomy and the limited range of EUS imaging. However, EUS still has an important role in the evaluation of hepatic metastasis in the left or caudate lobe and malignant ascites, both of which may be accessible by EUS-FNA (Fig. 30.2). Identification of liver metastases or malignant ascites by EUS-FNA may preclude surgical resection and is associated with poor survival following diagnosis [27].

Interpretation of data regarding the accuracy of EUS for vascular invasion is difficult for several reasons. First, there is little histologic correlation with intraoperative findings regarding vascular invasion in most studies. Second, there is no established consensus among endosonographers for the optimal criteria to utilize for the determination of vascular invasion. Consequently, multiple criteria have been proposed by various authors for this indication.

For overall vascular invasion, the accuracy of EUS ranges from 68 to 93 % [6, 21, 26, 28]. Sensitivity and specificity of EUS for malignant vascular invasion range from 42 to 91 % and 89–100 %, respectively [6, 21, 26, 28]. Although some have reported EUS as more accurate [6] than CT for vascular invasion, others report the opposite [21, 26]. Overall accuracy of MRI is reportedly equivalent [21] or superior [26] to EUS.

The sensitivity of EUS for tumor invasion of the portal vein (PV) or PV confluence is 60–100 % [2, 5, 19, 20, 29] with most studies demonstrating sensitivities over 80 % (Fig. 30.3). The sensitivity of EUS for PV invasion is consistently superior to that of CT [2, 5, 19, 20]. For the superior mesenteric vein (SMV), superior mesenteric artery (SMA), and celiac artery, the sensitivity of EUS is only 17–83 % [28], 17 % [30], and about 50 % [2, 20], respectively. The sensitivity of CT for staging the SMA [19, 30] and celiac artery [2, 20] appears to be better than that of EUS. EUS staging of the superior mesenteric vessels may be difficult due to either the inability to visualize the entire course of the vessel or the obscuring of these vessels by a large tumor in the uncinate or inferior portion of the pancreatic head [29]. This is in contrast to the splenic artery and vein which are generally easily seen and staged well by EUS [20, 29]. Until further conclusive data become available, the assessment of tumor resectability should be done by both EUS and CT (or MRI) rather than by EUS alone.

Several studies have attempted to describe the accuracy of various endosonographic features to assess vascular invasion by malignant pancreatic tumors. Using the criteria ‘abnormal contour, loss of hyperechoic interface, and close contact,’ Rosch et al [2] found a sensitivity, specificity, and accuracy of 91 %, 96 %, and 94 %, respectively for invasion of the portal vein using these criteria. In blinded videotape review [29], these same authors found that no single criterion (irregular mass–vessel relationship, visualization of tumor in vessel lumen, complete vascular obstruction, collateral vessels) could predict portal venous invasion with a negative predictive value exceeding 35 % while positive predictive value was over 80 % for only complete vascular obstruction and collateral vessels. The latter two criteria each demonstrated a specificity of 94 % for vascular invasion. There exists a tradeoff between various criteria for sensitivity and specificity for vascular invasion. However, criteria with the highest specificity are needed to optimize selection of patients most likely to benefit from surgical exploration. Therefore, the findings of complete vascular obstruction, venous collaterals, and visible tumor within the vessel are the preferred criteria for the assessment of vascular invasion.

Complete surgical removal of pancreatic cancer with negative histopathologic margins (R0 resection) is the only potentially curative treatment and is an independent predictor of postoperative survival [31]. Therefore, the principle role of preoperative evaluation is to accurately identify patients with resectable disease who may benefit from surgery while avoiding surgery in patients with suspected unresectable disease (Fig. 30.4).

In a pooled analysis of nine studies involving 377 patients, the sensitivity and specificity of EUS for resectability of pancreatic cancer was 69 % and 82 %, respectively [6, 8, 18, 21, 26, 28, 32–34]. Ranges of reported sensitivities and specificities were 23–91 % and 63–100 %, respectively. Overall EUS accuracy for tumor resectability was 77 %.

Since most studies have reported that EUS is similar to both CT and MRI for the assessment of resectability, some authors have proposed that optimal preoperative imaging of pancreatic cancer requires the use of multiple modalities. Using a decision analysis, Soriano et al [21] found that accuracy for tumor resectability was maximized and costs were minimized when CT or EUS was performed initially followed by the other test in those with potentially resectable neoplasms. Ahmad et al. [18] proposed that although EUS and MRI individually are not sensitive for tumor resectability, their use together may increase positive predictive value of resectability compared to either test alone. When surgery is performed only when MDCT and EUS agree on tumor resectability, DeWitt et al. [8] reported a nonsignificant trend toward improved accuracy of resectability compared to either study alone. However, a study by Bao et al [35] found that MDCT was a better predictor of resectability than EUS, although the performance of EUS improved in patients without biliary stents. From a practical standpoint, the actual role of EUS in staging of pancreatic cancer will depend on its availability, referral patterns, and local expertise.

EUS-guided fine needle aspiration (EUS-FNA) is the currently preferred method to sample pancreatic mass lesions and has largely replaced intraoperative sampling or biopsies under CT or US guidance. EUS-FNA is highly accurate and two recent meta-analyses have reported sensitivity and specificity in the range of 85–89 % and 96–98 %, respectively [36, 37]. However, the diagnostic accuracy of EUS-FNA may be impaired in the setting of chronic pancreatitis. Fritscher-Ravens et al. [38] found that in a series of 207 consecutive patients with focal pancreatic lesions, the sensitivity of EUS-FNA for the diagnosis of malignancy in patients with normal parenchyma (89 %) was superior to those with parenchymal evidence of chronic pancreatitis (54 %). The presence of chronic pancreatitis may impair the visualization of tumors endosonographically or hinder the cytologic interpretation of pancreatic biopsy, thus reducing sensitivity.

At most tertiary referral centers, rapid on-site evaluation (ROSE) with cytopathology review is available to provide immediate feedback to the endosonographer about the quality of EUS-FNA specimens obtained. On-site review was found to correlate highly with the final diagnosis and can improve diagnostic certainty [39]. We recommend that hospital and personnel resources be utilized when feasible to provide ROSE.

Occasionally, ROSE of a suspected pancreatic cancer demonstrates insufficient tissue to confirm malignancy. This may be due to tumor necrosis (particularly with larger tumors), fibrosis, or hypervascularity. Yield may be increased by ‘fanning the lesion’ using different angles of scope deflection in order to sample the peripheral parts of the lesion with more viable tumor [40]. Increasing the number of passes may also overcome this problem, but the additional yield typically plateaus at seven passes and the amount of blood in the aspirate may increase with additional passes [41]. In this situation, avoiding suction and switching to a smaller gauge needle could help limit the amount of blood in the specimen. Finally, EUS-guided core biopsy may be considered in cases when immediate cytology review reveals insufficient material or inconclusive diagnosis although evidence supporting this approach is limited.