Table 1. Kappa Statistics for Individual Features Standard Features Kappa (95% CI) Rosemont features Kappa (95% CI) Lobularity 0.45 (0.39,0.51) Lobularity with honeycombing (major B) 0.45 (0.31,0.59) Lobularity without honeycombing (minor) 0.16 (0.03,0.28) Hyperechoic foci 0.26 (0.22,0.30) Hyperechoic foci with shadowing (major A) 0.36 (0.19,0.53) Hyperechoic foci without shadowing (minor) 0.21 (0.15,0.26) Hyperechoic strands 0.28 (0.24,0.33) Stranding (minor) 0.29 (0.26,0.33) Cysts 0.38 (0.11,0.64) Cysts (minor) 0.35 (0.10,0.59) MPD dilation 0.60 (0.55,0.65) MPD dilation (minor) 0.61 (0.54,0.68) MPD irregularity 0.49 (0.43,0.54) Irregular MPD contour (minor) 0.50 (0.44,0.55) Visible side-branches 0.35 (0.29,0.42) Dilated side-branches (minor) 0.46 (0.36,0.56) Hyperechoic duct wall 0.26 (0.23,0.30) Hyperechoic duct margin (minor) 0.33 (0.26,0.41) Calcifications 0.74 (0.64,0.85) MPD calculi (major A) 0.64 (0.49,0.80) CIZ95% confidence interval, MPDZmain pancreatic duct Table 2. Kappa Statistics Based on Level of Experience 300-999 pancreatic EUS exams (nZ7) R1000 pancreatic EUS exams (nZ7) Global diagnosis 0.55 (0.41,0.68) 0.62 (0.51,0.79) Standard scoring 0.48 (0.36,0.62) 0.60 (0.49,0.73) Rosemont Scoring 0.64 (0.50,0.76) 0.66 (0.51,0.78) M1418 Endosonographic Features of Gastric Gastrointestinal Stromal Tumors (GISTs): Is It Possible to Differentiate GISTs from Leiomyomas By EUS? Gwang Ha Kim, Dong Uk Kim, Geun-Am Song, Jeong Heo, Kyung Yeob Kim, Seongho H. Choi Background and Aims: Several studies have attempted to differentiate benign and malignant stromal tumors on the basis of their EUS features. However, most of these studies were performed before the concept of gastrointestinal stromal tumors (GISTs) was introduced. The aim of this study was to evaluate the EUS features of gastric GISTs as compared with gastric leiomyomas and then to determine the EUS features that could predict malignant GISTs. Patients and Methods: We evaluated the EUS features in 53 patients with gastric mesenchymal tumors confirmed by the histopathologic diagnosis. The GISTs were classified into the benign and malignant groups according to the histological risk classification. Results: Immunohistochemical analyses demonstrated 7 leiomyomas and 46 GISTs. Inhomogenicity, hyperechogenic spots, a marginal halo and higher echogenicity as compared with the surrounding muscle layer appeared more frequently in the GISTs than in the leiomyomas (P ! 0.05). The presence of at least two of these four features had a sensitivity of 89.1% and a specificity of 85.7% for predicting GISTs. Except for the tumor size and irregularity of the border, most of the EUS features were not helpful for predicting the malignant potential of GISTs. On multivariate analysis, only the maximal diameter of the GISTs was an independent predictor. The optimal size value for predicting malignant GISTs was 35 mm. The sensitivity and specificity with using this value were 92.3% and 78.8%, respectively. Conclusions: The EUS features such as inhomogenicity, hyperechogenic spots, a marginal halo and higher echogenicity as compared with the surrounding muscle layer may help to differentiate GISTs from leiomyomas. Once GISTs are suspected by EUS, surgical treatment should be considered if the size of the tumor is greater than 3.5 cm. EUS features of gastric mesenchymal tumors. A, A gastric leiomyoma is homogenous and its echogenicity is similar to that of the surrounding normal muscle layer. B, A gastric GIST with low risk potential is inhomogenous and its echogenicity is somewhat higher. A marginal halo (arrow) and hyperechogenic spots are seen. M1419 Optimizing EUS Staging of Pancreatic Neuroendocrine Tumors (pNETs) Through FNA Confirmation of Nodal, Vascular, or Hepatic Metastasis and Determination of Adverse Prognostic Markers Ferga C. Gleeson, Michael J. Levy, Suresh T. Chari, Jonathan E. Clain, Amy C. Clayton, Michael Henry, Michael L. Kendrick, Randall K. Pearson, Bret T. Petersen, Elizabeth Rajan, Santhi Swaroop Vege, Naoki Takahashi, Geoffrey B. Thompson, Mark D. Topazian, Kenneth K. Wang Background: Little consideration has been given to the use of EUS for enhancing pNET staging, in particular FNA of nodal or vascular invasion and hepatic metastasis. Such information may be critical to enhance prognostic determination and clinical decision making. Aims: In patients undergoing EUS FNA for a clinically suspected pNET, to determine the: 1.) prevalence of nodal, vascular, or hepatic disease identified by FNA and 2.) clinical and EUS features associated with advanced disease, and adverse prognostic factors associated with disease progression. Methods: A prospectively maintained database was reviewed to identify all patients undergoing EUS FNA pNET staging from 01/01/99 to 10/01/08. Clinical, EUS, cytologic, surgical and follow-up data were abstracted and analyzed by univariate, multivariate and receiver operating curve (ROC) analysis. Results: EUS FNA was performed to stage 108 patients (57 Æ 14 years, 53% female) with suspected pNETs. Non-specific symptoms were seen in 47 (44%), functioning pNETs were diagnosed in 14 (13%), and MENI was present in 12 (11%) patients. EUS FNA demonstrated disease confined to the pancreas in 76 (70%), locally advanced in 24 (22%), and hepatic metastasis in 8 (7%) patients, respectively. Locally advanced disease was based on either nodal (nZ22) or vascular involvement (nZ2). EUS FNA of advanced (locally or hepatic) disease was associated with non-specific symptoms (OR 9.3, 95% CI 1.8-49, pZ0.008), and a dilated pancreatic duct, (OR 19, 95% CI 1.1- 362, pZ0.04). ROC curve analysis indicated that a cutoff pNET size of 2.1cm, (AUCZ0.72: 95% CI, 0.63-0.8) yielded the best power distinction for advanced disease providing a sensitivity and specificity of 70% and 69%. The traditional size threshold of 3.0cm provided a specificity of 87% and sensitivity of 37%. After a median 1 year (IQR 0.18-2.54) follow-up, disease progressed and death occurred in 10 (9%) and 12 (11%) patients, which were associated with tumor size (pZ0.003) and FNA confirmation of local lymphangio or hepatic metastases (pZ0.006). CT identified a pNET in only 49% of patients and was otherwise interpreted as normal (16%), inconclusive (15%), indeterminate pancreatic cyst (12%), or a false diagnosis of adenocarcinoma (9%). Conclusion: Our data demonstrate that EUS enhances pNET staging through FNA of nodal, vascular, and hepatic metastasis. In comparison, helical CT identified only 49% of pNET patients. Pancreatic duct dilation independently predicted advanced disease. Adverse prognostic factors included tumor size, lymphangio-, and hepatic metastasis. Our data suggest an even greater role for EUS FNA in the evaluation of pNET. M1420 Prevalence and Clinical Significance of Multiple Pancreatic Cysts Santo Maimone, Deepak Agrawal, Joseph Willis, Ashley L. Faulx, Gerard Isenberg, Richard C. Wong, Amitabh Chak Background: Diagnostic evaluation of pancreatic cysts may identify the presence of multiple cysts. We hypothesized that multiple pancreatic cysts are associated with branch duct type intraductal papillary mucinous neoplasia (IPMN), given the diffuse nature of the disease. Methods: A retrospective chart review was performed on consecutive patients who presented to a tertiary referral center from 2000-08 and underwent endoscopic ultrasound (EUS) for suspected pancreatic cysts. Cyst descriptions were gathered from EUS, CT, MRI/MRCP, surgical pathology, tumor marker, and cytology reports. All patients with more than one cyst on any imaging study were included in the multiple cyst category. Patients with non-cystic adenocarcinoma as well as those with known pseudocysts were excluded. The cysts were categorized as follows: 1) Definite IPMN: definite communication between the cyst and the pancreatic duct on any imaging or surgical pathology; 2) Not IPMN: when another definitive diagnosis was made, such as serous cystadenoma (SCA) based on CEA !20 ng/mL or pathology; mucinous cystic neoplasm (MCN) based on pathology; or other diagnoses on pathology such as neuroendocrine tumor, metastatic pancreatic lesions, or cystadenocarcinoma; 3) Possible IPMN: cysts of unknown status as well as those with CEA O192 ng/mL but no definitive communication described on imaging (uncategorized mucinous cysts). Cysts with CEA levels between 20 and 192ng/mL and without other identifying information were considered unknown. Results: A total of 145 patients met criteria for inclusion in the study. Forty-eight (33.1%) patients had more than 1 cyst detected in the pancreas. Of these 48 patients, 18 (37.5%) had IPMN, 8 (16.7%) SCA, 1 (2.1%) MCN, 3 (6.3%) uncategorized mucinous, 1 (2.1%) metastatic hypernephroma, and 17 (35.4%) unknown. The presence of multiple cysts on imaging was significantly associated with a diagnosis of IPMN compared to possible IPMN and not IPMN (chi square for trend, p Z 0.0417). The calculated odds ratio for the presence of multiple cysts was 2.96 for definite IPMN vs. not IPMN (95% CI: 1.12, 8.07) and 2.43 for possible IPMN vs. not IPMN (95% CI: 1.05, 5.69). Conclusion: Multiple pancreatic cysts are seen in up to one-third of cases on pancreatic imaging. The presence of multiple cysts is suggestive of a diagnosis of IPMN. This association can help guide the management of patients with multiple cystic lesions of the pancreas. Abstracts www.giejournal.org Volume 69, No. 5 : 2009 GASTROINTESTINAL ENDOSCOPY AB237