Mural nodules are predictors of malignancy in mucusproducing. Histologic and Imaging Features of Mural Nodules in Mucinous Pancreatic Cysts.

Transcription

1 CLINICAL GASTROENTEROLOGY AND HEPATOLOGY 2012;10: Histologic and Imaging Features of Mural Nodules in Mucinous Pancreatic Cysts NING ZHONG,*, LIZHI ZHANG, NAOKI TAKAHASHI, VLADISLAV SHALMIYEV,* MARCIA IRENE CANTO, JONATHAN E. CLAIN,* JOHN C. DEUTSCH, # JOHN DEWITT,** MOHAMAD A. ELOUBEIDI, FERGA C. GLEESON,* MICHAEL J. LEVY,* SHAWN MALLERY, MASSIMO RAIMONDO, ELIZABETH RAJAN,* TYLER STEVENS, and MARK TOPAZIAN* *Department of Internal Medicine, Division of Gastroenterology and Hepatology, and Departments of Pathology, and Radiology, Mayo Clinic, Rochester, Minnesota; Qilu Hospital of Shandong University, Shandong, China; Division of Gastroenterology and Hepatology, Sol Goldman Pancreatic Cancer Research Center, Johns Hopkins Medical Institutions, Baltimore, Maryland; # Department of Gastroenterology and Cancer Center, Essentia Health Systems, Duluth, Minnesota; **Department of Gastroenterology and Hepatology, Indiana University Medical Center, Indianapolis, Indiana; Division of Gastroenterology, University of Alabama in Birmingham, Birmingham, Alabama; Park Nicollet Clinic, Minneapolis, Minnesota; Department of Internal Medicine, Division of Gastroenterology and Hepatology, Mayo Clinic, Jacksonville, Florida; and Department of Gastroenterology and Hepatology, Cleveland Clinic, Cleveland, Ohio BACKGROUND & AIMS: Mural nodules predict malignancy within pancreatic cysts, but it is not clear whether endoscopic ultrasound (EUS) and computed tomography (CT) accurately identify nodules. We assessed images and the histology of mural nodules in branch duct intraductal papillary mucinous neoplasms (BD-IPMNs) and mucinous cystic neoplasms (MCNs) and identified criteria to distinguish mural nodules from mucus. METHODS: We reviewed pathology specimens and EUS and CT images from consecutive patients with resected BD-IPMNs or MCNs. A blinded interobserver study of the EUS images was then conducted to identify features that distinguished nodules from mucus. After education about these features, the raters interpreted the EUS images again. RE- SULTS: On the basis of histologic analysis, 22 of 57 cases had epithelial nodules. Cancer or high-grade dysplasia was found in 23% of cysts with nodules versus 3% without nodules (P.02). On the basis of reports, EUS detected epithelial nodules with 75% sensitivity and 83% specificity, whereas these values were 24% and 100%, respectively, for CT. Mucus accounted for 65% of intracystic lesions detected by EUS and was often diagnosed by using change in body position and fine-needle aspiration. Interobserver analysis identified 3 features that were detected by EUS (echogenicity, edge, and rim) that distinguished mucus from epithelial nodules. The diagnostic accuracy of the raters improved from a mean of 57% to 79% after education about these features (P.004); accuracy was 90% when all 3 features of mucus were present. CONCLUSIONS: Malignancy is associated with epithelial nodules in BD-IPMNs and MCNs, but most echogenic lesions detected in cysts by EUS are mucus. Knowledge of features that discriminate mucus from mural nodules improves the diagnostic accuracy of EUS. Keywords: Pancreas; Radiology; Pancreatic Cancer; Endosonography. Mural nodules are predictors of malignancy in mucusproducing cystic neoplasms of the pancreas, including intraductal papillary mucinous neoplasm (IPMN) and mucinous cystic neoplasm (MCN). 1,2 The presence of a mural nodule is a consensus criterion for resection of these cysts. 3,4 Although the histologic finding of an epithelial nodule in a resected cyst is strongly associated with malignancy, 5 mural nodules (also called protruding lesions, epithelial nodules, echogenic lesions, and epithelial projections) are diagnosed by imaging studies, and the accuracy of preoperative imaging modalities for diagnosis of these lesions varies. Studies relying on computed tomography (CT) and magnetic resonance imaging show a relationship between mural nodules and malignancy, 6,7 but studies primarily using endoscopic ultrasound (EUS) do not confirm this finding. 8,9 There are few studies directly comparing the imaging and histologic features of mural nodules. 10,11 Furthermore, it is uncertain which imaging modality best detects mural nodules. EUS is often used but may have low specificity, 12 and mucus in a pancreatic cyst might have a similar appearance. The aims of this study were to evaluate the histology of mural nodules in resected pancreatic branch duct (BD)-IPMN and MCN, to assess the accuracy of EUS and CT in the diagnosis of mural nodules in direct comparison with histology, and to identify imaging features that distinguish epithelial mural nodules from mucus in pancreatic cysts. Methods The first part of the study was a retrospective review of consecutive resected pancreatic cystic neoplasms. Pathology, EUS, and CT were separately and blindly reviewed using standardized data forms, and EUS and CT were compared with the histologic gold standard. By using EUS images from the resected cases, a prospective interobserver study was then conducted to identify specific echofeatures associated with epithelial mural nodules versus mucus. Patients From the pathology database at one institution (Mayo Clinic, Rochester, MN), we identified 57 consecutive cases (26 Abbreviations used in this paper: BD-IPMN, branch duct intraductal papillary mucinous neoplasm; CT, computed tomography; EUS, endoscopic ultrasound; FNA, fine-needle aspiration; HGD, high-grade dysplasia; IPMN, intraductal papillary mucinous neoplasm; MCN, mucinous cystic neoplasm by the AGA Institute /$36.00 doi: /j.cgh

2 February 2012 MURAL NODULES IN PANCREATIC CYSTS 193 Figure 1. Histologic criteria of epithelial mural nodules in BD- IPMN and MCN. (A) Nodule is grossly visible on the histology slide (original magnification, 1 ). (B) A stalk (arrow) is present connecting to the cyst wall, with glands on both sides of the stalk (IPMN only) (original magnification, 2 ). (C) Detached fragments of complex papillary mucosa are present, with stalk (arrow) or fibrotic cores (IPMN only) (original magnification, 2 ). (D) Intracystic nonpapillary excrescences (arrow) with broad bases connecting to the cystic wall, which also contain ovariantype stroma and entrapped glands instead of fibrovascular cores (MCN only) (original magnification, 2 ). male; mean age, years) of resection of BD-IPMN (37 cases) or MCN (20 cases) between January 1, 2006 and December 31, Cases with gross or microscopic main duct IPMN were excluded. Among the 57 cases, 41 (33 BD-IPMN; 22 male; mean age, years) underwent preoperative EUS, and 44 (28 BD-IPMN; 22 male; mean age, years) had preoperative contrast-enhanced multidetector CT images available for review. Histologic Evaluation One pathologist (L.Z.) reviewed all specimens for the presence and histologic features of epithelial mural nodules, while blinded to clinical, laboratory, and imaging findings and the original pathology reports. Epithelial nodules were generally impossible to discern in residual gross specimens because of collapse of the cyst lumens during fixation and prior sectioning of the specimens. As shown in Figure 1, we accepted any of 4 histologic features as proof of the presence of an epithelial mural nodule. Each case was assessed for the presence, number, and maximum diameter of nodules, maximum degree of dysplasia in the nodule and in flat cyst mucosa, the histologic subtype of IPMN, 13,14 and presence of acellular mucin pools. Cysts were deemed malignant if either invasive cancer or high-grade dysplasia (HGD) was present. Endoscopic Ultrasound and Computed Tomography Review One investigator (V.S.) reviewed the preoperative EUS reports for the presence of an echogenic lesion in the cyst, while blinded to other data. Echogenic lesions in cysts were categorized as suspected mural nodules when EUS reports used the terms mural nodule, solid component, mass in the cyst, thickened mucosa, or papillary projection and were categorized as mucus when EUS reports used the terms intracystic debris, mucus, or mucus nodule. The size and echofeatures of echogenic lesions in cysts were noted, as well as maneuvers performed during EUS to distinguish mural nodules from mucus, including change in patient body position and EUS fine-needle aspiration (FNA). A radiologist (N.T.) reviewed all available CT images, while blinded to all other data. Cystic lesions were rated for the presence of mural nodules, and the size and features of nodules were noted, as well as contrast enhancement of the nodule. The original CT reports were independently reviewed. After blind review, an unblinded review of histology, CT images, and EUS images was jointly conducted by L.Z. and M.T. for all cases in which an epithelial nodule (pathology), mural nodule (CT), or intracystic echogenic lesion (EUS) was identified. This review resulted in a change in pathology diagnosis in one case, in which EUS had identified a lesion, and an epithelial nodule was detected on re-review of histology slides. In 17 cases in which EUS showed an intracystic lesion but no epithelial nodule was seen histologically, the imaging findings were deemed to have been mucus. Endoscopic Ultrasound Interobserver Study Of the 57 consecutive resected cases, 24 had stored EUS images that were adequate for review, including 17 with an echogenic lesion. Five additional nonresected cases were identified from the same time period in which an intracystic echogenic lesion was proved to be mucus during EUS by patient body position change or EUS FNA (features validated during the retrospective review phase of the study). EUS images and video clips from these 29 cases formed the derivation set, which was blindly reviewed by 10 experienced endosonographers. They

3 194 ZHONG ET AL CLINICAL GASTROENTEROLOGY AND HEPATOLOGY Vol. 10, No. 2 Figure 2. Pathology and imaging review of 57 consecutive resected mucinous pancreatic cystic neoplasms. EN, epithelial nodule by histology review; malignancy includes both HGD and invasive carcinoma. rated each case for the presence and features of an echogenic lesion in the cyst, including the lesion s shape (round, oblong, irregular), echogenicity compared with adjacent soft tissue (hypoechoic, isoechoic, hyperechoic), edge (smooth, irregular), echotexture (heterogeneous, homogeneous), hyperechoic rim (absent, present), calcification (absent, present), and Doppler flow in the lesion (absent, present). Images showing EUS FNA or movement of echogenic lesions after patient body position change were not included. Raters also rated septa and the cyst wall and indicated their diagnosis of echogenic lesions (epithelial nodule, invasive malignancy, mucus, debris, indeterminate, other). The raters diagnoses were categorized as epithelial nodule, mucus, or unknown. Data from the derivation set were used to identify predictors of gold standard diagnosis for echogenic lesions in cysts. Subsequently, a validation set was constructed that included the 29 cases of the derivation set arranged in different order as well as 21 additional cases. Of the 21 additional cases, 16 were resected between 2003 and 2005, and preoperative EUS had shown an echogenic lesion in the cyst. Blinded histology review of all 16 cases identified an epithelial nodule in 6. An additional 5 nonresected cases in which an echogenic lesion within a cyst was proved to be mucus by body position change or EUS FNA were also included. Six months after rating the derivation set, the raters were educated about the results of the derivation set analysis. They then rated the validation set. Statistical Analysis All statistical analyses were performed with software JMP 8.0 and SAS 9.1 (SAS Institute Inc, Cary, NC). P values.05 indicated statistically significant differences. We used the 2 test or the Fisher exact test to compare rates between categorical factors, depending on the sample size. Interobserver reliability for various study parameters was assessed with a multirater kappa value, which estimates the level of interobserver agreement. Univariate logistic regression was used to test for EUS features significantly associated with the pathologic diagnosis of mucus versus epithelial nodule. The differences between 2 groups of paired continuous data were calculated by Wilcoxon signed rank test. Interobserver analyses were performed by the Mayo Department of Biomedical Statistics and Informatics. Results Results of the retrospective review of pathology, EUS, and CT are summarized in Figure 2. Pathology In the 57 consecutive resected cases (37 BD-IPMNs, 20 MCNs), at least 1 epithelial mural nodule was identified in 22 cases (39%). Malignancy was found in 23% of these 22 cases (1 invasive carcinoma and 4 HGD) versus 3% of the 35 cases without epithelial nodules (1 case of HGD) (P.024). In 3 of 4 cases with both HGD and a nodule, HGD was found only in flat cyst mucosa separate from the nodule. Altogether, 83% of cysts containing invasive cancer or HGD also contained an epithelial mural nodule. The median diameter of nodules (and number of cases) containing invasive cancer, HGD, intermediate-grade dysplasia, and low-grade dysplasia was 30 mm (1), 18

4 February 2012 MURAL NODULES IN PANCREATIC CYSTS 195 Figure 3. Epithelial nodules detected by EUS and CT. (A) The maximum diameter of epithelial nodules detected and missed by EUS and CT. (B) A 2-mm epithelial nodule missed by EUS (original magnification, 2 ). mm (1), 19 mm (4), and 5 mm (16), respectively. Additional pathology features of epithelial nodules are shown in Supplementary Table 1. In 2 additional cases an invasive adenocarcinoma was present in the resected specimen. These cancers had been identified preoperatively and were the indication for resection; the BD- IPMN lesions within both specimens did not contain mural nodules or malignancy. Preoperative Endoscopic Ultrasound For the 41 consecutive resected cases that underwent preoperative EUS, at least 1 intracystic echogenic lesion was identified in 26 EUS reports (63%) by using radial and/or linear echoendoscopes. Nine of these 26 (35%) proved to be epithelial nodules, and 17 (65%) were mucus. In an additional 3 cases EUS showed no intracystic echogenic lesion, but pathology demonstrated an epithelial nodule. All 3 missed epithelial nodules were less than 5 mm in diameter, significantly smaller than the 9 epithelial nodules detected by EUS (P.04) (Figure 3). Fourteen of the 26 echogenic lesions were suspected to be mural nodules by the original endosonographer, 9 of which actually correlated with epithelial nodules on pathology review. The other 12 intracystic echogenic lesions seen by EUS were suspected to be mucus by the original endosonographer, and none of these cases actually contained an epithelial nodule on pathology review. EUS features of cysts with and without mural nodules are shown in Supplementary Table 2. When compared with histology, the mere presence of an echogenic lesion in the cyst was a poor predictor of an epithelial nodule, with sensitivity and specificity of 75% and 41%, respectively. However, the sensitivity and specificity of the preoperative EUS diagnosis of a mural nodule were 75% and 83%, respectively. Cancer or HGD was found in 3 of 14 (14%) resected lesions with an EUS diagnosis of a mural nodule versus 0 of 27 (0%) of those without a mural nodule by EUS (P.03). For diagnosis of intracystic mucus, the sensitivity of EUS was 41%, and the specificity was 100%. Confidence intervals, positive predictive values, and negative predictive values are provided in Supplementary Table 3. The high specificity of preoperative EUS for diagnosis of mucus was associated with various strategies used to distinguish mural nodules from mucus during EUS examinations. Blood flow within an echogenic lesion was detected in 2 cases by color Doppler, and both were epithelial mural nodules by histology. In 6 cases, movement of the echogenic lesion within the cyst was noted during EUS after changing the patient s body position from left to right decubitus; all were deemed mucus after pathology review. A total of 17 patients underwent EUS FNA of their cystic lesions. In 2 cases, the intracystic echogenic lesion moved to a different part of the cyst after the needle puncture, in another case the lesion detached from the cyst wall and moved with the tip of the needle, and in another case the lesion disappeared after aspiration and saline lavage of the cyst. In all 4 of these cases no epithelial nodule was present in the cyst on pathology review. In total, movement of a lesion with body position change or FNA correctly diagnosed mucus in 10 of 10 cases (100%). Preoperative Computed Tomography In the 44 cases with available contrast-enhanced multidetector CT images, median slice thickness was 2 mm (range, 2 5 mm). At least 1 mural nodule was diagnosed in 4 cases by original CT reports, and a nodule was present on final pathology review in all 4 cases. The sensitivity and specificity of the original CT reading for diagnosis of an epithelial nodule were 24% and 100%, respectively (Supplementary Table 3). Cancer or HGD was found in 2 of 4 (50%) of resected cysts with a CT diagnosis of a mural nodule, compared with 1 of 40 (3%) of those without a nodule by CT (P.02). Retrospective blinded review of CT images by one radiologist identified mural nodules in 11 cases, 8 of which actually contained epithelial nodules on pathology review. In an additional 9 cases, retrospective CT review failed to detect epithelial nodules identified by pathology. The sensitivity and specificity of retrospective CT review for epithelial nodules were 47% and 89%, respectively (Supplementary Table 3). CT did not identify mucus in cysts. The agreement between retrospective CT review and EUS for diagnosis of epithelial nodules in the 31 cases where both tests were performed was limited (kappa 0.31). Endoscopic Ultrasound Interobserver Study Among 10 endosonographers, there was generally poor accuracy for diagnosis of intracystic echogenic lesions as epithelial nodule versus mucus when compared with histology (median rater accuracy, 57%; range, 38% 95%). Interobserver agreement was also poor (multirater kappa, 0.09; 95% confidence interval, ). Although overall diagnostic accuracy was poor, there was good agreement between raters regarding 5 sonographic features of echogenic lesions in pancreas cysts. Three of these features (lesion is hypoechoic compared with adjacent soft tissue, lesion is smooth-edged, lesion has a hyperechoic rim) were identified significantly more often in mucus

5 196 ZHONG ET AL CLINICAL GASTROENTEROLOGY AND HEPATOLOGY Vol. 10, No. 2 Figure 4. EUS features of mucus. (A, B) Two cases of mucus with typical features: hypoechoic compared with adjacent soft tissue, smooth-edged, with a hyperechoic rim. (C) Mucus with only 2 features: hypoechoic compared with adjacent soft tissue, partial hyperechoic rim. (D) Histology of the case in panel C shows acellular mucus. (E, F) Mucus with 3 typical features, which moved after change of patient position (F). Arrows indicate mucus. than in epithelial nodules (Figures 4 and 5, Supplementary Table 4). When nonresected cases were excluded, 2 of the 3 features retained statistical significance, with a trend for the third feature. Kappa ranged from for these 3 features, and on average, 85% 90% of raters agreed on the presence or absence of these features per case. The raters were informed of the 3 echofeatures of intracystic echogenic lesions that were associated with a diagnosis of mucus and then blindly rated the validation set. The mean diagnostic accuracy for echogenic lesions in cysts (epithelial nodule vs mucus) improved from 57% to 79% (range, 67% 95%) (P.004) for the 29 cases included in both the derivation and validation sets and was 76% for the 21 new cases (range, 63% 87%). Accuracy improved for 9 of the 10 raters and was unchanged for one. Interobserver kappa also improved from 0.09 ( ) for the derivation set to 0.44 ( ) for the validation set. When nonresected cases were excluded, the results were similar. Diagnostic accuracy for cases in both the derivation and validation sets improved from 59% to 76% (range, 67% 94%) (P.004) and was 71% for new cases, and interobserver kappa improved from 0.10 to In the validation set, 0, 1, 2, or 3 features of mucus were rated as present 23%, 21%, 17%, and 39% of the time, respectively, with mean rater diagnostic accuracy of 71% (range, 50% 85%), 59% (40% 82%), 71% (50% 100%), and 90% (71% 100%), respectively. Discussion Histologically identified mural nodules are predictors of malignancy in BD-IPMN and MCN 2,11 but the accuracy of clinical diagnosis of mural nodules varies. Although series that use CT, magnetic resonance imaging, or contrast-enhanced ultrasound for diagnosis of mural nodules in BD-IPMN have shown a strong association between mural nodules and cancer 6,10,11 series relying mainly on EUS have failed to show a correlation between nodules and malignancy. 8,9 A previous study comparing EUS with histology in IPMN found the EUS specificity for protruding lesions in pancreas cysts was only 33% 12 and in our study the mere presence of an intracystic echogenic lesion was also a poor predictor of an epithelial nodule on pathology, because most such lesions proved to be mucus. We found that specific maneuvers performed during EUS enhanced the accuracy of diagnosis and resulted in a specificity of 83%. Preoperative EUS diagnosis of a mural nodule was a statistically significant indicator of malignancy in this series. In addition, we identified 3 specific echofeatures that differentiate mucus from epithelial nodules in pancreatic cysts. These findings might refine the use of consensus criteria for resection of BD-IPMN 3 by limiting the number of resections performed for mural nodules that are actually mucus. Because mural nodules are diagnosed by imaging modalities, few articles have studied the pathology of mural nodules in BD-IPMN and MCN ,15 The histology of mural nodules differs significantly between BD-IPMN and MCN, but we included both types of cystic neoplasm in this study. Preoperatively, it might be difficult to distinguish BD-IPMN from MCN, and mural nodules in these neoplasms have similar imaging characteristics and clinical significance. As reported by others 6,10,11 we found that invasive cancer and HGD were associated with epithelial nodules, but we noted that HGD was usually present in flat cyst mucosa separate from the epithelial nodule itself. Mural nodules are usually identified by CT, magnetic resonance imaging, or EUS. Previous studies suggest that EUS is more sensitive and CT more specific for diagnosis of mural nodules. 12 We found that clinical CT interpretation has poor sensitivity for diagnosis of histologically confirmed epithelial

6 February 2012 MURAL NODULES IN PANCREATIC CYSTS 197 Figure 5. EUS features of mural nodules. EUS (A, C, E) and matching histology (B, D, F) from 3 cases. Epithelial mural nodules are usually isoechoic or hyperechoic compared with adjacent soft tissue, are not smooth-edged, and do not have a hyperechoic rim. Arrows indicate epithelial nodules. nodules. Retrospective interpretation of CT images by a radiologist interested in pancreatic disease was still not as sensitive as EUS. The relatively lower sensitivity of multidetector CT indicates that it might not be an ideal screening and surveillance method for patients with mucus-producing cystic neoplasms in whom identification of a mural nodule is likely to change management. We found that EUS missed some nodules less than 5 mm in diameter, perhaps because of tangential imaging or failure to image the entire cyst. In our series the specificity of preoperative EUS for diagnosis of epithelial nodules was substantially higher than previously reported. 12 This might relate to diagnostic maneuvers performed during EUS to distinguish mucus from epithelial nodules, including change in patient body position and EUS FNA. These maneuvers accurately distinguished mucus from epithelial nodules in histologically confirmed cases. Changes in body position can be easily performed during EUS with minimal risk and should be considered a diagnostic adjunct in patients with intracystic echogenic lesions. If FNA is performed, maneuvers such as dislodging or moving an echogenic lesion with the needle tip or performing saline lavage of the cyst to disperse mucus might also be warranted. EUS interpretation is often considered highly subjective. Nonetheless, we identified and validated 3 endosonographic features that differentiate epithelial nodules from mucus in pancreas cysts. Knowledge of these features improved the diagnostic accuracy of expert endosonographers and also mitigated the subjectivity of EUS interpretation, substantially improving interobserver agreement. These features were most accurate when all 3 features of mucus were present. By applying these criteria and using additional diagnostic maneuvers as needed (body position change and FNA), EUS is likely to yield high overall accuracy for diagnosis of epithelial nodules and might become the preferred test for mural nodule detection. There were several limitations to this study. We studied resected cystic neoplasms, and our results might not be generalizable to all pancreatic cysts. The histologic criteria we established for diagnosis of epithelial nodules have not been independently validated. It was not possible to independently confirm that the lesions identified on imaging and histology were the same. The data regarding diagnostic accuracy of imaging modalities came from one institution, and sample size was limited. CT scans were obtained by using different protocols and slice thickness; scans obtained with optimal current methods might yield better performance. We did not perform contrast-enhanced EUS, which has been shown by other investigators to improve diagnosis of mural nodules. 10 We included some nonresected cases in the interobserver portion of the study to increase the sample size; the criteria used to diagnose mucus in these cases (changes in body position and EUS FNA) were highly accurate when assessed in resected cases during the retrospective portion of the study. Finally, it seems likely that in some cases mucus might overlay or mix with epithelial nodules, compounding the difficulty of imaging diagnosis, and we were unable to distinguish these cases from those with epithelial nodules alone during our histology review. In conclusion, mural nodules in pancreatic cysts are predictors of malignancy, although HGD is often found in flat cyst mucosa separate from the mural nodule. Most echogenic lesions seen in cysts during EUS are mucus, not true epithelial mural nodules. Three EUS features discriminate mucus from mural nodules in pancreas cysts; mucus is usually hypoechoic compared with adjacent soft tissue, smooth-edged, and has a hyperechoic rim. In addition, body position change and EUS FNA are useful for distinguishing mucus from mural nodules during EUS.

7 198 ZHONG ET AL CLINICAL GASTROENTEROLOGY AND HEPATOLOGY Vol. 10, No. 2 Supplementary Material Note: To access the supplementary material accompanying this article, visit the online version of Clinical Gastroenterology and Hepatology at and at doi: / j.cgh References 1. Khalid A, Brugge W. ACG practice guidelines for the diagnosis and management of neoplastic pancreatic cysts. Am J Gastroenterol 2007;102: Yamao K, Yanagisawa A, Takahashi K, et al. Clinicopathological features and prognosis of mucinous cystic neoplasm with ovarian-type stroma: a multi-institutional study of the Japan Pancreas Society. Pancreas 2011;40: Tanaka M, Chari S, Adsay V, et al. International consensus guidelines for management of intraductal papillary mucinous neoplasms and mucinous cystic neoplasms of the pancreas. Pancreatology 2006;6: Rautou PE, Lévy P, Vullierme MP, et al. Morphologic changes in branch duct intraductal papillary mucinous neoplasms of the pancreas: a midterm follow-up study. Clin Gastroenterol Hepatol 2008;6: Rodriguez JR, Salvia R, Crippa S, et al. Branch-duct intraductal papillary mucinous neoplasms: observations in 145 patients who underwent resection. Gastroenterology 2007;133: Schmidt CM, White PB, Waters JA, et al. Intraductal papillary mucinous neoplasms: predictors of malignant and invasive pathology. Ann Surg 2007;246: Akita H, Takeda Y, Hoshino H, et al. Mural nodule in branch duct-type intraductal papillary mucinous neoplasms of the pancreas is a marker of malignant transformation and indication for surgery. Am J Surg 2011;202: Mimura T, Masuda A, Matsumoto I, et al. Predictors of malignant intraductal papillary mucinous neoplasm of the pancreas. J Clin Gastroenterol 2010;44:e224 e Lévy P, Jouannaud V, O Toole D, et al. Natural history of intraductal papillary mucinous tumors of the pancreas: actuarial risk of malignancy. Clin Gastroenterol Hepatol 2006;4: Ohno E, Hirooka Y, Itoh A, et al. Intraductal papillary mucinous neoplasms of the pancreas: differentiation of malignant and benign tumors by endoscopic ultrasound findings of mural nodules. Ann Surg 2009;249: Kurihara N, Kawamoto H, Kobayashi Y, et al. Vascular patterns in nodules of intraductal papillary mucinous neoplasms depicted under contrast-enhanced ultrasonography are helpful for evaluating malignant potential. Eur J Radiol [Epub ahead of print]. 12. Nakagawa A, Yamaguchi T, Ohtsuka M, et al. Usefulness of multidetector computed tomography for detecting protruding lesions in intraductal papillary mucinous neoplasm of the pancreas in comparison with single-detector computed tomography and endoscopic ultrasonography. Pancreas 2009;38: Bosman FT, Carneiro F, Hruban RH, et al. WHO classification of tumours of the digestive system. 4th ed. Lyon, France: International Agency for Research on Cancer, 2010: Sanada Y, Kunita S, Yoshida K. Comparison of histologic subtype and growth pattern in intraductal papillary-mucinous carcinoma of the pancreas. Oncol Rep 2008;19: Yamada Y, Mori H, Matsumoto S. Intraductal papillary mucinous neoplasms of the pancreas: correlation of helical CT and dynamic MR imaging features with pathologic findings. Abdom Imaging 2008;33: Reprint requests Address requests for reprints to: Mark Topazian, MD, Mayo Clinic, Department of Internal Medicine, 200 First Street Southwest, Rochester, Minnesota topazian.mark@mayo.edu; fax: (507) Conflicts of interest: The authors disclose no conflicts.

8 198.e1 ZHONG ET AL CLINICAL GASTROENTEROLOGY AND HEPATOLOGY Vol. 10, No. 2 Supplementary Table 1. Pathology Characteristics of Epithelial Mural Nodules in Mucus-Producing Cystic Pancreatic Neoplasms IPMN (n 37) MCN (n 20) Total (n 57) Epithelial nodule present 12/37 (32.4%) 10/20 (50.0%) 22/57 (38.6%) Multiple epithelial nodules 4/37 (10.8%) 3/20 (15.0%) 7/57 (12.3%) Maximum nodule diameter (mm) 6.5 (2 30) 6 (1.8 5) 6/57 (1.8 30) Calcification in nodule 3/37 (8.1%) 2/20 (10.0%) 5/57 (8.8%) Degree of dysplasia in nodule Invasive carcinoma 1/12 (8.3%) 0/10 (0%) 1/22 (4.5%) High grade 1/12 (8.3%) 0/10 (0%) 1/22 (4.5%) Intermediate grade 1/12 (8.3%) 3/10 (30%) 4/22 (18.2%) Low grade 9/12 (75.0%) 7/10 (70%) 16/22 (72.7%) Degree of dysplasia in flat cyst mucosa Invasive carcinoma 0/37 (0%) 0/20 (0%) 0/57 (0%) High grade 3/37 (8.1%) 1/20 (5.0%) 4/57 (7.8%) Intermediate grade 4/37 (10.8%) 1/20 (5.0%) 5/57 (9.8%) Low grade 29/37 (78.4%) 18/20 (90.0%) 47/57 (92.1%) Coexistent ductal adenocarcinoma elsewhere in the specimen 2/37 (5.4%) 0/20 (0%) 2/57 (3.9%) Supplementary Table 2. EUS Features of Resected Mucus-Producing Cystic Neoplasms With or Without Epithelial Nodules Identified by Pathology Epithelial nodule Present (n 12) Absent (n 29) P value Total (n 41) Overall cyst features Multifocal cysts 5/12 (41.7%) 18/29 (62.1%).31 23/41 (56.1%) Maximum cyst diameter (mm, 31 ( ) 30 (21 34).44 a median, IQR) Cyst communication with 4/12 (33.3%) 13/29 (44.8%).35 17/41 (41.5%) pancreatic duct Presence of septation 7/12 (58.3%) 24/29 (82.8%).12 31/41 (71.6%) EUS FNA performed 4/12 (33.3%) 13/29 (44.8%).35 17/41 (41.5%) Fluid CEA (ng/ml) 924 (204 27,630) 799 ( ).36 a 924 (113 27,630) Features of intracystic echogenic lesions Intracystic echogenic lesion 9/12 (75%) 17/29 (58.6%).73 26/41 (63.0%) present Size (mm, median, IQR) 15.5 ( ) 9.0 ( ).07 a Doppler blood flow positive 2/8 (25.0%) 0/17 (0%).09 2/25 (8.0%) Moved after body position change 0/8 (0%) 6/17 (32.3%).13 6/25 (24.0) EUS FNA: suspected mucus 0/4 (0%) 4/13 (23.5%).52 4/17 (23.5%) EUS diagnosis (original report) Mural nodule 9/12 (75.0%) 5/29 (17.2%) /41 (34.1%) Mucus 0/12 (0%) 12/29 (41.4%) /41 (29.3%) CEA, carcinoembryonic antigen; IQR, interquartile range. a Wilcoxon test.

9 February 2012 MURAL NODULES IN PANCREATIC CYSTS 198.e2 Supplementary Table 3. Diagnostic Performance of EUS and CT for Detection of Mural Nodules in Comparison With Histology Diagnosis of mucus Diagnosis of mural nodule EUS report (n 41) EUS report (n 41) CT report (n 44) CT expert review (n 44) Sensitivity (%) 41 (25 59) (12/29) 75 (43 95) (9/12) 24 (7 50) (4/17) 47 (23 72) (8/17) Specificity (%) 100 (73 100) (12/12) 83 (64 94) (24/29) 100 (87 100) (27/27) 89 (71 98) (24/27) PPV (%) 100 (73 100) (12/12) 56 (30 80) (9/16) 100 (40 100) (4/4) 73 (39 94) (8/11) NPV (%) 41 (25 59) (12/29) 89 (71 98) (24/27) 68 (51 81) (27/40) 73 (54 87) (24/33) Accuracy (%) 59 (42 74) (24/41) 80 (65 91) (33/41) 70 (55 83) (31/44) 70 (55 83) (31/44) NPV, negative predictive value; PPV, positive predictive value. Supplementary Table 4. Interobserver Analysis of EUS Features of Intracystic Echogenic Lesions Interobserver reliability a Gold standard diagnosis a excluding nonresected cases a Gold standard diagnosis EUS feature Kappa 95% CI Mean percentage of raters agreeing Epithelial nodule Mucus P value Epithelial nodule Mucus P value Hypoechoic compared /5 (20%) 7/7 (100%).01 1/5 (20%) 4/4 (100%).048 with adjacent soft tissue Hyperechoic rim /6 (0%) 8/10 (80%).007 0/5 (0%) 5/7 (71%).028 Smooth edge /5 (20%) 7/8 (87.5%).032 1/5 (20%) 4/5 (80%).21 Round or oblong shape /6 (33%) 7/9 (77.8%).13 2/6 (33%) 4/6 (67%).57 Calcification /6 (0%) 0/9 (0%) NA 0/5 (0%) 0/6 (0%) NA Blood flow within lesion to 0.67 NA NA NA NA NA NA by Doppler Heterogenous lesion to 0.27 echo texture Smooth cyst wall to 0.15 Hypoechoic cyst wall to 0.02 Thickened cyst wall to 0.07 CI, confidence interval. a Only cases in which all raters agreed on the presence or absence of an intracystic echogenic lesion and rated the feature in question are included.