Role of Endoscopic Ultrasound in the Diagnosis of Intraductal Papillary Mucinous Neoplasms: Correlation With Surgical Histopathology

Transcription

1 CLINICAL GASTROENTEROLOGY AND HEPATOLOGY 2007;5: Role of Endoscopic Ultrasound in the Diagnosis of Intraductal Papillary Mucinous Neoplasms: Correlation With Surgical Histopathology SHIREEN A. PAIS,* SIRIBOON ATTASARANYA,* JULIA K. LEBLANC,* STUART SHERMAN,* C. MAX SCHMIDT, and JOHN DEWITT* *Department of Gastroenterology and Hepatology, Department of Surgery, Indiana University Medical Center, Indianapolis, Indiana Background & Aims: Intraductal papillary mucinous neoplasms (IPMNs) are precancerous tumors characterized by dilation of the main pancreatic duct, its side branches, or both. The purpose of this study was to evaluate the role of endoscopic ultrasound (EUS) in differentiating benign and malignant IPMNs. Methods: We identified all patients between July 1996 November 2005 who underwent preoperative EUS for IPMNs. Malignancy was defined as the presence of invasive carcinoma; all other neoplasms were considered benign. The results of EUS and EUS-guided fine-needle aspiration (EUS-FNA) were compared with corresponding histopathology. Results: Seventy-four patients (38 male; mean age, 65 years) with 21 (28%) malignant and 53 (72%) benign IPMNs were identified. Sixty-five (88%) underwent EUS-FNA. Compared with benign tumors, patients with malignant IPMNs were more likely to be older (P.011), present with jaundice (P.03) or weight loss (P.03), and have EUS features of a dilated main pancreatic duct (P.0001), solid lesion (P.0001), pancreatic ductal filling defects (P.03), or thickened septa within any cyst (P.02). The sensitivity, specificity, and accuracy of EUS-FNA for the diagnosis of malignancy were 75% (95% confidence interval [CI], 53% 89%), 91% (95% CI, 79% 97%), and 86% (95% CI, 76% 93%), respectively. Cyst or pancreatic duct fluid carcinoembryonic antigen (CEA) and carbohydrate antigen (CA) 19-9 did not differ between groups. Conclusions: Older age, jaundice and weight loss, and EUS features of a solid lesion, dilated main pancreatic duct, ductal filling defects, and thickened septa are predictive of malignancy in patients with IPMNs. EUS-FNA cytology is helpful, but cyst fluid CEA and CA 19-9 are of limited value to differentiate malignant from benign IPMNs. Intraductal papillary mucinous neoplasms (IPMNs) are premalignant or malignant mucin-producing pancreatic tumors that might arise from the main pancreatic duct, the side branches, or both. They typically present with cystic dilation of the pancreatic duct associated with intraductal papillary growth, mucin-production, and variable cellular atypia. 1 Since Ohhashi et al 2 first described this tumor in 1982, they have been increasingly recognized. In 1996, the World Health Organization recognized IPMNs as a distinct entity and in 2000 classified them as IPM adenoma, IPM with moderate dysplasia, IPM carcinoma noninvasive and invasive carcinoma. 3 These tumors have a well-defined adenoma-to-carcinoma sequence and usually have a more favorable prognosis compared with pancreatic ductal adenocarcinoma. 4,5 Clinical management varies from observation to surgery, depending on tumor histology and the location within the pancreas. 6 Surgical resection is generally recommended for any IPMNs involving the main pancreatic duct because of the increased incidence of malignancy in these tumors compared with side-branched cysts 6 8 and the usual excellent prognosis after surgery before the development of invasive carcinoma. 4 6,9 Long-term prognosis significantly worsens with identification of invasive carcinoma, particularly with associated lymph node metastases. 4 6,9 12 Previous studies suggested that clinical features such as older age, jaundice, weight loss, and new onset of diabetes at presentation and imaging features of main duct dilation, mural nodules, and cysts greater than 30 mm in size are suggestive of malignancy. 5,9 However, it has been reported that up to 29% of asymptomatic patients might have malignant IPMNs. 6 Therefore, use of symptoms alone cannot reliably exclude malignancy and emphasizes the need for accurate preoperative staging. Traditionally, the diagnosis of IPMNs is made by endoscopic retrograde pancreatography (ERP), which might demonstrate a segmental or diffusely dilated duct with filling defects corresponding to mucin or mural nodules. Alternatively, a gaping papilla with visible mucin might be seen and, when present, is diagnostic for this condition and usually indicates an advanced neoplasm. 16 ERP allows for aspiration of pancreatic juice and biopsy of the pancreatic duct, but isolated cystic lesions and solid masses associated with IPMNs are infrequently recognized by this technique. Peroral pancreatoscopy might also assist in the diagnosis, preoperative localization, and the collection of pancreatic juice for cytology. Pancreatoscopic features suggesting malignancy include linear or spotty vascular marking, fish-egg like lesions, villous proliferations, and vegetative-type lesions. 17 Although high-resolution computed tomography (CT), magnetic resonance cholangiopancreatography, endoscopic ultrasound (EUS), and intraductal ultrasound (IDUS) are sometimes used for evaluation of these lesions, most comparative studies indicate EUS is the most accurate imaging study for the diagnosis of malignancy Abbreviations used in this paper: CA, carbohydrate antigen; CEA, carcinoembryonic antigen; CI, confidence interval; CT, computed tomography; ERP, endoscopic retrograde pancreatography; EUS, endoscopic ultrasonography; FNA, fine-needle aspiration; IDUS, intraductal ultrasound; IPMN, intraductal papillary mucinous neoplasm; NPV, negative predictive value; PPV, positive predictive value by the AGA Institute /07/$32.00 doi: /j.cgh

2 490 PAIS ET AL CLINICAL GASTROENTEROLOGY AND HEPATOLOGY Vol. 5, No. 4 The use of EUS to identify and preoperatively assess IPMNs has been increasingly recognized However, previous studies have been limited by small sample sizes, lack of uniform use of EUS fine-needle aspiration (EUS-FNA), and incomplete histopathologic correlation. The purpose of this retrospective, single-center case series was to investigate the utility of clinical presentation, EUS morphology, EUS-FNA cytology, and cyst fluid analysis to differentiate benign from malignant IPMNs in a large consecutive cohort who underwent surgical resection. Methods Study Population This study was approved by the Institutional Review Board of Indiana University/Clarian Health Partners. By use of surgical and pathology databases, consecutive patients who underwent surgical resection for IPMNs were identified. Patients who had a preoperative EUS comprised the study population. Hospital records, endoscopy, pathology, and surgical reports of these patients were reviewed retrospectively. The following clinical information was noted: age, gender, weight loss, diarrhea, abdominal pain, back pain, jaundice, and any history of previous acute pancreatitis. EUS features recorded were size and location (head, body, tail, multifocal) of any dilated main pancreatic duct or side branch, filling defects in main duct or side branches, communication between cyst and main pancreatic duct, filling defects in a cyst, cyst size, septations and papillary projections in cyst, and any associated solid mass. A dilated main pancreatic duct was defined as greater than 3 mm, 2 mm, or 1 mm in the head, body, and tail, respectively. EUS-FNA puncture site (main duct, cyst, or solid mass), number of passes, needle size, cytology results, cyst fluid carcinoembryonic antigen (CEA), and carbohydrate antigen (CA) 19-9 were also noted. ERP findings such as patulous papilla, duct size, associated cysts, stricture of duct, associated filling defects within the duct, and associated cysts were documented. Surgery and pathology findings recorded included type of surgery, surgical margins, associated lymph nodes, and final surgical histopathology. On the basis of EUS, ERCP, and surgical pathology reports, tumors were classified as involving the main pancreatic duct, its side branches, or both. If there was isolated main pancreatic duct dilation without associated cysts, the tumor was classified as IPMN main duct disease (IPMN-M). If isolated cystic lesions or side branches communicated with the main pancreatic duct without any main pancreatic duct dilation, it was classified as IPMN side branch tumor (IPMN-Br). When both features were present, it was classified as a mixed-type IPMN. Endoscopic Ultrasound Examination After written informed consent, patients received conscious sedation with various combinations of intravenous midazolam, meperidine, fentanyl, or propofol under appropriate cardiorespiratory monitoring. All procedures were performed by or under the supervision of 1 of 6 experienced attending endoscopists. EUS examinations were usually initiated with an Olympus GF-UM20, GFUM-130, or GF-UM160 radial echoendoscope (Olympus Corporation, Melville, NY). Curvilinear array endosonography was performed in selected patients with the Pentax 32-UA, Pentax 36-UX (Pentax Precision Instruments, Orangeburg, NY), Olympus GF-UC30P, or Olympus GF- UC140P (Olympus Corporation) endoscope. When performed, EUS-FNA was done by using a 22-gauge, 8-cm Wilson-Cook EUSN-1, EUSN-2, EUSN-3, or Echotip Ultra needle (Wilson- Cook Medical Inc, Winston-Salem, NC) or Olympus EZ-Shot needle (Olympus Corporation). When the definitive diagnosis of malignancy or IPMN was obtained before the EUS exam, further mucosal biopsies or FNAs were only obtained if the endosonographer believed the results would upstage malignancy or alter planned surgical resection. The site of any EUS- FNA performed was at the discretion of the endosonographer. Doppler color angiography was used to ensure the absence of intervening vascular structures along the anticipated FNA needle path. Depending on the amount of blood anticipated during tissue sampling, full, partial, or no suction was applied at the discretion of the endoscopist or cytopathologist. A cytopathologist was available on-site for preliminary interpretations on all procedures. Samples aspirated were expressed onto a glass slide, and 2 smear preparations were made. One slide was air-dried and stained with a modified Giemsa stain for rapid on-site interpretation. The other slide was alcohol-fixed and stained by the Papanicolaou method. EUS-FNA was repeated until a definitive diagnosis was made, or the endoscopist believed that further sampling would not likely increase yield. If EUS-FNA was performed, the patient received 1 dose of intravenous antibiotics before discharge followed by 3 7 days of oral antibiotics. Per department policy, all patients were telephoned within 48 hours after the procedure to assess for any short-term complications. Neither peroral pancreatoscopy nor IDUS was used in any patient in this study. Cytology Examination Within several days of each EUS exam, a final cytologic diagnosis was made by a staff cytopathologist. Cytology features considered diagnostic of an IPMN included mucin-producing goblet cells arranged in a papillary pattern within thick, glistening, viscous mucus. 29,30 If specimens were not considered to be diagnostic because of a cellularity or hypocellularity, other cytologic features recorded included the presence of mucinproducing epithelial cells within a mucin background or mucin alone. Malignant features were indicated by the presence of hyperchromasia, nuclear crowding, and loss of nuclear uniformity, nucleolar prominence, or chromatin abnormalities Cytology reports were categorized as diagnostic for malignancy, suspicious for malignancy or atypical, benign IPMN, or nondiagnostic for IPMN. Mucin stains were not performed on any specimens. The final diagnosis in each patient in this study was made by the results of surgical resection and corresponding histopathology. Surgery and Surgical Pathology All surgical consultations and operations were performed by 1 of 5 experienced pancreatobiliary surgeons. Decisions for surgery were based on a preoperative evaluation of the patient s fitness for operation coupled with the results of all preoperative imaging studies. Patients were considered inoperable on the basis of preoperative imaging when they had histologically confirmed distant metastatic disease (distant lymphadenopathy, liver or pulmonary metastases, peritoneal carcinomatosis, malignant ascites) or locally advanced tumors involving major upper abdominal arterial vessels or complete encasement of the portal vein and/or superior mesenteric vein with loss of blood flow. Patients with suspected involvement of

3 April 2007 EUS AND DIAGNOSIS OF IPMNS 491 the portal vein and/or superior mesenteric vein or locoregional unresectability were considered for surgery on a case-by-case basis. Before tumor resection, all patients had complete abdominal exploration by laparoscopy or laparotomy to rule out metastatic or locally advanced disease. Operative criteria for attempted curative resection were absence of: liver metastases; peritoneal dissemination or drop metastases in the pelvis; invasion of the transverse mesocolon; metastases to celiac or other distant lymph nodes; and involvement of major upper abdominal arteries. A standard pancreaticoduodenectomy or pylorus-preserving variant was done for tumors located in the head or uncinate process. A distal pancreatectomy and splenectomy were done for tumors located in the body or tail. No tumor enucleations were performed. Routine intraoperative histologic frozen section examinations were done on the pancreatic, bile duct, and retroperitoneal soft tissue margins. A positive pancreatic or bile duct margin mandated further resection until a negative margin was obtained. Persistently positive pancreatic margins for either dysplastic IPMN or cancer often resulted in a total pancreatectomy at the discretion of the surgeon. Regional lymph nodes were routinely resected en bloc with the tumor specimen. IPMNs were defined as mucinous lesions of the pancreas that lacked ovarian-type stroma and were in communication with the main pancreatic duct or its side branches. 9,10 As previously published by Wiesenauer et al, 9 the tumors were classified as (1) adenoma (dilated pancreatic duct lined by mucinous epithelium with 1 criterion for low-grade dysplasia); (2) low-grade dysplasia ( 2 of the following criteria: epithelial tufting, nuclear pseudostratification, nuclear atypia, mitotic figures; also labeled as borderline for malignancy); (3) high-grade dysplasia (cribriform or solid growth associated with high-grade nuclear atypia; also labeled as noninvasive intraductal carcinoma or carcinoma in situ); or (4) invasive carcinoma (cytologic atypia included at least one of the following: increased nuclear cytoplasmic ratio, increased nuclear size, nuclear crowding, and hyperchromasia). Tumors were considered as benign if classified as any of the following: (1) IPMN adenoma; (2) low-grade dysplasia/borderline for malignancy; or (3) high-grade dysplasia/carcinoma in situ/noninvasive carcinoma. Only invasive carcinoma was considered malignant. For the current study, the final diagnosis in each patient was made by the results of surgical resection and corresponding histopathology. Statistical Analysis A true-positive EUS-FNA for malignancy was defined as unequivocal cytologic evidence of malignancy. A false-negative aspirate was a nondiagnostic or benign specimen that was subsequently found to be malignant by intraoperative and histopathologic findings. A true-negative EUS-FNA for malignancy was defined as benign or nondiagnostic samples that were confirmed as benign by intraoperative examination, histopathology, and clinical follow-up. Aspirates from EUS-FNA that were interpreted as nondiagnostic, suspicious, or atypical for malignancy were considered negative for malignancy. For analysis, continuous variables were described with means and standard deviations, and dichotomous variables were expressed as simple proportions with or without 95% confidence intervals (CIs). Confidence intervals were calculated by using the modified Wald method. Continuous variable associations Table 1. Presenting Symptoms of Patients With Malignant (n 21) and Benign (n 53) IPMNs Characteristic Malignant (n 21) Benign (n 53) P value Age (y), mean standard deviation Gender (n, %) Male 12 (57) 26 (49).46 Female 9 (43) 27 (51) 1.00 Abdominal pain (n, %) 15(71) 38(72) 1.00 Jaundice (n, %) 6(29) 4(7).03 Weight loss (n, %) 11(52) 13(24).03 Diarrhea (n, %) 4(19) 4(8).21 Back pain (n, %) 5(24) 5(9).13 Preceding acute pancreatitis (n, %) 6(29) 23(43).29 were assessed with an unpaired t test. The association between categorical variables and malignancy was assessed with the Fisher exact test. A P value of less than.05 was considered significant. Results Clinical Features During the study period, 88 patients underwent surgery for IPMNs. Of these, 74 (84%) patients (38 male; median age, 65 years; range, years) had a preoperative EUS and therefore comprised the study population. Surgical pathology revealed 21 (28%) malignant and 53 (72%) benign tumors. No endoscopic complications were noted in any patient. Patients with malignant tumors were significantly older (mean age, 70 8 years) compared with those with benign disease (mean, years; P.011; Table 1). No significant difference was noted between the presence of malignancy and patient gender (P.61). The most common presenting symptoms included abdominal pain (41%), weight loss (32%), back pain (14%), jaundice (13%), and diarrhea (11%). A history of acute pancreatitis was present in 29 (39%) patients. Eight patients (7 benign and 1 malignant) were asymptomatic. Jaundice (P.03) and weight loss (P.03) were both significantly more likely in malignant compared with benign IPMNs. A history of acute pancreatitis, abdominal pain, diarrhea, and back pain did not correlate with the presence of malignancy. New onset diabetes was inconsistently recorded in reviewed charts, and hence it was not evaluated. Surgery and Surgical Pathology Operative procedures performed included 47 (63%) pancreaticoduodenectomies, 16 (22%) distal pancreatectomies, 8 (11%) total pancreatectomies, 2 (3%) central pancreatectomies, and 1 (1%) subtotal pancreatectomy. The tumors were located (Table 2) in the pancreatic head alone in 43 (58%), body alone in 17 (23%), tail alone in 5 (7%), or were multifocal in 9 (12%). Classification of tumors included IPMN-M, IPMN-Br, or mixed type in 21 (28%), 18 (25%), and 35 (47%), respectively (Table 2). Final histopathologic diagnoses were IPMN-adenoma (n 14, 19%), IPMN low-grade dysplasia (n 38, 52%), IPMN highgrade dysplasia (n 1, 1%), and invasive carcinoma (n 21, 28%). There was no significant association between the tumor location (head, body, tail, or multifocal), tumor origin (main

4 492 PAIS ET AL CLINICAL GASTROENTEROLOGY AND HEPATOLOGY Vol. 5, No. 4 Table 2. EUS and Clinical Features in Patients With Malignant (n 21) and Benign (n 53) IPMNs EUS or clinical feature Malignant (n 21) Benign (n 53) P value Location within pancreas (n, %) Head 14 (67) 29 (55).32 Body 3 (14) 14 (26).36 Tail 1 (5) 4 (7) 1.0 Multifocal 3 (14) 6 (11).7 Origin (n, %) Main duct (IPMN-M) 6 (29) 16 (31) 1.0 Branched (IPMN-Br) 2 (9) 17 (32).07 Mixed-type 13 (62) 20 (38).07 Maximum diameter of main pancreatic duct (mm), mean standard deviation Dilated side branches (n, %) 15 (71) 29 (55).29 Solid mass (n, %) 16 (76) 8 (15).0001 Largest cyst/mass size (sq cm) Cyst/mass 30 mm (n, %) 7 (33) 7 (13).09 Filling defect in main pancreatic 8 (38) 7 (13).02 duct (n, %) Filling defect in cyst (n, %) 3 (14) 6 (11).70 Thickened septa within largest cyst (n, %) 4 (19) 1 (2).02 duct, side branch, or both), or the type of surgery performed and the presence of malignancy. Endoscopic Ultrasound Morphology The clinical and EUS morphologic features of the study population are described in Table 2. Patients with malignancy had a larger mean main pancreatic duct diameter ( mm) compared with those with benign tumors ( mm; P.0001). Malignant IPMNs were more likely to have filling defects with the main pancreatic duct or side branches (P.03), an associated solid lesion (P.0001), and thickened septa (P.02). Cyst size was not significantly different between the 2 groups (P.37). The presence of main duct disease alone was not significantly associated with malignancy (P 1.0). Endoscopic Ultrasound Fine-Needle Aspiration Cytology EUS-FNA (Tables 3 and 4) was performed in 65 (88%) patients, including 20 of 21 (95%) with malignancy and 45 of 53 Table 3. Results of EUS-FNA Cytopathology for 45 Patients With Benign IPMNs Confirmed by Surgical Pathology a Cytologic diagnosis (n 45) n (%) IPMN 16 (36) Mucin-producing epithelium and extracellular mucin 7 (16) Extracellular mucin alone 6 (13) Adenocarcinoma 4 (9) Chronic pancreatitis 1 (2) Gastrointestinal contaminant 2 (4) Nondiagnostic 5 (11) Benign ductal epithelium 4 (9) a EUS-FNA performed in 45 of 53 patients with benign IPMNs. (85%) with benign disease (Figure 1). Needle puncture was transduodenal in 40 patients and transgastric in 25. There was no significant difference in the mean number of passes performed between malignant ( ) and benign groups ( ; P.42). Compared with surgical pathology, the sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and accuracy of EUS-FNA for malignancy were 75% (95% CI, 53% 89%), 91% (95% CI, 79% 97%), 79% (95% CI, 56% 92%), 89% (95% CI, 77% 96%), and 86% (95% CI, 76% 93%), respectively. Of the 5 false-negative aspirates for malignancy, 2 obtained from cysts alone were nondiagnostic, 1 obtained from a dilated pancreatic duct revealed only extracellular mucin, 1 showed gastric contaminant alone, and 1 was misdiagnosed as acinar cystadenoma. Among the 45 with benign tumors who underwent EUS- FNA, 16 (36%) were interpreted cytologically as diagnostic of IPMNs, 7 (16%) had mucin-producing epithelium and extracellular mucin, and 6 (13%) samples had only extracellular mucin. The remaining 16 benign IPMNs did not show any cytologic features of IPMNs and were interpreted as follows: adenocarcinoma (n 4), chronic pancreatitis (n 1), benign ductal cells (n 4), gastrointestinal contaminant (n 2), or nondiagnostic (n 5). All 4 patients with false-positive cytologic identification of carcinoma had a solid mass on EUS imaging. Final surgical pathology in these 4 patients included benign IPMN (n 2) or low-grade dysplasia (n 2). There was no difference in the frequency of gastrointestinal contaminant observed by needle puncture through a transgastric (n 1, 2%) or transduodenal (n 2, 4%) route. (Figure 1) Cyst Fluid Analysis Cyst fluid or pancreatic duct CEA was available in 23 of 65 (35%) patients who underwent EUS-FNA including 15 with benign and 8 with malignant tumors (Figure 2). There was no statistically significant difference between the mean CEA in the malignant ( ng/ml; range, ,765) compared with benign IPMNs ( ng/ml; range, ; P.28). An arbitrary cutoff of CEA 800 ng/ml has a sensitivity, specificity, PPV, NPV, and accuracy of 13% (95% CI, 0.01% 49%), 93% (95% CI, 68% 99%), 50% (95% CI, 9% 91%), 67% (95% CI, 45% 83%), and 62% (95% CI, 45% 81%), respectively, to distinguish benign from malignant tumors. Cyst or pancreatic duct fluid CA 19-9 was available in 9 of 65 (14%) patients who underwent EUS-FNA including 5 with malignant and 4 with benign tumors. The mean CA 19-9 in the malignant IPMNs (16, U/mL; range, ,490) was not significantly different compared with the benign tumors ( ,651 U/mL; range, 15 10,236; P.08). If a Table 4. Results of EUS-FNA Cytopathology for 20 Patients With Malignant IPMNs Confirmed by Surgical Pathology a Cytologic diagnosis (n 20) n (%) Adenocarcinoma 15 (75) Gastrointestinal contaminant 1 (5) Mucin-producing epithelium and extracellular mucin 1 (5) Nondiagnostic 2 (10) Acinar cystadenoma 1 (5) a EUS-FNA performed in 20 of 21 patients with malignant IPMNs.

5 April 2007 EUS AND DIAGNOSIS OF IPMNS 493 Figure 1. Transduodenal EUS-FNA of IPMN-Br in the head of the pancreas. The side branch is septated, but no solid mass or thickened wall is present. value of 10,000 U/mL is used as the arbitrary cutoff between benign and malignant tumors, cyst fluid CA 19-9 has sensitivity, specificity, PPV, NPV, and accuracy of 60% (95% CI, 23% 86%), 75% (95% CI, 29% 97%), 75% (95% CI, 29% 97%), 60% (95% CI, 23% 86%), and 67% (95% CI, 35% 88%), respectively. Discussion In patients with suspected IPMNs, preoperative evaluation attempts to define the presence or absence of malignancy, involvement of the main duct and/or side branches, and the longitudinal extent of involvement within the pancreas. The symptoms of a patient at presentation might help to differentiate benign from malignant tumors. Similar to other studies, we found that among the clinical features considered, advanced age and symptoms of jaundice and weight loss at presentation were significantly associated with malignant IPMNs. 4,5,9,10 Because clinical symptoms at presentation might vary and not always be predictive of malignancy, some studies have evaluated the role of imaging morphology to diagnose and help differentiate benign from malignant IPMNs. However, reports have found that variable features might 19,25 or might not 24 reliably differentiate malignant from benign IPMNs. In the current study, we found that malignant IPMNs were more likely to have a larger diameter pancreatic duct, thickened cyst septations, main pancreatic duct filling defects, or an associated solid mass compared with benign tumors. These findings collectively might help clinicians decide who might benefit from surgical resection in suspected or known IPMNs. In our study, the rate of invasive malignancy in patients with IPMN-M and mixed-type IPMNs was 27% and 40%, respectively. A recent review 6 found a mean rate of invasive malignancy of 43% in patients with IPMN-M or mixed-typed tumors. Thus, our findings support a recent international consensus statement that all main duct and mixed-type IPMNs be removed in surgically fit patients. 6 Previous studies have suggested that side branch disease is more likely to be benign. 6 8 Accordingly, only 2 of 19 (11%) patients in our study with IPMN-Br had invasive carcinoma. Although EUS morphology might be helpful in some cases, most studies indicate that morphology alone does not accurately differentiate either the type or malignant potential of pancreatic cystic lesions Therefore, EUS-FNA has played an increasing role in the evaluation of cystic neoplasms of the pancreas, including IPMNs. Some studies that describe the role of EUS-FNA for pancreatic cysts do not separate findings of IPMNs from other cysts 27,36 or have sparse histologic correlation with cytology findings, 27 thus making it difficult to ascertain its role for this condition. Maire et al 26 reported that EUS-FNA obtained a positive diagnosis in 10 of 10 patients with IPMN when compared with surgical pathology. Similarly, Frossard et al 36 found that EUS-FNA had a sensitivity and specificity of 100% and 98%, respectively, for 14 patients with histologically proven benign IPMNs. We found that for 45 benign IPMNs that underwent EUS-FNA, the correct cytologic diagnosis of an IPMN was made in only 16 (36%). An additional 13 of 45 (29%) had mucin-producing epithelium and extracellular mucin (n 7) or extracellular mucin alone (n 6) but insufficient cytologic material to render a diagnosis of IPMN. These findings demonstrate the limitations of EUS-FNA cytology alone for the diagnosis of a benign IPMN. Although we did not find that performance of either transgastric or transduodenal puncture increased the accuracy of the diagnosis of a benign IPMN, we prefer to perform needle puncture through the duodenum to avoid the possibility of a false-positive diagnosis of an IPMN as a result of the presence of abundant mucinproducing epithelium in the stomach. For the diagnosis of malignancy, Maire et al 26 found that the combination of transpapillary biopsy and EUS-FNA only offered a sensitivity of 44% and specificity of 100% for the diagnosis of malignant IPMNs. We found that among 21 patients with histologically proven cancer, EUS-FNA had a sensitivity, specificity, and accuracy of 75%, 91%, and 86%, respectively, for Figure 2. Cyst or pancreatic duct fluid CEA in patients with benign (n 15) and malignant (n 8) IPMNs confirmed by surgical resection. There was no statistically significant difference between the mean CEA in the malignant ng/ml; range: 2.5 to 12765) compared to benign IPMNs ( ng/ml; range 0.5 to 4878; P.28).

6 494 PAIS ET AL CLINICAL GASTROENTEROLOGY AND HEPATOLOGY Vol. 5, No. 4 the diagnosis of malignancy. This sensitivity is higher than that reported (47%) for 111 patients with various types of pancreatic cysts in a recent review of 11 studies. 37 This review, however, did not include patients with IPMNs. Our higher sensitivity might reflect the fact that 76% of the 21 patients with malignancy had a solid, hypoechoic mass associated with their malignancy, thus permitting easy targeting by EUS-FNA. Alternatively, our high sensitivity might reflect surgical referral bias after a positive cytologic diagnosis of malignancy. Our specificity was only 91% for the diagnosis of malignancy because 4 patients with histologically proven benign tumors were misclassified as malignant by EUS-FNA cytology. Interestingly, all 4 patients were found to have a solid mass associated with a cystic lesion. Similar to previous reports, we found no statistically significant difference in mean cyst or pancreatic duct fluid CEA and CA 19-9 between malignant and benign cysts. Kawai et al 38 reported that pancreatic duct juice CEA was significantly higher among malignant IPMNs; however, CA 19-9 values were not different between the 2 groups. For differentiating mucinous adenocarcinomas from mucinous cystadenomas, van der Waaij 37 reported a CEA 800 ng/ml had a sensitivity, specificity, and accuracy of 48%, 98%, and 79%, respectively. With the same cutoff of 800 ng/ml, we found a sensitivity, specificity, and accuracy of 13%, 93%, and 62%, respectively, for the diagnosis of malignancy. Therefore, from these limited data, it appears that cyst or duct juice tumors markers are of limited value in differentiating malignant from benign IPMNs. In the 65 patients in our study who underwent EUS-FNA, cyst or pancreatic duct fluid testing for CEA and CA 19-9 was performed in only 23 (35%) and 9 (14%) patients, respectively. There are several possible explanations for this. First, in our hospital, about 1 ml of fluid is required to perform these tests. For mucinous cystic tumors (such as IPMNs), the viscous fluid from aspirates often precludes obtaining a sufficient sample, and therefore only cytologic evaluation is performed. Second, when preliminary reports from the Cooperative Pancreatic Cyst Study 35 suggested that CA 19-9 was of limited value for differentiating mucinous from non-mucinous tumors, we stopped routinely using this test. Third, the high rate of malignancy and excellent sensitivity for its diagnosis in our series obviated the need for tumor marker evaluation in some of these patients. The current study is the largest series to date to review consecutive patients with histopathologically confirmed IPMNs who undergo preoperative EUS. Furthermore, the use of EUS- FNA in 89% of patients permits evaluation of the advantages and limits of this procedure in these patients. Nevertheless, our study has some limitations that merit some discussion. First, the evaluation of only patients with IPMNs limits application of our results to all patients with pancreatic cystic lesions. Second, selection bias might exist in this study population because patients with a preoperative diagnosis of malignancy would more likely have been referred for surgery. Therefore, the accuracy of EUS-FNA for the diagnosis of malignancy in these patients might be overstated. Third, the acquisition of tumor markers in only the minority of patients limits conclusions that might be made about its utility in these patients. Finally, IDUS was not used in any of these patients, and no comparison between EUS and any other preoperative test was performed in this study. In summary, older age, jaundice, and weight loss and EUS features of a solid lesion, dilated main pancreatic duct, ductal filling defects, and thickened septa are predictive of malignancy in patients with IPMNs. EUS-FNA cytology is relatively sensitive and specific for the diagnosis of malignancy, but cyst fluid CEA and CA19-9 appear to be of limited value to differentiate malignant from benign IPMNs. References 1. D Angelica M, Brennan MF, Suriawinata AA, et al. Intraductal papillary mucinous neoplasms of the pancreas: an analysis of clinicopathologic features and outcomes. Ann Surg 2004;239: Ohhashi K, Murakami Y, Maruyama M. Four cases of mucin producing cancer of the pancreas on specific findings of the papilla of Vater. Prog Dig Endosc 1982;20: Kloeppel G, Solcia EC, Longnecker DS, et al. Histological typing of tumors of the exocrine pancreas. In: World Health Organization international classification of tumors. 2nd ed. Berlin: Springer, 1996: Sohn TA, Yeo CJ, Cameron JL, et al. Intraductal papillary mucinous neoplasms of the pancreas: an updated experience. Ann Surg 2004;239: Jang JY, Kim SW, Ahn YJ, et al. Multicenter analysis of clinicopathologic features of intraductal papillary mucinous tumor of the pancreas: is it possible to predict the malignancy before surgery? Ann Surg Oncol 2005;12: Tanaka M, Chari S, Adsay V, et al. International Association of Pancreatology: international consensus guidelines for management of intraductal papillary mucinous neoplasms and mucinous cystic neoplasms of the pancreas. Pancreatology 2006;6: Kobari M, Egawa S, Shibuya K, et al. Intraductal papillary mucinous tumors of the pancreas comprise 2 clinical subtypes: differences in clinical characteristics and surgical management. Arch Surg 1999;134: Sugiyama M, Izumisato Y, Abe N, et al. Predictive factors for malignancy in intraductal papillary-mucinous tumours of the pancreas. Br J Surg 2003;90: Wiesenauer CA, Schmidt CM, Cummings OW, et al. Preoperative predictors of malignancy in pancreatic intraductal papillary mucinous neoplasms. Arch Surg 2003;138: Suzuki Y, Atomi Y, Sugiyama M, et al. Cystic neoplasm of the pancreas: a Japanese multiinstitutional study of intraductal papillary mucinous tumor and mucinous cystic tumor. Pancreas 2004;28: Fernandez-del Castillo C, Targarona J, Thayer SP, et al. Incidental pancreatic cysts: clinicopathologic characteristics and comparison with symptomatic patients. Arch Surg 2003;138: Salvia R, Fernandez-del Castillo C, Bassi C, et al. Main-duct intraductal papillary mucinous neoplasms of the pancreas: clinical predictors of malignancy and long-term survival following resection. Ann Surg 2004;239: Dabezies MA, Campana T, Friedman AC. ERCP in the diagnosis of ductectatic mucinous cystadenocarcinoma of the pancreas. Gastrointest Endosc 1990;36: Azar C, Van de Stadt J, Rickaert F, et al. Intraductal papillarymucinous tumors of the pancreas: clinical and therapeutic issues in 32 patients. Gut 1996;39: Loftus EV, Olivares-Pakzad B, Batts KP, et al. Intraductal papillary-mucinous tumors of the pancreas: clinicopathologic features, outcome, and nomenclature. Gastroenterology 1996; 110: Shyr YM, Su CH, Tsay SH, et al. Mucin-producing neoplasms of the pancreas: intraductal papillary and mucinous cystic neoplasms. Ann Surg 1996;223: Hara T, Yamaguchi T, Ishihara T, et al. Diagnosis and patient management of intraductal papillary-mucinous tumor of the pan-

7 April 2007 EUS AND DIAGNOSIS OF IPMNS 495 creas by using peroral pancreatoscopy and intraductal ultrasonography. Gastroenterology 2002;122: Cellier C, Cuillerier E, Palazzo L, et al. Intraductal papillary and mucinous tumors of the pancreas: accuracy of preoperative computed tomography, endoscopic retrograde pancreatography and endoscopic ultrasonography, and long-term outcome in a large surgical series. Gastrointest Endosc 1998;47: Sugiyama M, Atomi Y, Saito M. Intraductal papillary tumors of the pancreas: evaluation with endoscopic ultrasonography. Gastrointest Endosc 1998;48: Baba T, Yamaguchi T, Ishihara T, et al. Distinguishing benign from malignant intraductal papillary mucinous tumors of the pancreas by imaging techniques. Pancreas 2004;29: Yamao K, Ohashi K, Nakamura T, et al. Evaluation of various imaging methods in the differential diagnosis of intraductal papillary-mucinous tumor (IPMT) of the pancreas. Hepatogastroenterology 2001;48: Williams DB, Sahai AV, Aabakken L, et al. Endoscopic ultrasound guided fine needle aspiration biopsy: a large single centre experience. Gut 1999;44: Aithal GP, Chen RY, Cunningham JT, et al. Accuracy of EUS for detection of intraductal papillary mucinous tumor of the pancreas. Gastrointest Endosc 2002;56: Song MH, Lee SK, Kim MH, et al. EUS in the evaluation of pancreatic cystic lesions. Gastrointest Endosc 2003;57: Kubo H, Chijiiwa Y, Akahoshi K, et al. Intraductal papillary-mucinous tumors of the pancreas: differential diagnosis between benign and malignant tumors by endoscopic ultrasonography. Am J Gastroenterol 2001;96: Maire F, Couvelard A, Hammel P, et al. Intraductal papillary mucinous tumors of the pancreas: the preoperative value of cytologic and histopathologic diagnosis. Gastrointest Endosc 2003;58: Brandwein SL, Farrell JJ, Centeno BA, et al. Detection and tumor staging of malignancy in cystic, intraductal, and solid tumors of the pancreas by EUS. Gastrointest Endosc 2001;53: Farrell JJ, Brugge WR. Intraductal papillary mucinous tumor of the pancreas. Gastrointest Endosc 2002;55: Recine M, Kaw M, Evans DB, et al. Fine-needle aspiration cytology of mucinous tumors of the pancreas. Cancer 2004; 102: Layfield LJ, Cramer H. Fine-needle aspiration cytology of intraductal papillary-mucinous tumors: a retrospective analysis. Diagn Cytopathol 2005;32: Stelow EB, Stanley MW, Bardales RH, et al. Intraductal papillarymucinous neoplasm of the pancreas: the finding and limitations of cytologic samples obtained by endoscopic ultrasound guided fine needle aspiration. Am J Clin Pathol 2003;120: Solcia EC, Kloeppel G. Tumors of the pancreas, fascicle 20. 3rd series ed. Washington: Armed Forces Institute of Pathology, Ahmad NA, Kochman ML, Lewis JD, et al. Can EUS alone differentiate between malignant and benign cystic lesions of the pancreas? Am J Gastroenterol 2001;96: Ahmad NA, Kochman ML, Brensinger C, et al. Interobserver agreement among endosonographers for the diagnosis of neoplastic versus non-neoplastic pancreatic cystic lesions. Gastrointest Endosc 2003;58: Brugge WR, Lewandrowski K, Lee-Lewandrowski E, et al. Diagnosis of pancreatic cystic neoplasms: a report of the cooperative pancreatic cyst study. Gastroenterol 2004;126: Frossard JL, Amouyal P, Amouyal G, et al. Performance of endosonography-guided fine needle aspiration and biopsy in the diagnosis of pancreatic cystic lesions. Am J Gastroenterol 2003; 98: van der Waaij LA, van Dullemen HM, Porte RJ. Cyst fluid analysis in the differential diagnosis of pancreatic cystic lesions: a pooled analysis. Gastrointest Endosc 2005;62: Kawai M, Uchiyama K, Tani M, et al. Clinicopathological features of malignant intraductal papillary mucinous tumors of the pancreas: the differential diagnosis from benign entities. Arch Surg 2004;139: Address requests for reprints to: John DeWitt, MD, FASGE, FACP, Assistant Professor of Medicine, Division of Gastroenterology, Indiana University Medical Center, 550 N University Blvd, UH 4100, Indianapolis, IN jodewitt@iupui.edu; fax: (317)