SS

AT

Ab

stra

cts

961

Outcomes of Primary Surveillance for Intraductal Papillary MucinousNeoplasmChristy E. Cauley, Joshua A. Waters, Ryan P. Dumas, Juliana E. Meyer, Mohammad A. Al-Haddad, John M. DeWitt, Keith D. Lillemoe, C. Max Schmidt

BACKGROUND: Limited data is available regarding the natural history of patients undergoingprimary surveillance for intraductal papillary mucinous neoplasm (IPMN). We examine theoutcome of patients selected for primary surveillance in a multidisciplinary pancreas cystclinic. We hypothesize that symptoms, radiologic characteristics, and cytology will predictrisk of developing pancreatic cancer in patients undergoing primary surveillance for IPMN.METHODS: Between January 2002 and March 2010, 522 patients were diagnosed withIPMN at a single, high volume institution. Patients were prospectively stratified as low orhigh oncologic risk based on analysis of demographic, clinical, radiologic, and cytopathologicdata. Any patient who underwent primary operative management or less than 3 monthsof surveillance was excluded. RESULTS: 174(33%) patients underwent primary operativemanagement for IPMN. Of these, 46(26%) were found to have invasive cancer. Alternatively,292(56%) patients underwent primary surveillance for IPMN. Of these, 244(84%) wereclassified as low oncologic risk. Mean duration of primary surveillance was 35 (3-99) months.28(11%) patients initially stratified as low-risk ultimately underwent pancreatic resectionwith a mean preoperative surveillance of 11 (4-42) months. Indications for resection were8(29%) new main duct dilation or mural nodule, 6(21%) new or worsening symptoms,5(18%) increasing lesion size, 2(7%) concerning cytopathology, and 7(25%) patient prefer-ence. Of the 28 patients resected after surveillance, 27(96%) demonstrated low-grade IPMNand 1(4%) high-grade dysplastic (HGD) IPMN. The patient with HGD had a family historyof pancreatic cancer, but was asymptomatic, and no radiographic or cytologic indicators ofmalignancy. Of non-operated patients, 2(1%) developed invasive cancer at 18 and 51 monthsof surveillance. Neither of these patients demonstrated increasing cyst size or new concerningradiographic features prior to the diagnosis of invasive cancer. 48(16%) patients stratifiedas high oncologic risk were initially managed non-operatively due to age, comorbidity orpatient preference. Five-year survival for this group was 62%. Of these, 13(27%) patientshave died during follow-up: 2(15%) from pancreas cancer, 7(54%) from other causes, and4(31%) unknown. CONCLUSIONS: For IPMN initially classified as low-risk, progressionto pancreas cancer during surveillance was rare. Current accepted indications for resectiondid not forecast malignancy in this group. More accurate markers are needed to better guideIPMN surveillance. For poor operative candidates with high risk IPMN, progression toinvasive cancer during surveillance was more common, though a substantial portion succumbto non-IPMN related death.

962

Updated Results for Dual Modality Versus Percutaneous Drainage for theTreatment of Symptomatic Walled off Pancreatic NecrosisMichael Gluck, Andrew S. Ross, Shayan Irani, S. Ian Gan, Mehran Fotoohi, Robert Crane,Justin Siegal, Ellen Hauptmann, Richard A. Kozarek

Background Treatment of symptomatic walled off pancreatic necrosis (WOPN) by dualmodality-endoscopic and percutaneous-drainage (DMD) has been shown to decrease lengthof hospitalization (LOH), use of CT scans and drain studies by radiology in comparison topercutaneous drainage alone. Aim To demonstrate the durability of the initial conclusionsas the cohort of DMD patients expanded. Methods A prospective database of patientsundergoing DMD was analyzed and compared to patients who had standard percutaneousdrainage from 2006 to the present time. Results 41 patients had undergone DMD with 39completing therapy, defined as removal of percutaneous drains, definitive surgery, or death.43 patients underwent percutaneous drainage alone. Patient characteristics including age,sex, etiology of pancreatitis, and severity of pancreatitis based on computed tomographyseverity index were indistinguishable between the two groups. Initial endoscopic access tothe necroma was obtained by endoscopic ultrasound in 30 of the 41 patients undergoingDMD. The DMD cohort had shorter mean LOH (27 vs 55 days), time until removal ofpercutaneous drains (78 vs 188 days), fewer CT scans (8.0 vs 14.3), drain studies (6.0 vs13), and number of drains per patient (1.29 vs 2.0), all statistically significant. The DMDcohort also had fewer total ERCP's (2.0 vs 2.6, p<0.026). There have been 3 total deathsin the DMDgroup: 1 fromMRSA pneumonia during therapy for incidentally found esophagealcancer 6 months after removal of the percutaneous drain; 1 from congestive heart failure 2weeks after discharge from hospitalization for WOPN and electing hospice; and 1 patientwith multi-system organ failure during treatment for DMD who was found to have an occultpancreatic adenocarcinoma at autopsy. Three patients in the standard drainage cohort diedwith drains in place while in the hospital undergoing therapy. No DMD patient neededsurgery or had a pancreatico-cutaneous fistula (PCF). Conclusions Compared to standardpercutaneous drainage, DMD of WOPN reduces LOH and the use of radiological andendoscopic resources. Surgery and PCFs were avoided in patients undergoing DMD whilesingle digit mortality was maintained.

963

Is It Worth Looking? Abdominal Imaging After Pancreatic Cancer Resection: ANational StudyElan R. Witkowski, Jillian K. Smith, Elizaveta Ragulin-Coyne, Sing Chau Ng, Shimul A.Shah, Jennifer F. Tseng

BACKGROUND: Sequential followup imaging is often performed after pancreatic resectionfor cancer. We attempted to quantify the volume and cost of complex abdominal imagingafter pancreatic resection nationwide, and determine whether their frequent use confersbenefit. METHODS: Patients with pancreatic adenocarcinoma who underwent surgical resec-tion were identified in SEER-Medicare (1991-2005). Claims for abdominal imaging (CT/CTA, MRI/MRA, PET) ≤5 years after resection were analyzed. After initial screen, CT/CTAwas used for longitudinal analyses. Univariate and multivariate analyses were performed byKruskal-Wallis, logistic regression, and Cox. CT utilization was calculated by dividing total

S-1008SSAT Abstracts

scans by months of available postoperative data. Routine annual CT scanning was definedas at least one CT/CTA performed within each 12-month block, excluding year of death/censoring. To assess frequency of annual CT scanning in patients with superior survival,the top decile were further analyzed. RESULTS: Within 5 years of pancreatic resection,39316 studies were performed on 2792 patients. The majority of these were CT scans(36521, 92.9%), and the remainder MRI (2425, 6.2%) and PET (370, 0.9%). A minorityreceived no imaging after resection, both when analyzing all patients (168/2792, 6.0%), aswell as the subset with >5 year-survival (11/265, 4.2%). Mean monthly CT utilization perpatient increased from 0.4 in 1991 to 1.1 in 2005 (p<0.0001), including the immediatepostoperative period. Overall mean utilization was 0.6 CT scans per month, but only 0.2scans per month for patients in the top survival decile. Among all patients, scans were notevenly distributed over the years of the study: many of the scans were clustered. Patientsreceived a mean of 2.8 scans within 3 months of surgery, 4.7 scans between 3 months and1 year, and 5.6 scans between 1 year and study termination. Among 1127 patients withsufficient survival to allow for analysis, 569 (50.5%) received annual CT scans as previouslydefined. Interestingly, only 81 (28.9%) of the top-performing patients received annual CTscans. Among all patients, postoperative complications, non-white race, advanced stage, andreceipt of chemotherapy or radiation were predictive of receiving routine annual CT scans(p<0.05). Routine annual CT scans were associated with negative rather than positive impacton survival (HR for death 1.2, p=0.02). Based on current Medicare line-item payments, thebare minimum mean imaging cost incurred per patient would be $3736, or $5287 over 5years of survival. CONCLUSIONS: Most patients undergo abdominal imaging, usually CT,after pancreatic cancer resection. CT utilization has increased in recent years. Administrativedata from a large national database suggests that performance of routine annual CT scansafter resection does not confer a survival benefit.

964

Activation of the BMP4 Pathway and Early Expression of CDX2 Characterizethe Development of Nonspecialized Columnar Metaplasia in a Human Modelof Barrett EsophagusManuel R. Pera Roman, Daniel Castillo, Sonia Puig, Carme de Bolós, Mar Iglesias,Agustin Seoane, Laura Comerma, Vicente Munitiz, Pascual Parrilla, Richard Poulsom, LuisGrande

BACKGROUND: The cardia type epithelium, a nonspecialized columnar type of metaplasia(NSCM), has been proposed as an intermediate stage in the transdifferentiation process fromnormal squamous epithelium into the specialized intestinal type of differentiation presentin Barrett esophagus. Recent studies suggest that the BMP4 pathway is involved in thetransition of squamous epithelium into nonspecialized epithelium, and that CDX2 and CDX1transcription factors seem critical for acquisition of the intestinal phenotype. Using an InVivo human model of pathologic reflux disease, we assessed prospectively the developmentof NSCM and the involvement of the BMP4 pathway andCDX2 expression in this intermediatephenotypic change. METHODS: Biopsy samples from the remnant esophagus of patientshaving an esophagectomy with gastric preservation were taken at different time periods (6,12, 18, 24 and 36 months) after their operation and examined for the activation of the BMPpathway (BMP4 / pSMAD 1/5/8) and expression of CDX2 and CDX1. Samples were preparedfrom squamous esophageal mucosa just above the esophagogastric anastomosis, from colum-nar-appearing mucosa and from squamous esophageal mucosa 2 cm below the upperesophageal sphincter level and were assessed by immunohistochemistry, quantitative real-time PCR (qRT-PCR) and Western blot (WB). RESULTS: Since June 2006, 18 patients (16male) were included in the study. Thirteen had an intrathoracic anastomosis. A short segment(mean length: 15.6 mm; longest 30 mm) of NSCM was detected in 10 (56%) patients, withan increasing prevalence along the time periods (17%, 25%, 38%, 42% and 71% at 6, 12,18, 24 and 36 months, respectively). All cases except one were detected in patients havingan intrathoracic anastomosis. Seven of 10 NSCM segments were first detected at 12-18months postoperatively. Immunohistochemistry detected nuclear expression of pSMAD 1/5/8 in the squamous epithelium close to the anastomosis with strong expression in allepithelial cells of the NSCM areas. These results were confirmed by WB analysis. Scatterednuclear expression of CDX2 was observed in 44/59 biopsies (75%) with NSCM. Two casesshowed isolated glands at 18 and 36 months that fully expressed CDX2 and co-expressedCDX1. BMP4 mRNA and CDX2 mRNA expression were significantly greater in NSCMcompared with squamous epithelium. CONCLUSIONS: In this human model of columnarmetaplasia, we identified early activation of BMP4 in squamous epithelium and NSCM andobserved that CDX2 but not CDX1 expression was detected in NSCM before the appearanceof the intestinal phenotype.

965

Cholinesterases Predict Outcome in Patients Undergoing Hepatic Resection forHepatocellular Carcinoma. Results From a Retrospective AnalysisMatteo Donadon, Matteo M. Cimino, Fabio Procopio, Emanuela Morenghi, AngelaPalmisano, Daniele Del Fabbro, Marco Montorsi, Guido Torzilli

Background. Estimation of functional liver reserve in patients with hepatocellular carcinoma(HCC) is of paramount importance to properly select candidates for surgical resection.Together with the value of bilirubin, the presence/absence of ascites and esophageal varices,and the rate of residual liver volume, which are our current parameters to measure functionalliver reserve, we sought to investigate the value of preoperative cholinesterases (CHE) inpredicting postoperative adverse outcomes after hepatic resection for HCC. Methods. Wereviewed the records of 181 consecutive patients who underwent hepatic resection for HCCin our Unit between 2001-2009. The value of preoperative CHE was analyzed against theoccurrence of postoperative adverse events. Receiver-operator characteristic curve analysiswas used to identify cut-off values of CHE that predicted adverse outcomes. Univariate andmultivariate analyses on clinically relevant variables were performed. P<0.05 was consideredstatistically significant. Results. Forty-nine (27%) of 181 patients had complications, ofwhich 36 (20%) were liver-related. Major morbidity occurred in 11 (6%) patients. The 30-day postoperative mortality was 1% (2 patients). A value of CHE≤5,900 UI/L had a sensitivityof 73% and a specificity of 67% in predicting liver-related postoperative complications (P=

0.001). The multivariate analysis revealed that blood transfusion, major resections, and avalue of CHE≤5,900 UI/L independently predicted the risk of morbidity. Conclusions. Thevalue of CHE contributed important information in predicting postoperative adverse out-comes after hepatic resection for HCC. Thus, it should be included in the selection processof candidates to surgery for such disease. [ClinicalTrials.gov ID=NCT00883454]

966

Comprehensive Perioperative Geriatric Assessments May Predict SurgicalOutcomes in a Prospective Study of Older Patients UndergoingPancreaticoduodenectomyKevin K. Roggin, Joshua A. Hemmerich, Jeffrey B. Matthews, Mitchell C. Posner, WilliamDale

Introduction: Older patients with pancreatic cancer are often not offered pancreaticoduod-enectomy (PD) due to potentially high perioperative risk and prolonged post-operativerecovery. These patients may have undetected vulnerabilitities related to frailty that mayadversely affect surgical outcomes. Methods: PD-eligible patient over 50 were enrolledonto this IRB-approved prospective observational study. Extensive perioperative clinicalinformation including Vulnerable Elder Survey (VES-13), short physical performance battery(SPPB), and Fried's Criteria for frailty were collected. Complications were prospectivelyrecorded and graded according to established guidelines (Clavien scores). Quality of life(QoL) measures were taken at baseline and at 1, 3, and 6-months post-PD. Results: Seventy-eight patients were enrolled from 10/1/2007-9/1/2010: 53 had a PD (median age 68years;range 46-85) and 25 were inoperable or declined PD. Significant preoperative vulnerab-ility was identified in resected patients; VES-13 >3 in 24%; SPPB <10 in 49%; Fried'sexhaustion in 40%. The perioperative morbidity rate was 70% and three patients died (8.6%).Twenty-three patients (43%) had severe complications (Clavien grade≥ III). Abnormal VES-13 showed predictive promise for post-operative admission to the SICU rs(53)=.34, p<.05,longer SICU stays rs(53)=.42, p<.01, and total hospital days rs(53)=.26, p=.09. Self-reportedexhaustion predicted complications and post-operative admission to the SICU, χ2(1, N =53) = 4.1, p<.05. Older age was predictive of not being discharged to home rs(53)= -.35,p<.05 and perioperative death, rs(53)=.31, p<.05. 22 patients have complete QoL data thatshows a wide variance at baseline, a significant drop-off at 1 month and return to baselineby 6 months. Conclusion: We identified significant vulnerability in older patients undergoingPD. Early analyses suggest that comprehensive perioperative geriatric measures can predictpost-surgical outcomes. If sought for and identified, these issues could be managed expect-antly, leading to more accurate preoperative counseling and in-hospital care.

967

Trends in the Palliative Surgical Management of Patients With UnresectablePancreatic Adenocarcinoma: Lessons Learned From a Large, Single InstitutionExperiencePeter J. Kneuertz, Steven Cunningham, Sergio Lopez, John L. Cameron, Joseph M.Herman, Martin A. Makary, Frederic E. Eckhauser, Kenzo Hirose, Barish H. Edil, MichaelA. Choti, Richard D. Schulick, Christopher L. Wolfgang, Timothy M. Pawlik

Introduction: Traditionally, routine palliative bypass (PBP) has been advocated for palliationof patients (pts) with pancreatic adenocarcinoma (PAC) who are explored with curativeintent but have inoperable disease discovered at the time of surgery. We examined trendsin the relative use of PBP over time with a particular emphasis on identifying changes insurgical indications, type of bypass performed, as well as perioperative outcomes associatedwith surgical palliation. Methods: Between Jan 1996 and Jun 2010, 1913 pts with PAC inthe head of the pancreas were surgically explored. Data regarding preoperative symptoms,intraoperative findings, type of surgical procedure performed, as well as perioperative andlong-term outcomes were collected and analyzed using univariate and multivariate analyses.Results: Of the 1913 pts, 1330 (70%) underwent a pancreaticoduodenectomy, while 583(30%) did not. Among the 583 pts who did not undergo curative intent surgery, mostpresented with either preoperative nausea/vomiting (18%) or jaundice (72%). The majorityof pts had been evaluated by CT scan (97%), while a minority had an MRI (5%) or PETscan (1%). On preoperative imaging, median size of the pancreatic lesion was 3.2cm witha median CA19-9 of 351 ng/dl. At surgery, 553 (95%) pts had PBP and 30 (5%) patientshad laparotomy+biopsy without further surgical intervention. Among the 553 pts who hadPBP, most (65%) underwent double bypass, while a minority had either gastrojejunostomy(28%) or hepaticojejunostomy (7%) alone. While the number of pancreaticoduodenectomiesremained relatively stable over time, there was a temporal decrease in the utilization ofPBP (Figure). Unanticipated locally advanced disease vs. liver/peritoneal metastasis as theindication for PBP also changed over time (1996-2001: 46% vs. 54%; 2002-2007: 49% vs.51%; 2008-2010: 20% vs. 80%, respectively)(P=0.009). Following PBP, median length ofstay was 8 d and 37% pts had a complication (major: 14%) with a readmission rate of 18%.

S-1009 SSAT Abstracts

Postoperative mortality after PBP was 2%. Few pts developed late obstruction (enteric: 3%;biliary: 4%) or needed a postoperative stent (enteric: 2%; biliary:5%). Overall survivalfollowing PBP was 6 months; pts with unsuspected metastatic disease as the indication forPBP had a significantly worse survival compared with pts who had locally unresectabledisease (median survival: 5 vs. 8 months, respectively; HR=1.43, P=0.001). Conclusion:Utilization of PBP for inoperable pancreatic cancer has significantly decreased over time.The indication for PBP has also changed over time, with the discovery of unsuspectedmetastatic disease now being the main indication for PBP. While PBP remains effective, themorbidity can be significant and survival following PBP especially among pts with metastaticdisease is extremely short.

968

Management and Perioperative Morbidity Among Patients With SurgicallyManaged Pancreatic Adenocarcinoma: A Population-Based Analysis UsingSEER-Medicare DataSkye C. Mayo, Marta M. Gilson, John L. Cameron, Hari Nathan, Joseph M. Herman,Martin A. Makary, Frederic E. Eckhauser, Kenzo Hirose, Barish H. Edil, Michael A. Choti,Richard D. Schulick, Christopher L. Wolfgang, Timothy M. Pawlik

Introduction: Surgical resection remains the only potentially curative therapeutic option forpatients(pts) with pancreatic adenocarcinoma(PAC). Over the last several decades, advancesin surgical technique and perioperative care have reduced perioperative mortality. Changesin perioperative management and morbidity associated with PAC surgery, however, remainpoorly characterized. We sought to define the utilization patterns of perioperative andoperative procedures for pts with PAC, as well as evaluate population-based temporal trendsin morbidity and mortality. Methods: Using Surveillance, Epidemiology and End Results(SEER)-Medicare linked data, we identified 2461 pts with PAC who underwent pancreaticresection from 1991-2005. We collected data to assess morbidity, mortality, and survivalwhile adjusting with the Elixhauser comorbidity index. Trends in preoperative comorbidityindices, perioperative management, type of surgical procedures performed, as well as changesin morbidity and mortality were examined. Results: Preoperative evaluation includedERCP(59%), CT(92%), MRI(14%) and PET(2%) with a temporal increase in the use of allfour diagnostic modalities (all P<0.005)(Table). Use of diagnostic laparoscopy increasedfrom 3% to 16%(P<0.001) over time. The proportion of pts who underwent total pancreatec-tomy(n=28; 1%) or pancreaticoduodenectomy(n=1945; 79%) did not change (both P>0.05)whereas distal pancreatectomy(n=333; 14%) increased over the study period(P=0.04). Therewas a temporal increase in median pt age(1991-1996: 71yo; 1997-2000: 72yo; 2001-2002:73yo; 2003-2005: 74yo;P<0.05) and number of pts with multiple preoperative comorbidities(Elixhauser comorbidities ≥3: 1991-1996, 10%; 1997-2000, 17%; 2001-2002, 25%; 2003-2005, 26%;P<0.001). Despite the increase in pt age and comorbidities over time, overallperioperative morbidity(53%) did not change during the study period(P=0.97). The mostcommon postoperative complications were bleeding and need for re-exploration, both ofwhich decreased over time(9% to 4% and 11% to 7%, respectively;both P<0.05). In contrast,there was a temporal increase in the number of percutaneous interventional procedures (8%to 12%;P=0.005). Perioperative mortality decreased by half over the study period(1991-1996: 6% vs. 2003-2005: 3%;P=0.04). The overall 1- and 5-yr survival was 53% and 13%.A modest improvement in median survival(12 to 16 months;P=0.01) was noted over thetime periods examined. Conclusions: Mortality associated with pancreatic resection for PAChas decreased by one-half. Despite surgical resection for PAC being offered to older pts withmore preoperative comorbidities, the incidence of perioperative complications remainedstable. Resection for PAC in an aging population with more medical comorbidities can beperformed safely, however further progress is necessary to decrease morbidity.Trends in Perioperative Staging, Morbidity, and Complications 1991-2005 for Patients withResected Pancreatic Adenocarcinoma

SS

AT

Ab

stra

cts