Abstracts

pancreatitis (nZ3), and other benign lesions (nZ3). The sensitivity, specificity, PPV,NPV, and accuracy rate of EUS elastography determined by qualitative score methodwere 100%, 70%, 89.28%, 100%, and 84.21%, respectively. Using strain ratio at 6.04 asa cut-off value; sensitivity, specificity, PPV, NPV and accuracy rate were at 84.6%,83.3%, 91.7%, 71.4%, and 84.2% respectively. Interestingly, among 31 patients withavailable histologies by EUS-FNA; the negative cytology and strain ratio!6.04 pro-vided a NPV at 100% and from these results malignant SPMs were able to beexcluded. Conclusion: In this prospective single-blind study, EUS elastography wasnot superior to EUS-FNA alone for the diagnosis of SPMs. However, the combinationof the results from FNA and elastography was able to exclude malignant SPMs.

Result of EUS elastography and FNA for diagnosis of solid pancreaticmasses

AB336 GASTRO

Strain ratio

INTESTINAL ENDO

Qualitative score

SCOPY Volume 79, N

FNA alone

Sensitivity

84.6% 100% 90% Specificity 83.3% 70% 100% PPV 91.7% 89.28% 100% NPV 71.4% 100% 80% Accuracy 84.2% 84.21% 92.86%

Mo1455Accuracy of the Hue Histograms in the Evaluation of Patients WithPancreatic Masses. Do We Get More With Histogram Ratio?Nadan Rustemovic1, Milorad Opacic*1, Katja Grubeli�c Ravi�c1,Zvonimir Ostoji�c2, Dalibor Opa�cI�c3, Marko Brinar1, Iva Ledinsky4,ANA VI�sNji�c2, Matea Majerovi�c11Dept of Gastroenterology, University Hospital Center, Zagreb, Croatia;2Medical Faculty University of Zagreb, Zagreb, Croatia; 3UniversityHospital Center, Zagreb, Croatia; 4University Hospital Center "Sisters ofMercy", Zagreb, CroatiaEUS quantitative elastography methods, hue histogram (HH) and the strain ratio arethe methods developed for non-invasive analysis and differentiation of pancreaticmasses . Technical details of the methods are described elsewhere (1,2, 3).Aims: To evaluate the diagnostic value of the HH in the patients with pancreaticmasses, and to determine the cut-off value between the pancreatic cancer andfocal pancreatitis. As the next step we calculated new variable hue histogram ratio(HHR) in attempt to improve sensitivity, specificity and accuracy of the method.Methods: In a prospective single center study 149 patients were examined, 105 withthe pancreatic masses and 44 controls. Elastography images were recorded usingPentax EUS linear probe FG 38 UX and EG- 3870 UTK in combination with Hitachi8500 and Hitachi Avius. Histograms were calculated by Hitachi software in regions ofinterest as Mode 1 (over the mass) and Mode 2 over the adjacent part of homoge-nous pancreatic tissue representing reference area. Calculation was performed bydividing the value of Mode 2 with value of Mode 1. After the final diagnosis wasestablished two groups were formed: pancreatic cancer group with positive cytologyor histology after surgery (58patients) and focal pancreatitis group with negativecytology and follow up after 3 and 6 months (47 patients). All statistical analysis hasbeen made in SPSS 14.0 (SPSS Inc., Chicago, IL, USA). Results are shown in table 1and table 2. Discussion: In the two studies published recently (1,2) post-processinganalysis was performed on the previously recorded qualitative elastography videos,and average value HHs were calculated by special computer program. Our resultswere achieved by software for HH integrated in Hitachi platforms with reversed huescale (0 represents the hardest and 255 the softest tissue structure). Mode 1 histo-gram measured over the tumor with cut-of value 86 showed high sensitivity incancer detection, but also disappointingly low specificity, especially if comparedwith results of previous studies (1,2) and studies using strain ratio calculation,including our results published separately . Slight improvement in specificity wasachieved using Mode2/Mode 1 ratio. Conclusion: Statistical analysis showed that HHwith a cut-off value of R86 reaches 100% sensitivity and just 45% specificity withoverall accuracy of 66 % (CI 61%-66%) in detection of pancreatic cancer. HistogramMode 2/Mode1 ratio with cut-of R 1.153 is slightly better with 98% sensitivity and50% specificity, with overall accuracy of 69% (CI 63%-70%). More studies on Hitachiplatforms are needed. References: 1. Saftoiu A et al. Gastrointest Endosc2008;68:1086-942. Saftoiu A et al. Endoscopy 2011;43:596-603 3. Iglesias-Garcia J etal. Gastroenterology 2010;139:1172-80.

Analysis of diagnostic validity of the Mode 1 histogram

Pancreaticcancer

Mode 1

Indicative!86

True positive58/58

Falsepositive 50/91

PPV54% (50%-54%)

Nonindicative

R86

Falsenegativey 0/58

True negative 41/91

NPV100%(90%-100%)

o. 5S : 2014

Pancreaticcancer

Sensitivity100%

(93%-100%)

Specificity 45%(41%-45%)

www

Analysis of diagnostic validity of the histogram ratio(Mode2/Mode1)

Pancreatic cancer

.giejo

Mode 2/ Mode 1

IndicativeR1.153

Truepositive57/58

Falsepositive46/91

PPV92%(87%-92%)

Nonindicative!1.153

Falsenegativey

1/58

Truenegative45/91

NPV100%(96%-100%)

Sensitivity98%(91%-100%)

Specificity50%

(45%-50%)

Mo1456‘Seeing Is Not Believing’- HistopathologicalCorrelation of Endoscopic Ultrasound (EUS) in Non-CalcificChronic Pancreatitis (NCCP)Guru Trikudanathan*1, Ahmad Malli2, Satish Munigala3, Yusheng Han4,Melena Bellin5, T.Y. Dunn6, Timothy L. Pruett6, Greg Beilman6,Jose Vega-Peralta1, Mustafa a. Arain1, Shawn Mallery1, David Sutherland6,Martin L. Freeman1, Rajeev Attam1

1Gastroenterology, Univerisity of Minnesota, Minneapolis, MN; 2InternalMedicine, University of Minnesota, Minneapolis, MN; 3Gastroenterology,St Louis University School of Medicine, St Louis, MN; 4Pathology,University of Minnesota, Minneapolis, MN; 5Endocrinology, The SchulzeDiabetes Institute, Minneapolis, MN; 6Surgery, University of Minnesota,Minneapolis, MNBackground and Aim: EUS is widely regarded as a sensitive and fairly specific test fordetection of chronic pancreatitis (CP). However prior studies correlating EUS withhistopathology for CP are limited by small sample size, and/or inclusion of manypatients without CP, limiting applicability to patients with painful CP. The aim of thisstudy was to assess correlation of standard EUS features for CP with surgical histo-pathology in a large cohort of patients with NCCP. Methods: Adult patients withNCCP undergoing total pancreatectomy and islet autotransplantation (TPIAT), be-tween 2008 and 2013 were identified from our institutional database. Patients wereincluded if they underwent EUS at our center within a year before surgery. Thepresence or absence of standard EUS (4 ductal and 5 parenchymal) features for CPwas determined by expert endosonographers using linear endosonography. Histol-ogy was obtained at time of TPIAT from the resected pancreas by wedge biopsy ofhead, body and tail. All histopathology was reviewed by a GI pathologist blinded toendosonographic features and clinical outcomes. Available pancreatic tissue wasgraded for severity of intralobular and perilobular pancreatic fibrosis by the Ammannclassification system. A fibrosis score (FS) R 2 was considered abnormal. A multi-variate regression analysis was performed for EUS features predicting fibrosis, aftertaking age, sex, smoking and BMI into consideration. Receiver operating charac-teristic (ROC) curve analysis and Spearman rank correlation coefficients (r) werecalculated. Results:68 patients (56 females & 12 males) underwent TPIAT for NCCPand had preoperative EUS. Sensitivity of EUS for NCCP was poor, with a sensitivity of83% for two features and 75% for 3 features. ROC curve showed that 4 or more EUSfeatures provided the best balance of sensitivity (61%), specificity (75%) and accu-racy (63%). 6 or more EUS features offered 100% specificity and 100% positivepredictive value in diagnosing NCCP. Although significant, correlation betweenstandard EUS features and degree of fibrosis was poor (rZ0.24, p!0.05). Multivar-iate regression analysis of all the EUS features after taking age, sex, smoking and BMIinto consideration showed that main pancreatic duct (MPD) irregularity was the onlyindependent EUS feature (pZ0.02) which predicted CP. Conclusions: This studydemonstrated that standard EUS criteria correlate poorly with histopathology ofNCCP. Up to four features of CP had a sensitivity of only 61%, suggesting that EUSalone cannot be used to rule out CP. Although a lower cut-off provides bettersensitivity, a threshold of 6 or more EUS features is needed for a definitive diagnosisof NCCP. MPD irregularity is the only feature of EUS with strong correlation withhistopathological NCCP.

urnal.org