S1582: Accuracy of Predicting Distal Colorectal Histology in Real-Time Using Narrow Band Imaging Without Optical Magnification

Abstracts histological correlation (59 adenomas, 59 HPs). Blinded to histology, each endoscopist independently scored the images as adenomas or HPs, and assigned a level of confidence (high, HC or low, LC) to each prediction. RESULTS: HC predictions were made for 86.4% (74.6-94.9) of adenomas and 73.2% (50.9-89.8) of HPs. For adenomas, the overall (all raters) sensitivity of the criteria was 93.3% (88.1-100), specificity 97.6% (94.9-100), accuracy 95.9% (range 92.4-97.5). The positive predictive value (PPV) of adenoma prediction was 97.5% (95.2-100) and 94.4% for HPs (89.0-100). For those polyps rated with HC, the sensitivity for adenomas was 98.0% (95.8-100), specificity 100% (100-100), accuracy 98.9% (98.0-100). The PPV for a HC adenoma prediction was 100% (100-100) and HC HP prediction 97.7% (96.4-100). The weighted multirater kappa coefficient for distinguishing adenomas from HPs was 0.87, and when limited to HC polyps was 0.97 indicating almost perfect agreement. CONCLUSIONS: Using simplified criteria, experienced endoscopists can determine polyp histology with a very high degree of accuracy, which is further enhanced by the use of confidence levels (achieving perfect specificity). They have almost perfect interobserver agreement when classifying a lesion with high confidence. The NBI classification system can allow the clinical application of endoscopic diagnosis, which would reduce overall colonoscopy costs by limiting pathologic examination of small colorectal polyps.

Methods: We prospectively evaluated 1247 patients with PET-CT for staging of primary cancer or suspected relapse of non-gastrointestinal tumor origin. In 56 cases (4.5%) suspect asymptomatic PET-spots in the GI-Tract were described. Among them in 35 patients the findings were accessible for endoscopic evaluation.Results: Upper GI-tract: 14 patients (40%). In 3 cases (21.4%) falsepositive PET-findings. 11 patients with pathologic findings (79%): 9 patients (82%) with malign findings, 2 patients (18%) with benign diseases (GERD). Lower GI-tract: 21 patients (60%). In 3 cases (14.3%) false-positive PET-findings. 18 patients with pathologic findings (86%): 7 patients (39%) with malign findings, 11 patients (61%) with benign diseases (adenomas, IBD). Overall in the GI tract in 5 patients (14.3%) false-positive PET-CT findings were detected. In 29/35 cases (83%) a pathologic intraluminal lesion was visible. Among them in 16/29 cases (55%) an asymptomatic second primary cancer was found and in 13/29 cases (45%) a benign disease respectively.Conclusion: Accidental findings in the gastrointestinal tract in the context of PET-CT-examinations are rare (4.5%), but the detection of asymptomatic GIT-Tracer-spots is accompanied by the prevalence of relevant comorbidities. Surprisingly endoscopic examination revealed in 46% of the patients with positive accidential findings a second primary cancer. Especially asymptomatic PET-CT tracer spots in upper GI tract need further endoscopic evaluation.

Table 1: Simplified NBI criteria for prediction of colorectal polyp histology Vascular intensity Vessels Pit pattern

Hyperplastic(TYPE 1)

Adenoma(TYPE 2)

Same or less than background None, or isolated lacy vessels coursing across the lesion Brown or black pits (dots, star or round shaped) surrounded by white

Increased Thick brown vessels White pits (oval, tubular or branched) surrounded by brown vessels

S1582 Accuracy of Predicting Distal Colorectal Histology in Real-Time Using Narrow Band Imaging Without Optical Magnification David G. Hewett, Douglas K. Rex Accurate colonoscopic assessment of histology could avoid resection of distal colorectal hyperplastic polyps, reduce costs and risk. AIMS: To assess the accuracy of predicting histology using narrow band imaging (NBI) in real-time for distal colorectal polyps, and to validate endoscopic criteria for predicting polyp histology. METHODS: We prospectively evaluated the real-time histology of 235 rectosigmoid polyps using high definition colonoscopy and NBI without optical magnification (CFH180AL, Olympus, Japan). For each polyp, in real-time, the endoscopist described location, size, Paris classification, and surface characteristics (vascular and pit pattern, overall color and presence of pseudodepression).1 Histology was predicted and a level of confidence (high or low) was assigned. All polyps were then resected and retrieved separately for blinded histological examination. The primary outcome was the accuracy of endoscopic histology prediction. Logistic regression was used to analyse specific endoscopic features associated with histology. RESULTS: We evaluated 235 polyps from the rectosigmoid colon (sigmoid, n⫽125, 61 patients; rectum, n⫽110, 41 patients): 38 adenomas (5.3mm, range 1-20mm), 188 hyperplastic polyps (mean 3.1mm, 1-8mm) and 9 other (lymphoid follicle, granulation tissue etc). The overall sensitivity of real-time prediction for adenomas was 89.5% (rectum 81.8, sigmoid 92.6) and specificity 94.4% (rectum 97.0, sigmoid 91.8). The overall accuracy was 93.6% (sigmoid, 92.0%; rectum, 95.4%). High confidence (HC) predictions were made for 216 (91.9%) polyps (sigmoid 110, rectum, 106). For HC adenomas, the sensitivity was 93.9%, specificity 98.4%, accuracy 97.7%. The positive predictive value for a HC rectosigmoid adenoma prediction was 91.2%, for HC hyperplastic predictions was 98.9%, and for rectal HC hyperplastic predictions was 99.0%. The NBI criteria significantly associated (P⬍.001) with adenoma histology were brown color, thick vessels and tubular or oval pits; and with hyperplastic histology were bland color, black dot pattern and lacy vessels. CONCLUSIONS: NBI without optical zoom is sufficiently accurate to allow distal hyperplastic polyps to be left in place without resection, and small distal adenomas to be discarded without pathologic assessment. These findings validate NBI criteria for predicting real-time colorectal polyp histology. 1. Rex Gastro 2009;136(4):1174-81

S1583 Endoscopic and Histological Evaluation of Asymptomatic Accidental Findings of the Gastrointestinal Tract in PET-CTImaging Thomas Kratt, Philip Aschoff, Christina Pfannenberg, Lena Minkley, Falko Fend, Alfred Ko¨nigsrainer Background: Due to the increasing spread of PET-CT-technique, ambiguous, asymptomatic, surprising findings in projection on the gastrointestinal tract occur more frequently. The clinical relevance of these findings remained unclear so far.

AB200 GASTROINTESTINAL ENDOSCOPY

Volume 71, No. 5 : 2010

S1584 Fluorescence-Guided Endoscopic Submucosal Dissection Using 5-Aminolevulinic Acid-Induced Protoporphyrin IX in Upper GI Tumors Yoshiro Kawahara, Hiroyuki Okada, Kazuhide Yamamoto BACKGROUND: To evaluate the clinical performance of photodynamic diagnostic (PDD) after oral administration of 5-aminolevulinic acid (5-ALA) for assessment of tumor type and surgical margins in Endoscopic submucosal dissection(ESD) in upper GI tumors. MATERIALS AND METHODS: This is a prospective, non-randomized single-center study. A total of 77 patients with a gastric cancer lesion and 4 patients with six Barret’s esophageal cancer lesions underwent ESD. For photosensitization, 1.5 g 5-ALA was administered orally 4 h prior to therapy. During the procedure, the resection site and the outer tumor border were inspected under excitation light for characteristic fluorescence using fluorescence endoscopy(Olympus FQ260Z,Olympus Japan) . The results of PDD were matched up to the histological findings. RESULTS: In all 83 lesions, 72(86.7%) lesions demonstrated fluorescence-positive endoscopic images and could be cofirmed negative margins during ESD. EGC lesions resected by ESD method showed PPIX fluorescence under the PDD examination. The intensity of the 635-nm emission peak of PpIX was quantified in 54 patients by spectroscopy.Sensitivity, specficity, and positive and negative predictive values was 96.4%, 82.1%, 84.4%, and 95.8% respectively. Furthermore, PDD with 5-ALA also identified all cases with negative resection margins, which were confirmed on histological examinations. No side effects of systemic 5-ALA administration were observed. CONCLUSIONS: PDD after systemic administration of 5-ALA is a reliable tool to assess the resection status of upper GI tumors during endoscopic resection.

S1585 Clinical Experience With a Second Generation, 6 Month Duration, Endoscopic, Duodenal-Jejunal Bypass Liner Alex Escalona, Luis I. Anrique, Frenando Pimentel, Allan C. Sharp, Manoel Galvao, Keith S. Gersin Background: The endoscopically delivered and retrieved duodenal-jejunal bypass liner (DJBL, EndoBarrier™, GI Dynamics, Inc, Lexington, MA) has been shown to promote weight loss and improve Type 2 diabetes. Here we report results from two ongoing studies of a second generation DJBL tailored to improve its stability for a minimum duration of 6 months.Methods: A modified DJBL was implanted in 61 subjects in two prospective, single arm studies at two institutions in Brazil and Chile. One study enrolled 39 subjects who were obese with a mean BMI of 43.8 ⫾ 0.9 kg/m2 (range: 35.4-58.2 kg/m2) while a second study enrolled 22 subjects who were obese with a mean BMI of 44.9 ⫾ 1.6 kg/m2 (range: 35.6-59.6 kg/m2) and with type 2 diabetes. These subjects had a mean baseline HbA1c of 8.9 ⫾ 0.4% (range: 6.7-11.6%). Assessments in both trials were similar and included monthly weights and safety evaluations. The diabetes trial also assessed HbA1c at all post-implant follow-up visits. All subjects were placed on a liquid diet for the first week post-implant, followed by a low-calorie diet for the remainder of the 24 weeks.Results: The change in subject weights at 24 weeks (n⫽48) were -15.0 ⫾ 1.1Kg (-27.0 ⫾ 1.7%EW, -13.1.0 ⫾ 0.9%TBW). In the diabetes study at week 24 (n⫽16), the change in HbA1c was -1.5 ⫾ 0.4% while 58% of the subjects (7/12 with data) achieved HbA1c of 7% or less. Implant delivery times were 26.0 ⫾ 1.6 minutes and used 8.0 ⫾ 0.5 minutes of fluoroscopy. Removal procedures took 11.0 ⫾ 3.0 minutes and used 1.9 ⫾ 1.0 minutes of fluoroscopy. There was one migration of the DJBL prior to 24 weeks (1.6%), a significant improvement (p⬍ 0.05) over the first generation design (31%). There were an additional 8 early removals due to device-related adverse

www.giejournal.org