AGA Abstracts infliximab. We describe the first case of disseminated mucocutaneous granulomas in a patient with UC receiving infliximab. Case: 19-year-old female with UC in clinical remission, who had been treated with infliximab (10mg/kg/every 8 weeks) for 2 years, presented with multiple skin, nasal and oral lesions developing over the course of three months. The skin lesions were erythematous annular plaques on the right hip, inner thighs and forearms. Multiple ulcers were found in the nasal turbinates and one on the hard palate. Repeat esophagogastroduodenoscopy and colonoscopy at the time of evaluation for these lesions showed mild chronic esophagitis and mild chronic proctitis. She had a normal MRI enterogra- phy. Her IBD serology (pANCA positive; ASCA IgA, IgG negative) was also supportive of a diagnosis of UC. Histological examination of the skin biopsy revealed a diffuse interstitial granulomatous infiltrate with lymphocytes and histiocytes. Biopsy of the nasal and oral mucosa revealed a marked submucosal inflammatory infiltrate composed of lymphocytes, plasma cells and eosinophils with granulomas and foci of necrosis. Special stains and cultures for bacteria, fungus and mycobacteria were negative. Evaluation was negative for Cryptococcus, Bordetella henselae, Blastomyces and Histoplasma. Angiotensin I converting enzyme was normal. CT scan of the sinuses and chest radiograph were normal as well. Discussion: Metastatic Crohn's disease, opportunistic infections and granulomatous reaction to infliximab were considered in the differential diagnosis. IGD was the most likely cause of these lesions in our patient. Metastatic Crohn's disease was unlikely given the quiescence of her intestinal disease compared to the activity of the mucocutaneous involvement. Oppor- tunistic infections, which can be fatal in the face of anti-TNF therapy were ruled out by negative cultures. Furthermore, in our patient, her skin and mucosal lesions had significantly improved with no other intervention except for the discontinuation of her anti-TNF therapy. Conclusion: Granulomatous diseases, including IGD, should be considered in the differential diagnosis of annular or ulcerated lesions in the setting of anti TNF- therapy in a patient with IBD. T1165 Generation of a New Classification Scheme for Probe-Based Confocal Laser Endomicroscopy By International Experts Including Their Interobserver Agreement and Accuracy Frank J. van den Broek, Alexander Meining, Michael B. Wallace, Anna M. Buchner, Susanne van Eeden, Jessica A. van Es, Paul Fockens, Evelien Dekker Introduction: Probe-based confocal laser endomicroscopy (pCLE) enables In-Vivo histology during colonoscopy for instant differentiation of neoplasia and non-neoplastic tissue. This technology may allow endoscopists to make real time diagnoses and treatment decisions without biopsies. Since pCLE (Cellvizio, Mauna Kea Technologies, Paris, France) is a relatively new technique, an international collaboration of pCLE users was erected to generate a uniform classification for colonic use. The aims of this study were to assess the interobserver agreement and accuracy of this new classification scheme. Methods: Patients undergoing surveillance colonoscopy for a history of ulcerative colitis (UC) or polyps were included. Detected lesions and randomly chosen areas of normal colonic mucosa were inspected by pCLE before taking biopsies. Intravenous fluorescein was used for tissue contrast. All video sequences were stored on a personal computer for post hoc analysis. First, a subset of 10 video sequences was used to generate a uniform classification by an international collaboration of pCLE users. Subsequently, 5 blinded pCLE users scored a different set of video sequences by the new classification. Histopathology was used as reference standard. Results: The new pCLE classification comprised 3 categories (1-3) concerning vessel type and 7 categories (1, 2a-e, 3) concerning crypt type. In total, 23 patients (13 UC) underwent colonoscopy during which 102 pCLE video sequences were made of 47 lesions (5 UC-associated neoplasia, 5 adenomas, 10 sessile serrated adenomas, 27 non-neoplastic) and 55 control areas (non- neoplastic). The interobserver agreements on vessel and crypt type were ‘fair' with kappa values of 0.29 and 0.28 respectively. When reducing the classification to neoplasia (i.e. vessel/crypt type 3) vs. non-neoplastic, the agreements became ‘moderate' (kappa 0.37 and 0.42). For differentiating neoplasia from non-neoplastic tissue, the consensus diagnosis on the new classification had an overall sensitivity, specificity and accuracy of 80%, 84% and 83%. When analyzing crypt type separately, only in case at least 4 of the 5 assessors fully agreed on their crypt diagnosis (in 34 cases), the sensitivity and specificity became 100% and 97%. Conclusion: A new classification for pCLE has been generated by an international collaboration group, having a ‘moderate' interobserver agreement for differentiating neoplasia from non-neoplastic tissue. The overall accuracy (84%) was acceptable for our learning phase; and was excellent (97%) for crypt type with high agreement only. Future research should focus on simplification and validation of the classification. T1166 Interobserver Agreement and Accuracy Among International Experts of Probe- Based Confocal Laser Microscopy (pCLE) in Predicting Colorectal Neoplasia Victoria Gomez, Anna M. Buchner, Evelien Dekker, Frank J. van den Broek, Alexander Meining, Muhammad W. Shahid, Marwan Ghabril, Paul Fockens, Michael B. Wallace Introduction and aims: The high resolution probe-based confocal laser endomicroscopy system (pCLE) has recently been developed that may allow imaging of surface colonic epithelium In Vivo during any ongoing endoscopy with the potential of immediate diagnosis of GI premalignant and malignant lesions. The aims of this study were to assess the interobserver agreement and accuracy of this novel technology for prediction of colorectal neoplasia. Methods: Patients undergoing surveillance and screening colonoscopies were enrolled in this study. All detected lesions were inspected by pCLE before sampling or performing polypectomy. Intravenous fluorescein was used to enhance tissue contrast. All pCLE video sequences were stored on a personal computer for post hoc analysis. First, a subset of 10 video sequences was used to generate a uniform classification by an international collaboration of 5 pCLE users. Subsequently, the 3 pCLE users (excluding the 2 users whom generated the videos) reviewed a different set of video sequences using the crypt architecture and vessels features for colorectal neoplasia prediction in a fashion blinded to histopathologic and endoscopic findings and compared to a gold standard histopathologic diagnosis from the corresponding biopsy sites. Results: 53 patients underwent colonoscopy during which 81 pCLE video sequences were made of 81 colorectal lesions (50 neoplastic, 31 benign). A-514 AGA Abstracts The interobserver agreements on vessel and crypt type were moderate with Kappa coefficient values of 0.58 and 0.45 respectively. The interobserver agreements for the classification of neoplasia vs non neoplasia was moderate (Kappa coefficient values of 0.55 ( 0.47-0.62)). The overall consensus diagnosis for the differentiation neoplastic from non- neoplastic lesions based on crypt architecture and vessel morphology had a sensitivity of 75.7 %, specificity 70.5% and accuracy of 74.0%. Conclusion: The prediction of neoplasia using pCLE system by an international collaboration group was found to have a ‘moderate' interobserver agree- ment for differentiating neoplasia from non-neoplastic tissue. The overall accuracy rate (74%) was acceptable for the experts' learning phase of this new technology. T1167 Diagnostic Value of Endomicroscopy in Celiac Disease Ute Günther, Severin Daum, Frank Heller, Michael Schumann, Christoph Loddenkemper, Maria Grünbaum, Martin Zeitz, Christian Bojarski Background: The diagnostic follow up in patients with celiac disease (CD) is essential to identify refractory CD and to early detect enteropathy- associated T-cell lymphoma. Targeted biopsies facilitate the yield of conventional histopathology. Endomicroscopy is a non-invasive method and allows In Vivo histology during ongoing endoscopy. Methods: Patients with known CD with or without clinical improvement under a gluten-free diet and 25 controls were examined by endomicroscopy and conventional histology and were evaluated for villous atrophy (VA), crypt hyperplasia (CH) and intraepithelial lymphocytes (IEL > 40/100 enterocytes). Sensitivity, specificity and interobserver variability of endomicroscopy were assessed. Marsh classification score determined by endomicroscopy was compared to Marsh score of conventional histology within the last 24 months. Results: 25 control patients and 25 patients with known CD were examined. Histology detected 19 of 25 CD patients with VA and CH and 23 of 25 CD patients with increased IEL. Earlier performed random biopsies in these patients showed VA, CH and increased IEL in 13, 10 and 15 patients, respectively. With endomicroscopy VA, CH and increased IEL were identified in 15, 10 and 19 patients. The consistency of endomicroscopy with conventional histology was good for villous atrophy (sensitivity 78%) and IEL (82%) and insufficient for crypt hyperplasia (52%). IELs were semiquantitatively assessed. The kappa values of intraobserver variability were 0.95 for VA, 1.00 for CH and 0.85 for IEL detection. In 25 control patients normal duodenal architecture was found in histology and endomicroscopy indicating an overall specificity of 100%. The preselection of biopsy sites led to an increase in the Marsh score in 3 of 6 patients with refractory CD. Discussion: Assessment of duodenal histology by endomicroscopy in patients with CD is highly sensitive and specific in determining increased IEL and VA, but insufficient for CH. Endomicroscopy in CD may lead to a higher detection rate of patients with refractory CD. Preselection of biopsy sites contribute to a more effective histology in CD patients with persistent symptoms under a gluten-free diet. T1168 In Vivo Endomicroscopy and Gastroesophageal Reflux Disease Kerry B. Dunbar, Eun Ji Shin, Patrick I. Okolo, Anne Marie Lennon, Elizabeth A. Montgomery, Marcia I. Canto Diagnosis of GERD can be challenging, particularly in patients with atypical symptoms or nonerosive disease. Endomicroscopy (EM) can evaluate the esophageal mucosa In Vivo to visualize the cellular and vascular changes associated with chronic reflux injury in this otherwise difficult population. AIM: To evaluate the EM features of GERD and correlate them with esophageal pH measurement. METHODS: Patients with typical and atypical GERD symptoms were recruited and completed a validated GERD questionnaire. Fluorescein-aided EM was performed in all patients to examine normal-appearing squamous mucosa at 3-5, 13 and 18cm above the GEJ. Dilation of intracellular spaces (DIS), increased number of intrapapillary capillary loops(IPCLs) per screen,'halos' or a glow around IPCLs, and fluor- escein leakage on EM were recorded and blindly reviewed. Subsurface optical sectioning was performed to determine the distance from the mucosal surface to the proximal level of the esophageal papillae by identifying the IPCLs. A Bravo wireless pH capsule was then placed in each patient and 48 hour pH data were collected. Rank sum test was used to compare the EM GERD features between patients with abnormal(Demeester score >14.72) and normal Bravo results. RESULTS: To date, there is complete data on 13 patients (69% female, mean age 49, range 24-74) in this ongoing study. 9 (69%) had heartburn (HB), 8(62%) had regurgitation, 6 (46%) had both and 2 (15%) had atypical symptoms (cough). 9 patients (69%) had abnormal Bravo results. In the distal esophagus, there was no difference in the prevalence of DIS, halos, fluorescein leakage, or IPCL density between the patients with abnormal and normal Bravo results. In the midesophagus, patients with an abnormal Bravo had a higher mean IPCLs of 5.8 vs. 3.75 (p=0.04) in patients with normal Bravos. In the proximal esophagus, a mean of 5.6 IPCLs were found in patients with an abnormal Bravo vs. 3 (p=0.07) in Bravo normal patients. Patients with an abnormal Bravo were more likely to have >5 IPCLs by EM (p=0.03) in the mid and upper esophagus. The distance of the top of the IPCLs from the surface was less in the upper esophagus in patients with a positive Bravo (124 um vs 192 um, p = 0.04), but there was no significant difference in the distal esophageal measurements. CONCLUSIONS: In Vivo EM imaging features of the normal-appearing mid and proximal squamous esophagus in patients with suspected GERD are associated with abnormal Bravo results. Preliminary data for this ongoing study show that increased number of IPCLs and IPCLs close to the mucosal surface are present in patients with acid reflux confirmed by wireless pH monitoring.