the detection of 1) patients with endoscopically visible lesions, 2) number of endoscopicallyvisible lesions, 3) patients with HGD/EAC, and 4) number of HGD/EAC areas Methods:Data from a prospective, multi-center, randomized controlled trial evaluating the role ofnovel imaging techniques [High Definition White Light Endoscopy (HD-WLE), narrow bandimaging (NBI), and probe based confocal endomicroscopy (pCLE)] were reviewed. As partof the protocol, coordinators at each study site recorded the time spent inspecting the BEmucosa using HD-WLE (Olympus 180 HD, with clear cap, no magnification) using a stopwatch prior to biopsies being obtained. All visible lesions were described using the Parisclassification system. To determine each patient's final histologic diagnosis, every patientwas examined with HD-WLE, NBI, and pCLE. All suspicious areas as well as four quadrantrandom locations were biopsied then reviewed by a central pathologist. Fisher's exact testand an unpaired t-test were used to compare categorical and continuous variables, respect-ively. Results: 112 patients (mean age 65.5; 83.9% men, 100% Caucasian) with a mean BElength of 3.7cm were enrolled. The mean BIT with HD-WLE was 3.8min (SD 2.5min). 57patients had a visible lesion seen on HD-WLE examination and 38 patients had a finaldiagnosis of HGD/EAC. Patients with a BIT with HD-WLE < 5 min were less likely to havea visible lesion (32.4% vs. 82.9%, p<0.001) and less likely to have a final diagnosis of HGD/EAC (22.5% vs. 53.7%, p=0.002) compared to patients with a BIT ≥ 5 min. In addition,patients with a BIT of < 5 min had fewer visible lesions (0.51 vs. 1.95, p<0.0001) and fewerareas with HGD/Ca (0.51 vs. 2.29, p=0.004). This was despite no significant difference inmean BE length between patients with BIT < 5 and ≥ 5 min (3.3cm vs. 4.4cm, p=0.11).When patients with Paris I and III lesions were excluded, patients with a BIT < 5 minuteswere still less likely to have a visible lesion (29.4% vs. 79.4%, p<0.001) less likely to havea final diagnosis of HGD/EAC (19.1% vs. 50.0%, p=0.002), had fewer visible lesions (0.43 vs.1.82, p<0.0001), and had fewer areas with HGD/EAC (0.45 vs. 1.91, p=0.003). Conclusion: Alonger inspection time of the BE mucosa is associated with a higher rate of detection ofendoscopically visible lesions, patients with HGD/EAC, and areas with HGD/EAC. Endoscop-ists practicing BE surveillance should spend on average approximately 1 min per cm of BEusing HD-WLE prior to obtaining biopsies.

1167

Perception of High Esophageal Cancer (EC) Risk is Associated WithDecreased Quality of Life (QOL) in Patients With Barrett's Esophagus (BE)Neil Gupta, Srinivas Gaddam, Benjamin R. Alsop, April D. Higbee, Tracy Shipe, Sachin B.Wani, Amit Rastogi, Ajay Bansal, Prateek Sharma

Background: Prior studies have demonstrated that BE patients have decreased QOL comparedto normal healthy subjects without BE. However the cause of reduced QOL in these patientsremains unclear. Aim: To investigate whether personal perception of EC risk is associatedwith reduced QOL in patients with BE Methods: Patients with BE undergoing surveillanceor endoscopic therapy were prospectively enrolled between March 2010 and October 2010at a tertiary VA referral center. Demographic information, anti-reflux medication use, endos-copic history, and co-morbidities via the Charlson co-morbidity index were recorded. Allpatients completed the validated gastroesophageal reflux questionnaire (GERQ) to assessfrequency and severity of gastroesophageal reflux (GER) symptoms. Overall QOL was assessedusing a previously validated, paper based, standard gamble method. All patients were askeda single question to assess their perceived cancer risk (what do you think is your chanceof developing cancer of the esophagus in the next year?). They were asked to select fromthe following options: less than 1 in 1000, 1 in 1000, 1 in 500, 1 in 200, 1 in 100, 1 in50, 1 in 25, or more than 1 in 25. Those who estimated their annual cancer risk to be >1 in 500 (0.2%) were considered to have a high perceived cancer risk. The Mann-WhitneyU test and Spearmans correlation coefficient were used for univariate analysis and multivariatelinear regression was used to identify independent predictors of QOL scores. Results: 63BE patients (100% male, 100% Caucasian, mean age 64.1 years, mean BMI 29.6, mean GERduration 23.8 years, mean duration of BE diagnosis 8.8 years) were enrolled. The grade ofdysplasia was as follows: 25 (39.7%) no dysplasia, 16 (25.4%) low grade dysplasia, 19(30.2%) high grade dysplasia, and 3 (4.8%) esophageal cancer. 45 (71.4%) patients hadGER symptoms and 48 (77.4%) had a perceived annual cancer risk ≤ 0.2%. For overallQOL, the mean utility was 0.79 (SD 25.7) for the entire cohort. Patients with a perceivedhigh risk of EC had lower utility scores compared to those with a perceived low risk forEC (0.60 vs. 0.84, p=0.0005). There was no significant association between other factorssuch as age, BMI, GER symptoms, Charlson score, and dysplasia grade and overall QOL onunivariate analysis. By multivariate linear regression analysis, a perceived high risk of ECremained the only factor associated with decreased QOL (Beta coeff -26.4, 95%CI -13.3 --39.5). Conclusions: In this cohort of BE patients, patient perception of EC risk was theonly factor found to be associated with decreased QOL. This assumes special importancewhen one considers prior studies reporting that a significant majority of BE patients over-estimate their true EC risk. Future studies are needed to determine whether patient educationprograms can improve the QOL in BE patients.

1168

Technical Factors Cannot Explain the Limited Effectiveness of Colonoscopy inReducing Mortality From Right-Sided Colorectal CancerSameer D. Saini, Philip S. Schoenfeld, Laurence F. McMahon, Sandeep Vijan

Background: A recent large population-based study showed that while colonoscopy is effectivein reducing mortality from colorectal cancer (CRC) of the left colon (L-CRC) (OR 0.33,95% CI: 0.28-0.39), it has at best a modest effect in reducing mortality from CRC of theright colon (R-CRC) (OR 0.99, 95% CI: 0.86-1.14, suggesting at best a 14% mortalityreduction). This finding has been corroborated by other large population-based studies. Thepotential mechanism of this observed discrepancy in mortality reduction is unclear, thoughtechnical factors unique to colonoscopy in the right colon have been proposed (e.g., morefrequently missed or incompletely resected neoplasia). The purpose of this study was to usemodeling techniques to determine whether technical factors alone are likely to explain thelimited benefit of colonoscopy in R-CRC. Methods: We modified a previously-published,SEER-calibrated Markov cohort model of CRC to examine whether technical factors couldexplain the limited effectiveness of colonoscopy in reducing R-CRC mortality (at best, a

S-199 AGA Abstracts

14% mortality reduction). The Markov cohort consisted of average-risk 50 year-old patientswithout prior CRC screening, followed for 10 years. We modeled the effectiveness of one-time screening colonoscopy versus no screening. Primary outcomes were CRC incidence,CRC mortality, and the CRC mortality reduction due to colonoscopy. Sensitivity analysiswas performed on key variables, with a focus on the advanced adenoma (AA) miss rate andthe probability of incomplete AA resection, to see if technical factors could explain thelimited observed reduction in R-CRC mortality. Results: The model was well calibrated forL-CRC, with a cumulative incidence of 1.4 cancers per 1000 patients in the interventiongroup and 2.9 per 1000 in the control group (RRR 53%). Similarly, the cumulative mortalityrelated to L-CRC was 0.4 per 1000 in the intervention group and 1.1 per 1000 in thecontrol group (RRR 63%). For model output to match reported data for R-CRC, either themiss rate for AAs or the probability of incomplete AA resection had to exceed 80% (substan-tially higher than that reported in published literature). Assuming a lower AA miss rate of50%, the probability of incomplete adenoma resection still had to exceed 70% to matchobserved data. Conclusions: The limited effectiveness of colonoscopy in reducing mortalityfrom R-CRC cannot be the result of technical factors alone, given the implausibly high ratesof missed or incompletely resected lesions that would be necessary to explain observed data.This suggests that biological factors must play a critical role in the observed differences incolonoscopy effectiveness. Future efforts should focus on better understanding the riskfactors for interval R-CRCs and performing targeted interval screening for R-CRCs in patientswho are at risk for these cancers.

1169

Narrow Band Imaging With Optical Magnification: High Accuracies With HighConfidence in Predicting Histology in Colorectal LesionsRajvinder Singh, Nazree Nordeen, Swee Lin Chen Yi Mei, Garry Nind, Biju George,William Tam

Introduction: Narrow Band Imaging with optical magnification (NBI-Z) (Olympus Co.Ltd.Australia; available commercially) enables mucosal morphology to be assessed in real timewith magnification of upto 115X. Methods: Colorectal lesions detected were assessed withNBI-Z. Histology was predicted using the modified Sano's classification based on capillarynetwork patterns (cn); Type I: absent cn (hyperplastic polyp), Type II: cn present, surroundingmucosal glands (adenoma), Type IIIa: high density cn with tortuosity and lack of uniformity(intramucosal cancer) and Type IIIb: nearly avascular cn (invasive cancer). Each lesion wasalso graded with a confidence level (low/high). High definition videos (mean 28.2sec; range12-55) of each lesion assessed with NBI-Z were then taken. This was followed by polypectomy,endoscopic or surgical resection. NBI-Z diagnosis was then compared to the final histopatho-logy. To test for interobserver agreement, an endoscopist blinded to the video acquisitionprocess/histology was invited to grade the videos. Results: A total of 50 lesions (2 assessors:100 studies, with an average size of 8.4mm (range 3-30), in 32 patients were assessed. 20were hyperplastic, 25 adenomas, 2 intramucosal cancers and 3 invasive cancers of which19 were located in the right and 31 in the left colon. The overall accuracy of NBI-Z inpredicting histology was 90% which increased to 95% (88/93) when lesions were predictedwith high confidence. The Sensitivity (Sn), Specificity (Sp), Positive (PPV) and Negativepredictive values (NPV) in differentiating neoplastic from non neoplastic lesions with highconfidence were 98%, 89%, 93% and 97% respectively whilst the Sn, Sp, PPV and NPV inpredicting endoscopic resectability [Type II, IIIa vs. Type I, IIIb] was 100%, 90%, 93%and 100% respectively. The interobserver agreement between both assessors (kappa) wassubstantial at 0.89. Conclusions: Using confidence levels, NBI-Z permits prediction of col-orectal neoplasia with high accuracies and may allow prompt decisions to be made if alesion should be left in situ, resected and discarded or biopsied. This approach may leadto substantial time/cost savings and could potentially reduce complications associated withpolypectomy and endoscopic resections.

1170

Increased Adenoma Find Rate (AFR) and Proportionately Greater SerratedAdenoma Find Rate (SAFR) Following a Systematic Quality ImprovementProgram in a Large Community Gastroenterology PracticeJeffrey M. Rank

We report the result of an ongoing process to improve AFR within Minnesota Gastroenterol-ogy (MNGI), a private community based practice which performs > 30,000 colonoscopiesin 5 ambulatory surgery centers annually. We initiated a quality improvement program in2004 to track and improve AFR. This was a four-part program that was reported at DDW2009 (1028912). This included; discussion and education, monitoring of AFR and 6 minutewithdrawal time (WT), reporting individual and group data, and finally pay for performancebased on 6 minute WT. WT compliance, (based on time when the ileal-cecal valve is passedto time of extubation) increased from 78% in 2007 to 99% in 2010 (Graph 1). During thattime practice-level AFR rose from 28% in all males over 50 in 2007 to 42% in 2010 andin all females over 50 from 20% to 29% (Graph 2). AFR and SAFR is calculated bytotal number of colonoscopies and adenomas/serrated adenomas as determine by in-housepathology results entered directly into our EMR. Since 2006, there has been an increasedfocus on sessile serrated adenomas (SSA) (often flat, right-sided lesions) since they may beassociated with missed cancers. We instituted a program of education including writtenmaterial and two 1-hour physician education seminars in 2008 and 2009 focused on SSAdetection A program to compare indigo carmine exam after standard colonoscopy was alsoimplemented. The relative value of this was questioned and we have variable use throughoutthe practice. Numbers are too small to examine value at this time. Between 2007 and 2010,serrated adenomas found increased 100% from 590 to 1190 in absolute numbers, and SAFRincreased 86% from 3.8% to 7.1% (Inset Graph 2.1). Conclusion: The systematic qualityimprovement program instituted within MNGI was effective in increasing both AFR andSAFR. This demonstrates the value of reporting AFR and SAFR in a structured mannerwithin a practice that emphasizes physician feedback, performance measurement and finan-cial incentives.

AG

AA

bst

ract

s