AG

AA

bst

ract

sSu1228

Increased Risk of Colorectal Cancer in Patients With Prior Ovarian CancerDiagnosis: A Population Based US StudyYezaz A. Ghouri, Sachin Batra, Nirav C. Thosani, Sushovan Guha

Background: Increased incidence of developing colorectal cancer (CRC) among patients withprevious diagnosis of ovarian cancer (OC) were described in a few population based studiesfrom Europe and a US-based study published over a decade ago. Thus, we performed apopulation based study evaluating the risk of developing subsequent CRC in OC patientsusing Surveillance, Epidemiology, and End Results (SEER-9) registry of US population.Methods: SEER-9 database from NCI was queried to identify adult patients from 1973-2000with an index diagnosis of ovarian adenocarcinoma (ICD-O-3 code C56.9), after excludingany patient with cancer prior to being listed with SEER and allowing sufficient time forsubsequent CRC to develop. The outcome of interest was colorectal cancer (ICD-O-3 sitecode C18.0 - C18.9, C19.9). Patients developing lesions within 12 months of each otherwere regarded as synchronous lesions and were therefore excluded. Relative risk (RR) ofdeveloping CRC was determined comparing the observed incidence of subsequent CRCamongst OC survivors to that in general population with similar demographics and timeperiod. Byar's accurate approximation to the exact Poisson probability was used to calculate95% confidence intervals (CI) and determine statistical significance. All analysis was per-formed using SEER*STATA software. Results: Patients with history of OC had 1.33 foldgreater incidence of CRC than controls (RR 1.37, 95% CI: 1.2-1.48) with an absolute riskexcess of 3.85 cases per 1000 as observed over 242, 404 person years of follow-up. Themean ages for developing OC and for subsequent CRC were 58 and 70 years, respectively.This increase in risk was greatest in the population diagnosed with OC aged between 18-49 years (RR 3.82, 95% CI: 2.44-5.68, risk excess 3.37 cases per 1000) with the shortestlatency period between OC and CRC (mean age ovarian: 39 year and mean age CRC: 43years). The OC survivors who received radiation had a marginal increase in risk (RR 1.65;95% CI: 1.24-2.15) when compared to those who had not been treated with radiation (RR1.29; 95% CI: 1.15-1.44). The RR for subsequent CRC was minimally affected by changesin CRC screening practices over time with 34% risk during 1974-1990 (RR-1.35; 95% CI:1.19-1.52) and 30% risk during 1973-1990 (RR-1.30; 95%CI: 1.06 - 1.58). A reciprocalassociation for risk of OC amongst patients with index primary CRC was also observed withhighest relative risk amongst ages 20-49 years (RR 4.32; 95% CI: 2.64-6.67) with an excessrisk of 4.22 cases per 10000. Conclusions: Our study showed higher risk of developingCRC among patients who have been diagnosed with OC and support early CRC screening inyounger OC patients (<50 years). Several factors may contribute to this observed bidirectionalassociation including genetics, radiation exposure, and aggressive screening in cancer survi-vors.Table showing site specific risk of colon cancer in patients with preexisting ovarian cancer

Su1229

Interval Cancers After Screening Colonoscopy Performed in the Setting of aContinuous Quality Improvement Program With Monitoring of EndoscopistPerformance Are Associated With Low Endoscopist Adenoma Detection RateRobert J. Hilsden, S. Elizabeth McGregor, Ronald Bridges, Steven J. Heitman, AlaaRostom, Catherine Dube

Purpose: To describe the frequency and characteristics of interval colorectal cancers (CRC)occurring after screening colonoscopy at an endoscopy unit that regularly monitors endoscop-ist performance indicators. Methods: All patients aged 18-75 years who underwent a CRCscreening-related colonoscopy at the Forzani & MacPhail Colon Cancer Screening Centre,Calgary, Canada, from 2008-2010 were linked to the provincial cancer registry in October2013. Patients were allocated to endoscopists (gastroenterologists or colorectal surgeons)from a common queue. All procedures were performed with high definition colonoscopes.Endoscopist's withdrawal time (WT) was calculated for procedures where no polypectomywas performed and adenoma detection rate (ADR) was calculated as the proportion of caseswith at least one adenoma, cancer or sessile serrated adenoma detected. An interval cancerwas defined as an invasive CRC occurring more than six months after the index colonoscopy.Data on the index procedure and on the interval cancer was independently abstracted fromthe clinical record by three gastroenterologists. The cause of the interval cancer was classifiedby consensus using the Polyp Prevention Trial's algorithm for cancer occurrence after colonos-copy (GIE 2005). Results: The study included 21,351 patients (53% female, mean age 56)who underwent colonoscopy for screening (89%), post-polypectomy surveillance (5%) orfor investigation of a positive screening test (5%). Of those undergoing primary screening,45% were at average risk and 55% were at increased risk due to a family history. The 51endoscopists had a mean WT of 5 minutes and a mean ADR of 26%. The index colonoscopywas complete to the cecum in 98% and 25% had a neoplastic lesion detected. A cancer wasdetected at the index colonoscopy in 116 (28% proximal to splenic flexure). An intervalCRC was diagnosed in 10 patients (0.05%) at a mean interval of 38 months (range 9 - 66)and 50% were proximal. In 7 patients the interval CRC was asymptomatic, and in 6 ofthese it was found at a planned surveillance colonoscopy. Six of the CRCs occurred inpatients with no index neoplasia. The cause of the interval CRC was incomplete colonoscopy(1), missed cancer (2), incomplete polyp removal (1), failed biopsy diagnosis (1), new cancer(4) and other (1). Patient non-compliance with post-colonoscopy recommendations was acontributor in 2 cases. No interval CRC occurred in patients whose index colonoscopy wasperformed by an endoscopist with an average WT > 6 minutes (P=NS) or an ADR > 25%(P=0.002). Conclusions: An early interval cancer was detected in 0.05% of patients after

S-408AGA Abstracts

screening colonoscopy. All cases occurred after colonoscopies performed by endoscopistswith lower WT and ADR. Monitoring and enforcing mandatory WT and ADR and surveillanceguidelines are warranted in an organized screening program.

Su1230

Colon Capsule Endoscopy Versus Colonoscopy in Familial Colorectal CancerScreening: a Randomized Controlled TrialZaida Adrián de Ganzo, Onofre Alarcón, Antonio Z. Gimeno-García, Laura Ramos,Inmaculada Alonso-Abreu, Marta Carrillo-Palau, Enrique Quintero

Introduction: The efficacy of screening colonoscopy in first-degree relatives (FDR) of patientswith colorectal cancer (CRC) is limited by poor adherence (<40%). Colon capsule endoscopy(CCE) is a minimally invasive procedure that could improve participation in CRC screeningin this population. Aim: to compare adherence to CCE versus colonoscopy in familial CRCscreening. Methods: asymptomatic FDR aged ≥40 years or ten years younger than the case-proband were randomized (1:1) to 2nd generation CCE (PillCam colon2, Given Imaging)or colonoscopy, before signing informed consent (Zelen design). FDR who underwentprevious CRC screening, had a personal history of colorectal neoplasia, inflammatory boweldisease or hereditary CRC syndrome were excluded. Crossover between groups was permittedand both procedures were free of charge. Subjects assigned to CCE who had advancedcolorectal neoplasia (at least one polyp ≥10 mm, ≥3 polyps <10 mm or CRC) were invitedto undergo colonoscopy. Participants completed a questionnaire to clarify the reasons forcrossover. Assuming a participation rate of 38% and 58% for colonoscopy and CCE screening,respectively, with an alpha risk of 0.05 and a beta risk of 0.20 in a two-sided test, 108 FDRin each arm of the study were needed to find a statistically significant difference. Results:Overall, 329 FDR were randomized to undergo either CCE or colonoscopy screening. Ofthese, 41 could not be contacted and 55 were excluded (27.3% in the CCE group and31.1% in the colonoscopy group). Finally, 120 FDR in the CCE group and 113 in thecolonoscopy group were included. According to the intention-to-screen analysis, adherencewas similar in the CCE group (55.8%) than in the colonoscopy group (52.2%) (OR 1.15;95% CI, 0.69 to 1.93; P=0.57). The crossover rate was significantly higher in the CCE group(57.4%) than in the colonoscopy group (30.2%) (OR 3.11; 95% CI, 1.51 to 6.40; P<0.01).The reason for changing the assigned strategy in 35 (89%) subjects in the CCE group was"to avoid a second bowel preparation in case of a positive result" whereas in 19 (100%) inthe colonoscopy group it was "fear of colonoscopy". In the per-protocol analysis, adherencewas significantly higher in the colonoscopy group (37.2%) than in the CCE group (23.3%)(OR 1.85; 95% CI, 1.10 to 3.43; P<0.03). Advanced colorectal neoplasia was detected in13 (28.9%) subjects undergoing CCE and in 16 (19.8%) receiving colonoscopy (OR 2.17;95% CI 0.85 to 5.50; P=0.10). Conclusions: 1) Adherence to screening CCE was inferiorto that of screening colonoscopy in this familial risk population; and 2) CCE may be a validrescue strategy for subjects rejecting screening colonoscopy as it may increase screeningadherence up to 52%.

Su1231

Implementation of a Brief Familial Cancer Assessment Tool ImprovesSurveillance Recommendations and Identification of Patients At IncreasedRisk for Colorectal CancerTannaz Guivatchian, Erika S. Koeppe, Caitlin Foor-Pessin, William D. Chey, Shanti L.Eswaran, Joseph C. Kolars, Stacy B. Menees, Michael Rajala, Michael D. Rice, Rafat Rizk,Joel H. Rubenstein, Pratima Sharma, Andrea Todisco, Elena M. Stoffel

Background: Family history of cancer is one of the primary criteria used for colorectal cancer(CRC) risk stratification. Although 30% of individuals diagnosed with CRC have a familyhistory of the disease, most endoscopy units do not have a standardized method for collectingand reviewing patients' family history. Our aim was to implement a familial cancer riskassessment tool among patients presenting for outpatient colonoscopy and to test its effecton CRC risk stratification and surveillance recommendations. Methods: We conducted chartreviews of patients referred for outpatient colonoscopy at a large academic medical centerbefore and after implementation of a cancer risk assessment tool. The 5 yes/no questionsabout the patient's family and personal history of CRC and polyps were adapted from avalidated tool and completed by patients (Table 1). Subjects were stratified as average,increased, or high risk based on the definitions provided by the 2008 American Gastroenterol-ogy Association (AGA) CRC screening guidelines. Endoscopist recommendations for CRCscreening/surveillance were assessed for adherence to these guidelines. Results: Charts from800 sequential colonoscopy patients were reviewed (200 before and 600 after implementationof the CRC risk assessment tool). Baseline characteristics including age, gender, procedureindication, and prevalence of adenomas were similar between the two groups. Overall, 45%of patients were average risk for CRC, 40.3% were at increased risk, and 14.6% were highrisk based on their personal and/or family history. More patients at increased/high risk forCRC on the basis of family history alone were identified in the intervention group (33% vs22%, p=0.003) and patient referrals for genetic evaluation increased from 6 (3%) pre-intervention to 32 (5.3%) post-intervention. Surveillance recommendations were consistentwith AGA guidelines in 91% and 95% of cases before and after implementation of the riskassessment tool, respectively. In the subset of cases in which the endoscopists confirmedthey reviewed the tool, the number of inappropriate surveillance recommendations wassignificantly lower compared to the pre-intervention group (1.9% vs 9%, p=0.02). Among21 endoscopists surveyed, 85% found the CRC risk assessment tool helpful, 71% indicatedit influenced their surveillance recommendations, and 28.5% said it prompted them to refera high risk patient for genetic evaluation. Conclusion: Use of a 5 question cancer riskassessment tool during outpatient colonoscopy facilitates identification of patients with familyhistory conferring increased risk for CRC and reduces discrepancy between endoscopists'recommendations and published guidelines for CRC screening and surveillance.Table 1: 5 Question Cancer Risk Assessment Tool

Adapted from Kastrinos F, et al., Development and validation of a colon cancer risk assessmenttool for patients undergoing colonoscopy. Am J Gastroenterol, 2009 June. 104(6):1508-1518.

Su1232

Sessile Serrated Polyp Prevalence Determined by a High Level Detector and anExpert PathologistKhaled H. Abdeljawad, Krishna C. Vemulapalli, Charles J. Kahi, Oscar W. Cummings,Dale Snover, Douglas K. Rex

Background: The prevalence of sessile serrated adenomas/polyps (SSA/Ps) is uncertain.Objective: Determine the prevalence of SSA/P and SSA/P with cytological dysplasia (SSA/P-CD) using a known high level detector and an experienced pathologist. Design: Cross-sectional screening colonoscopy study. Settings: Academic endoscopy unit. Patients andInterventions: There were 1910 average risk asymptomatic patients ≥ 50 years old whounderwent screening colonoscopy between August 2005 and April 2012 by a single colonos-copist with a proven high detection rate. Slides of all lesions in the serrated class proximalto the sigmoid, and all rectal and sigmoid serrated lesions > 5mm in size were reviewed byan experienced GI pathologist. Main outcome measurements: Prevalence of SSA/P. Results:There were 1910 patients with 656 lesions in the serrated class (as defined above) removedfrom 389 patients. Review by the experienced GI pathologist determined a prevalence ofSSA/P of 7.2%, and SSA/P-CD of 0.9% (total SSA/P prevalence 8.1%). SSA/P and SSA/P-CD comprised 5.9% and 0.5% of all resected polyps. The mean size of SSA/P was 7.13mm(SD 4.66) and 51/77 (66.2%) of polyps ≥ 10mm in the serrated class were SSA/Ps. Limitation:Retrospective design. Conclusion: Using a proven high detecting colonoscopist and anexperienced pathologist, we identified a high prevalence (8.1%) of SSA/P in a screeningpopulation. SSA/Ps are more common than previously believed.Polyp distribution according to size ranges per initial pathology review

SSA/P: sessile serrated adenoma/polyp. HP: hyperplastic. Mixed TA: HP: mixed tubularadenoma/hyperplastic.Polyps distribution according to size ranges per the experienced GI pathologist review

SSA/P: sessile serrated adenoma/polyp. TSA: traditional serrated adenoma. HP: hyperplastic.

Su1233

Increased Efficiency in Adenoma Detection With Screening and SurveillanceColonoscopies Over the Course of Scheduled BlockRoy D. Yen, Sachin Wani, Lindsay Hosford, Brian C. Brauer, Dennis J. Ahnen, Gregory L.Austin

Background: Physician (MD) factors associated with missing adenomas during colonoscopyinclude suboptimal technique, shorter withdrawal time (WT) and possibly MD fatigue. Whilesome studies suggest that the effect of queue position (Qp) impacts adenoma detection(AD), conflicting results have been reported. Aim: To determine the effect of Qp on AD andWT over the scheduled block in patients (pts) undergoing screening/surveillance colonoscopy(SSC). Methods: Pts age >40yrs who underwent screening or surveillance colonoscopy from10/2010 - 12/2012 at an academic center were included in this analysis. Exclusion criteriaincluded: h/o colorectal cancer, IBD, and procedures performed in inpatient setting. Data

S-409 AGA Abstracts

collected included pt demographics, procedure characteristics, and whether any adenomaswere detected. Each MD's colonoscopies were ordered by Qp. Multiple regression modelingwas used to adjust WT for endoscopic maneuvers (such as biopsy, polypectomy) that didnot contribute to examination time, specific for MD performing SSC. A multivariate mixed-effect logistic model was used to test the effect of Qp on outcome of AD; a multivariatelinear mixed model was used to test the effect of Qp on outcome of adjusted (adj) WT.Random intercept terms were used to account for the multiple SSCs performed by each MD.Test statistics were considered significant at α=.05. Results: 4,879 eligible SSCs performed by35 MDs (mean 4 colonoscopies per MD per day, SD =3; max 17) were included. 46.1%pts were male with mean age of 59.8 (SD 8.9). Adenomas were detected in 34.2% of pts.The mean adj WT was 10.4 min (SD 5.1). Qp was not found to be a significant predictorfor AD; however, results suggest a trend towards an association of increased AD withincreasing Qp (for increase in Qp by 1 colonoscopy: OR 1.03, 95% CI 0.99-1.06, p=.052).Independent positive predictors of AD included pt age, pt male gender, pt h/o polyps, fellowinvolvement, poor bowel prep, technically difficult procedure, and adj WT (Table 1). Qpwas found to be a significant predictor for adj WT (Table 2). A linear negative associationof adj WT with increasing Qp was seen (p=.0012), and for a difference of (Δ) 5 in Qp, adjWT showed a -3.6% change. Other independent predictors of adj WT included pt age, ptgender, pt h/o polyp, and quality of bowel prep. Conclusions: There was a trend towardsa linear effect of increased adenoma detection with increasing Qp. A significant negativeassociation was seen with adj WT with increasing Qp. This may suggest there is improvedMD efficiency in performing SSC with sustained (possibly increased) AD over the courseof procedure blocks. These results need to be validated in future prospective studies.Table 1. Independent Predictors of Adenoma Detection

Δ = Difference ofTable 2. Independent Predictors of Adjusted Withdrawal Time (Adj WT)

Queue Position (Qp): Δ5 Qp represents difference in Qp of 5 colonoscopies (eg, betweencolonoscopy #3 and #8)

Su1234

A Systematic Review on Diagnostic Test Accuracy of Fecal ImmunochemicalTests for Colorectal Cancer ScreeningAafke H. van Roon, Leonie van Dam, Lidia R. Arends, Ann G. Zauber, Graeme P. Young,Marjolein van Ballegooijen, Dik F. Habbema, Monique E. van Leerdam, Ewout W.Steyerberg, Ernst J. Kuipers

Introduction: The more recently developed fecal immunochemical tests (FITs) have severaladvantages over guaiac-based fecal occult blood tests for colorectal cancer (CRC) screening,but evidence of its accuracy is largely uncharted. The aim of this study was to review andcompare the diagnostic test accuracy measures (i.e., positivity rate (PR), detection rate (DR),and positive predictive value (PPV)) of published data concerning different FITs. Methods:Studies were identified with use of the electronic databases Medline, Embase, the CochraneLibrary, BIOSIS Citation Index, and Science Citation Index-expanded (through September2011). All (randomized and/or comparative) accuracy studies in which participants (i.e.,asymptomatic average-risk individuals ≥ 40 years of age) with a positive FIT result werereferred for colonoscopy were included. There were no restrictions on date or language.Results: We analyzed data from 44 studies; 20 evaluated the performance characteristics of12 qualitative FITs, and another 24 papers evaluated 5 quantitative FITs. A large variationwas seen between the diagnostic test accuracy measures; for 1-sample FIT screening the PRvaried between 4.2% and 35.0%, the DR of CRC varied between 0.2% and 0.8%, and foradvanced adenomas the observed DR was between 0.7% and 5.5%. None of the investigatedFITs dominated others for these outcome parameters. Only for quantitative FITs a significantlyhigher DR of advanced adenomas was seen for 2-sample FIT screening; 1.1% (CI 0.8-1.6)vs. 2.7% (CI 1.7-4.5). However, this was at the cost of a significantly higher PR (4.8% (CI3.9-5.9) vs. 9.3% (CI 6.8-12.5), respectively). At last, we repeated the analysis correctingall data for the CRC prevalence in the population under investigation, which did not alterour main conclusions. Conclusions: This systematic review of accuracy studies of FIT testsin average-risk individuals does not provide evidence that one FIT outperforms others interms of PPV and DR of advanced neoplasia. Choosing between FIT brands is hindered bytoo few studies investigating the same test and by the heterogeneity in study design, useddefinitions, variation in FIT methodology, target population, and CRC prevalence rates.

AG

AA

bst

ract

s