Risk factors for adverse events related to polypectomy in the English Bowel Cancer Screening Programme *

Transcription

1 9 Original article Risk factors for adverse events related to polypectomy in the English Bowel Cancer Screening Programme * Authors Matthew D. Rutter 1,5, 6, Claire Nickerson 2, Colin J. Rees 3,5, 6, Julietta Patnick 2, Roger G. Blanks 4 Institutions Institutions are listed at the end of article. submitted 6. March 213 accepted after revision 1. August 213 Bibliography DOI /s Endoscopy 214; 46: 9 97 Georg Thieme Verlag KG Stuttgart New York ISSN X Corresponding author Matthew D. Rutter, MD University Hospital of North Tees Stockton-on-Tees Cleveland, TS19 8PE, UK Fax: Matt.rutter@nth.nhs.uk Background and study aims: The English National Health Service Bowel Cancer Screening Programme (NHSBCSP) is one of the world s largest organized screening programs. Minimizing adverse events is essential for any screening program. Study aims were to determine rates and to examine risk factors for adverse events. Patients and methods: Bleeding and perforations in NHSBCSP colonoscopies between August 26 and January 212 were examined. Logistic regression was used to examine risk factors for adverse events, including age, gender, polyp size, morphology, and location. For accurate attribution of adverse events, procedures with resection of only one polyp ( single-polypectomy ) were analyzed in detail. Results: colonoscopies ( polypectomies) were analyzed, including singlepolypectomies. Overall bleeding rate was.65 %, rate of bleeding requiring transfusion was.4 % Introduction Colonoscopy is the gold standard investigation for colorectal cancer and premalignant polyps. It is an invasive procedure which carries a risk of bleeding, perforation and even death [1 3]. Although the risk is small with diagnostic colonoscopy it is essential to keep the risk to a minimum, particularly in screening as patients are asymptomatic. A significant advantage of colonoscopy over other colorectal investigations is the ability to perform polypectomy, which is necessary to break the adenoma carcinoma sequence and reduce subsequent colorectal cancer development [4]. However polypectomy increases the risk of bleeding and perforation. The English NHS Bowel Cancer Screening Programme (NHSBCSP), based on fecal occult blood * Work undertaken on behalf of the National Health Service (NHS) BCSP Evaluation Group and perforation rate was.6 %. Polypectomy increased bleeding risk fold and perforation risk 2.97-fold. Cecal location (but not elsewhere in the proximal colon) and increasing polyp size were the two most important risk factors for bleeding and perforation. After adjustment for polyp size, the odds ratio (OR) relative to the distal colon for bleeding requiring transfusion after cecal snare polypectomy was 13.5 (95 %CI ) and for perforation after cecal nonpedunculated polypectomy it was 12.2 (95 %CI ). Conclusion: This is the largest study focusing on polyp-specific risk factors. We have confirmed that the greatest risk factor for both post-polypectomy bleeding and perforation is polyp size. This is the first demonstration of substantial and significantly increased risk for both noteworthy bleeding (requiring transfusion) and perforation from cecal polypectomy for a given polyp size, compared with elsewhere in the colon. testing, commenced in 26. By January 212, over 12 million people had been invited to participate, resulting in over 13 diagnostic colonoscopies. Details of the NHSBCSP structure have been published previously [5]. Colonoscopies are only performed by screening-accredited colonoscopists, who have undertaken a written and practical examination [2]. We have previously demonstrated high values in quality indicators for the screening colonoscopists. [2]. Previous peer-reviewed publications on colonoscopic adverse events have either examined adverse event rates for colonoscopy per se or looked at risk factors for adverse events, mainly following polypectomy. Overall colonoscopic perforation rates in four recent large series (from 97 to 277 colonoscopies) ranged from.3 % to.85% [6 9]. A review of recent studies calculated an overall perforation rate of.7 % (1 in 14) and a therapeutic rate of.1 % (1 in 1) [1]. For colonoscopy-

2 Original article 91 related bleeding, a recent large series reports a rate of.164% [9]. Bleeding is uncommon following diagnostic colonoscopy, but is however the most frequent significant adverse event following polypectomy: bleeding rates of.26 to 6.1% are reported for polypectomies [3, 11, 12]. Studies that have looked specifically at risk factors for adverse events have usually been smaller, mainly focusing on polypectomy. The two main risk factors appear to relate to the size and location of polyps [13]. Other reported risk factors include cardiovascular or chronic renal disease [14], age [11, 14,15], anticoagulant use [11, 14], and endoscopist experience [11]. Studies assessing polyp morphology are inconclusive, with some showing increased risk for pedunculated [15] or sessile/thick-stalked polyps [11], but others showing no effect [16]. To improve training and reduce occurrence of adverse events, it is important to understand which factors influence risk. This is challenging when multiple polypectomies have been performed as it is often unclear which polypectomy led to the complication. Arbitrary decisions, for example to attribute the complication to the largest or most proximal lesion, introduce bias and may be inaccurate or misleading. This study of data from the NHSBCSP examined a large number of colonoscopies. We aimed to determine: 1. The overall rates of colonoscopic bleeding and perforation 2. Risk factors for such adverse events. This would include: a) comparison of diagnostic versus therapeutic procedures b) multivariable analysis for patients undergoing polypectomy with resection of a single polyp ( single-polypectomy ), thus enabling accurate attribution of the adverse event to a specific polyp. Patients and methods The Bowel Cancer Screening Programme and adverse events All colonoscopies performed in the NHSBCSP between the start of the program in 26 and January 212 were studied. Within the NHSBCSP, all data from procedures and complications are recorded on a single national web-based database (Bowel Cancer Screening System, BCSS) by dedicated specialist screening practitioners (SSPs) who attend all procedures. All patients are contacted by the SSP the day after the procedure and also receive a 3-day questionnaire which includes questions about complications and whether the patient sought additional medical advice post-procedure (current questionnaire return rate is 83.7 %). Any adverse event is flagged on the BCSS and the event is reviewed by the regional screening colonoscopy lead person. We have previously demonstrated that this system is a reliable means of capturing colonoscopy-related adverse events [2]. An adverse event is defined by the NHSBCSP as an event that prevents the completion of the procedure (excluding technical failure or poor bowel preparation) and/or results in: (i) admission to hospital or prolonged hospital stay; (ii) another intervention (endoscopic, radiological, or surgical); or (iii) subsequent medical consultation. Our study focused on colonic perforation and bleeding. Periprocedural bleeding which is controlled during the procedure and which does not prevent completion of the procedure is not considered to be an adverse event. The NHSBCSP definition of colonic perforation is evidence of air, luminal contents, or instrumentation outside the gastrointestinal tract. Bleeding was analyzed for all grades of severity and subanalyzed for bleeding requiring transfusion [17]. Potential explanatory factors examined were age, gender, colorectal location, polyp morphology (pedunculated versus nonpedunculated), size of polyp, therapy device and American Society of Anesthesiologists (ASA) grade. Size of polyp was recorded as actual size in millimeters measured by the histopathologist; where this was not documented the colonoscopist-estimated size was used. Location was categorized as rectum, sigmoid, descending colon, splenic flexure, transverse colon, hepatic flexure, ascending colon, and cecum. For analyses examining rarer events (perforation and bleeding requiring transfusion) location was categorized into three areas (dictated by the results of initial data analysis, explained in the Results section): the distal colon, namely rectum to splenic flexure; the proximal colon excluding cecum (PCxC), namely the transverse colon to ascending colon; and lastly the cecum. The number of polyps is recorded on the BCSS as number seen, number resected, and number retrieved. For our study, we took the number of polyps resected as the variable of interest. Statistical analysis Logistic regression models were produced using STATA version 11 (StataCorp, Texas, USA) logistic and logit commands and modelled the log odds of the occurrence of an event against the explanatory variables. A statistical significance level of P <.5 was used in all models and pseudo R 2 and deviance were used to assess goodness of fit. Polyp morphology was analyzed as a dichotomous explanatory variable, that is, pedunculated or nonpedunculated. In the final model, size of polyp was entered as log size; this produced a better fit as the distribution of polyp sizes was positively skewed (log-normal). In a small number of cases (about 5%) an individual screenee contributed more than one procedure and to allow for this the models calculated robust standard errors. Results During the study period, colonoscopies were performed on participants (mean 1.17 per participant), during which polyps were identified and polyps resected. Of the colonoscopies, 6.7 % were on men. The mean age was 65.7 years, range years. The NHSBCSP invites 6 74-yearolds but older people may opt in; 98.7% were 6 74.) Rates of adverse events overall, for colonoscopies with and without polypectomy, and according to number of polyps resected The overall rate of adverse events was 1.42 % (1 in 71 procedures). The overall rate of bleeding was.65 % (1 in 154 procedures), the rate of bleeding requiring transfusion was.4 % (1 in 2469) and the overall rate of perforation was.6% (1 in 1596). In a small number of procedures there was more than one adverse event. In procedures no polyps were resected. The diagnostic colonoscopy perforation rate was.3 % (1 in 3253 procedures), bleeding rate.1 % (1 in 981), and rate of bleeding requiring transfusion was.2 % based on only 1 case (1 in 61 83). In procedures polypectomy was performed. The therapeutic colonoscopy perforation rate was.9 % (1 in 196 procedures), bleeding rate was 1.14 % (1 in 88) and the rate for bleeding requiring transfusion was.8% (1 in 1189).

3 92 Original article Table1 Polypectomy adverse event rate: for procedures with single-polyp resection, by polyp location for full colonoscopies. Location Procedures, n Adverse events Bleeding Perforation n % (95 %CI) n % (95 %CI) Rectum ( ) 2.4 (.4.13) Sigmoid colon ( ) 1.7 (.34.13) Descending colon ( ) 1.5 (.1.29) Splenic flexure ( ) 2.29 ( ) Transverse colon (.21.74). Hepatic flexure (.3.9). Ascending colon (.37 1.) 2.7 (.9.27) Cecum ( ) 3.13 (.26.37) Total ( ) 2.6 (.4.1) CI, confidence interval. Polypectomy increased the risk of bleeding by a factor of compared with no polypectomy, whereas for perforations the relative risk was Appendix 1 shows the observed adverse event rates by total number of procedures, grouped by number of polyps resected. Adverse events from single polypectomies, by location Analysis of procedures where a single polyp was resected ( single-polypectomy procedures) enables the most accurate attribution of adverse events. There were single-polypectomy colonoscopies, from which 15 were excluded because of incomplete data; thus procedures from participants (mean of 1.5 per participant) were analyzed. For single-polypectomy procedures the mean age of participants was 65.7 years and 61.3 % were men. " Table 1 shows the single-polypectomy adverse event rates by location. The absolute risk was U -shaped, being higher in the distal colon, becoming lower in the PCxC, and then higher again in the cecum, although the number of perforations was small. For bleeding, the risk in the distal colon was 1.5 % and the risk in the PCxC was.49%; the difference was highly statistically significant (P<.1). The risk for the cecum of 1.13 % was not statistically different from the distal colon (P=.71; test of two proportions). Based on this evidence, and particularly from the data on all bleeding adverse events, we grouped the data into the three sections of distal colon, PCxC and cecum as detailed in the Patients and methods section. Values according to location for the additional potential explanatory or confounding variables of gender, age, polyp size, and polyp morphology (pedunculated or nonpedunculated) are shown in Appendix 2; these also relate to single-polypectomy procedures. Both the size of polyps and the proportion of pedunculated polyps decreased with more proximal location in the colon (Appendix 2). Size and proportion with pedunculation were largest for polyps resected from the sigmoid colon, and only 4.4 % of polyps in the cecum were pedunculated. Neither mean age nor gender showed much variation over polyp location. Risk of adverse events from single polypectomies by size and other factors Log size was used in the logistic regression models because size was highly skewed with most polyps being small; this improved data presentation (see, for example, Appendix 2), provided better-fitting models and minimized the effect of outliers. The log size median value was 1.8 and the maximum value 5.6. Univariate logistic regression analysis of log polyp size demonstrated that polyp size was a highly significant predictor of bleeding (P <.1) and perforations (P <.2). Age, gender, polyp morphology and ASA grade were not significant predictors of risk. Risk of bleeding and of perforation as adverse events: logistic regression analyses " Table 2 shows the logistic regression analysis for significant factors affecting the risk of bleeding as an adverse event. Model 1 used only location as a categorical explanatory variable. Model 2 showed that after adjustment for polyp size, location of polyp in the cecum was highly significant (P <.1). This model gave a much better fit to the data, the pseudo R 2 increased to 1.6 %, and the deviance reduced considerably. The difference in the deviance measure compared with the chi-squared distribution indicated a highly significant increase in model fit (P<.1). In model 3, using all variables, resection by hot snare was also significant. No other factor was significant. " Table 3 shows similar data for the more severe bleeding subgroup of bleeding requiring transfusion. When bleeding requiring transfusion was examined, these risks were considerably magnified, with an unadjusted OR of 5.2 (P <.1) in the cecum relative to the distal colon (model 1), increasing to an OR of 1.6 (95 %CI ) after adjustment for polyp size, that is, in model 2. This model therefore suggested that, for any polyp of a given size, the risk of bleeding requiring transfusion was over ten times greater for cecal polyps compared with those in other colonic locations. " Table 4 shows the analysis for perforations. The effect of location in the cecum became significant only after the inclusion of polyp morphology in the model, after adjustment for size Risk of bleeding and perforation for hot snare resections only We subsequently restricted the analysis to hot (diathermy) snare resections, as all but diminutive polyps were removed by hot snare alone and the risk of adverse events from diminutive polypectomy was small. Whilst these results were only applicable to hot snare resections, the internal validity of the study was increased by removing potential confounding factors. These models are shown in " Table 5. In all four models location in the cecum was significant and in models 1 and 2 (risks of bleeding and of bleeding requiring transfusion, respectively, both including polyp location and log size) it was highly significant (P <.1). The results for model 1 are shown in " Fig. 1.

4 Original article 93 Table2 Risk of bleeding as an adverse event at polypectomy: logistic regression models, with location as a univariate factor and multivariate model including other factors. Data from single-polyp resections only. Logistic regression models, odds ratio (95%CI) Table3 Risk of bleeding requiring transfusion as an adverse event at polypectomy: logistic regression models, with location as a univariate factor and a multivariate model. (Data from single-polyp resections only. No other variables contribute to the multivariate model apart from polyp size. All other variables were not significant predictors of risk in analysis.) Logistic regression models, odds ratio (95 %CI) Univariate analysis (Model 1) Multivariate analysis (Model 2) Distal colon PCxC 1.2 ( ) 2.78 ( ) Cecum 5.2 ( ) ( ) 2 Log size* 5.54 ( ) 2 Pseudo R 2 (Deviance) 2.4% (282.2) Univariate (Model 1) 15.1 % (245.6) CI, confidence interval; PCxC, proximal colon excluding cecum. 1 P<.1 2 P<.1 * Risk based on change of polyp size by one unit in log size (e.g. 1 log unit change represents a size increase from 7mm to 2mm, or from 2mm to 54mm); see Figure 2. Multivariate with size only added (Model 2) Regarding perforation, model 4 (including location and log size) restricted the data to nonpedunculated polyps only (pedunculated polyps are rare in the cecum) and provided convincing evidence for a very high risk in the cecum (OR (95 %CI ) for the resection of nonpedunculated polyps. Multivariate full model (Model 3) Distal colon PCxC.46 (.31.67) 1.98 ( ) 1.8 ( ) Cecum 1.8 ( ) 2.13 ( ) ( ) 1 Log size* 3.9 ( ) ( ) 1 Hot snare 2.2 ( ) 2 Pseudo R 2 (Deviance) 6.1 % (3273.8) 1.6 % (2942.8) 1.9 % (2932.6) CI, confidence interval; PCxC, proximal colon excluding cecum. 1 P<.1 2 P<.1 * Risk based on change of polyp size by one unit in log size (e.g. 1 log unit change represents a size increase from 7mm to 2mm, or from 2mm to 54mm); see Figure 2. Risk of bleedings as adverse event per 1 procedures Polyp size, (mm) Fig. 1 Bleeding as an adverse event of polypectomy: modelled risk per 1 procedures by polyp size for cecal (upper line) and noncecal (lower line) locations. This analysis used data from only hot snare procedures where only a single polyp was resected. Discussion This study assessing all polypectomies performed within a national screening program is the largest to focus on polypspecific risk factors. Uniquely, we have been able to focus on colonoscopies where only one polypectomy was performed, allowing precise analysis of what intervention led to the complication. We have examined the risk of the adverse events of bleeding or perforation against a number of potential factors including age, gender, polyp size, morphology, location, and therapy device. Understanding of the risk factors for adverse events is important for enabling clinicians to select appropriate management strategies and for allowing patients to make informed decisions regarding risk. It also permits more refined quality assurance, targeted training and support for colonoscopists, and further improvements in therapeutic technique. In turn this should lead to safer procedures. Polyp size This study demonstrates that it is principally the size of the polyp that determines the risk of adverse events of bleeding and perforation. This was demonstrated on both univariate logistic regression analysis and multivariate analysis. With the exception of polyps located in the cecum, small polypectomies (< 5mm) confer very little risk of an adverse event whereas larger polypectomies can confer very high levels of risk, with a 4-mm polypectomy from the cecum having a predicted risk of bleeding as an adverse event that is as high as in 1 in every 8 procedures. It would be interesting to look at factors such as surface area for sessile polyps and volume for pedunculated polyps but this is beyond the scope of this study. Our study supports the findings of smaller studies. The Munich Polypectomy Study (MUPS) looked prospectively at 4 snare polypectomies [13]; on multivariate analysis, polyp size proved to be the main risk factor for significant adverse events (OR 31.1, 95 %CI ). Other studies have come to similar conclusions [11, 14 16,18 21].

5 94 Original article Table4 Factors affecting risk of perforation as an adverse event at polypectomy: logistic regression models. Data from single-polyp resections only. Logistic regression models, odds ratio (95%CI) Table5 Factors affecting risk of bleeding, bleeding requiring transfusion, and perforation as adverse events at polypectomy: multivariate logistic regression models, from single-polyp resections only and for hot snare resections only. Logistic regression models, odds ratio (95 %CI) All bleeding (Model 1) Bleeding requiring transfusion (Model 2) Perforation (Model 3) Perforation (sessile polyps only) (Model 4) Distal colon PCxC 1.17 ( ) 3.34 ( ).96 ( ) 2.65 ( ) Cecum 2.49 ( ) ( ) ( ) ( ) 2 Log size* 4.17 ( ) ( ) ( ) 2.2 ( ) Pedunculated 2.1 ( ) N/A Pseudo R 2 (Log likelihood) Univariate (Model 1) 8.% ( 134.5) Multivariate with size only added (Model 2) 12.3 % ( 119.5) 3.2 % ( 125.4) Multivariate full model (Model 3) Distal colon PCxC.48 ( ).83 ( ) 1.11 ( ) Cecum 1.87 ( ) 3.8 ( ) 4.49 ( ) 1 Log size* 2.42 ( ) ( ) 1 Pedunculated 2. ( ) Pseudo R 2 (Deviance).7 % (331.4) 3.9 % (32.6) 4.3 % (319.3) CI, confidence interval; PCxC, proximal colon excluding cecum. 1 P<.5.All other variables were not significant predictors of risk in analysis 2 P<.1 * Risk based on change of polyp size by one unit in log size (e.g. 1 log unit change represents a size increase from 7mm to 2mm, or from 2mm to 54mm); see Figure % ( 31.52) CI, confidence interval; PCxC, proximal colon excluding cecum. 1 P<.1 2 P<.5.All other variables were not significant predictors of risk in analysis. * Risk based on change of polyp size by one unit in log size (e.g. 1 log unit change represents a size increase from 7mm to 2mm, or from 2mm to 54mm); see Figure 2. Cecal location Because of the size of our study we have been able to assess adverse events by colorectal segment for the first time. Other studies have demonstrated that polypectomy in the proximal colon carries a greater risk, but have not had enough power to analyze according to colonic segment. For example the MUPS study reported that proximal polyp location was a significant risk factor for major complications (OR 2.4, 95 %CI ) [13]. The study of 216 polypectomies by Choo et al. also showed that right-colon polypectomies had a higher tendency to lead to post-polypectomy syndrome and bleeding (P =.2) [22]. Buddingh and colleagues case control study of 39 cases of delayed hemorrhage following polypectomy showed that polyps located in the right hemicolon had an OR of 4.67 ( ; P =.1) for delayed hemorrhage [16]. They also suggested that the cecum seemed to represent an especially high risk in univariate analysis (OR 13.82, 95 %CI ), but could not assess this in multivariate analysis as the number of cases was too small. The study of 4719 patients by Saraya et al. found that rectal location was a significant risk factor for noteworthy complications (OR 3.55) [15]. We have shown that after adjustment for size, it is cecal location that confers a higher risk for all adverse events. For example the risk of bleeding requiring transfusion following snare polypectomy is up to 14 times higher in the cecum. Other sites in the proximal colon do not confer an increased risk, and a simple dichotomous proximal/distal division is inadequate for studying this. The idea is biologically plausible as the cecum is the thinnest and most distensible part of the colon with saccular haustrations between the linear teniae coli. Insufflation during colonoscopy thins the wall further, leaving the cecum more prone to damage. A second explanation is that as cecal therapy is more often performed perpendicularly to the mucosa, compared with a more tangential approach elsewhere in the colon, there may be more potential for snaring deeper submucosa or muscularis propria, thus increasing the risk of complications. A third potential explanation is that in order to reduce the risks of diathermy damage and risk of perforation, in the thinnest part of the colon, clinicians may underutilize the coagulation element of diathermy thus increasing the risk of bleeding. This important new finding regarding cecal location needs to be considered by clinicians and patients when making an informed choice for therapeutic options. Polyp morphology Our study indicates that although pedunculation in a polyp confers an increased risk of bleeding, this is not an independent risk factor. Rather it is simply that pedunculated polyps tend to be larger, thus the risk relates to polyp size rather than morphology.

6 Original article 95 Low rates overall The overall perforation rate of.6 % compares favorably with other series, particularly when it is considered that over half of all colonoscopies in our series were therapeutic. Although the overall bleeding rate in our study was.65%, our definition for bleeding as an adverse event was wide, including many clinically insignificant bleeds. When the analysis is limited to bleeding requiring transfusion, the rate is low at.4%. In our previous analysis of adverse events of bleeding stratified according to severity, for intermediate or major severity (hemoglobin drop of 2 g, transfusion, admission to intensive treatment unit, unplanned hospital admission for 4 or more nights, interventional radiology or endoscopy, or surgery), the rate was.13% [2], similar to the figure of.164% in the recent series of Rabeneck et al. Study limitations The study database is observational and great care has been undertaken to avoid analyses which may lead to false-positive findings. Whilst the database is very large, like all observational data it is still subject to major limitations in studying extremely rare events such as death. Furthermore cold biopsy forceps, hot biopsy forceps and cold snare are only used to remove small polyps which have a very low risk of adverse events and therefore there is little power to study these factors. A potential drawback of our study is that it was performed on patients undergoing colonoscopy within a national fecal occult blood-based screening program, and excluded any patient below the age of 6 years. This may reduce the generalizability of the results. However as most patients undergoing colonoscopy are over the age of 6, we feel the results remain broadly representative. Moreover, within our study population we did not see an effect of age on adverse event rates. Future research In any endoscopy service it is essential to monitor the quality of endotherapy and to maximize safety. This is particularly pertinent in a screening program. Further to our current analyses, it should be possible to develop mathematical models of risk to estimate the risk of adverse events by screening unit or individual colonoscopist. Further work will determine risk models for multiple polypectomies that can be applied to such measures. In so doing we should be able to compare an individual colonoscopist s risk of an adverse event, weighted for procedure complexity, against the average risk in the English BCSP, thus aiding quality assurance, enabling review and support for underperforming colonoscopists, and maximizing safety. Summary Our study on the adverse events of bleeding, bleeding requiring transfusion, and perforation is the largest to focus on polypspecific risk factors. Uniquely, we have been able to focus on colonoscopies where only one polypectomy was performed, allowing a detailed examination of the characteristics of the polyp associated with the adverse event. We have confirmed that the greatest risk factor for an adverse event is polyp size. For the first time we have demonstrated a substantial significantly increased risk for both bleeding requiring transfusion and perforation for polyps in the cecum, but not elsewhere in the proximal colon, when compared with polyps in the distal colon for a given polyp size. Endoscopists need to take particular care when performing cecal polypectomy. Appendix 1: Bleeding and perforations in the English National Health Service Bowel Cancer Screening Programme (NHSBCSP) " Table 6 shows the relation between observed adverse events and performance of polypectomy. Table6 Adverse events related to number of polyps resected per procedure 1 for full colonoscopies. Number of polyps resected Procedures, n Bleeding, n (%) Perforation, n (%) (.1) 19 (.3) 1 or more (1.14) 63 (.9) (.95) 2 (.6) (1.15) 25 (.1) (1.57) 1 (.13) (1.53) 8 (.15) Total (95%CI) (.65) (.6.69) 82 (.6) (.5.8) 1 Procedure with bleeding or perforation recorded. Note that in a small number of procedures more than one adverse event was recorded (e.g. both perforation and bleeding). Thus the total number of adverse events was actually slightly higher than the total number of procedures with an adverse event. Appendix 2: Patient and polyp data by location Data on polyp size and other potential explanatory variables for adverse events at polypectomy are shown according to polyp location in " Table7. These relate only to those procedures (with or without adverse events) where a single polyp was resected. " Fig. 2 shows the distribution of polyp sizes in each colorectal location.

7 96 Original article Table7 Values of potential explanatory variables for adverse events at polypectomy, according to polyp location. These relate only to those procedures (with or without adverse events) where a single polyp was resected. Location Procedures, n Polyp size, median (mean), Male, % Age, mean, years Pedunculated, % mm Rectum (8.1) Sigmoid colon (11.2) Descending colon (7.7) Splenic flexure (5.7) Transverse colon (5.) Hepatic flexure 83 4 (6.7) Ascending colon 71 3 (5.5) Cecum (5.8) Total Frequency Rectum Sigmoid colon Descending colon Splenic flexure Transverse colon Hepatic flexure Ascending colon Cecum log e (Polyp size in mm) Fig. 2 Histograms showing distribution of polyp sizes by location, in procedures where only a single polyp was resected. Polyp sizes are expressed as log size, and the size range is up to 28 mm which is 5.63 on the log scale (base e). Competing interests: None Institutions 1 Gastroenterology, University Hospital of North Tees, Stockton on Tees, UK 2 NHS Cancer Screening Programmes, Sheffield, UK 3 Gastroenterology, South Tyneside NHS Foundation Trust, South Shields, UK 4 Cancer Epidemiology Unit, Oxford University, Oxford, UK 5 Durham University, Durham, UK 6 Northern Region Endoscopy Group (NREG), UK Acknowledgment This study was commissioned by and performed on behalf of the Bowel Cancer Screening Evaluation Group. References 1 Bowles CJ, Leicester R, Romaya C et al. A prospective study of colonoscopy practice in the UK today: are we adequately prepared for national colorectal cancer screening tomorrow? Gut 24; 53: Lee TJ, Rutter MD, Blanks RG et al. Colonoscopy quality measures: experience from the NHS Bowel Cancer Screening Programme. Gut 212; 61: Gavin DR, Valori RM, Anderson JT et al. The national colonoscopy audit: a nationwide assessment of the quality and safety of colonoscopy in the UK. Gut 213; 62: Epub 212 Jun 1 4 Zauber AG, Winawer SJ, O Brien MJ et al. Colonoscopic polypectomy and long-term prevention of colorectal-cancer deaths. N Engl J Med 212; 366: Logan RFA, Patnick J, Nickerson C et al. Outcomes of the Bowel Cancer Screening Programme (BCSP) in England after the first 1 million tests. Gut 212; 61: Epub 211 Dec 7

8 Original article 97 6 Korman LY, Overholt BF, Box T et al. Perforation during colonoscopy in endoscopic ambulatory surgical centers. Gastrointest Endosc 23; 58: Arora G, Mannalithara A, Singh G et al. Risk of perforation from a colonoscopy in adults: a large population-based study. Gastrointest Endos 29; 69: Iqbal CW, Cullinane DC, Schiller HJ et al. Surgical management and outcomes of 165 colonoscopic perforations from a single institution. Arch Surg 28; 143: ; discussion Rabeneck L, Paszat LF, Hilsden RJ et al. Bleeding and perforation after outpatient colonoscopy and their risk factors in usual clinical practice. Gastroenterology 28; 135: Panteris V, Haringsma J, Kuipers EJ. Colonoscopy perforation rate, mechanisms and outcome: from diagnostic to therapeutic colonoscopy. Endoscopy 29; 41: Rosen L, Bub DS, Reed JF et al. Hemorrhage following colonoscopic polypectomy. Dis Colon Rectum 1993; 36: Nelson DB, McQuaid KR, Bond JH et al. Procedural success and complications of large-scale screening colonoscopy. Gastrointest Endosc 22; 55: Heldwein W, Dollhopf M, Rösch T et al. The Munich Polypectomy Study (MUPS): prospective analysis of complications and risk factors in 4 colonic snare polypectomies. Endoscopy 25; 37: Kim HS, Kim TI, Kim WH et al. Risk factors for immediate postpolypectomy bleeding of the colon: a multicenter study. Am J Gastroenterol 26; 11: Saraya T, Ikematsu H, Fu KI et al. Evaluation of complications related to therapeutic colonoscopy using the bipolar snare. Surg Endosc 212; 26: Epub 211 Sep Buddingh KT, Herngreen T, Haringsma J et al. Location in the right hemi-colon is an independent risk factor for delayed post-polypectomy hemorrhage: a multi-center case-control study. Am J Gastroenterol 211; 16: Chilton A, Rutter M. Quality assurance guidelines for colonoscopy. NHS BCSP Publication No 6. Sheffield, UK: NHS Cancer Screening Programmes; Watabe H, Yamaji Y, Okamoto M et al. Risk assessment for delayed hemorrhagic complication of colonic polypectomy: polyp-related factors and patient-related factors. Gastrointest Endosc 26; 64: Consolo P, Luigiano C, Strangio G et al. Efficacy, risk factors and complications of endoscopic polypectomy: ten year experience at a single center. World J Gastroenterol 28; 14: Sawhney MS, Salfiti N, Nelson DB et al. Risk factors for severe delayed postpolypectomy bleeding. Endoscopy 28; 4: Gimeno-Garcia AZ, de Ganzo ZA, Sosa AJ et al. Incidence and predictors of postpolypectomy bleeding in colorectal polyps larger than 1 mm. Eur J Gastroenterol Hepatol 212; 24: Choo WK, Subhani J. Complication rates of colonic polypectomy in relation to polyp characteristics and techniques: a district hospital experience. J Interv Gastroenterol 212; 2: 8 11