Supported by grants from the Doris Duke Foundation (to R.G.), and National Institutes of Health (K-08-DK069393) (to T.U.). These organizations had no role in the study design, collection, analysis, or interpretation of data.
Dysplasia is more common in the distal than proximal colon in ulcerative colitis surveillance†
Article first published online: 7 JUL 2011
Copyright © 2011 Crohn's & Colitis Foundation of America, Inc.
Inflammatory Bowel Diseases
Volume 18, Issue 5, pages 832–837, May 2012
How to Cite
Goldstone, R., Itzkowitz, S., Harpaz, N. and Ullman, T. (2012), Dysplasia is more common in the distal than proximal colon in ulcerative colitis surveillance. Inflamm Bowel Dis, 18: 832–837. doi: 10.1002/ibd.21809
- Issue published online: 12 APR 2012
- Article first published online: 7 JUL 2011
- Manuscript Accepted: 16 MAY 2011
- Manuscript Received: 17 JAN 2011
- ulcerative colitis;
- surveillance colonoscopy;
- biopsy location
In patients with long-standing ulcerative colitis (UC), current dysplasia surveillance guidelines recommend four-quadrant biopsies every 10 cm throughout the colon. However, this may be inefficient if neoplastic lesions are localized in particular segments of the colorectum. The aim was to determine whether a difference exists in the anatomic distribution of dysplasia discovered in UC patients undergoing colonoscopic surveillance.
From an institutional database of over 700 patients with UC who underwent two or more surveillance colonoscopies between 1994–2006, we identified all patients with flat (endoscopically invisible) low-grade dysplasia (fLGD) or advanced neoplasia (colorectal cancer [CRC] or high-grade dysplasia [HGD]). Pathology reports were reviewed regarding the anatomic location of all dysplastic lesions. Fisher's exact test was used to compare the frequencies of neoplasia among the different colonic segments.
We identified 103 patients who progressed to any neoplasia (fLGD, HGD, or CRC). These patients underwent a total of 396 colonoscopies. The mean age at first surveillance colonoscopy was 48.6 years, with a mean UC disease duration of 18.2 years; 100% had extensive disease. Fifty-five patients developed advanced neoplasia. The rectosigmoid was found to have a significantly greater number of biopsies positive for advanced neoplasia and for any neoplasia compared to all other colonic segments (P < 0.0007); 71.2% of all advanced neoplasia was in the rectosigmoid.
The majority of dysplastic lesions identified in a surveillance program was detected in the rectosigmoid. Endoscopists should consider taking a greater percentage of biopsies in these segments as opposed to more proximal areas. (Inflamm Bowel Dis 2011;)
Patients with extensive, long-standing ulcerative colitis (UC) have a heightened risk for developing colorectal cancer (CRC).1 In an effort to minimize unnecessary colectomies and improve CRC-related mortality and morbidity, it is recommended that such patients undergo a program of dysplasia surveillance in which periodic colonoscopies are performed and both targeted and nontargeted biopsies are taken.2 When dysplasia is detected but cannot be completely removed, it is recommended that such patients consider a colectomy.1 Patients who are dysplasia-free are advised to continue with periodic colonoscopies with surveillance biopsies.1
A consensus conference sponsored by the Crohn's and Colitis Foundation of America (CCFA) recommended that surveillance colonoscopy include 4-quadrant nontargeted biopsies every 10 cm of colon, with a minimum of 32 biopsies for patients with extensive disease, as well as biopsy or removal of all suspicious lesions. The CCFA's Consensus Statement also suggests that gastroenterologists “consider” reducing the 10-cm interval to 5 cm in the lower sigmoid and rectum due to the higher frequency of CRC in this region.2, 3 A recent AGA Technical Review supported this strategy.4
The recommendation to intensify biopsy sampling in the distal colon, however, is based on limited knowledge of where in the colon dysplasia is more likely to occur. In a series looking at the natural history of low-grade dysplasia (LGD), Woolrich et al3 noted that dysplasia was more commonly identified in the distal colorectum, with 19 of the 28 (68%) cases in their series coming from the rectosigmoid. Similarly, in a pooled analysis, Choi5 reported that 52% (range: 47%–62%) of colitis-related CRCs occurred in the rectosigmoid. Additionally, a 20-year prospective study on surveillance colonoscopy by Lindberg et al6 noted that 44% of dysplastic lesions occurred within the rectosigmoid, and a 30-year analysis of surveillance colonoscopy by Rutter et al7 reported that 58.6% of CRCs were identified within the rectosigmoid. The purpose of this study was to determine whether a difference exists in the anatomic distribution of dysplasia discovered in a cohort of UC patients undergoing colonoscopic surveillance at a large IBD Center.
MATERIALS AND METHODS
Identification of Subjects
This study was conducted with the approval of the Mount Sinai School of Medicine Institutional Review Board and in accordance with Health Insurance Portability and Accountability Act (HIPAA) regulations. UC patients who received a minimum of two colonic evaluations (colonoscopic or surgical) were enrolled in the Mount Sinai Ulcerative Colitis Surveillance database. Details of this cohort of patients have been described previously.8, 9 Briefly, patients were identified by both computerized and hand searches of records in the Department of Pathology between 1994 and 2006 and patients with at least left-sided or more extensive UC of more than 7 years duration were included. Patients with primary sclerosing cholangitis (PSC, n = 8) were included. Patients were excluded for a diagnosis of Crohn's disease, ulcerative proctitis only, or indeterminate colitis. The medical records of the included patients were then reviewed and information including gender, date of birth, date of UC diagnosis, anatomic extent of disease, extraintestinal manifestations of UC, and UC-related medication use were recorded. Additionally, pathology reports and endoscopic notes for all colonoscopies were reviewed and abstracted to record anatomic extent of inflammation, location and number of all biopsies, and location and degree of dysplasia if present.
Endoscopic and pathology reports for all biopsy specimens that demonstrated LGD, high-grade dysplasia (HGD), or CRC were reviewed to determine the anatomic location of the detected neoplasia. Anatomic locations included cecum, ascending colon, hepatic flexure, transverse colon, splenic flexure, descending colon, sigmoid, rectosigmoid, and rectum. For our analyses, rectosigmoid consisted of all biopsy jars labeled rectum, sigmoid, and rectosigmoid. When pathology reports used distances from the anal verge instead of specific colonic locations, conversions to locations were made according to the following scheme: a distance of 0–20 cm was considered rectum, 21–40 cm was considered sigmoid colon, 41–60 cm was considered descending colon, 61–80 cm was considered transverse colon, and >80 cm was considered ascending colon or cecum.
For the present study, additional inclusion criteria needed to be met. In addition to having UC greater than 7 years and extensive colitis (histological or endoscopic inflammation proximal to the splenic flexure) prior to, or at the time of, dysplasia detection, patients needed to have a finding of neoplasia (LGD, HGD, or CRC) during a surveillance examination, and could not have had a partial or more extensive colonic resection in an area of colitis prior to study inclusion. Of the LGD identified, only subjects with flat LGD (fLGD) were included in this study because of the difficulty with retrospectively distinguishing among sporadic adenomas and colitis-associated dysplastic lesions. LGD was determined to be flat if a histologic diagnosis of LGD was made in the absence of documentation by the endoscopist or pathologist of a mass or polypoid lesion at the biopsy site. Similarly, patients with HGD that was not endoscopically apparent were determined to have flat HGD (fHGD). Patients were excluded if they had an uncertain diagnosis of UC, insufficient follow-up or clinical detail, or non-UC-related CRC (CRC that was located proximal to the maximal extent of colitis).
Classification of Dysplasia
All biopsy specimens were reported as LGD, HGD, or CRC according to the criteria of the Inflammatory Bowel Disease Morphology Study Group.10 The diagnoses of dysplasia or neoplasia were made prospectively by faculty of the Mount Sinai Gastrointestinal Pathology Department and confirmed during conferences with the department's director (N.H.). These diagnoses were not altered retrospectively for the purpose of this study. The term “any flat neoplasia” was defined as fLGD, fHGD, or CRC; “any neoplasia” was defined as fLGD, HGD, or CRC; and “advanced neoplasia” was defined as HGD or CRC. Dysplasia yield was limited to first time discovery; colonoscopies performed and dysplasia discovered following index cases of dysplasia were not included in analyses.
Previous studies have reported that the sensitivity of dysplasia detection is highly dependent on the number of biopsies performed.11 While the endoscopy and pathology reports reviewed provided limited information on the total number of biopsies taken per colonoscopy, accurate information was available regarding the number of biopsy jars used and the locations of the biopsy specimens. Thus, in our study the frequency of dysplasia detection per anatomic segment was calculated from the number of positive biopsy jars from a particular colonic location divided by the total number of biopsy jars from the location. In general, Mount Sinai gastroenterologists take very similar, if not equal, numbers of nontargeted biopsies per biopsy jar.
Means with standard deviations were calculated for all continuous variables. Proportions were calculated for all binary and ordinal variables. The frequencies for which each segment of the colon were found to be positive for advanced neoplasia and any neoplasia were calculated. Furthermore, the percent of biopsy jars positive for advanced neoplasia or any neoplasia were calculated for each anatomic segment of the colon by dividing the total number of biopsy jars positive for dysplasia at a particular location by the total number of biopsy jars obtained from that location across all patients who met the inclusion criteria. The percentage of biopsy jars positive for advanced neoplasia for each segment of the colorectum were compared using Fisher's exact testing with 1 degree of freedom and P < 0.05 for statistical significance with SPSS Statistics 17.0 for Windows (Chicago, IL). The same test was performed for comparison of the percentage of biopsies positive for any neoplasia for each segment of the colon.
Patient characteristics are shown in Table 1. One hundred and three patients (68 male, 35 female) developed fLGD, HGD, or CRC and underwent a total of 396 examinations. Of these subjects, 55 (37 male, 18 female) developed HGD or CRC at some time during surveillance. A total of 396 colonoscopic examinations were performed and 34 (8.6%) were considered to have not reached the cecum due to absence of biopsies from the cecum or ascending colon. The mean age of the patients at first surveillance colonoscopy was 48.6 years. These patients had a mean disease duration of 18.2 years and 100% had extensive disease (colitis extending proximal to the splenic flexure) before or at the time of initial dysplasia detection.
|Any Neoplasia (fLGD, HGD, or CRC)||Advanced Neoplasia (HGD or CRC)||Any Flat Neoplasia (fLGD, fHGD, or CRC)|
|Age at dysplasia discovery||52.7 ± 15.1 years||52.9 ± 14.2 years||52.9 ± 15.3 years|
|Male||68 (66%)||37 (67%)||57 (66%)|
|Duration of UC at dysplasia Discovery||22.6 ± 12.0 years||24.8 ± 11.5 years||22.5 ± 11.8 years|
|Extent of UC at diagnosis|
|• Extensive||27 (26%)||13 (24%)||23 (26%)|
|• Left-sided||14 (14%)||8 (15%)||12 (14%)|
|• Proctosigmoiditis||16 (16%)||8 (15%)||14 (16%)|
|• Unknown||46 (46%)||26 (48%)||38 (44%)|
|Extent of UC at dysplasia Discovery|
|• Extensive||93 (90%)||51 (93%)||78 (90%)|
|• Left-sided||7 (7%)||2 (4%)||6 (7%)|
|• Proctosigmoiditis||1 (1%)||0||1 (1%)|
|• Unknown||2 (2%)||2 (4%)||2 (2%)|
|Number of previous surveillance examinations||3.8 ± 2.9||3.9 ± 2.9||4.0 ± 3.0|
A total of 3091 biopsy jars taken during surveillance were analyzed. The total number of biopsy jars per colonic segment, as well as the number and percentage of biopsy jars positive for any neoplasia, advanced neoplasia, or any flat neoplasia per colonic segment are shown in Table 2. The greatest number of biopsy jars were taken in the transverse colon.
|Segment||Any Neoplasia||Advanced Neoplasia||Any Flat Neoplasia|
|Rectosigmoid||76/1030 (7.38%)||42/691 (6.08%)||63/920 (6.85%)|
|• Rectum||45/527 (8.54%)||27/366 (7.38%)||39/468 (8.33%)|
|• Sigmoid||26/489 (5.32%)||13/315 (4.13%)||20/439 (4.56%)|
|Descending||12/429 (2.8%)||2/287 (0.7%)||12/385 (3.12%)|
|Transverse||21/973 (2.16%)||9/571 (1.58%)||18/883 (2.04%)|
|Ascending/Cecum||9/659 (1.37%)||6/393 (1.53%)||7/586 (1.19%)|
|P value (*)||<0.0001||0.0002||<0.0001|
For any neoplasia the rectum demonstrated the greatest percentage of positive biopsy jars (8.54%), while the cecum/ascending colon had the lowest percentage of positive biopsy jars (1.37%). The rectosigmoid had a significantly greater number of biopsy jars positive for any neoplasia compared to the descending colon (P = 0.0006), transverse colon (P < 0.0001), and ascending colon/cecum (P < 0.0001). Figure 1A illustrates the distribution of any neoplasia according to colonic segment. Of all biopsy jars showing any neoplasia, 64.4% were in the rectosigmoid and 10.2% were in the descending colon; thus, 74.6% of any neoplasia occurred in the distal colon, whereas 25.4% were positive in the proximal colon.
When only biopsy jars positive for advanced neoplasia were analyzed, very similar trends were noted (Table 2). The rectosigmoid had a significantly greater number of biopsy jars positive for advanced neoplasia compared to the descending colon (P < 0.0001), transverse colon (P < 0.0001), and ascending colon/cecum (P < 0.0003). As such, 74.6% of all advanced neoplasia was located distal to the splenic flexure (Fig. 1B). Likewise, for any flat neoplasia, similar trends were observed (Table 2, Fig. 1C), with 75% of flat neoplastic lesions detected distal to the splenic flexure.
Among our patient population we identified nine CRCs. Of these nine, six were located in the rectum (67%), one was located in the hepatic flexure (11%), and two were located in the cecum (22%). No cancers were observed within the sigmoid colon, descending colon, splenic flexure, or transverse colon.
There were eight subjects with PSC in this study. Three of them had proximal dysplasia only. Omitting these patients did not appreciably alter the overall results.
The present retrospective cohort study addressed the anatomic distribution of neoplasia in UC patients undergoing periodic surveillance examinations. Our results demonstrated that biopsies of the rectum and sigmoid colon in UC patients undergoing routine colonoscopic surveillance are more likely to harbor fLGD, HGD, or CRC than biopsies of more proximal segments of the colon. Even though only 33.3% of all biopsy jars came from the rectosigmoid, the great majority (64.4%) of dysplasia was located in these distal regions.
Currently, many gastroenterology associations recommend that numerous nontargeted biopsies be obtained during each UC surveillance colonoscopy. The CCFA Consensus states that four-quadrant biopsies be taken every 10 cm from the cecum to the rectum, but that “consideration should be given to taking four-quadrant biopsies every 5 cm in the lower sigmoid and rectum, because the frequency of CRC is higher in this region.”2 A new AGA Technical Review states that although no prospective studies have determined the optimal number of biopsy specimens for reliable dysplasia detection, representative biopsies should be obtained from each anatomic section of the colon.4 Our study supports the recommendation to increase biopsy sampling of the sigmoid colon and rectum. These results are similar to previous reports regarding the location of incident CRC in UC. Connell et al,12 for example, found that carcinoma was located in the rectosigmoid in 36 of 50 (72%) proctocolectomy specimens with UC-related CRC in a study performed in the UK. Similarly, Choi5 reported in a pooled analysis that 52% (47%–62%) of colitis-associated colorectal cancers occurred within the rectosigmoid. Higgins and colleagues11 recently modeled dysplasia in the colon and reported that current nontargeted biopsy techniques will miss a number of moderate-sized lesions based on the density of sampling. Thus, an increased sampling pattern in the distal colorectum may be able to overcome some of the limitations of current techniques. The ability to find dysplasia in the distal colorectum has additional importance when one considers recent observations from our research group that distal dysplasia appears to progress more rapidly than proximal dysplasia to advanced neoplasia (Goldstone et al, submitted). While we report a predominance of dysplasia within the distal colon, we must also consider that our results may suggest a different possibility, namely, that proximal dysplasia is inherently more difficult to detect endoscopically than distal dysplasia, although this explanation would be purely speculative.
Our study has a few limitations. First, our UC population comes from a single tertiary referral center. These patients may be viewed as having a greater inflammatory burden than UC in the general nonreferred population, and might therefore have a greater risk of advanced neoplasia.9, 13 While the overall incidence might be greater, it is not known whether location of dysplasia varies by severity of inflammation. A second potential limitation relates to the identification and classification of dysplasia, as previous studies have reported low rates of observer agreement.14 For this study, all slides and histopathologic diagnoses were made by a single group of gastrointestinal pathologists and reviewed by one expert pathologist (N.H.). In our previous experience,9 there was minimal interobserver variation, so we see no reason why any particular segment of the colon would be more prone to any minor interobserver variation. Another potential limitation is the nonprotocolized nature of the surveillance exams, with each exam having the possibility for a different number of biopsies in variable locations within the colon. While a nonstandardized colonoscopic approach may have limited dysplasia yield, it may also more accurately reflect the variation in current surveillance practice and improve the external validity of our findings. Although we were unable to obtain actual total biopsy numbers per anatomic segment of the colon, we did utilize biopsy jar number in our analyses. Assuming Mount Sinai gastroenterologists take relatively equal numbers of nontargeted biopsies per anatomic segment, biopsy jar number should correlate well with total biopsy number. Moreover, in this study we analyzed the percent of positive biopsy jars per anatomic segment in an attempt to limit the influence of sampling error, as fewer total biopsy jars were obtained from the proximal colon than the distal colon.
In conclusion, our results showed that neoplastic lesions occur with greater frequency within the rectosigmoid region. We suggest that greater attention be paid to taking biopsies of the rectosigmoid colon, whether by random biopsies or biopsies in the context of chromoendoscopy, as this could have important clinical implications regarding the practice and effectiveness of surveillance colonoscopy. Future prospective studies are required to evaluate the efficacy of this proposed approach, as well as to determine the impact of evolving adjunctive techniques such as chromoendoscopy.
The authors thank the many attending gastroenterologists at Mount Sinai who supported this effort, as well as the Clinical Research in IBD (CRIB) group.