OUP user menu

Journal of Crohn's and Colitis: 10 (10)


Laurence J. Egan, Ireland

Associate Editors

Shomron Ben-Horin, IsraelSilvio Danese, ItalyPeter Lakatos, HungaryMiles Parkes, UKJesús Rivera-Nieves, USABritta Siegmund, GermanyGijs van den Brink, NLSéverine Vermeire, Belgium


Published on behalf of

Random biopsies during surveillance colonoscopy increase dysplasia detection in patients with primary sclerosing cholangitis and ulcerative colitis

Udayakumar Navaneethan, Gursimran Kochhar, Preethi G.K. Venkatesh, Ana E. Bennett, Maged Rizk, Bo Shen, Ravi P. Kiran
DOI: http://dx.doi.org/10.1016/j.crohns.2013.02.009 974-981 First published online: 1 December 2013


Background and aim: Patients with primary sclerosing cholangitis (PSC) and ulcerative colitis (UC) are at increased risk of colon dysplasia. The role of random vs. target biopsies in these patients has not been investigated. Our aim was to evaluate the yield and clinical impact of random biopsies during surveillance colonoscopies in patients with PSC–UC.

Methods: Data from 71 patients (267 colonoscopies) with PSC and UC, who underwent surveillance colonoscopies and followed-up from 2001 to 2011 was obtained. Colonoscopy and pathology reports were reviewed to assess the yield of random biopsies.

Results: A total of 3975 (median 12) random biopsies were taken during surveillance colonoscopies. Overall, neoplasia was detected in 22 colonoscopies (16 patients): in 8 colonoscopies (36.4%) by targeted biopsies only and in 4 (18.2%) by both targeted and random biopsies. Neoplasia was detected in random biopsies only in 10 (45.5%) colonoscopies in 8 patients. On multivariate analysis, duration of UC (Odds ratio [OR] = 1.40; 95% confidence interval [CI], 1.08–1.81; P = 0.01), number of random biopsies (per increase by 8) (OR = 1.64; 95% CI, 1.18–2.28; P = 0.003) and target biopsies during colonoscopy (OR = 9.08; 95% CI, 3.18–26.0; P < 0.001) independently predicted the presence of dysplasia; endoscopic features of prior inflammation did not.

Conclusions: Random biopsies significantly increase the yield of dysplasia in patients with PSC and UC even in the absence of endoscopic features of prior inflammation and significantly impact clinical outcomes.

  • BMI
  • body mass index
  • CD
  • Crohn's disease
  • CRC
  • colorectal cancer
  • IBD
  • inflammatory bowel disease
  • LGD
  • low grade dysplasia
  • HGD
  • high grade dysplasia
  • aOR
  • adjusted odds ratio
  • UC
  • ulcerative colitis
  • Colonoscopy
  • Dysplasia
  • Random biopsies
  • Inflammatory bowel disease
  • Ulcerative colitis

1 Introduction

Primary sclerosing cholangitis (PSC) is a cholestatic liver disease seen in association with underlying inflammatory bowel disease (IBD), most commonly ulcerative colitis (UC).1,2 Approximately 70% to 80% of patients with PSC have associated with concomitant IBD.3 Patients with IBD appear to be at an increased risk of colorectal neoplasia; the risk increases when the duration of IBD exceeds 10 years, patients are younger at the onset of the disease and patients have extensive colitis.4 Previous studies have shown that in patients with IBD, the presence of concomitant PSC increases the risk of developing CRC almost 4 fold.511 In fact, the 10-year and 20-year risks of developing colorectal neoplasia have been reported as 14% and 31% respectively.7 It is recommended that colonoscopy with biopsy should be performed at the time of diagnosis in patients with PSC–IBD with yearly surveillance thereafter.11 We recently reported the high incidence of colon dysplasia in patients with PSC–UC soon after these 2 diseases are discovered and the lack of any association of severity of PSC with dysplasia risk.12

During surveillance colonoscopies, we perform targeted biopsies of suspicious appearing lesions along with random biopsies for dysplasia surveillance.13 There have been developments in advanced endoscopic imaging techniques that have improved neoplasia detection compared with conventional white light endoscopy.14 Also the advent of high definition endoscopy has questioned the yield of random biopsies in surveillance of patients with UC. The arguments are that even forty biopsies sample only a tiny proportion, estimated at less than 0.05%, of the large bowel epithelium.15 Certain endoscopic features predict an increased risk of neoplasia as a result of severe inflammation in the past (e.g., pseudopolyps) and it has been suggested that random biopsies may only be useful in this subset of patients.16 A recent Dutch study reported a lack of impact of random biopsies in routine UC surveillance and suggested that these biopsies may be important in patients with prior inflammation.17 However the study included only 23 PSC patients.17 Since PSC is associated with quiescent IBD; these patients often lack endoscopic features of previous inflammation.

No previous study has specifically evaluated the yield and clinical impact of random biopsies specifically in patients with PSC–UC. We sought to study the yield of neoplasia from random biopsies in patients with PSC–UC and the impact of these on subsequent clinical management.

2 Materials and methods

2.1 Patients

This historical cohort study was approved by the Cleveland Clinic Institutional Review Board. Patients with PSC and UC were identified from a prospectively maintained Electronic Data Interface for Transplantation (EDIT) database that has information on all PSC patients who underwent OLT at the Cleveland Clinic. Patients with PSC and UC who did not require OLT were identified using the PSC and IBD database. We have previously described the details of data collection into this database.12 We identified a total of 98 patients with PSC–UC who underwent surveillance colonoscopies in our institution after 2001 since pathological information on the number of biopsies could only be obtained since that time at our institution. We finally included 71 patients with PSC–UC who underwent more than two surveillance colonoscopies and had regular follow-up after their colonoscopies over the time period 2001 to 2011.

2.2 Inclusion and exclusion criteria

Inclusion criteria were as follows: 1) age older than 18 years 2) UC 3) presence of PSC with or without OLT and 4) regular surveillance colonoscopies at our institution (≥ 2). Patients with dysplasia diagnosed at their first surveillance colonoscopy and those referred with a diagnosis of dysplasia in colonoscopies performed at outside institutions or those with any prior diagnosis of dysplasia or neoplasia were excluded. Colonoscopies that were performed for a symptomatic indication (as patients did not get standardized surveillance biopsies) and patients without subsequent follow-up at our institution were similarly excluded.

2.3 Demographic and clinical variables

We defined PSC based on the finding of intra- and/or extra-hepatic bile duct abnormalities such as beading, duct ectasia, and stricturing of the intra- or extrahepatic bile ducts based on endoscopic retrograde cholangiopancreatography and/or magnetic resonance cholangiopancreatography.1

The diagnosis of UC was confirmed based on the characteristic endoscopic findings of inflammation combined with compatible histological examination as previously described.18

Demographic and clinical variables obtained by review included age, gender, ethnicity, smoking and alcohol history, and family history of IBD, PSC, or liver/colon cancer in first degree relatives. The definitions for the clinical variables including duration of UC, smoking and alcohol use and the use of long-term medical therapy including corticosteroids, immunomodulators including azathioprine/6-mercaptopurine and biologics (infliximab and adalimumab) have been previously described.12 The use of ursodeoxycholic acid (UDCA) was defined by the use of this medication for at least 50% of the follow-up period.

All patients underwent colonoscopic surveillance for the detection of colonic neoplasia or colon cancer.19 Bowel preparations was performed using 2–4 L of hypertonic polyethylene glycol solution. At colonoscopy, random biopsies were taken every 10 cm of the colon with targeted biopsies of suspicious areas (an irregular fold, mass or stricture) in addition. Details relating to the number of surveillance colonoscopies for each patient were obtained from the electronic medical records. We adopted the study design for reporting the yield of random biopsies from an earlier Dutch study.17

Information on endoscopic features of previous inflammation including backwash ileitis, shortened colon, tubular colon, featureless colon, scarring, segment of severe inflammation, post-inflammatory (pseudo) polyps, and colonic strictures was obtained. All these were based on the endoscopist's subjective assessment at the time of colonoscopy. Both general gastroenterologists and IBD specialists performed these colonoscopies. The number of random biopsies was calculated from the pathology report while the number of suspicious lesions was obtained from both the endoscopy and pathology reports. Since it is not standard practice to routinely count the number of biopsies obtained at colonoscopy at our center, we calculated this based on the pathologist's interpretation in the biopsy report. If any specific lesion was specifically targeted and mentioned in the endoscopy or pathology report, this was considered a suspicious lesion. When there was no mention in the endoscopy or pathology reports of a specific targeted biopsy, these were considered random biopsies.

All biopsy specimens were evaluated by experienced gastrointestinal pathologists at our institution. The slides were not re-reviewed for the study. It is routine practice in our institution that two different gastrointestinal pathologists review the biopsies if there is any evidence of dysplasia. Definite dysplasia was graded as low or high grade (unifocal or multifocal) based on criteria established by the Inflammatory Bowel Disease/Dysplasia Morphology Study Group.20 The highest grade of dysplasia in each patient was selected for analysis. Biopsies which were “indefinite for neoplasia” were excluded for this analysis.

We obtained information on the management of patients with neoplasia after surveillance colonoscopy. Patients either underwent endoscopic removal of the neoplasia or instead underwent surgery (proctocolectomy). Patients who were not eligible or did not elect to undergo either of the two were followed with surveillance which was classified either as intensified or regular. Patients with HGD or cancer routinely undergo colectomy in our institution. We also recommend surgery in patients with LGD in the setting of PSC. However, some patients elect to have non-surgical management and surveillance was hence performed in those patients. Patients with biopsies indefinite for dysplasia undergo a repeat colonoscopy within 3–6 months. Regular surveillance was defined as an interval between two surveillance colonoscopies of > 1 year < 2 years. If “indefinite for neoplasia” was diagnosed during colonoscopy, a subsequent interval of more than 6 months was defined as “regular” surveillance. Patients who underwent surveillance colonoscopy earlier than their regular surveillance colonoscopy were considered as undergoing “intensified surveillance”. We adopted these definitions from an earlier Dutch study.17

2.4 Outcomes

Our primary outcome of interest was to determine the yield of random biopsies for neoplasia. We performed per colonoscopy and per patient analysis separately. Per colonoscopy analysis was based on the number of colonoscopies in which neoplasia was detected by random biopsies only (biopsy not described as suspicious for neoplasia by the endoscopist) as compared to those when neoplasia was detected by either targeted and/or random biopsies. Per patient analysis was also performed in relation to the number of patients in whom random biopsies detected neoplasia in relation to those detected by either targeted and/or random biopsies. The secondary outcome of interest was to study the impact of neoplasia detection on random biopsies on subsequent clinical management.

2.5 Statistical analysis

Descriptive statistics were computed for all factors. These include medians, 25th and 75th percentiles, range or mean and standard deviation for quantitative variables and frequencies and percentages for categorical factors. Normally distributed continuous variables were analyzed by using a t test, and continuous variables that were not normally distributed were analyzed by using the nonparametric (Wilcoxon) rank sum test. Comparisons between categorical variables were performed by Pearson's chi-square test. A multivariable logistic regression model was constructed for detection of neoplasia during colonoscopy including variables that had significant univariable associations with development of colon dysplasia or cancer, P < 0.10. We used generalized estimation equations for multivariable logistic regression analysis as some patients may have multiple surveillance colonoscopies confounding the results. R 2.10.1 software (The R Foundation for Statistical Computing, Vienna, Austria) was used to perform all analyses.

2.6 Sensitivity analysis

We observed abnormalities in the endoscopy report which were reported but were not explicitly mentioned as targeted in pathology reports. These were noted as random biopsies which could overestimate the yield of random biopsies. We hence performed a sensitivity analysis excluding these reports and the results were unaltered as these numbers were very small (N = 12).

3 Results

3.1 Demographic and clinical characteristics

The basic demographic and clinical information of our entire cohort including age, sex, UC duration from diagnosis, and extent of UC is summarized in Table 1. Patients who underwent surveillance colonoscopy from 2001–2011 were included in the analysis. Backwash ileitis was documented in 13 colonoscopies, shortened colon in 3, tubular colon in 1, featureless colon in 28, scarring in 62, and post-inflammatory polyps in 56. Severe inflammation in at least one colonic segment was documented in 19 colonoscopies and colonic strictures in 1 colonoscopy.

View this table:
Table 1

Demographics and clinical characteristics of patients who underwent surveillance colonoscopy.

Variable (N = 71) Colonoscopies (267)Number (%)
Mean age at first colonoscopy in this study, years (range)48 (20–79)
Males47 (66.2%)
Smoking10 (14.1%)
Family history of colon neoplasia13 (18.3%)
Alcohol use18 (25.4%)
Follow-up duration, months (range)72 (10–384)
Body mass index, kg/m225.6 (18.0–41.5)
Extension of UC
Extensive colitis71 (100%)
Median duration of UC before first colonoscopy, years (range)9 (0–50)
Mean number of colonoscopies per patient, mean (SD)3.8 (1.2)
5-Aminosalicylate use59 (83.1%)
Ursodeoxycholic acid use57 (80.2%)
Azathioprine/6-mercaptopurine4 (5.6%)
Biologics2 (2.8%)

A total of 3975 (median 12) random biopsies were taken during 267 surveillance colonoscopies. Neoplasia was detected in random biopsies only in 10 (0.3%) of the 3975 random biopsies. However, in an additional 4 patients in whom neoplasia was detected by target biopsies, random biopsies also demonstrated neoplasia.

Thus, a total of 14 random biopsies demonstrated neoplasia (10 flat LGD, and 4 HGD). Furthermore, 32 suspicious (target) lesions were detected of which 12/32 (37.5%) were neoplastic (6 LGD, 4 HGD and 2 with colon carcinoma). The suspicious target lesions were described as abnormal folds (6, 50%), and lesion or mass in 6 (50%). Overall, 26 neoplastic sites were found in this study of which 12 were detected by targeted biopsies only, 10 by random biopsies only and 4 by both random and target biopsies.

3.2 Per-colonoscopy analysis

Neoplasia was detected by targeted or random biopsies in a total of 22/267 colonoscopies (8.2%). Neoplasia was detected by only random biopsies in 10 (3.7%) colonoscopies. When evaluated based on the number of colonoscopy procedures performed, 5 of the biopsies showed LGD and the remaining 5 revealed HGD. Two of the colonoscopies that detected LGD and subsequent surgery led to an intensified surveillance at shorter intervals but did not lead to any further dysplasia over a mean follow-up period of 4 years. Two other colonoscopies which detected LGD revealed multifocal HGD on subsequent colonoscopies during intensified surveillance performed after 3 months and led to proctocolectomy. The remaining one colonoscopy detected LGD and the colonoscopies showing HGD were managed by proctocolectomy.

Random biopsies detected neoplasia in four additional surveillance colonoscopies (4/267 = 1.5%) in which neoplasia was detected by targeted biopsies as well. The random biopsies showed LGD in three and multifocal HGD in one colonoscopy. Proctocolectomy was performed in two patients (HGD in random and targeted biopsies in one; multifocal LGD in random and HGD in target biopsies in one), while 2 colonoscopies which detected LGD led to intensified surveillance.

In 8 colonoscopies, neoplasia was only detected by targeted biopsies of suspicious lesions. Two of these led to endoscopic resection and 6 to proctocolectomy.

3.3 Per-patient analysis

Neoplasia was detected in at least one colonoscopy in 16/71 patients (22.5%). Five of these 16 patients had visible neoplasia (31.3%). Of the 16 patients with neoplasia, 8 patients had neoplasia which was detected by random biopsies only. In three additional patients, neoplasia was detected by both targeted and random biopsies (Fig. 1).

Figure 1

Distribution of colon neoplasia in PSC–UC patients with random and/or target biopsies.

3.4 Follow-up of patients with neoplasia detected by only random biopsies

A total of 11 patients had neoplasia detected by random biopsies of which 8 had neoplasia detected by random biopsies alone. One patient had unifocal LGD and 3 other patients had HGD resulting in proctocolectomy in all the 4 patients. Two patients had an endoscopically removable raised lesion which was resected (biopsies negative around the lesion) and random biopsies from the rest of the colon showed LGD. These patients were counseled for surgery. However the patients elected intensive surveillance and the subsequent surveillance colonoscopies have been negative for dysplasia. Two patients who had LGD on random surveillance colonoscopies and did not elect to have surgery had subsequent colonoscopies repeated within 3 months which showed multifocal HGD and the patients underwent proctocolectomy. Only one of the eight patients had endoscopic evidence of scarring in the colon. None of the other 7 patients had other endoscopic features of previous inflammation and had quiescent disease (Fig. 2).

Figure 2

Shows the management of PSC–UC patients diagnosed with colon neoplasia on random biopsies alone.

3.5 Multi-variable analysis of risk factors for colon neoplasia

Tables 2 and 3 demonstrate the univariable and multivariable regression analysis for the detection of neoplasia during colonoscopy. Univariable analysis did not show any association between endoscopic features of previous inflammation and the detection of neoplasia. On multivariable analysis using generalized estimation equations, duration of UC (Odds ratio [OR] = 1.40; 95% confidence interval [CI], 1.08–1.81; P = 0.01), number of random biopsies (per increase by 8) (OR = 1.64; 95% CI, 1.18–2.28; P = 0.003) and target biopsies during colonoscopy (OR = 9.08; 95% CI, 3.18–26.0; P < 0.001) independently predicted the detection of any dysplasia.

View this table:
Table 2

Univariable analysis using generalized estimation equations to evaluate associations between patient risk factors of neoplasia and the detection of neoplasia during colonoscopy.

VariableOdds ratio (95% CI)P-value
Duration of UC (per 5 years)1.35 (0.99–1.85)0.05
Number of random biopsies (per increase by 8)1.59 (1.14–2.23)0.006
Feature-less colon0.84 (0.19–3.81)0.99
Scarring0.50 (0.14–1.74)0.28
Segment of severe inflammation0.60 (0.08–4.72)0.66
Inflammatory pseudo-polyps2.35 (0.93–5.91)0.21
Target biopsy7.64 (2.75–21.2)< 0.001
View this table:
Table 3

Multivariable logistic regression analysis using generalized estimation equations to evaluate associations between patient risk factors of neoplasia and the detection of neoplasia during colonoscopy.

VariableOdds ratio (95% CI)P-value
Duration of UC (per 5 years)1.40 (1.08–1.81)0.01
Number of random biopsies (per increase by 8)1.64 (1.18–2.28)0.003
Target biopsy9.08 (3.18–26.0)< 0.001

4 Discussion

To our knowledge, no previous study has attempted to study the impact and clinical outcomes of dysplasia detected in random biopsies in patients with PSC and UC. This study highlights that random biopsies in PSC–UC patients significantly impact the clinical outcome. Our study also supports the current surveillance guidelines that suggest the performance of multiple random biopsies since the yield of dysplasia increased proportionately with the number of random biopsies taken. Also, the duration of UC and targeted biopsies of suspicious lesions increased the yield of dysplasia. None of the endoscopic features of previous inflammation predicted dysplasia in patients with PSC and UC highlighting further the importance of random biopsies in patients with concomitant PSC.

The role of random biopsies during colonoscopic surveillance in UC patients remains controversial. Come studies suggest that random biopsies have very little utility since most neoplasia is endoscopically visible with white light endoscopy revealing 61–88% of neoplastic sites in the colon.17,2123 The argument against random biopsies is that even forty biopsies sample only a tiny proportion, estimated at less than 1% of the colon mucosa.15 Thus, 64 biopsies have been reported to be required to achieve a 95% sensitivity for the detection of dysplasia.24,25 In addition, obtaining multiple random biopsies is time-consuming, and hence gastroenterologist's compliance with the number of biopsies is poor.2628 In addition, there exists interobserver variability in the diagnosis and reporting of dysplasia. Although there has not been any randomized controlled trial in UC patients, there is currently an ongoing trial comparing target vs. random and target biopsies in Japan.29 PSC–UC patients are at a higher risk for colon neoplasia compared to UC alone and no study has specifically studied these associations in this high-risk population of PSC patients.

In UC patients, a yield of 7.5–17% for neoplasia detection by random biopsies alone has been reported.17,2123 We found the yield of random biopsies to be significantly higher in patients with concomitant PSC, for neoplasia (8/71 (11.7%)). These findings support a previous study that reported that although unlike in patients with UC, in the subset of patients with UC- PSC (n = 23), random biopsies still had a significant impact.17 In addition to the detection of dysplasia, random biopsies resulted in changes in clinical management; six of the 8 (75%) patients with dysplasia detected on random biopsies underwent proctocolectomy within 3 months. Of these, only one patient did not have neoplasia in the proctocolectomy specimen. These observations differ from previously published studies in UC alone which have reported a low impact of random biopsies on clinical decision-making.17,30

The routine screening/surveillance protocol at our institution is the use of white light endoscopy for dysplasia surveillance for all UC patients. Previous studies have demonstrated that the yield of random biopsies decreases when using advanced endoscopic techniques.25,30,31 In fact, none of our patients underwent advanced imaging techniques including chromo-endoscopy in our cohort. Identifying patients most at risk for colon neoplasia in whom random biopsies may then be useful has also been investigated. Rutter et al.16 evaluated endoscopic features of severe inflammation and found that “post-inflammatory polyps” and “strictures” were associated with neoplasia. In another study, “tubular colon” and “strictures” were associated with the detection of neoplasia.17 Thus, the presence of features of previous inflammation as described above may warrant random biopsies. However such a strategy is questionable in patients with PSC since such patients have endoscopically quiescent disease and we have demonstrated that neoplasia can occur even in the presence of endoscopic quiescent disease.32

A previous Dutch study that included 23 PSC patients reported that random biopsies were significant for the detection of dysplasia17 and hence proposed that random biopsies should be considered in the presence of visible suspicious lesions, colonic features of severe inflammation in the past or PSC. The results of our study that report the findings in a large population of PSC patients undergoing routine surveillance colonoscopies at our institution confirm this strategy.

We have previously demonstrated that patients with PSC–UC are at a high risk for the detection of dysplasia as soon as the co-existence of the 2 diseases is discovered.12 The number of random biopsies increased dysplasia detection in these patients. However, the median number of biopsies performed per colonoscopy was only 12, which could be related to either errors in pathological interpretation of the number of biopsies as this was assessed retrospectively. The other possibility is that compliance with the guidelines may be suboptimal as a combination of general and IBD specialists perform surveillance colonoscopies in our institution. More than 32 biopsies was performed in 15/71 (21.1%) of patients in our study. Previously published studies have also shown poor compliance with the recommendations for surveillance.2628

The findings of our current study suggest that random biopsies significantly impact the yield of dysplasia and change clinical management in PSC-UC patients. The number of random biopsies was directly related to dysplasia detection in these patients. Also none of the endoscopic features predicted dysplasia in these patients. This suggests that patients with PSC–UC should have multiple random biopsies performed even in the presence of endoscopically normal mucosa.

There are certain limitations of our study. The study population was recruited from a subspecialty tertiary care referral center, which potentially contributes to a selection bias. The yield of random biopsies in a general practice may therefore be different. Second, whether biopsies were taken in a targeted or random manner was assessed retrospectively. In order to counter this potential drawback a sensitivity analysis excluding the patients who had discrepancy in the endoscopy and pathology reports was performed and this suggested that the results were unaltered. Although we did not specifically address the risk of colon dysplasia and cancer in relation to the use of UDCA in this study, we have earlier demonstrated its lack of efficacy in our population in a previous study.12 Irrespective of these drawbacks, this study is one of the largest studies that evaluates the yield of random biopsies during colonoscopic surveillance in patients with PSC and UC and the findings are clinically relevant.

Random biopsies significantly increase the yield of dysplasia in patients with PSC and UC even in the absence of endoscopic features of previous inflammation. Both random and target biopsies should be pursued during surveillance colonoscopy in all patients with PSC and UC.

Conflict of interest

None of the authors declared commercial conflict of interest.

Financial support



We thank Jeffrey Hammel MS for doing the statistical analysis for the paper.


View Abstract