Yield of Random Biopsies During Colonoscopies in Inflammatory Bowel Disease Patients Undergoing Dysplasia Surveillance

Abstract

Background

The development of chromoendoscopy (CE) and high definition endoscopy (HDE) has improved detection of subtle colonic dysplasia in patients with inflammatory bowel diseases (IBDs). The role of random biopsies for dysplasia surveillance is unclear.

Methods

We reviewed patients with IBD who underwent a CE or HDE colonoscopy and had colonic dysplasia detected. Detection of dysplasia was classified as either visible or random and graded as low grade dysplasia (LGD), high grade dysplasia (HGD), or indefinite for dysplasia. Multivariable regression adjusted for relevant confounders examined the predictors of dysplasia detectable on random biopsies alone.

Results

The study included 300 patients (203 ulcerative colitis, 97 Crohn’s disease with colonic involvement) contributing 442 colonoscopies; the mean disease duration was 24.5 years; 7.2% had primary sclerosing cholangitis (PSC). Three hundred sixty-two colonoscopies (82%) had only visible dysplasia, 52 (12%) had only random dysplasia, and 28 (6%) had both visible and random dysplasia. Longer disease duration (odds ratio, 1.04; 95% CI, 1.01–1.07), active inflammation (odds ratio, 2.89; 95% CI, 1.26–6.67), and concomitant PSC (odds ratio, 3.66; 95% CI, 1.21–11.08) were associated with detecting dysplasia on random biopsies compared with visible lesions. Patients with random dysplasia (21%) or both random and visible dysplasia (21%) were more likely to undergo surgical resection compared with those with only visible dysplasia (5%; P < 0.001) and have subsequent development of colorectal cancer (15%, 7%, 1%, respectively; P < 0.0001).

Conclusion

Nearly one fifth of dysplasia detected in patients with IBD was found on random biopsies. Patients with high risk characteristics may benefit from continuing the practice of random biopsies during surveillance examinations.

INTRODUCTION

Patients with long-standing ulcerative colitis (UC) or Crohn’s disease (CD) with colonic involvement have a higher risk of colorectal cancer (CRC) compared with the general population.^1–3 Early studies estimated this risk to be 18% at 30 years after diagnosis of UC, though subsequent studies have determined this excess risk to be of a smaller magnitude and temporally declining over the past 3 decades.^3–5 Societal guidelines recommend surveillance colonoscopies for patients with inflammatory bowel disease (IBD) every 1 to 3 years depending on duration of disease, extent of disease activity, and comorbid conditions such as primary sclerosing cholangitis (PSC).^6–9

In IBD, CRC develops through a chronic inflammation-dysplasia-carcinoma sequence in which dysplasia can present on flat mucosa.^{8, 10} Historically, detection of colonic neoplasia required multiple random biopsies throughout the colon.¹¹ The development of chromoendoscopy (CE) and high definition endoscopy (HDE) has improved detection of subtle mucosal abnormalities compared with standard definition endoscopy (SDE) such that most dysplasia in patients with IBD may now be visible. This has led to expert opinions favoring the viewpoint that targeted biopsies (and removal) of visible lesions are sufficient, obviating the need for random biopsies for surveillance.⁸ However, despite the limited yield, cost, and time associated with random biopsies when using CE and HDE^12–14 in an individual patient, random biopsies may detect dysplasia at sites distinct from that of visible neoplasia. This becomes increasingly important as management of visible dysplasia has evolved from total proctocolectomy to bowel-sparing segmental resections.^{15, 16} Previously, studies have reported 45% of patients who underwent total proctocolectomy or subtotal colectomy for high grade dysplasia (HGD) to have remote invasive cancer and dysplasia in the resected specimen.¹⁷ Thus, as surgical strategies continue to evolve, the ability to identify synchronous dysplasia is imperative and has important implications for treatment.

The aims of this study were (1) to determine the number of patients in whom neoplasia was detected by random biopsies as opposed to biopsies of visible lesions, (2) to assess the impact of neoplasia detected by random biopsies on subsequent treatment including risk for colorectal cancer, and (3) to determine the risk factors for identifying dysplasia only on random biopsies when compared with visible lesions.

MATERIALS AND METHODS

Study Population

This was a retrospective study approved by the Partners Healthcare institutional review board. We identified eligible patients through the Partners Healthcare Research Patient Data Registry (RPDR). Partners Healthcare System comprises Massachusetts General Hospital (MGH) and Brigham and Women’s Hospital (BWH); it provides care for the Boston and greater Boston metropolitan area. The RPDR is a central data warehouse containing patient information including demographic data, billing codes, problem lists, inpatient notes, and outpatient notes.¹⁸ Details of this use have been reported in our previous publications.^19–21 First, we used the RPDR to identify unique patients from 2011 to 2019 who had a diagnosis of CD, UC, or IBD-unclassified (IBD-U) and had a colonoscopy demonstrating a dysplastic lesion. We began the analysis in 2011 to capture only the time period when endoscopic procedures were being performed using high definition endoscopes. Colonoscopy procedures at the included institutions are performed by board-certified gastroenterologists or gastroenterology trainees (supervised directly by an attending). The choice of white light endoscopy (WLE) or chromoendoscopy (CE) and the choice of random biopsies varied across endoscopists, but during much of the study period, standard of care suggested 32 random biopsies from the colon for surveillance in this population.¹¹ From among patients who underwent a colonoscopy during the study period, we performed a text search of pathology reports using the terms “low-grade dysplasia,” “high-grade dysplasia,” “dysplasia,” and “adenoma.” The list generated by RPDR was manually screened to confirm dysplasia. Patients were included if they met one of the following criteria: (1) history of UC with extension past the rectum for at least 8 years, (2) history of Crohn’s colitis for at least 8 years, (3) history of IBD with concomitant PSC at the time of diagnosis irrespective of duration of disease, or (4) history of IBD with concomitant colorectal malignancy at the time of diagnosis irrespective of time since diagnosis. Patients were excluded if they had undergone subtotal colectomy or proctocolectomy before colonoscopy. We also excluded patients with isolated UC proctitis as they are not recognized to be at a higher risk of colorectal cancer. For each patient, the index colonoscopy was the first one during this study period where dysplasia was detected. Information was retrieved for every colonoscopy for the included patients during the study period.

Covariates

Manual chart review was performed for each patient that met inclusion criteria. We noted information on age, gender, duration of disease at the time of the colonoscopy, disease extent in UC, location and behavior in CD, smoking status, and concomitant diagnosis of PSC. We also obtained information on personal history of CRC, prior dysplasia, and family history of CRC. For each colonoscopy, we noted whether it was performed using WLE or CE, quality of bowel preparation (excellent, good, fair, adequate, or poor), and whether active disease was noted on exam. Where possible, disease activity was scored using the Mayo endoscopic score for UC or simple endoscopic score (SES-CD) for Crohn’s disease. Mayo score of ≥2 or SES-CD >3 was considered to demonstrate active inflammation. We also noted whether the exam was performed by an IBD specialist gastroenterologist or a nonspecialist gastroenterologist.

The biopsies were then analyzed by pathologists with gastrointestinal (GI) training for the presence of inflammation and degree of dysplasia. Dysplasia was graded as low grade dysplasia (LGD), which also included tubular and serrated adenomas, indefinite for dysplasia, high grade dysplasia, and cancer.²² For each colonoscopy, we classified whether dysplasia was identified on random biopsies alone (random dysplasia), as a visible lesion (visible dysplasia), or both. Visible dysplasia was typically labeled as being associated with an endoscopically detectable abnormality (polyp or other irregularity) associated with a recognition of this mucosal abnormality on the pathology label. In contrast, random dysplasia was identified by reference to surveillance or random biopsies during the colonoscopy without endoscopically visible abnormality denoted on either the endoscopy report or pathology label.

Statistical Analysis

Categorical variables were presented as counts and percentages, and continuous variables were presented as means with standard deviations. The χ ² test was used to compare the categorical variables, and t test was used to analyze continuous variables. The unit of analysis was each colonoscopy. However, to account for correlation of findings within an individual, we used generalized linear models accounting for repeated measures in the same individual. Univariate logistic regression was used to study the relationship between each covariate and the risk of colonic neoplasia. Variables that met statistical significance on univariate analysis at P < 0.10 or those that were robustly identified in the literature to be risk factors for dysplasia in IBD were included in the multivariate model. In our multivariate analysis, we identified independent predictors of identifying colonic dysplasia only on a random biopsy compared with those with visible dysplasia. Secondarily, we examined the predictors of any random dysplasia (either alone or in conjunction with a visible lesion on the same biopsy) when compared with those without any random dysplasia. In addition, colonoscopies after the index examination with dysplasia were labeled as being in patients with prior dysplasia, and this was adjusted for in the multivariable model. We performed subgroup analyses separately for UC and CD. All analysis was performed in Stata 15.0 (StataCorp, College Station, TX).^{23, 24} The study was approved by the institutional review board of Partners Healthcare.

RESULTS

Study Cohort

A total of 1790 patients with at least 1 diagnosis code for CD, UC, or unspecified colitis were identified through the RPDR as having had a colonoscopy during the study period. After applying the exclusion criteria, our final cohort included 300 patients with longstanding IBD who had colonic involvement resulting in a higher risk of colorectal cancer (left-sided ulcerative colitis, UC pancolitis, ileocolonic CD, or Crohn’s colitis), contributing 442 colonoscopies. The reasons for exclusion were (1) diagnosis other than IBD (n = 875), (2) CD but no colonic involvement (n = 84), (3) UC with proctitis (n = 64), (4) diagnosis of IBD less than 8 years without PSC (n = 180), (5) subtotal colectomy (n = 34), (6) no dysplasia detected on MGH colonoscopy (n = 121), (7) no follow-up (n = 54), (8) flexible sigmoidoscopies not colonoscopies (n = 12), (9) and hyperplastic polyps and no colonic dysplasia (n = 66).

The demographic details of the patients are summarized in Table 1. Two thirds (n = 203, 66.7%) had UC, and the remaining had CD with colonic involvement (n = 97, 33.3%). The mean disease duration was 24.5 years (range 1–61). Just under half of the cohort were women (45.8%). There was a total of 32 patients (7.2%) with primary sclerosing cholangitis, and 10 patients (3.3%) had prior colorectal cancer. Random biopsies were taken in a majority of colonoscopies (88.9%), whereas the remainder only included targeted biopsies. Out of the 442 colonoscopies, there were 362 colonoscopies in which visible dysplasia was identified, 52 colonoscopies in which dysplasia was only on random biopsies, and 28 colonoscopies in which both a visible lesion and random dysplasia were identified (Fig. 1). Among the visible lesions, 1% was indefinite for dysplasia, 94% were LGD including tubular and serrated adenomas (irrespective of previous endoscopic involvement of that segment), 4% were HGD, and 1% was colorectal cancer (CRC). Among the random dysplasia, 6% were indefinite, 83% were LGD, 8% were HGD, and 3% were CRC. In this group, 11 colonoscopies yielded multifocal dysplasia, and 38 were unifocal. There were 3 colonoscopies that did not identify the segment of the colon with dysplasia.

TABLE 1.

Characteristics of Included Colonoscopies with History of Inflammatory Bowel Disease and Dysplasia

	Visible Dysplasia (N (%)) (N = 362)	Random Dysplasia (N (%); N = 52)	Visible and Random Dysplasia (N (%); N = 28)	P
Age	59.8 ± 13.5	57.2 ± 14.1	64.3 ± 13.6	0.007
Disease Duration	23.8 ± 12.1	28.7 ± 14.8	26.5 ± 17.1	0.012
Male Sex	198 (54.7)	31 (59.6)	15 (53.6)	0.788
IBD Type				0.028
Ulcerative Colitis	246 (68.0)	27 (51.9)	22 (78.6)
Crohn’s Disease	116 (32.0)	25 (48.1)	6 (21.4)
Smoker
Current or Former	13 (3.6)	1 (1.9)	0	0.499
Never	349 (96.4)	51 (98.1)	28 (100)
Location of CD				0.027
Ileocolonic	49 (43.8)	18 (72)	4 (66.7)
Colonic	63 (56.3)	7 (28)	2 (33.3)
Extent of UC				0.101
Left-sided	93 (41)	9 (34.6)	4 (18.2)
Pancolitis	134 (59)	17 (65.4)	18 (81.8)
Primary sclerosing cholangitis	19 (5.3)	8 (15.4)	5 (17.9)	0.003
Personal history of colorectal cancer	5 (1.4)	1 (1.9)	3 (10.7)	0.003
Prior dysplasia on index colonoscopy	117 (32.3)	10 (35.7)	15 (28.9)	0.807
Family history of colonic neoplasia	53 (14.6)	5 (9.6)	3 (10.7)	0.548
Presence of active disease	35 (9.8)	12 (26.1)	2 (7.4)	0.004
Year of Endoscopy				0.431
2010–2014	140 (38.7)	24 (46.2)	9 (32.1)
2015–2019	222 (61.3)	28 (53.9)	19 (67.9)
Endoscopic Modality				0.303
White Light	310 (87.1)	40 (81.6)	22 (78.6)
Chromoendoscopy	46 (12.9)	9 (18.4)	6 (21.4)
Presence of IBD specialized endoscopist	192 (54.1)	34 (69.4)	15 (53.6)	0.126
Presence of pseudopolyps	50 (14.1)	8 (16.3)	5 (17.9)	0.806
Prep Quality
Poor	35 (9.7)	3 (5.8)	2 (7.1)	0.615

FIGURE 1.

Flowchart of colonoscopies with inflammatory bowel diseases (IBD) with dysplasia on targeted biopsies and random biopsies.

There was a slightly higher proportion of those with CD among those with random dysplasia (48%) compared with those with only visible (32%) or both random and visible dysplasia (21%; P = 0.028). In the entire cohort, 14% of all colonoscopies were chromoendoscopy. This proportion was similar among those with random dysplasia (18%) as in those with visible dysplasia (13%; P = 0.30). The presence of pseudopolyps was similar among all 3 groups (P = 0.81). Interestingly, the proportion of those with PSC was higher among those with random dysplasia alone (15%) or in addition to visible dysplasia (18%) compared with those with only visible dysplasia (5%; P = 0.003). There was also a larger proportion of those with endoscopically active disease among those with random dysplasia only (26%) compared with those with only visible dysplasia (10%; P = 0.04). Just over half the colonoscopies were performed by IBD specialists (55%); this was similar across all 3 groups (P = 0.13).

On multivariable analysis, compared with those who had visible dysplasia alone, presence of dysplasia on random biopsies alone was more likely in those with longer duration of disease (odds ratio [OR], 1.04; 95% confidence interval [CI], 1.01–1.07) when active disease was noted on colonoscopy (OR, 2.89; 95% CI, 1.26–6.67) and in those with PSC (OR, 3.66; 95% CI, 1.21–11.08; Table 2). None of the other variables were independently significant on multivariable analysis. When expanding the analysis to compare those with any random dysplasia (either alone or along with a visible lesion on the same scope) against those with only visible dysplasia, the analysis demonstrated similar associations with longer duration of disease (OR, 1.03; 95% CI, 1.00–1.04), PSC (OR, 4.49; 95% CI, 1.82–11.07), and a history of prior colorectal cancer (OR, 5.98; 95% CI, 1.76–45.90). In addition, having an endoscopy by an IBD specialist was associated with identifying dysplasia on random biopsies (OR, 1.88; 95% CI, 1.02–3.49 95% CI), but the association with active disease was borderline statistically significant (P = 0.084; Supplemental Table 1).

TABLE 2.

Multivariate Analysis of Predictors of Random Dysplasia Compared With Visible Dysplasia

Risk Factors	Odds Ratio	95% Confidence Interval	P
CD	1.37	0.67–2.77	0.389
Disease Duration	1.04	1.01–1.07	0.004
Presence of active disease	2.89	1.26–6.67	0.013
Presence of PSC	3.66	1.21–11.08	0.022
Ever smoker	1.16	0.14–9.85	0.89
Age	0.99	0.96–1.02	0.505
Female	0.75	0.36–1.54	0.429
Presence of IBD specialized endoscopist	2.14	0.98–4.66	0.056
Prior dysplasia	0.79	0.37–1.70	0.549
Prior GI cancer	5.99	0.54–66.42	0.145

Follow-up of Patients with Dysplasia

The proportion of patients who underwent surgical resection was higher in those who had dysplasia on random biopsies (21%) or both random and visible dysplasia (21%) compared with those with only visible lesions (5%; P < 0.001). In contrast, polypectomy or endoscopic mucosal resection was more common in the latter group. Patients with any random dysplasia were more likely to have a higher frequency of surveillance exams as a result of the index colonoscopy (P < 0.001). Importantly, patients with dysplasia detected on random biopsies alone had the highest risk of subsequently developing colon cancer (15%) compared with those with dysplasia detected on both random and visible lesions (7%) or visible lesions alone (1%; P < 0.0001).

Follow-up of Patients with Both Random and Visible Dysplasia

We further studied patients who had both random and visible dysplasia to see if the findings on random biopsy altered management. Of the 28 colonoscopies (28 patients), 7 patients had CD, and 21 had UC (Table 3 and Supplemental Table 2). There was broad correlation between degree of dysplasia on random and polypoidal lesions (Table 4); however, the random biopsies often yielded dysplastic lesions at other sites in two thirds of the patients. Among the 7 CD patients, 2 patients had dysplasia detected in the same area of the colon on both random and targeted biopsies, whereas the other 5 (71.4%) patients had dysplasia detected in different segments of the colon. Among the 21 UC patients, 7 patients had dysplasia detected in the same segment of the colon on both random and targeted biopsies, and the majority (14 patients, 66.7%) had dysplasia detected in different segments of the colon by random and targeted biopsies.

TABLE 3.

Characteristics of Random Dysplasia and Visible Dysplasia Found in Ulcerative Colitis Patients Detected on Both Random and Targeted Biopsies

	Random Dysplasia		Visible Dysplasia
Patient	Site	Grade of Dysplasia	Site	Grade of Dysplasia
1	Rectum	1 Sessile	Rectum	1 Sessile
2	Ascending	1 LGD	Ascending	3 LGD
3	Rectum	1 LGD	Sigmoid	1 LGD
4	Ascending	1 LGD	Ascending Cecum	3 LGD
5	Rectum	1 LGD	Cecum	1 LGD
6	Descending	1 LGD	Sigmoid	1 LGD
7	Cecum Ascending	2 LGD	Transverse	1 LGD
8	Transverse Descending	1 indefinite 1 LGD	Rectum Ascending	2 LGD
9	Rectum Sigmoid Descending	1 HGD 2 LGD	Rectum	1 HGD 1 LGD
10	Transverse	1 Sessile	Transverse	1 LGD
11	Transverse	1 LGD	Transverse	1 LGD
12	Rectum	1 LGD	Rectum Descending	2 LGD
13	Transverse	Indefinite	Transverse Ascending	1 LGD 1 HGD 1 Adenocarcinoma
14	Transverse	1 LGD	Ascending	1 LGD
15	Ascending	1 LGD	Transverse Cecum	4 LGD
16	Ascending	1 LGD	Transverse	1 LGD 1 HGD
17	Transverse	1 LGD	Rectum Sigmoid	2 indefinite
18	Transverse	1 Sessile	Descending	2 Sessile
19	Transverse Ascending Cecum	3 LGD	Ascending	1 LGD
20	Rectum	1 LGD	Transverse	1 LGD
21	Rectum Descending	1 LGD 1 HGD 1 Indefinite	Rectum	1 HGD

Abbreviation: Indefinite: indefinite for dysplasia

TABLE 4.

Association Between Degree of Dysplasia on Random Biopsies Compared With Visible Lesions in Patients With Both Types of Dysplasia

		Highest Degree of Random Dysplasia
		Indefinite	LGD	HGD
Highest degree of visible dysplasia	indefinite	0	1	0
	LGD	2	19	0
	HGD	0	2	2
	CRC	1	0	0

DISCUSSION

The incidence of CRC in patients with IBD and surveillance practice for dysplasia have evolved over the past 2 decades. Paralleling the secular decrease in rates of CRC in this population, surveillance has shifted to use of high definition colonoscopes, more frequent use of dye spray–assisted chromoendoscopy or narrow band imaging, and endoscopic or segmental resection rather than total proctocolectomy for visible dysplasia. However, there remains active debate about whether there is a role for random biopsies in the surveillance of cancer. In this retrospective study encompassing nearly a decade and 442 colonoscopies with at least 1 dysplastic lesion, we demonstrated that 18.1% of such procedures identified dysplasia on random biopsies, including 11.8% of procedures in which dysplasia was detected only on random biopsy. Concomitant PSC, longer disease duration, and endoscopically active disease increased the likelihood of a dysplastic lesion being detected on random biopsies. Importantly, patients who had dysplasia detected on random biopsies were more likely to undergo surgical resection and had higher rates of subsequent development of CRC.

Our results confirm that some colonic dysplasia is not visible on CE or HDE, and random biopsies are needed for detection. Prior studies showed that CE and HDE improve dysplasia detection compared with SDE,^{8, 14, 25–27} and targeted biopsies are noninferior to and preferred over random biopsies.^28–30 These studies enrolled IBD patients requiring dysplasia surveillance colonoscopies in which the proportion of patients with dysplasia detected was between 5.6% to 19.1%. This resulted in a small number of patients with dysplasia in whom they concluded limited utility for random biopsies. However, it is difficult to draw conclusions on the results because these studies were underpowered to assess the role of random biopsies. Similarly, in a recent prospective study by Moussata et al, 1000 patients were enrolled for surveillance colonoscopies, and dysplasia was detected in 94 patients (9.4%). Nineteen patients (20.2%) had dysplasia detected by both random and targeted biopsies or only random biopsies.³¹ Ten patients (10.6%) had several dysplastic sites requiring surgical intervention, and omitting random biopsies would have changed clinical management in 7 patients. In comparison with that study, by including all patients with IBD with an identified dysplastic lesion, our sample size of 442 colonoscopies in 300 patients with dysplasia is significantly larger. Of these, 80 colonoscopies (18.1%) had dysplastic sites identified by random biopsies, an estimate similar to the French study.³¹ Thus, the yield of random biopsies in an individual patient is not trivial.

We identified several independent risk factors for random dysplasia among patients with IBD.^{28, 32–40} We found in our study that on multivariate analysis comparing random and visible lesions, longer disease duration, presence of active inflammation, and—importantly—concomitant PSC were independent risk factors for random dysplasia. Some of these risk factors are in keeping with prior studies³¹ in which they determined the presence of PSC, prior history of dysplasia, and disease duration on univariate analysis were predictive for colonic dysplasia, though the number of events were fewer. The presence of PSC in both Moussata et al study and our study has the highest OR for predicting colonic dysplasia. In a retrospective study looking specifically at PSC and IBD patients,³⁹ 71 patients were included, and 16 patients (22 colonoscopies) had dysplasia. Of those colonoscopies with dysplasia, 4 (18.2%) had dysplasia detected on both targeted and random biopsies, and 10 (45.5%) had dysplasia detected on only random biopsies. Random biopsies increased the yield for dysplasia detection by 40%. The association between active inflammation and random dysplasia is interesting, as well. A possible explanation is that active colonic inflammation resulting in mucosal irregularity may mask dysplastic lesions that may have otherwise been visible due to a similar nodular, friable appearance. Alternately, presence of active inflammation on surveillance examination may indicate longstanding poor disease control and higher cumulative burden of inflammation, which, in turn, has also been linked to higher risk of colorectal dysplasia in IBD.^{41, 42} The higher rate of colon cancer or colectomy in those with random dysplasia may reflect both the challenges in surveying this population and the inability to endoscopically resect affected sites, but it could also reflect biologic differences between visible and random dysplasia.

There are several strengths of our study. First, this is one of the largest studies to date in the high definition colonoscopy era that specifically attempts to identify factors predicting random compared with visible dysplasia in patients with IBD undergoing surveillance examination. The large number of included patients allowed us to robustly estimate the predictors of identifying dysplastic lesions on random biopsies. Secondly, we included both CD and UC patients, as there is evidence that the surveillance intervals for Crohn’s colitis and UC should be the same, given the same inflammation-dysplasia-carcinoma sequence. Third, rather than rely on administrative databases, all cases and outcomes were confirmed through manual record review by a board-certified gastroenterologist. In addition, unlike studies that utilize pathology database data, we were also able to define endoscopic appearance (visible or random) and account for endoscopic intervention such as polypectomy or mucosal resection.

We acknowledge several limitations to our study. First, our study was retrospective, and thus we could not identify the number of biopsies taken at each surveillance colonoscopy. We could not confirm that at least 33 biopsies were taken. However, this would only strengthen our argument that random biopsies are beneficial in current practice, as fewer biopsies in practice would bias our findings toward the null. Secondly, the procedures were performed by many GI operators with varying levels of expertise and practice techniques. However, this allows the results to be more generalizable. Bias introduced by non-IBD specialist endoscopists would only make our estimates more conservative. Further, globally, the vast majority of surveillance exams in patients with IBD are likely to be performed at nonreferral centers by endoscopists of varying experience, making our findings more generalizable and relevant as opposed to a clinical trial examining efficacy of an intervention where few consistent specialist operators would be important. Thirdly, though the procedures are typically scheduled for 30-minute blocks, we do not know the actual withdrawal time, which could decrease adenoma detection for targeted biopsies. However in our study, the neoplastic sites detected by random biopsies alone was 11.7%, which is similar to the SCENIC meta-analysis when comparing random and targeted biopsies.⁸ Lastly, we were limited in our ability to adjust for other potential risk factors. Although we had access to the endoscopy report, the data may have been incomplete regarding structural changes in the colon, inflammation appearance, and presence of pseudopolyps. Further prospective studies are needed to create a standardized protocol for assessing potential risk factors.

In conclusion, our findings suggest that nearly one fifth of dysplastic lesions in patients with IBD is detected by random biopsies. We also identified certain high risk groups may be most likely to benefit from this strategy, including those with longer disease duration, concomitant PSC, and those with active inflammation. Furthermore, in patients where dysplasia was detected as both random and visible lesions, the site of dysplasia detected on random biopsies was separate from the visible lesion, which would have had an impact on surgical management such as segmental resection. Patients with random dysplasia had worse outcomes, including higher rates of colorectal cancer, than those with visible lesions alone. Adequately powered, longitudinal, prospective studies are needed to robustly quantify the benefit of random biopsies for detection of dysplasia in patients with IBD.

Author Contribution: AH contributed to the conception and design of study, acquisition and compilation of data, statistical interpretation, and drafting of manuscript. KB and BK contributed to the drafting of manuscript. AA contributed to the study design, acquisition and compilation of data, statistical analysis, and drafting of the manuscript. All authors gave final approval of the manuscript.

Supported by: This work is supported by the National Institutes of Health (P30 DK043351) to the Center for Study of Inflammatory Bowel Diseases.

Conflicts of Interest: AA is funded by the Crohn’s and Colitis Foundation, National Institutes of Health (R03 DK112909), and the Chleck Family Foundation; has served on scientific advisory boards for Kyn therapeutics and has research funding from Pfizer. BK is funded by Crohn’s and Colitis Foundation (567635).