Assessment of the Degree of Variation of Histologic Inflammation in Ulcerative Colitis

Abstract

Introduction

Treatment of ulcerative colitis (UC) now includes mucosal healing. Adoption of histologic end points is hindered by a lack of evidence guiding optimal sampling, interpretation, and reproducibility of results.

Methods

We analyzed biopsy fragments from colonoscopies in 92 patients with UC. Fragments were scored using 6-point histologic inflammatory activity (HIA) scale. Variability was determined using ordinal representations of HIA scores. The most frequently observed score and percentage of biopsy fragments with that score were determined for each biopsy, each segment, and across all 3 segments for each colonoscopy. Mean percentages and 95% confidence intervals (CIs) were calculated.

Results

We reviewed 1802 biopsy fragments. The mean percentages of intrasegment biopsy fragments with the same HIA score were 85.5% (95% CI, 80.9% to 92.9%), 79.6% (95% CI, 76.0% to 87.3%), and 82.7% (95% CI, 79.1% to 90.0%) for the rectum, left colon, and right colon, respectively. The mean percentage of intersegment biopsy fragments with the same HIA score was 70.2% (95% CI, 65.7% to 82.5%). The mean percentages of intrabiopsy fragments with the same HIA score were 83.3% (95% CI, 77.6% to 93.5%), 83.6% (95% CI, 80.1% to 89.7%), and 90.2% (95% CI, 87.6% to 94.7%) for the rectum, left colon, and right colon, respectively. All 3 analyses revealed increased variation when a greater degree of histologic inflammation was present in the biopsies (mean HIA score ≥2).

Conclusions

We found minimal variability between degree of inflammation among biopsy fragments within and among different colorectal segments in UC, suggesting that even a single biopsy would adequately reflect the inflammation of the entire colorectum. These findings have significant implications for the use of histology as a clinical target and trial end point in UC.

Introduction

Traditionally, clinical trials in ulcerative colitis (UC) have utilized scoring systems that depend upon patient-reported symptoms. However, it has long been recognized that endoscopically and histologically evident active inflammation can still be present despite a patient being in clinical remission.^1-4 In addition, endoscopic and histologic healing have been linked to improved long-term outcomes in UC patients.^6,7 Recently, endoscopic confirmation of mucosal healing has been invoked as a potential therapeutic goal in UC and is routinely included as an end point in clinical trials. However, it is suggested that endoscopy may underestimate disease extent and/or resolution, as the correlation between histology and endoscopy seems to be stronger when the samples are obtained during active inflammation and less so in endoscopically quiescent disease.⁵ As such, microscopic features of active inflammation may persist in macroscopically inactive UC.^1,8-11

Endoscopic remission is defined as the absence of blood, erosions, and ulcers in all visualized segments of gut mucosa,⁵ whereas histologic healing is variously defined as the microscopic absence of residual mucosal inflammation with distinctive changes in crypt architectural distortion and/or atrophy, absence of or less than 5% neutrophils, or completely normal mucosa.^12,13 It has been previously demonstrated that despite symptomatic and endoscopic remission, persistent active histologic inflammation in patients with symptomatic and endoscopic remission is associated with an increased risk of disease relapse.¹⁴ Furthermore, histologic remission has been shown to be associated with lower rates of hospitalizations and a reduction in epithelial neoplasia risk.^15-19 In light of these findings, there is interest in including histologic remission as a therapeutic goal in UC; and histologic remission—or a composite of histologic and endoscopic healing—are now secondary end points to assess therapeutic efficacy in clinical trials.^{5, 20}

The Food and Drug Administration (FDA) Clinical Trial Guidelines for Industry discusses the limitations of current histological inflammatory activity grading systems and details methods for the inclusion of histology as an end point (eg, requiring centralized evaluation of histology).^{21, 22} However, the number and location of biopsies is not specified, and best practice for such assessment has not been evidence-based. Practical details such as the quantity and location of biopsies necessary for appropriate assessments are also needed before widespread application can occur. We analyzed intrasegment, intersegment, and intrabiopsy variation of histologic scores of colonic mucosal inflammation in patients with UC.

Methods

Patients

This retrospective study included patients with a histologically confirmed diagnosis of ulcerative colitis with no evidence of colonic neoplasia or dysplasia. The study population was derived of subjects with UC included in the control arm of a previous case-control study assessing inflammation as a risk factor for colon cancer.¹⁸ Biopsies were obtained from subjects included in the University of Chicago Medicine Inflammatory Bowel Disease Endoscopy Database and University of Chicago Medicine Inflammatory Bowel Disease Registry, 2 databases that include subjects with a diagnosis of inflammatory bowel disease who are seen at our center. This study was approved by the University of Chicago Institutional Review Board. Subjects had previously provided consent for their medical information to be used for research purposes.

Pathology

Hematoxylin-eosin stained slides from all diagnostic, screening, and surveillance endoscopy biopsies for subjects in the control group obtained between May 1994 and July 2005 were included in this study. The left colon was defined as the colon between the sigmoid colon and the splenic flexure; the right colon was defined as spanning from the ileocecal valve to the splenic flexure. Biopsies were excluded from the data analysis if they were not labeled in a way that clearly described the location from which they were obtained. Each colonoscopy was considered separately, and each biopsy fragment was assessed as an individual specimen and had been assigned a score from the 6-point histologic inflammatory scale (HIA) adapted from the Geboes score and used in our previous study.^{18, 23} In our work, adoption of the traditional score was created to address the category of histologic normalization (see Supplemental Table 1). A score of 0 was given to normal tissue completely uninvolved by disease with no architectural distortion or inflammatory infiltrates. An HIA score of 1 was given to tissue with quiescent disease exhibiting architectural distortion and increased lamina propria mononuclear infiltrates. Tissue with an increased number of granulocytes in the lamina propria but no intraepithelial granulocytes was assigned an HIA score of 2. An HIA score of 3 denoted intraepithelial granulocytes without crypt abscesses, whereas the presence of crypt abscesses in less than 50% of crypts was rated as an HIA score of 4. If crypt abscesses were present in more than 50% of crypts or erosions or ulcerations were present, an HIA score of 5 was assigned (Figure 1).

Figure 1.

Six point inflammatory activity (HIA) scale.¹⁸

As described in our previous study, 2 pathologists independently reviewed samples using the novel HIA scoring system for the first 30 surveillance endoscopy procedures (429 biopsies total) and reviewed together to develop consensus. The weighted kappa values were 0.49, 0.54, and 0.60 for the first, second, and third sets of 10 procedures reviewed, respectively. This indicated moderate agreement without training, with improvement to substantial agreement after training, which demonstrated that our novel scale was reproducible.¹⁸ The remaining tissue samples were reviewed by only 1 of the 2 study pathologists. The biopsy scores were organized in the database by colonoscopy (all the biopsies from a given colonoscopy were assigned to 1 of the 2 pathologists).

Data Analysis

Variability was determined using ordinal representations of the HIA scores. We examined the extent to which values clustered together within and across segments, irrespective of where the clustering occurred. All analyses were performed in R version 3.0.2 (2013-09-25).

In the ordinal intrasegment analysis, the most frequently observed score and the percentage of pieces with that score were determined for each segment in a colonoscopy. In the ordinal intrabiopsy analysis, the most frequently observed score and the percentage of pieces with that score were determined for all slides with more than 2 pieces. The ordinal intersegment analysis identified the most frequently observed score and the percentage of biopsies with that score in all 3 segments. In each analysis, if a subject had more than 1 colonoscopy, an average of the percentages from each colonoscopy was calculated for that patient. Mean percentages and 95% confidence intervals were then calculated for all of the patients. In order to determine the distribution of the scores not included in the percentage of the most frequently observed score, the average percentage of HIA scores within one (+/- 1) of the most frequently observed scores was also calculated. Finally, patients were classified as “high” or “low” inflammation if the mean HIA scores (as continuous outcomes) across a biopsy, segment, or colon was ≥2.0 or less than 2.0, respectively. The most frequently observed scores and the percentage of biopsies with these scores were then calculated for the 2 groups, and permutation tests were used to assess the differences between the average percentages.

Results

We analyzed 1802 pieces of 854 biopsies obtained from 182 colonoscopies performed on 92 patients with UC. We excluded 562 biopsy pieces from the analysis because the description of the location in the endoscopy report was not clearly labeled. Subject demographic and clinical characteristics are described in Table 1, and the HIA scores included in the analyses are shown in Table 2.

Table 1.

Patient Demographics and Clinical Characteristics

Characteristic	Value
Male—n (%)	48 (51)
Mean age—y ± SD	45.8 ± 10.7
Mean age of diagnosis—y ± SD	26.6 ± 9.4
Mean duration of disease—y ± SD	19.1 ± 9.3
Smoking history
Never—n (%)	70 (76)
Active—n (%)	11 (12)
Past—n (%)	11 (12)
Disease extent
Pancolitis	88 (96)
Left sided colitis	4 (4)
Mayo score at colonoscopy—mean ± SD	1.16 ± 1.01
Treatments at time of colonoscopy
5 ASA—n (%)	92 (100)
Thiopurines—n (%)	40 (43)
Corticosteroids—n (%)	42 (46)
Total number of colonoscopies—n	182
Total number of biopsies—n	854
Total number of biopsy pieces—n	1802

Table 2.

The Number of Biopsy Fragments With Each HIA Score Per Segment.

	HIA score
Segment	0	1	2	3	4	5	Total
Rectum	1	126	27	32	5	12	203
Left Colon	17	503	72	107	20	18	737
Right Colon	168	530	71	67	17	9	862
Total	186	1159	170	206	42	39	1802

For the ordinal analysis, the mean percentages of intrasegment biopsy pieces with the same HIA score were 85.5% (95% CI, 80.9% to 92.9%) for the rectum, 79.6% (95% CI, 76.0% to 87.3%) for the left colon, and 82.7% (95% CI, 79.1% to 90.0%) for the right colon. The mean percentages of intrasegment biopsy pieces with an HIA score within 1 value of the most frequently observed HIA score were 9.0% (95% CI, 95% CI, 5.0% to 13.5%) for the rectum, 11.4% (95% CI, 8.3% to 14.6%) for the left colon, and 9.6% (95% CI, 6.6% to 12.7%) for the right colon. The mean percentage of intersegment biopsy pieces with the same HIA score (in 38 subjects) was 70.2% (95% CI, 65.7% to 82.5%), and an additional 16.0% of pieces were within one of the most frequently observed score (Figure 2). The intrabiopsy ordinal analysis, which included 164 biopsies, revealed the mean percentages of biopsy pieces with the same HIA score: 83.3% (95% CI, 77.6% to 93.5%) for the rectum, 83.6% (95% CI, 80.1% to 89.7%) for the left colon of the colon, and 90.2% (95% CI, 87.6% to 94.7%) for the right colon. The mean percentages of biopsy pieces with an HIA score within one of the most frequently observed HIA scores was 9.68% (95% CI, 5.4% to 14.0%) for the rectum, 7.2% (95% CI, 5.2% to 9.3%) for the left colon, and 5.7% (95% CI, 4.1% to 7.5%) for the right colon.

Figure 2.

Heat maps characterizing agreement of the histologic inflammatory activity scores for each colonoscopy in the ordinal intrasegment and intersegment analyses.

When the subjects’ mean intrasegment HIA scores were grouped into low and high inflammation categories, the ordinal analysis revealed that the segments in the higher inflammation group had significantly more intrasegment variation for the left and right colon (P = .0013 and .025, respectively), but not for the rectum (P = .28; Table 3). When the subjects’ mean intersegment HIA scores were divided into low and high inflammation groups, the ordinal analysis revealed that the average percentage of samples with the same score in the high inflammation category was 59.1% (n = 8); and in the low inflammation category, it was 73.2% (n = 30; difference of 14.1%, P = .05). Finally, in the low and high inflammation groups, there was a statistically significant difference between the degrees of intrabiopsy variation for the left and right colon (P = .0067 and 0.021, respectively), but not the rectum (P = .063; Table 4).

Table 3.

The Differences in the Degrees of Intrasegment Variability in Subjects With an Overall Mean HIA Score <2 and in Those With an Overall Mean HIA score ≥2.

Segment	High Inflammation	Low Inflammation	Difference (P)
Rectum (%)	80.6	87.6	-6.9 (0.28)
Left colon (%)	69.7	83.6	-14 (0.0013)
Right colon (%)	68.7	84.5	-16 (0.025)

Table 4.

The Differences in the Degrees of IntraBiopsy Variability in Subjects With a Mean Intrabiopsy HIA Score < 2 and in Those With an Overall Mean HIA score ≥2.

Segment	High inflammation	Low inflammation	Difference (P-value)
Rectum (%)	73.5	87.4	−14 (0.063)
Left colon (%)	73.8	87.9	−4 (0.0067)
Right colon (%)	80.2	91.3	−11 (0.021)

Discussion

We analyzed the variability of histologic inflammation within individual subjects with UC. In a large number of biopsy fragments from a large number of patients, we have demonstrated minimal variability within a colonic segment and across different, segments as well as across a single biopsy itself. These findings provide necessary information for standardization of the number and location of biopsies for evolving clinical trials and clinical practice end points.

The ordinal intrasegment analysis revealed a biopsy fragment obtained from 1 location within a colonic segment had the same or within 1 HIA score as a fragment obtained from a second location within a colonic segment more than 90% of the time. Furthermore, 86% of the time, a biopsy fragment from 1 colonic segment had the same or within 1 HIA score as a biopsy fragment from within a different colonic segment. These results suggest that a single biopsy fragment obtained in the distal rectum from a flexible sigmoidoscopy performed on a UC patient with pancolitis is likely to have the same degree of histologic inflammation in the rectum and in other colonic segments. This minimal degree of histologic variability suggests that fewer biopsies may be needed to establish the meaningful end point of histologic remission. Thus, from a practical standpoint, it would make the inclusion of this end point in future clinical trials and in clinical practice more convenient/less burdensome for the patients, researchers, and endoscopists, as well as potentially less expensive. One can further imagine a clinic based blind biopsy performed of the rectum to inform treatment adjustments.

A previous study comparing histologic sections of the ascending colon and rectosigmoid colon from colectomy specimens in patients with UC revealed both micro heterogeneity and ascending colon vs rectosigmoid colon discordance.²⁴ This is in contrast with our study thta describes minimal variability across the various colonic segments; this is possibly because the study by Harpaz et al included mostly colectomy specimens from patients with medically (and therefore quite active) refractory disease.²⁵ In contrast, in our study the mean Mayo endoscopic score was 1.16 ± 1.01, indicating less active disease in a population of patients mostly undergoing colonoscopic disease surveillance. One possibility is that in patients with endoscopic remission or less active disease, the histologic degree of inflammation in UC is more homogenous. Indeed, areas with a higher inflammatory burden showed significantly more intrasegment and intrabiopsy variability, particularly in the left colon. We also found significantly more intersegment variability across colons with higher degrees of inflammation, as well as more intrasegment and intrabiopsy variability in biopsies from more inflamed right colons, but these observations were in a smaller number of patients. Correlating this finding to endoscopic data is important, as it may indicate that in patients with low degrees of inflammation, fewer biopsies may be able to represent larger colonic areas. This may be particularly useful in clinical trials assessing histologic remission. Future studies will further explore the issue. Whether this is true or not, the real issue will be whether it affects clinical decision-making differently.

There are several limitations to this study: First is the large number of biopsy fragments with a lower degree of inflammation. Because these patients were sampled during surveillance colonoscopies, this is expected (to a degree), but it does limit broader generalization of the findings. A second limitation is potential sampling bias of the endoscopists, as there were no prerequisite colonic locations that needed to be biopsied; it is possible that the endoscopists were obtaining biopsies in areas that have less histologic variability. Third, these analyses rest upon the assumption that there is minimal intrarater variability in the determination of histologic scores by a histopathologist. However, a study by Mosli et al²⁴ demonstrated strong intrarater agreements with several histologic scales, including elements of our own. Finally, it is notable that most of these patients had pancolitis, and none were treated with biological agents. This is the nature of the data set used for this analysis.

The findings and implications of this study suggest that a single nontargeted biopsy of the rectum may be sufficient to inform the degree of inflammation in the remainder of the colon. An exception to this presumption would be in patients with concomitant primary sclerosing cholangitis, inasmuch as our group had previously identified greater histological activity in the proximal (right) colon.²⁶ In conclusion, we have demonstrated that the assessment of histologic degree of inflammation in UC can be performed with moderate to very good accuracy with a limited number of biopsies. These findings have considerable importance for the evolving role of histology as an end point for clinical trials and clinical management of UC. Future studies are needed to clarify the value of these findings in larger, prospective cohorts with predefined biopsy protocols.

Conflicts of Interest

A.E.M., S.W., M.A., S.R.G., and J.H. have no relevant disclosures.

N.A.C. has served as a consultant for Seres Therapeutics.

J.R.T. is a cofounder of Thelium Therapeutics and has served as a consultant for Entrinsic, Immunic, and Kallyope.

D.T.R. has received grant support from Takeda; and has served as a consultant for Abbvie, Altrubio, Arena Pharmaceuticals, Bristol-Myers Squibb, Genentech/Roche, Gilead Sciences, Iterative Scopes, Janssen Pharmaceuticals, Lilly, Pfizer, Prometheus Biosciences, Takeda, and Techlab Inc.