Endoscopic and Histological Assessment, Correlation, and Relapse in Clinically Quiescent Ulcerative Colitis (MARQUEE)

Abstract

Objective

It is difficult to predict relapse in quiescent ulcerative colitis (UC), but newer endoscopic and histological indices could improve this. This study aimed to determine in UC patients in clinical remission (1) the prevalence of active endoscopic and histological disease; (2) the correlation between endoscopic and histological scores; and (3) the predictive power of these scores for clinical relapse.

Design

This multicenter prospective cohort study conducted by the Crohn’s and Colitis Foundation Clinical Research Alliance included 100 adults with UC in clinical remission undergoing surveillance colonoscopy for dysplasia. Endoscopic activity was assessed using the Mayo endoscopic score (MES), ulcerative colitis endoscopic index of severity (UCEIS), and ulcerative colitis colonoscopic index of severity (UCCIS). Histology was assessed with the Riley index subcomponents, total Riley score, and basal plasmacytosis.

Results

Only 5% of patients had an MES of 0, whereas 38% had a score of 2 to 3; using the UCEIS, the majority of patients had at least mild activity, and 15% had more severe activity. Many patients also had evidence of histological disease activity. The correlations among endoscopic indices, histological subcomponents, and total score were low; the highest correlations occurred with the subcomponent architectural irregularity (ρ = 0.43–0.44), total Riley score (ρ = 0.35–0.37), and basal plasmacytosis (ρ = 0.35–0.36). Nineteen patients relapsed clinically over 1 year, with the subcomponent architectural irregularity being the most predictive factor (P = 0.0076).

Conclusions

This multicenter prospective study found a high prevalence of both endoscopic and histological disease activity in clinically quiescent UC. The correlations between endoscopy and histology were low, and the power to predict clinical relapse was moderate.

INTRODUCTION

Endoscopic mucosal healing has become an important end point in clinical practice and clinical trials for ulcerative colitis (UC), as it is associated with lower rates of symptomatic and corticosteroid-free remission,^{1, 2} colectomy,^1–4 hospitalization,³ and use of immunosuppressive medications.³ Because UC is largely a superficial mucosal disease, true mucosal healing can be assessed by histology.⁵ Recently, histological normalization in UC was shown to be associated with reduced rates of clinical relapse,⁶ and histological remission was superior to endoscopic remission as a predictor of long-term corticosteroid use and hospitalization.⁷ Not surprisingly histological disease activity has become an important outcome measure in recent UC clinical trials of ozanimod⁸ and ustekinumab.⁹

It was recognized as early as 1956 that clinical symptoms often underrepresent objective measures of UC disease activity.¹⁰ Although a number of older studies have reported prevalences of endoscopic and histological disease activity in UC patients in clinical remission,^10–12 more modern estimates are relatively lacking. Moreover, the historical correlation between endoscopic and histological disease activity assessment has been modest at best.⁵ Furthermore, there are a number of novel endoscopic and histological indices, and it is unclear which should be used preferentially in various situations, especially for predicting disease relapse. Though the Mayo endoscopic score (MES)¹³ is commonly used in clinical trials, the ulcerative colitis endoscopic index of severity (UCEIS)¹⁴ and the ulcerative colitis colonoscopic index of severity (UCCIS)¹⁵ have been developed but not studied with respect to correlation with histology or prediction of disease relapse. Also unclear is which specific aspect of histology is most useful in terms of correlation with endoscopy or ability to predict clinical relapse during quiescent disease. The Riley index¹⁶ stratifies histology into separate subcomponents that measure multiple aspects of microscopic disease activity.

Given the lack of clarity in the multiple areas mentioned previously, the following prospective multicenter observational study was undertaken. The primary aims of this study were (1) to determine the prevalence of active endoscopic disease defined by the MES, UCEIS, and UCCIS, active histological disease defined by the Riley index subcomponents and total Riley score, and basal plasmacytosis in patients with clinically quiescent UC undergoing routine surveillance colonoscopy for dysplasia; and (2) to correlate the endoscopic findings with the histological findings. The secondary aims were to determine the accuracy of these endoscopic and histological scores in predicting clinical relapse within 1 year.

METHODS

Study Design and Population

This was a multicenter prospective cohort study performed by the Crohn’s and Colitis Foundation Clinical Research Alliance, which included 5 IBD centers: University of Pennsylvania, University of North Carolina–Chapel Hill, University of Michigan, Beth Israel Deaconess Medical Center, and University of Maryland. Adults aged 18 to 75 years with an established diagnosis of UC for at least 6 months based on characteristic clinical symptoms, endoscopy, and histology were enrolled between April 2013 and January 2016. Only patients with quiescent disease who were off corticosteroids for at least 90 days and who were undergoing routine surveillance colonoscopy for dysplasia were included. Quiescent disease was defined as a 6-point partial Mayo score (PMS) ≤1.¹⁷ Patients with a history of Crohn’s disease or indeterminate colitis, dysplasia or colorectal cancer, or partial colectomy were excluded. Patients treated with 5-aminosalicylates (5-ASA) agents (oral and/or topical) or antibiotics for their UC were included, provided that their medication dose was stable for at least 30 days before study entry. Patients treated with immunomodulators or antitumor necrosis factor (TNF) agents were included if their medication dose was stable for at least 3 months before study start.

Data Collected

Clinical, endoscopic, and histological data were collected at the time of surveillance colonoscopy, and follow-up clinical data were collected every 3 months either during an office visit or by telephone. Clinical data included demographic information, specifically age, sex, race/ethnicity, maximum level of education, smoking status, alcohol use, history of appendectomy, and family history of UC in a first-degree relative; disease-related factors, specifically duration and extent of UC, duration of clinical remission at baseline, use of corticosteroids in the 1 year before study start, and history of extraintestinal manifestations; and medications, including UC medications and certain non-UC medications—aspirin for cardioprotection, nonsteroidal anti-inflammatory drugs (NSAIDs), and oral contraceptives. Of note, clinical status (relapse/remission) was assessed every 3 months using the 6-point PMS, with clinical relapse defined as a PMS ≥2 or need for change in treatment regimen based upon symptoms consistent with clinical relapse. Importantly, the research coordinators ascertaining clinical relapse status were blinded with respect to the endoscopy and histology scores.

Endoscopic data were captured as follows. Each subinvestigator performing surveillance colonoscopies recorded 30-second video clips representative of the most endoscopically active areas of the 5 segments of the colon: cecum/ascending, transverse, descending, sigmoid, and rectum. The 5 videos for each patient were scored by a central endoscopy reader specializing in the clinical care of and research in UC but not otherwise involved in the study according to the 3 endoscopic indices of interest: MES (total score range, 0–3), UCEIS (using the original total score range of 3–11), and UCCIS (total score range, 0–162; Supplementary Table 1). Given when this study was started, the scoring for the UCEIS was based on the original classification rather than the newer classification with score range 0–8, in which the lowest score for each category is 0 instead of 1.

With respect to histological data, patients underwent multiple dysplasia surveillance biopsies throughout the colon by the sub-investigator as part of routine standard of care. All biopsies were scored by a central pathologist at the University of Pennsylvania specializing in the clinical care of and research in UC but not otherwise involved in the study according to the Riley index subcomponents, total Riley score (Supplementary Table 2), and the degree of basal plasmacytosis graded 0–3 (0, none; 1, mild; 2, moderate; 3, severe). The Riley index was selected, given its clear, independent, categorical separation of histological components of acute and chronic inflammation and architectural integrity, and thus was preferred to the Geboes index¹⁸; also, this study was conducted before the advent of the Nancy¹⁹ or Robarts²⁰ Histological Indices. Basal plasmacytosis was selected due to its potential utility in predicting clinical relapse.²¹ Of note, there was variation in clinical practice with respect to biopsy location designation, as some endoscopists separated the biopsies into all 5 colon segments or 4 colon segments (with sigmoid and descending combined), whereas others put all biopsies into the same jar not separated by segment. For patients with biopsies separated into segments, maximum subcomponent and total scores were listed for each colon segment. In those with only 1 jar of biopsies, the maximum subcomponent and total scores across all biopsies was used.

Outcomes

There were 3 primary outcomes in this study: (1) the prevalence of endoscopic disease activity per MES, UCEIS, and UCCIS; (2) the prevalence of histological disease activity per Riley index subcomponents, total Riley score, and basal plasmacytosis; and (3) the correlation between endoscopic and histological scores. Secondary outcomes included prediction of clinical relapse within 1 year by each of these endoscopic and histological scores/components.

Statistical Analysis

This study was originally planned as a pilot for a future randomized controlled trial of patients with clinically quiescent UC but with active endoscopic disease on 5-ASA therapy to be randomized to biologic therapy vs continuing 5-ASA. As such, it was determined that 100 patients would be a reasonable sample size for a pilot to determine the prevalence of endoscopically active disease.

Descriptive statistics included patient demographics and categorization of disease phenotypes and medication use. The prevalence of endoscopic scores were reported by strata of each index as follows: MES (0, 1, 2, 3); UCEIS (3, 4–6, 7–9, 10–11), and UCCIS (0–40, 41–80, 81–120, 121–162). For histological scores, the prevalence of the Riley index subcomponents and basal plasmacytosis were reported by strata (0, 1, 2, 3), and total Riley score was described by a mean with standard deviation and range. Spearman correlation coefficients were calculated to assess the association of degree of endoscopic activity with degree of acute and chronic inflammatory histological activity for each component of the Riley index and basal plasmacytosis. For patients whose biopsies were separated by colon segment, correlation between endoscopic and histological scores was calculated by segment (primary outcome). For those whose biopsies were not separated by colon segment, correlation was calculated using the maximum endoscopic scores across all 5 colon segments and the maximum histological subscores or total score (secondary outcome). Additionally, correlation was computed for all patients using the maximum endoscopic scores and the maximum histological scores across all 5 colon segments (secondary outcome).

The rates of clinical relapse were calculated per endoscopic and histological score strata. To conduct an exploratory analysis to determine whether any endoscopic index or histological subscore or total score were associated with clinical relapse, the Cochran-Armitage exact test for trend was used in univariable analysis. Similarly, this test was used to determine whether any baseline demographic, disease-related, or medication variables were associated with clinical relapse. A multivariable logistic regression predictive model for clinical relapse was also developed, incorporating all variables with P < 0.1 on univariable analysis to determine whether a combination of variables was more predictive than any single variable alone.

RESULTS

One hundred patients were enrolled in this study. Baseline characteristics are shown in Table 1. The mean age was 50 years, 45 were women, most were white, and 58% had never smoked. More patients had extensive colitis instead of left-sided colitis; mean disease duration was 19 years with a mean of 5 years since last UC flare. Seventy-nine patients had a PMS of 0, and 21 had a score of 1, with 96 having a rectal bleeding subscore of 0 and 83 having a stool frequency subscore of 0. At baseline, 71 patients were receiving oral 5-ASA, with 14 receiving topical 5-ASA. Immunosuppressant medication use included 33 patients taking immunomodulators, 26 taking TNF inhibitors, and 1 taking cyclosporine. Only 12 patients had received corticosteroids in the year before enrollment.

Endoscopy scores were available for 99 patients, as videos were not recorded for 1 patient. A surprisingly high number of patients had evidence of endoscopic disease even though they were in clinical remission. Using the MES, 38 patients had a score of 2 to 3, and only 5 had a score of 0 (Table 2). Similarly with the UCEIS, only 5 patients had the lowest score of 3, whereas 15 had a score of 7 to 9. Using the UCCIS, 12 patients had a score >40.

Histological scores were available for 99 patients, as slides were unable to be located for 1 patient. A large number of patients also exhibited evidence of histological disease activity (Table 3). The mean total Riley score in this cohort was 3.34 (standard deviation 2.97, range 0–14). Specifically, 44 patients had some degree of acute inflammation, 26 had crypt abscess, 12 had mucin depletion, 68 had some degree of chronic inflammation, 23 had disruption of surface epithelial integrity, and 77 had crypt architectural irregularity. Additionally, 65 patients had some degree of basal plasmacytosis.

Correlation between endoscopy and histology could be calculated for 98 patients, as the patients missing endoscopic and histological scores were different; the UCCIS was not used in correlation analyses, as it is a total composite score, not a score based on a maximum from one segment of the colon. Fifty-eight patients had endoscopic and histological scores recorded in the same colon segments in a total of 275 samples, with an average of 4.74 colon segments scored per patient. Correlation—although significant—was generally low for all components of the Riley index, total Riley score, and basal plasmacytosis with either the MES or UCEIS, with very similar correlations regardless of endoscopic index (Table 4). The highest correlations occurred with the subcomponent architectural irregularity (ρ = 0.43–0.44), total Riley score (ρ = 0.35–0.37), and basal plasmacytosis (ρ = 0.35–0.36).

In the other 40 patients in which biopsies were not separated by colon segment, the maximum endoscopic and histological scores across all 5 segments were used. In these patients, correlations were also generally low but with a wider range across instruments, with the highest correlations observed for the subcomponent acute inflammatory cell infiltrate (ρ = 0.31–0.49), total Riley score (ρ = 0.34–0.46), and basal plasmacytosis (ρ = 0.33–0.49; Supplementary Table 3A). When calculating correlation using the maximum endoscopic scores and histological scores across all colon segments for all 98 patients, the results were similar with the highest correlations seen for the subcomponent acute inflammatory cell infiltrate (ρ = 0.33–0.43), total Riley score (ρ = 0.39–0.44), and basal plasmacytosis (ρ = 0.37–0.42; Supplementary Table 3B).

During the 12-month follow-up period, 19 patients experienced a clinical relapse: 4 during months 0 to 3, 8 during months 4 to 6, two during months 7 to 9, and 5 during months 10 to 12. When examining the association between baseline endoscopic activity scores and clinical relapse during the follow-up period, there was a suggestion that the MES was predictive of relapse, but the UCEIS and UCCIS were not (Table 5). With respect to baseline histology and clinical relapse, the Riley subcomponent architectural irregularity and basal plasmacytosis were significantly predictive of relapse in univariable analysis (Table 6). The most predictive multivariable model, which incorporated the MES, architectural irregularity, and basal plasmacytosis, had the following formula: risk of clinical relapse score = (MES + 1) x (architectural irregularity score + 1) x (basal plasmacytosis score + 1). This model was significantly predictive of clinical relapse (P = 0.0079), with area under the ROC curve of 0.67. A cutoff score of 16 was associated with a sensitivity of 78%, whereas a cutoff score of 8 was associated with a specificity of 81%. Of note, no demographic, disease-related, or medication variable was significantly predictive of relapse in univariable analysis.

DISCUSSION

At the present time, it is unclear which endoscopic and histological activity indices or components are most useful in assessing quiescent UC. In this multicenter prospective observational cohort study in the anti-TNF era, we found evidence of both endoscopic and histological disease activity in the majority of patients with clinically quiescent UC undergoing dysplasia surveillance. Overall, the correlations between endoscopic indices and histological scores were low. Nearly one fifth of the patients relapsed clinically over 1 year, with the Riley subcomponent architectural irregularity being the most predictive factor.

Historical data from the 1950s and 1960s showed evidence of active endoscopic and histological disease in 39%–60% and 38%–90%, respectively, of UC patients in clinical remission.⁵ More recently, a single-center study by Zenlea et al demonstrated that 55% and 52% of UC patients in clinical remission had evidence of active endoscopic and histological disease, respectively.²² Our multicenter prospective study provides further evidence of objective subclinical disease activity in the modern treatment era. Specifically, only 5% of patients had an MES of 0, whereas 38% had a score of 2 to 3. Using the UCEIS, the majority of patients had at least mild endoscopic activity (score of 4–6), and 15% had a more severe score of 7 to 9. We also observed high rates of histological disease activity, with almost half of the patients having evidence of acute inflammation, two thirds having chronic inflammation or basal plasmacytosis, and three quarters having architectural irregularity. We suspect that the higher rates of endoscopic and histological disease activity in our study compared with Zenlea et al were due to a higher proportion of patients with moderate to severe UC, as 33% were taking immunomodulators and 26% were receiving anti-TNF therapy in our cohort, compared with 18% and 6%, respectively, in the previous study.²²

Given that endoscopy may underestimate the degree of inflammation in UC, it was not surprising that the correlation between endoscopic and histological scores was imperfect; however, it was surprising that the correlations were so low. Prior data examining this question with these specific indices are limited. A single-center study from Leuven revealed a modest correlation between the MES and Riley index but did not examine each subcomponent of the Riley index.²³ A single-center study from Oxford found a high correlation between the UCEIS and both Nancy and Robarts Histological Indices; however, this study included patients with active disease, which may have contributed to the high correlations observed.²⁴ In our study of quiescent UC patients, the highest correlations between endoscopy and histology occurred with architectural irregularity, total Riley index, basal plasmacytosis, and chronic inflammatory infiltrate. Although the newer Nancy and Robarts Histological Indices have identified the presence of active inflammation, especially neutrophils, as being the most important histological element in patients with predominantly active UC,^{19, 20} it is possible that chronic inflammatory elements, including basal plasmacytosis and architectural changes, may be important in quiescent UC, as we observed. Interestingly, the correlations with histology were very similar for the MES and UCEIS for all histological subcomponents.

With respect to predicting clinical relapse in UC, a recent meta-analysis investigated subcomponents of histology in a subset of included studies and observed that the acute and chronic inflammatory cell infiltrates, including crypt abscesses but not basal plasmacytosis or architectural distortion, were associated with higher risks of clinical relapse.²⁵ These results are in contrast to those of our study in which architectural integrity and basal plasmacytosis were the most predictive histological features for clinical relapse. As our study was not powered to investigate prediction of clinical relapse as a primary end point, it is possible that our results differed due to statistical chance with a lower number of relapse events. Our study also differed in that ascertainment of clinical relapse was performed by research coordinators blinded to the endoscopy and histology scores, whereas this was not the case for the majority of studies in the meta-analysis. Although our multivariable predictive model incorporating MES with architectural irregularity and basal plasmacytosis had moderate predictive power for clinical relapse, it is possible that with a larger number of relapses its predictive power could have been improved.

Our study has a number of notable strengths, beginning with its prospective multicenter design in the modern treatment era. Importantly, all endoscopic scoring was performed by a single central reader using videos, and all histological scoring was done centrally by a single expert IBD pathologist. Additionally, both central readers were blinded to the clinical status and other scores (ie, endoscopy reader was blinded to the histology scores and vice versa), and ascertainment of clinical relapse was blinded with respect to the endoscopy and histology scores. In this way, inter-rater variability was eliminated, and bias was limited. Also, there was a large number of samples in which to correlate endoscopic and histological findings in the same colon segments, thereby minimizing misclassification bias. Furthermore, multiple endoscopic indices were evaluated, including some of the newest that have been developed but not studied extensively in this regard. Similarly, histology was examined at a very granular level to help discern which specific aspects of histology might be the most pertinent.

Limitations included lack of examination of the newer Nancy and Robarts Histological Indices, as this study was conceived before these were developed. Also, this study was not designed primarily to assess predictors of relapse, as the number of relapses observed was relatively small. However, histological components that were significantly predictive of relapse on univariable analysis were identified, and a multivariable predictive model that also incorporated endoscopic scoring was developed. Finally, this study included patients from tertiary care centers in which a relatively high proportion received immunosuppressive agents, and thus the results may not be generalizable to a community setting or to those with predominantly mild UC.

In conclusion, our multicenter prospective study showed a high prevalence of both endoscopic and histological disease activity in clinically quiescent UC. The correlations between endoscopic and histological scores were low, with the MES and UCEIS performing similarly in this regard. The most predictive factor for clinical relapse was the Riley subcomponent architectural irregularity. Larger prospective studies are needed to assess histological predictors of clinical relapse in quiescent UC.

Abbreviations

Abbreviations
 
• 5-ASA

  5-aminosalicylates

 
• IBD

  inflammatory bowel disease

 
• MES

  Mayo endoscopic score

 
• PMS

  partial Mayo score

 
• ROC

  receiver operating characteristic

 
• TNF

  tumor necrosis factor

 
• UC

  ulcerative colitis

 
• UCCIS

  ulcerative colitis colonoscopic index of severity

 
• UCEIS

  ulcerative colitis endoscopic index of severity

Supported by: Crohn’s Colitis Foundation, grant #287390 (PI Osterman)

REFERENCES