Validation of the Ulcerative Colitis Colonoscopic Index of Severity and Its Correlation With Disease Activity Measures

Abstract

Background & Aims

Endoscopic healing is likely to become an important goal for treatment of patients with ulcerative colitis (UC). A simple validated endoscopic index is needed. We validated the previously developed UC Colonoscopic Index of Severity (UCCIS).

Methods

In a prospective study, 50 patients with UC were examined by colonoscopy; we analyzed videos of rectum and sigmoid, descending, transverse, and cecum/ascending colon. Eight gastroenterologists blindly rated 4 mucosal lesions (for vascular pattern, granularity, friability, ulceration) and severity of damage to each segment and overall. The global assessment of endoscopic severity (GAES) was based on a 4-point scale and 10-cm visual analogue scale. Correlation of the UCCIS score with clinical indexes (clinical activity index and simple clinical colitis activity index), patient-defined remission, and laboratory measures of disease activity (levels of C-reactive protein, albumin, and hemoglobin and platelet counts) were estimated by using the Pearson (r) or Spearman (r_s) method.

Results

Interobserver agreement was good to excellent for the 4 mucosal lesions evaluated by endoscopy and the GAES. The UCCIS calculated for our data accounted for 74% (R² = 0.74) and 80% (R² = 0.80) of the variation in the GAES and visual analogue scores, respectively (P < .0001). The UCCIS also correlated with clinical activity index (r = 0.52, P < .001), simple clinical colitis activity index (r = 0.62, P < .0001), and patient-defined remission (r = 0.43, P < .01). The UCCIS also correlated with levels of C-reactive protein (r_s = 0.56, P < .001), albumin (r = −0.55, P < .001), and hemoglobin (r = −0.39, P < .01). A rederivation of the equation for the UCCIS by using the data from a previous study combined with those of the current study (n = 101) yielded similar results.

Conclusions

The UCCIS is a simple tool that provides reproducible results in endoscopic scoring of patients with UC.

The physician's impression of active disease in ulcerative colitis (UC) may not correlate well with endoscopic disease activity and is a poor predictor of mucosal inflammation.¹ Endoscopic evaluation is therefore increasingly used in the evaluation and management of patients with UC.²

New drugs in the form of biologics and immunomodulators are able to achieve clinical response and endoscopic (mucosal) healing in patients with moderate-to-severe disease.³ Although endoscopic and histologic disease activity in UC lags behind symptom improvement,⁴ there is now epidemiologic evidence to suggest medications that achieve mucosal healing could change the natural history of UC⁵ and improve disease outcomes (colectomy).⁶ Mucosal healing is therefore likely to become an important treatment goal in patients with UC.^7,8 The only validated UC endoscopic assessment tool available to date was reported 4 decades ago by Baron et al,⁹ who used rigid sigmoidoscopes. No similar validated score exists in the current period of high-definition videoendoscopes, and none have evaluated the colon by segment by using full colonoscopy, rather they have just assessed the distal colon, usually with a flexible sigmoidoscope. Several scoring systems have been used in clinical trials of UC to measure disease outcomes,¹⁰ and the Disease Activity Index or Mayo score is one of the most widely used.¹¹ Wide variation exists in assessing the endoscopic disease activity in UC.^12,13 We recently reported good interobserver agreement (IOA) for granularity, vascular pattern, ulceration, and bleeding/friability among a group of experienced gastroenterologists, and a weighted sum score of these 4 parameters (Ulcerative Colitis Colonoscopic Index of Severity [UCCIS]) significantly predicted the overall endoscopic severity.¹⁴

The aim of our current study was to validate the UCCIS score in a prospective cohort of UC patients and study its correlation to clinical indexes and laboratory parameters of disease activity.

Methods

Patients and Examination

Patients with UC and a spectrum of disease severity were prospectively recruited to this study, and written informed consent was obtained from all participants before inclusion. The study was approved by the Mayo Clinic Institutional Review Board. The patients included in the study had an established diagnosis of UC for ≥3 months and underwent colonoscopic assessment of their disease between December 2009 and June 2010. Those patients with prior colon surgery, indeterminate colitis, Crohn's colitis, colorectal neoplasia, or concomitant infectious colitis were excluded. The patients were interviewed before their endoscopy to collect their demographic data and clinical information pertaining to disease diagnosis, severity, treatment and extent, indication for colonoscopy, and to calculate the physician global assessment (0 = quiescent, 1 = mild, 2 = moderate, 3 = severe).¹⁵ Laboratory parameters of disease activity where available, which included C-reactive protein (CRP), erythrocyte sedimentation rate, hemoglobin, platelet count, and albumin, were recorded. All patients received the standard bowel preparation with Moviprep (Salix Pharmaceuticals, Raleigh, NC) (PEG-3350, sodium sulfate, sodium chloride, potassium chloride, sodium ascorbate, and ascorbic acid for solution).

Acquisition and Assessment of Digital Video Files

The colonoscopic procedures for the first 3 patients were recorded by using the Sony DVO1000MD medical DVD recorder (Sony Electronics, Inc, New York, NY) in the high-quality mode. For all the subsequent patients, an automated colonoscopy video capture system called EndoCapture, which is currently in use in the endoscopy rooms at Mayo Clinic in Rochester, Minnesota, was used. This automatically detects the start and end of individual endoscopic procedures and records them in high-digital quality without any human interaction.¹⁶ Both methods of acquisition yield MPEG2 video files of similar quality. Olympus Evis Exera II High Definition 180 series (Tokyo, Japan) colonoscopes with variable stiffness were used for the study. To avoid bias, one coinvestigator (S.S.) who did not take part in the scoring of the video clips interviewed study patients to assimilate clinical parameters and acquired and collated all their colonoscopy video files. This coinvestigator (S.S.) made 30-second clips representing the 5 colonic segments of the bowel (cecum/ascending, transverse, descending, sigmoid, and rectum). Thirty-second clips that were most representative of the individual segments of the large bowel were made from the video files of each patient in the study and uploaded onto a central database for the assessors. The endoscopic scorers comprised 8 gastroenterologists (W.J.S., D.H.B., W.J.F., D.S.P., S.K., W.J.T., E.V.L., K.W.S.) specializing in the care of patients with inflammatory bowel disease, who independently and blinded to all clinical and laboratory parameters rated the mucosal lesions and endoscopic severity for each segment and the overall disease severity. The different scores obtained from various assessors were entered into a spreadsheet for further analysis. For all patients, the withdrawal phase of the colonoscopy was used to make the video clips for scoring.

For the validation phase of the UCCIS score, we used the video files of 50 new UC patients recruited prospectively for this study (not the patients from the previous study¹⁴); the 4 endoscopic mucosal lesions (granularity, vascular pattern, ulceration, bleeding/friability) and their severity grading that had good IOA in our previous study were chosen to be validated in this new set of patients (Table 1).¹⁴ Segmental assessment of endoscopic severity (SAES) and global assessment of endoscopic severity (GAES) were the endoscopist's judgment of severity based on a predefined 4-point severity scale (Table 1). SAES was graded on the basis of the endoscopic severity for each of the colonic segments, and GAES was graded on the basis of the overall endoscopic assessment of severity, taking into account ratings from all 5 colonic segments. The GAES was also assessed on a 10-cm scale (0 for normal to 10 cm for extremely severe), whereby the raters placed an “X” mark on the scale depending on the severity of endoscopic activity. A score was then derived for each study patient based on the measured distance from 0 to the marked point.

Table 1. The UCCIS.

Lesion	Score	Definition
1. Vascular pattern	0	Normal, clear vascular pattern
	1	Partially visible vascular pattern
	2	Complete loss of vascular pattern
2. Granularity	0	Normal, smooth and glistening
	1	Fine
	2	Coarse
3. Ulceration	0	Normal, no erosion or ulcer
	1	Erosions or pinpoint ulcerations
	2	Numerous shallow ulcers with mucopus
	3	Deep, excavated ulcerations
	4	Diffusely ulcerated with >30% involvement
4. Bleeding/friability	0	Normal, no bleeding, no friability
	1	Friable, bleeding to light touch
	2	Spontaneous bleeding
5. Grading of SAES and GAES (4-point scale)	0	Normal/quiescent: visible vascular pattern with no bleeding, erosions, ulcers, or friability (includes altered vascular pattern seen in quiescent disease)
	1	Mild: erythema, decreased or loss of vascular pattern, fine granularity, but no friability or spontaneous bleeding
	2	Moderate: friability with bleeding to light touch, coarse granularity, erosions, or pinpoint ulcerations
	3	Severe: spontaneous bleeding or gross ulcers
6. GAES VAS 10-cm scale		| - - - - - | - - - - - | - - - - - | - - - - - | - - - - - | - - - - - | - - - - - | - - - - - | - - - - - | - - - - - |
		Normal	Extremely severe

Statistical Analysis

IOA was estimated for the ratings of all lesions including grading of endoscopic severity by segment and globally using Lin's concordance correlation coefficient (CCC). Strength of correlation was categorized as excellent (0.81–1.00), good (0.61–0.80), moderate (0.41–0.60), and fair (0.21–0.40).

To identify lesions predictive of GAES, the average scores among raters for the lesions within each colonic segment were calculated, and for each separately, the SAES was regressed on the average (over raters) scores of these characteristics (5 separate regression models, 1 for each segment). The regression coefficient for each characteristic was then averaged to yield a common mean coefficient estimate, which was then multiplied by the corresponding sum of the lesion characteristic scores over the 5 colonic segments separately in each patient to yield a weighted total score. Last, this single total patient lesion score was used as a single predictor in univariate linear regression models to predict both the GAES and separately the visual analogue scale (VAS). An approximate weighted total score was calculated by smoothing the average coefficient estimate of lesions to produce an index score more amenable to calculation. With the above, the previously developed equation for the UCCIS score [3.1 × Sum (vascular pattern across 5 segments) + 3.6 × Sum (granularity across 5 segments) + 3.5 × Sum (ulceration across 5 segments) + 2.5 × Sum (bleeding/ friability across 5 segments)] was used to calculate the score for this prospective cohort of 50 patients. In addition, we also studied the model by using regression analysis for the combined set of patients from the current study (50 patients) and the previous study by Thia et al¹⁴ (51 patients). This was, however, not part of the validation of UCCIS.

Clinical indexes of disease activity in the form of clinical activity index (CAI),¹⁷ simple clinical colitis activity index (SCCAI),¹⁸ and patient-defined remission (PDR)¹⁹ were calculated by using the data abstracted at the time of patient interview. These indexes (CAI, SCCAI, PDR) measure disease activity solely on the basis of clinical and biochemical parameters. In the current study, GAES served as the dependent variable, whereas the 4 identified endoscopic mucosal variables (used to develop the UCCIS) were the explanatory variables in the model. Correlation of the UCCIS score with these clinical indexes and the laboratory parameters of disease activity were estimated by using Pearson (r) or Spearman (r_s) correlation test as appropriate. All tests were carried out by using two-tailed P values, with significance levels of 5%.

Results

Patient Demographics and Video Clip Assessment

We recruited 53 patients to the study. Three patients were excluded because of poor bowel preparation or failed acquisition of digital video files. The remaining 50 patients had complete video recording of their colonoscopic assessment, and Table 2 shows their demographics and clinical information. The median age of the study patients was 46.5 years, and nearly half were men (48%). The most common indication for colonoscopy in our cohort was for assessment of activity of UC in 23 (46%), followed by surveillance for neoplasia in 17 (34%). All study patients had complete colonoscopy, with the endoscopists reaching the cecum in all patients. The prior extent of UC was extensive in slightly over half (52%). The edited video files from all patients produced a total of 250 video clips (each approximately 30 seconds in duration) representative of an equal number of colonic segments for assessment. Each of the 8 coinvestigators graded 250 video clips for a total of 2000 evaluations on 50 patients.

Table 2. Patient Demographics and Clinical Information (N = 50).

Characteristics	Frequency
Median age, y (range)	46.5 (20–77)
Males, n (%)	24 (48)
Indications, n (%)
Surveillance	17 (34)
Activity assessment	23 (46)
Monitoring response to treatment	4 (8)
Confirmation of diagnosis	6 (12)
Prior known extent of involvement, n (%)
Proctitis	2 (4)
Proctosigmoiditis	9 (18)
Left-sided up to splenic flexure	7 (14)
Pancolonic disease	26 (52)
Unknown	6 (12)
Physician's global assessment, n (%)
Quiescent disease	19 (38)
Mild disease activity	10 (20)
Moderate disease activity	16 (32)
Severe disease activity	5 (10)

Interobserver Agreement

Similar to the previous study by Thia et al¹⁴ the IOA was good to excellent for vascular pattern, granularity, and ulceration in all segments of the colon (Table 3). Bleeding/friability also showed good IOA except in the sigmoid (CCC, 0.58; 95% confidence interval [CI], 0.46–0.68) and descending colon (CCC, 0.56; 95% CI, 0.44–0.66), where the agreement was only moderate. The IOA for GAES based on a 4-point scale (CCC, 0.69; 95% CI, 0.59–0.77) and also on a VAS (CCC, 0.80; 95% CI, 0.73–0.86) was good among the assessors (Table 3).

Table 3. IOA for Mucosal Lesions and Segmental and Global Grading of Endoscopic Severity.

	CCC by location (95% CI)
Lesion	Rectum	Sigmoid	Descending	Transverse	Ascending/cecum
Vascular pattern	0.75 (0.66–0.82)	0.81 (0.73–0.86)	0.74 (0.64–0.81)	0.86 (0.80–0.90)	0.85 (0.78–0.90)
Granularity	0.70 (0.60–0.78)	0.78 (0.69–0.84)	0.73 (0.63–0.80)	0.88 (0.83–0.92)	0.82 (0.75–0.87)
Ulceration	0.80 (0.71–0.86)	0.75 (0.65–0.82)	0.72 (0.62–0.79)	0.73 (0.63–0.80)	0.73 (0.63–0.80)
Bleeding/friability	0.68 (0.57–0.76)	0.58 (0.46–0.68)	0.56 (0.44–0.66)	0.73 (0.63–0.80)	0.77 (0.68–0.84)
SAES	0.79 (0.71–0.85)	0.78 (0.69–0.84)	0.71 (0.61–0.79)	0.84 (0.77–0.89)	0.85 (0.78–0.89)
GAES
4-point scale	0.69 (0.59–0.77)
VAS	0.80 (0.73–0.86)

Predictors of Global Assessment of Endoscopic Severity

The UCCIS score based on weighted sum scores of the 4 mucosal lesions accounted for 74% (R² = 0.74) of the variation in the 4-point GAES score (P < .0001) and 80% (R² = 0.80) of the variation in the VAS-based GAES score (P < .0001). We combined the 50 patients from the current study with the 51 patients from our earlier study¹⁴ and rederived the weighted sum scores for the 101 patients. This yielded similar results, with 77% (R² = 0.77) and 82% (R² = 0.82) of variation in GAES and VAS, respectively, accounted for by this sum score (P < .0001).

Calculation and Validation of the Ulcerative Colitis Colonoscopic Index of Severity Score

The UCCIS scoring system is a composite index that uses weighted linear combination of the 4 characteristics used as prediction variables in the model (eg, vascular pattern sum), and these characteristics have a range of values. The “weights” (model coefficient estimates) are thus not integer-valued points but require a simple calculation. “Rounded” values of the coefficient estimates gave similar results but would still require a simple calculation rather than a simple sum of points corresponding to absence or presence of the specific characteristics (data not shown). Supplementary Table 1 illustrates the calculation of the UCCIS score in a hypothetical patient, and Supplementary Table 2 shows the median UCCIS scores of patients categorized into different disease severity on the basis of their full Mayo Clinic scores.

The UCCIS score was also validated by comparing its correlation with various clinical indexes and laboratory measures of disease activity. This calculated score showed moderate correlation with SCCAI (r = 0.62, P < .0001), CAI (r = 0.52, P < .001), and PDR (r = 0.43, P < .01) (Figure 1). The UCCIS score showed excellent correlation with the GAES, both 4-point scale (r = 0.86, P < .0001) and VAS (r = 0.89, P < .0001) (Figure 2). Serum concentration of CRP correlated significantly with the UCCIS score (r_s = 0.56, P < .001), whereas both hemoglobin and albumin were inversely correlated (r = −0.39, P < .01 and r = −0.55, P < .001, respectively) (Table 4). No significant correlation was found with platelet count (r = 0.19, P > .05).

Figure 1.

Correlation of UCCIS to clinical indexes of disease activity: SCCAI (A) and CAI (B); r = Pearson coefficient.

Figure 2.

Correlation between UCCIS and GAES: 4-point scale (A) and VAS (B); r = Pearson coefficient.

Table 4. Correlation Between UCCIS and Laboratory Parameters of Inflammation.

	UCCIS
Laboratory parameter	Correlation coefficient (95% CI)	P value
CRP	0.56a	<.001
Albumin	−0.55	<.001
Hemoglobin	−0.39	<.01
Platelets	0.19	NS

NS, not significant.

^aSpearman r was used to calculate correlation for CRP.

Discussion

Mucosal healing has become an important measure of treatment efficacy in UC clinical trials, and hence there is an urgent need for a simple, reproducible, and validated scoring index of severity in UC. We found good IOA (reproducibility) among a group of experienced gastroenterologists for granularity, vascular pattern, bleeding/friability, and ulcerations. A composite score that used these 4 mucosal variables significantly accounted for the variations in the GAES. We validated the UCCIS score, which showed significant correlation with the clinical indexes and laboratory measures of active disease and the endoscopists' overall assessment of disease activity.

For the widely used Mayo Clinic score,¹¹ complete agreement for the endoscopic subscore among physicians was present only in 20 of 96 (21%).²⁰ Thia et al¹⁴ studied IOA for 10 endoscopic mucosal descriptors of inflammation adapted from the Baron score and the Mayo endoscopic subscore. By using multivariate regression modeling, a full colonoscopy severity index (UCCIS) that accounted for significant GAES variance was developed by using just 4 mucosal lesions.¹⁴ Our current study is the extension of this previous work to correlate with clinical and laboratory parameters. In the previous study, some patients had the now discontinued Fleet Phospho-soda for bowel preparation, whereas in the current study, all patients received Moviprep as standard. We validated the consistently good IOA among our scorers of the previously identified mucosal descriptors and the GAES in this new set of 50 patients (reliability). The agreement was good to excellent for all 4 mucosal lesions except for bleeding/friability, in particular for sigmoid and descending colon, where the correlation was only moderate (CCC, 0.58 and 0.56, respectively). This may be due to either the difficulty in defining the concept of friability or the lack of understanding of this abnormality. Despite this, all 4 mucosal lesions had fairly equal contribution to the final weighted UCCIS score, which reflected the overall endoscopic activity. Recent study by Travis et al²¹ found contact friability to have the lowest level of agreement (kappa = 0.30) when compared with incidental friability (kappa = 0.40), which is bleeding from the mucosa during endoscope withdrawal. In our current study, the video clips of each colonic segment were made during the withdrawal phase of colonoscopy, and hence incidental friability was measured, and this showed overall good agreement throughout the 5 segments of the large bowel. Endoscopic disease activity assessment during insertion vs withdrawal could likely affect friability but not vascular pattern, granularity, or ulcerations. No data currently exist comparing the assessment of friability during endoscope withdrawal rather than endoscope insertion, and this may need further study.

We found the combined sum of the 4 mucosal variables accounted for 80% of the variance in the GAES (with the VAS) for the new set of 50 patients in this study. In a recent study by Travis et al,²¹ videos of flexible sigmoidoscopy (selected from a library of 670 videos acquired during clinical trials for the treatment of moderately active UC²²) were scored by 30 investigators from 13 different countries. This model used just 3 endoscopic descriptors (vascular pattern, bleeding, erosions and ulcers) and captured more than 90% of variability in the overall disease severity. A similar but earlier study by Lange et al¹³ found vascular pattern, erosions, ulcerations, and friability to account for 92% of the variance in the overall endoscopic activity index among experienced endoscopists. Among the inexperienced endoscopists, besides the above mucosal lesions, granularity was also important in explaining the variations in the endoscopic activity index.¹³ However, our study differs from these 2 studies in that all patients had a full colonoscopy, and every segment of their colon was assessed and scored by the coinvestigators. By using regression modeling, every patient had a single total lesion score, which was a composite index, calculated after taking into account the disease burden in every segment of his/her colon. We believe this gives a more comprehensive assessment of the UC disease activity, especially because patients can have patchy disease,²³ rectal sparing,²⁴ and right-sided inflammation in primary sclerosing cholangitis patients with UC.²⁵ The segmental colonic assessment in UC is becoming increasingly important with the United States Food and Drug Administration, and the recent proposed definition of mucosal healing by the International Organisation for the Study of Inflammatory Bowel Disease comprises absence of friability, blood, erosions, and ulcers in all segments of the gastrointestinal mucosa.¹⁰ Hence, the UCCIS score that uses the 4 mucosal descriptors from the entire colon is increasingly relevant.

The calculated UCCIS score correlated strongly with the GAES, both the 4-point scale and the VAS (r = 0.86 and 0.89, respectively; P < .0001), reflecting its ability to capture the endoscopists' overall assessment of UC severity. In addition, the score correlated significantly with the clinical indexes of disease activity (SCCAI,¹⁸ CAI,¹⁷ and PDR¹⁹) and laboratory parameters of active inflammation (hemoglobin, CRP, albumin). For the practicing gastroenterologist, this study provides a number of advances in knowledge. First, this study shows that vascular pattern, granularity, friability, and ulcerations are reproducible lesions with good IOA, and a weighted combination of these 4 parameters could predict the global endoscopic severity of the disease. Second, the use of such validated instruments is likely to provide more accurate information on endoscopic disease activity and mucosal healing, which can indirectly help clinical practice. Finally, once further research ascertains the predictive value of endoscopic activity as measured by this instrument, this index could be used in clinical practice (and in clinical trials) as part of the various endoscopic reporting systems (eg, Endosoft) to provide a more comprehensive and standardized assessment of endoscopic disease activity than is possible with other instruments, which use flexible sigmoidoscopy.

There are several limitations in our study. First, this is a single-center study in which the raters were experienced gastroenterologists subspecializing in the care of inflammatory bowel disease. We have not evaluated this index across different centers, physicians with varying experience, or in other parts of the world. Second, most of our patients were recruited from the outpatient department, and thus the responsiveness of this index in hospitalized patients with acute severe colitis is unclear. Third, there were minor variations in the IOA for the scored mucosal variables between the current and the previous study. We believe this represents sampling error, and much larger patient samples will be needed to quantify this. Finally, although we attempted to define thresholds (Supplementary Table 2) for the various severities of UC, much larger studies are needed to validate the usefulness of this index in research and clinical decision making. Nonetheless, this is a prospective study in UC in which validated endoscopic mucosal descriptors of inflammation were used to develop an index and its correlation studied with clinical indexes and laboratory biomarkers of inflammation.

In conclusion, we have validated a simple endoscopic activity index based on full colonoscopy in UC that gives an excellent overall assessment of endoscopic severity as judged by the VAS.

Supplementary Material

Supplementary Table 1. Example of Calculation of UCCIS in a Hypothetical UC Patient With Left-sided Involvement

Supplementary Table 2. The Calculated UCCIS Score in the Current Study Patients (N = 50) Categorized Into Varying Disease Activity Based on Their Full Mayo Clinic Score

Abbreviations used in this paper

CAI

clinical activity index

CCC

concordance correlation coefficient

CI

confidence interval

CRP

C-reactive protein

GAES

global assessment of endoscopic severity

IOA

interobserver agreement

PDR

patient-defined remission

SAES

segmental assessment of endoscopic severity

SCCAI

simple clinical colitis activity index

UC

ulcerative colitis

UCCIS

Ulcerative Colitis Colonoscopic Index of Severity

VAS

visual analogue scale