Background Patients with long-standing UC have an increased risk for the development of colonic neoplastic lesions. Chromoendoscopy (CE) has been proven to enhance neoplasia detection while the role of virtual chromoendoscopy (VC) is still to be defined.
Objective To compare the performance of CE to VC for the detection of neoplastic lesions in patients with long-standing UC.
Design A multicentre prospective randomised controlled trial. 131 patients with long-standing UC were randomised between CE with methylene blue 0.1% (n=66) or VC with narrow band imaging (NBI) (n=65). Biopsies were taken from visible lesions and surrounding mucosa. No random biopsies were performed. The primary outcome was the difference in total number of neoplastic lesions detected in each group.
Results There was no significant difference between NBI and CE for neoplasia detection. Mean number of neoplastic lesions per colonoscopy was 0.47 for CE and 0.32 for NBI (p=0.992). The neoplasia detection rate was not different between CE (21.2%) and NBI (21.5%) (OR 1.02 (95% CI 0.44 to 2.35, p=0.964). Biopsies from the surrounding mucosa yielded no diagnosis or dysplasia. The per lesion neoplasia detection was 17.4% for CE and 16.3% for NBI (OR 1.09 (95% CI 0.59 to 1.99, p=0.793). The total procedural time was on average 7 min shorter in the NBI group.
Conclusion CE and NBI do not differ significantly for detection of colitis-associated neoplasia. Given the longer withdrawal time for CE and easier applicability, NBI may possibly replace classical CE.
Trial registration number NCT01882205; Results.
- Ulcerative colitis
- dysplasia surveillance
- narrow band imaging
Statistics from Altmetric.com
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.
Significance of this study
What is already known on this subject?
Long-standing and extensive UC is associated with an increased risk of colorectal cancer.
Chromoendoscopy (CE) increases detection of dysplasia during colonoscopy surveillance in patients with long-standing colitis.
CE is the gold standard for UC surveillance.
The role of narrow band imaging (NBI) is not clear and has only been assessed in back-to-back studies. Large prospective trials are lacking.
What are the new findings?
We performed a prospective randomised controlled trial between methylene blue CE and NBI, only targeting suspicious areas.
There is no significant difference in neoplasia detection rate, per lesion neoplasia detection rate or median number of lesions per colonoscopy between CE and NBI.
The total procedural time with NBI is significantly shorter than with CE.
Biopsies around visible lesions detected with NBI or CE did not reveal any dysplasia.
More than 80% of the targeted lesions do not contain neoplastic changes.
How might it impact on clinical practice in the foreseeable future?
Given the shorter procedural time and easier applicability, NBI may replace CE in the future for surveillance of long-standing UC.
In patients with long-standing UC undergoing surveillance colonoscopy, it has been shown that 33 biopsies are required to detect the presence of dysplasia with 90% sensitivity.1 Guidelines therefore suggest that four quadrant undirected biopsies should be taken every 10 cm, as well as biopsies from visibly abnormal areas.2 Gastroenterologists, who perform the biopsies and pathologists who report them, have endeavoured to reduce the large number of biopsies per patient without compromising the dysplasia detection rate.
Chromoendoscopy (CE) uses topically applied dyes to highlight abnormal tissues, enabling a limited number of targeted biopsies, without compromising the detection of dysplasia.3 4 In patients with UC undergoing surveillance colonoscopy, the commonly used dyes are methylene blue and indigo carmine.3 4 CE has been shown to improve the detection of dysplasia and has been recommended by the European Crohn’s and Colitis Organisation (ECCO) and the European Society of Gastrointestinal Endoscopy (ESGE) to be the preferred endoscopic method in this clinical setting.3–6
Technological improvements in standard white light endoscopy (WLE), including greater magnification, high-definition (HD) systems and dye-less virtual chromoendoscopy (VC) like narrow band imaging (NBI), also augment the endoscopic recognition of abnormal mucosa.7 NBI is easier to perform than CE as the function is activated during the procedure by the flick of a switch. Since it neither involves coating of the colonic mucosa with dye nor the reinsertion of the scope to the caecum to start the mucosal inspection, it also has the potential to save procedural time. Limited published data are controversial regarding the potential for NBI to replace CE as the method of choice for colorectal surveillance in IBD.8–11
The aim of the study was to evaluate whether NBI would perform differently compared with CE in patients with chronic UC who undergo colonoscopy for the surveillance of dysplasia.
Study design and patients
We performed a multicentre international prospective non-blinded randomised controlled trial (clinical trial.gov ID: NCT01882205). The study subjects were recruited at two centres in Belgium (University Hospitals Leuven and AZ Delta, Roeselare, Belgium) and one in Canada (Royal Victoria Hospital, McGill University Health Center, Montreal, Canada). Patients were recruited between 2008 and 2013. The study was approved by the ethics committee of UZ Leuven (ML4291) and the McGill University Health Centre (ref number 12–213-GEN).
We included all adult patients (age >18 years) with long-standing UC (8 years after onset of symptoms for patients with extensive or pan-colitis, and 10 years after onset of symptoms for patients with left-sided colitis) who could sign the informed consent form and had not had a surveillance colonoscopy within the previous year.
Exclusion criteria were subjects unwilling to consent to the study protocol, pregnant or nursing women, patients with a history of colorectal cancer or referred with known dysplasia, inadequate bowel preparation (defined as stool remnants that could not be washed off, corresponding to Boston Bowel preparation Score12 (BBPS) 2 in at least one segment), active UC (defined as Mayo score >1) noted on colonoscopy to extend over 20 cm from the anal verge13 and allergy/intolerance to methylene blue dye.
The primary end point of the study was to compare the difference in total number of neoplastic lesions detected by CE (using 0.1% methylene blue) and NBI in patients with long-standing UC.
Secondary end points were
the duration of total endoscopic procedure time and of endoscopic procedure time during retraction for each technique
the difference in neoplasia detection rate (ie, the number of patients with at least one neoplastic lesion) between CE and NBI
the difference in the ratio number of neoplastic lesions/total number of lesions between CE and NBI
the difference in total number of non-neoplastic lesions detected by CE and NBI
the number of biopsies per colonoscopy taken in the different groups.
Demographic and clinical data were collected from each patient through direct interview or chart review performed by a study investigator. Variables collected included age, sex, smoking status (active, ex-smoker or never-smoker), date of diagnosis, duration of disease in years, medication at the time of endoscopy as well as those previously used (dosage, dates of start and end of therapy), number of flares in previous two years, quality of bowel preparation (BBPS),12 withdrawal time and total duration of endoscopy.
A standard split-dose polyethylene glycol-based bowel preparation was used prior to the procedure. The colonoscopy was performed in a routine fashion after appropriate consent. The commercially available H180Q series colonoscope from Olympus Corporation, Japan, was used to carry out all procedures. The endoscope was connected via an Excera II processor to an HD screen, using the HD serial digital interface (SDI) signal. Endoscopies were performed by five dedicated endoscopists. RB had a long-standing experience in CE and NBI. The four other endoscopists (FB, TB, ID and GVA) were systematically trained by RB prior to study participation. The minimum training consisted of a theoretical introduction in CE and NBI and examples of CE in UC. A minimum of three CE were supervised by RF before endoscopists could participate in the trial. Visible mucosal abnormalities (seen during CE or NBI) were either biopsied (if resection is not feasible) or resected and two biopsies from surrounding mucosa were performed using disposable biopsy forceps (Boston Scientific Radial Jaw 4 standard capacity forceps). All lesions were classified according to the Kudo pit pattern classification. Only typical pseudopolyps with pit pattern 1 were not biopsied or resected.
Chromo-endoscopy with 0.1% methylene blue
After passing the colonoscope up to the caecum and adequate water cleansing, a 7 Fr spray catheter (Olympus Corporation, Japan) was used through the biopsy channel to coat the mucosa with 0.1% methylene blue (diluted with saline), while the scope was slowly withdrawn. The concentration of methylene blue used and the technique have been validated in previous studies.3 After 1 min, the excess of methylene blue was removed by suction and the endoscope was reinserted in the stained colonic segment to inspect the mucosa rigorously for the presence of suspicious circumscribed lesions that could be flat, depressed or polypoid. Typical pseudopolyps with a type I pit pattern were not regarded as a lesion. Targeted biopsies of visible lesions were carried out and biopsies were also taken from mucosa surrounding the target biopsies for histological assessment. The spraying and examination of the colon was performed segmentally; spraying and examination of the ascending colon, followed by the transverse colon, and finally the left colon. Besides dying the entire colonic mucosa, no other manipulations were needed for this technique. The endoscopes used did not differ from the ones used in the other randomisation arm (Olympus H180Q colonoscope).
NBI with high definition
NBI was performed using the commercially available Olympus H180Q colonoscope. WLE was used during progression of the scope to the caecum. Once the caecum was reached and adequate cleansing performed, the NBI mode was switched on for the entire duration of scope withdrawal. Examination of the mucosa and targeted biopsies was performed during withdrawal. Suspicious lesions on NBI were defined as circumscribed lesions (polypoid, depressed or flat) or areas with increased vascular intensity. Typical pseudopolyps with a type I pit pattern were not regarded as a lesion. Two biopsies were also taken from areas surrounding the target biopsies.
Randomisation in a 1:1 ratio was used. The endoscopic surveillance technique to be used (NBI or CE) was marked and placed in batches of 20 sealed (opaque and unresectable) envelopes that were created by an independent research assistant. These batches were sent to the different research sites. After inclusion and prior to the procedure, one envelope was drawn by an independent research assistant, otherwise not involved in the procedure, and opened just before the colonoscopy.
Pathology specimens were placed in separate jars, in concordance with the guidelines on the use of CE-directed screening colonoscopy for UC.5 The handling of tissue samples and histological evaluation was performed by an expert GI pathologist in each centre, according to international standards.13 In case of doubt or dysplasia, cases were discussed with a second pathologist together to reach a consensus diagnosis. A lesion was considered to be neoplastic if it belonged to any of the following pathological types: adenocarcinoma, any grade of dysplasia, indefinite for dysplasia, serrated adenoma, tubular adenoma, dysplasia-associated lesion or mass (DALM) or adenoma-like mass (ALM).
Sample size calculation
At the time of study initiation, no data were available on the potential of lesion detection with HD CE or HD NBI for colitis surveillance. Therefore, we chose to apply a similar power calculation used in previous studies on CE in long-standing UC surveillance,3 and calculated that 67 patients per group had to be enrolled in the study (a total of 134 patients). This would allow for a threefold superior neoplasia detection rate of either technique with a power of 80% (beta error 0.2; alpha error 0.05), assuming a neoplasia incidence of 10%.
The rate of neoplastic lesions detected with either endoscopic techniques was analysed using a statistics software program (SPSS V.20.0). The analyses were conducted on a ‘per patient’ and ‘per lesion’ basis. The ‘per patient’ analysis was done by calculating the percentage of patients in whom true neoplastic lesions were detected from biopsies of endoscopically suspicious lesions. On the other hand, in ‘per lesion’ analysis, the percentage of neoplastic lesion among all endoscopically suspicious lesions was calculated. Fisher’s exact test and χ2 test, where deemed appropriate, were applied for dichotomous variables. A Mann-Whitney U test or Student’s t-test was used for continuous variables. p -Values<0.05 were considered significant for the analysis.
Between July 2008 and November 2013, we enrolled 157 patients into the study (figure 1). After randomisation, 26 patients were excluded, according to the predefined inclusion and exclusion criteria (6 in the CE group and 17 in the NBI group): 3 because the disease duration was <8 years and 23 because significant inflammation was noted during colonoscopy. For the final intention-to-treat analysis, there were a total of 131 patients, 66 in the CE group and 65 in the NBI group. Male patients represented 56% (n=73) of the cohort. Baseline patient characteristics did not differ between the groups (table 1). All data presented are based on an intention-to-treat analysis, unless otherwise mentioned.
Primary end point
Overall 31 neoplastic lesions were identified in 14 patients in the CE group (n=66) and 21 lesions identified in 14 patients in the NBI group (n=65). An example of typical neoplastic lesions that were detected by NBI and CE is shown in figure 2.
The mean (SD) number of neoplastic lesions detected per colonoscopy procedure was 0.47 (1.38) in the CE group and 0.32 (0.68) in the NBI group (p=0.992, Mann-Whitney U test).
All endoscopically visible lesions were resected or biopsied and the histology was classified according to the degree of dysplasia (low-grade dysplasia, high-grade dysplasia, carcinoma). The low-grade dysplasia group was further subdivided into DALM (endoscopic lesion and surrounding biopsies showing low-grade dysplasia), ALM (endoscopic lesion with low-grade dysplasia with surrounding biopsies showing chronic inflammatory changes in the field of colitis but no dysplasia) and sessile-serrated adenoma or sporadic adenoma (endoscopic lesion with low-grade dysplasia not in the field of colitis). A single outlier patient in the CE group had nine serrated polyps and five hyperplastic polyps. Since several lesions were >20 mm in size, this patient had a serrated polyposis syndrome, unrelated to the presence of long-standing UC. If we discount this patient, the number of neoplastic lesions was 22 among the 66 patients in the CE group (table 2).
We calculated the neoplasia detection rates for the different sites: this was 21/107 in centre 001 (19.6%), 3/6 in centre 002 (50%) and 4/18 (22.2%) in centre 004. There was no statistically significant difference between the three centres (Fisher’s exact test).
Secondary end points
The total procedure time and the withdrawal time were significantly shorter in the NBI group. CE added on average 7 min to the entire procedure (table 3). This difference in withdrawal time remained statistically significant when patients were grouped according to the total number of biopsies performed during the procedure (figure 3). This indicates that the difference in withdrawal time is related to the technique of dye spraying and not to a higher lesion detection, resulting in more biopsy sampling.
Fourteen patients in each group were found to have at least one neoplastic lesion. The neoplasia detection rate was therefore 21.2% in the CE group and 21.5% in the NBI group, p=0.964, OR 1.02 (95% CI 0.44 to 2.35) (χ2 test) (table 2).
The per lesion neoplasia detection rate can serve as a measure for the false positive rate of endoscopically targeted areas (clear polyps and flat or slightly elevated). This parameter was not statistically significant between CE and NBI (table 4). In the CE group, a total of 178 lesions were identified, of which 31 (17.4%) were considered neoplastic. In the NBI group, a total of 129 lesions were identified, of which 21 (16.3%) were considered neoplastic (OR (95% CI) 1.09 (0.59 to 1.99), p=0.793).
The median (range) number of lesions identified per patient was not statistically different between the groups: 2 (0–14) in the CE group and 2 (0–7) in the NBI group (p=0.185 Mann-Whitney U test). The median (range) number of non-neoplastic lesions was 2 (0–8) in the CE group and 1 (0–7) in the NBI group (p=0.141) (tables 2 and 5).
A per protocol analysis omitting the outlier with multiple serrated polyps reduced the per lesion neoplasia detection rate in the CE group to 13.4% (22/164), p=0.492.
As per study protocol, biopsies were taken from adjacent normal appearing mucosa in addition to endoscopically visible abnormalities. In all, 301 endoscopically abnormal looking sites and 241 adjacent normal appearing sites were biopsied. In total, 6 biopsies (in four patients) out of the 241 (2.5%) biopsies of normal mucosa yielded discrete abnormalities of regeneration or inflammation, but none was neoplastic in nature. A total of 308 biopsies were performed in the CE group (mean number of biopsies per patient (95% CI), 4.7 (3.5 to 5.8)) and 234 in the NBI group (mean number of biopsies per patient (95% CI), 3.6 (2.7 to 4.5), p=0.295, Mann-Whitney U test).
No abnormalities were noted and therefore no biopsies were carried out in 15 of the 66 (22.7%, 95% CI 12.6% to 32.8%) patients in the CE group and 21 of the 65 (32.3%, 95% CI 20.9% to 43.7%) patients in the NBI group, p=0.219.
In this randomised controlled trial, we did not detect a significant difference between CE and NBI for the detection of colitis-associated neoplasia in patients with long-standing UC in remission. To our knowledge, this is the first parallel designed study to compare HD CE to NBI. In addition, we could not find a significant difference in detection of non-neoplastic lesions or in the ratio number of neoplastic lesions to total number of lesions between the two groups. We did however find a significant reduction in withdrawal time in favour of NBI.
Though, different in design, our data concur with two other prospective studies comparing CE and NBI showing a similar lack of difference in yield between these two modalities.9 14 Current ECCO and ESGE guidelines recommend that CE with targeted biopsies should be used as the method of choice in neoplasia surveillance of UC.5 6
Although our study was not designed as a non-inferiority trial, the similar rates of dysplasia detection achieved by NBI in a large number of patients suggests that NBI may replace classical CE in the future. In particular, its ease of use and shorter procedure time are advantages and may facilitate a more widespread use.
In spite of its convenience compared with CE, NBI has not yet found widespread use. In part this may be due to the fact that studies which have not shown superiority of NBI over WLE.8 11 On the other hand, studies which compared CE and WLE showed that CE was superior in detecting neoplastic lesions.4 15 16 This may to some extent be explained by the fact that the comparative studies with NBI have used HD endoscopes, while in most comparative studies with CE standard definition scopes were used. As such the neoplasia detection rate by HD WLE endoscopy appears to be higher (12.5%–18.8%)8 11 compared with standard resolution WLE (2%–12.3%).4 15 16
It is conceivable that the learning curve for detecting NBI is achieved through using classical CE. Once the operator is more familiar with the aspect of subtle lesions, an increased detection of abnormalities by NBI may be possible. However, another strength of our study is the unique fact that endoscopists with little experience in CE and a secondary care centre were included. Indeed, only one of the endoscopists had a long-standing experience with NBI and CE, while the others were on-site trained during 1 day, with three procedures and video material. Therefore, a learning effect for NBI through previous use of CE is less likely in our study. This is in line with a recent Spanish multicentre trial, where the investigators could not detect a learning effect for CE in less experienced endoscopists, nor could they detect a significant difference in CE-based neoplasia detection between experts and non-experts.17
On average, in our study, neoplastic lesions were detected once in every five patients. In the literature, the neoplasia rate in a similar clinical setting is around 1 in 8. We do not know whether early dysplastic lesions which are picked up by these more sensitive techniques have the same malignant potential and whether use of CE and NBI translates into better patient outcomes and cost savings.
In addition, both CE and NBI have a high false positive rate: only 16%–17% of the lesions that were biopsied were neoplastic. The reason for this is probably twofold. First, since we opted not to take any random biopsies in our study, we applied a very low threshold to perform targeted biopsies of any circumscribed lesion. Second, it is hard to predict the neoplastic nature of lesions in the setting of a quiescent colitis.18 19 As such our findings are in concordance with early data on NBI in lesion characterisation by Matsumota et al, showing only 10% per lesion neoplasia detection rate for protruding lesions.19 We are however convinced that it is necessary to maintain a low threshold to biopsy any circumscribed lesion: in an interobserver study we have shown that the agreement for differentiation between non-neoplastic pit patterns (I, II) and neoplastic pit patterns (IIIL, IIIS, IV or V) was moderate (κ 0.587) and even significantly better for NBI compared with CE. The agreement becomes better when a diagnosis is made with high confidence.20 In this ‘offline’ setting with selected images, the diagnostic accuracy to distinguish neoplastic from non-neoplastic changes based on the presence or absence of pit pattern I/II was 70%. So, at this time, the diagnostic accuracy of HD CE or HD NBI without magnification is insufficient and endoscopists should maintain a low threshold to take biopsies of any suspicious site, as was instructed to the endoscopists in this trial. Within its new terminology, the Scenic working group included the term ‘resectable or non-resectable’ as an important descriptor for lesions found during colitis surveillance. They also agreed that amenable lesions can be resected and that patients can be followed up after resection if the lesion was removed completely and did not contain cancer, which was confirmed to be safe in a recent meta-analysis.21 22 However, we found that >80% of the lesions that were targeted for biopsies did not contain neoplasia. Taking into account the difficulty of optical diagnosis in patients with colitis, we believe that it is wise to take biopsies first prior to engaging in extensive and multiple endoscopic mucosal resections.
NBI has several clinical advantages compared with CE. First, it is easier and less messy to apply. The potential—although clinically not proven relevant—effect of DNA damage by methylene blue23 does not apply to NBI. It decreases the time of the procedure significantly. Finally, even in case of a suboptimally prepared colon, NBI is easier to apply than CE.
Older guidelines, dating back before the era of HD endoscopies, recommending to take random biopsies every 10 cm failed to detect a significant number of lesions as shown on colectomy specimens.24 Our study protocol did not include random undirected biopsies. It has been shown that the yield from such undirected biopsies is low when using CE. Rutter et al found no dysplastic tissue in 2904 non-targeted biopsies.4 In our study too, only 6 out of the 241 biopsies of normal looking mucosa yielded abnormalities, but none was neoplastic in nature. This finding supports the view that, with the use of enhanced imaging techniques, in perfect conditions of a quiescent colitis and good bowel preparation, the practice of taking random biopsies every 10 cm during surveillance colonoscopy can be abandoned. In addition, the fact that random biopsies of the surrounding mucosa did not reveal any dysplasia challenges the classical concept of DALM and adds to the recently proposed consensus to abandon this old terminology.22
Our study has some limitations. Although the number of missed lesions may be negligible with the use of advanced imaging techniques, it cannot be fully excluded that random biopsies would have picked up additional lesions. Nevertheless, no cancer developed in the patients that were included in the beginning of the study during follow-up. We opted to perform a parallel study and can therefore not calculate the neoplasia miss rates for each technique. We do believe however that a back-to-back protocol also has an inherent bias. Due to the selection criteria, our data only apply to patients without residual disease activity and with a good to excellent bowel preparation. Finally, the power calculation of our study was set up as a superiority design. The presumed difference between CE and NBI was not reached in our study sample and therefore a type II error cannot be excluded. Although the absence of evidence of a statistical difference is not a proof of absence, our sample size of >130 patients makes it very unlikely that any possible difference would be clinically relevant considering the high number needed to treat. From a clinical viewpoint, the current difference would be clinically quite irrelevant (21.2% vs 21.5%), even if it becomes statistically significant. To prove that a 1% difference would not be inferior, 56 854 patients need to be included in a non-inferiority design. One of the explanations why the detection was higher in our trial compared with available data at the start of the study may be found in the use of HD endoscopes. Indeed, a recent meta-analysis has shown that the difference between CE and WLE mainly applies to standard resolution endoscopes, but not to HD endoscopes.25
In conclusion, in patients with UC undergoing surveillance colonoscopy, we could not demonstrate a significant difference between NBI and CE. However, NBI significantly reduces procedure time and is technically easier to apply (table 6).
Contributors RB and TB participated in the study conception, patient recruitment, data collection, data analysis, writing of the manuscript and final approval of the manuscript. FB, MF, ID, SV, GV and PR participated in patient recruitment, data analysis, critical revision and approval of the final manuscript. VB and HW participated in patient recruitment and data collection. KG and GDH performed the pathological analysis, contributed to the writing and approval of the final manuscript. JAJ and MF participated in data analysis, writing of the manuscript and final approval of the manuscript.
Funding RB, MF and GVA are supported by a grant of Research Foundation – Flanders (FWO). RB has received a study grant from the Belgian Society of Gastrointestinal Endoscopy (BSGIE).
Competing interests RB has received speaker’s fee and research support from Olympus, not related to this trial.
Patient consent Obtained.
Ethics approval Ethics committee of UZ Leuven (ref nr ML4291), and the McGill University Health Centre (ref number 12-213-GEN).
Provenance and peer review Not commissioned; externally peer reviewed.
Data sharing statement The authors declare that their are no additional unpublished data from this study.