Narrow band imaging at colonoscopy: seeing through a glass darkly or the light of a new dawn?

Narrow band imaging (NBI) is a novel ‘blue light’ endoscopic imaging technology that enhances contrast for superficial mucosal capillaries and has been studied in the upper and lower GI tract, oro–pharynx, lung, and bladder [1–6]. In the colon, adenomatous polyps have an increased microvascular density, and therefore may be highlighted by NBI [7]. The superficial microcapillary plexus that surrounds each colonic crypt may be seen with NBI and magnification and can lead to an appearance very similar to the ‘pit pattern’ that can be seen using chromoendoscopy ‘dye spray’. Indeed, NBI has been described as ‘electronic chromoendoscopy’ although the patterns seen may not be exactly identical [8]. More simply, NBI with magnification allows a direct appreciation of the increased microvascular density seen in adenomas described as ‘meshed brown capillary vessels’, ‘strong vascular pattern intensity’ and ‘brown hue’ allowing easy differentiation from non-neoplastic polyps [2,8,9]. Clinical data on the use of NBI at colonoscopy for lesion detection, differentiation and assessment of potential invasion in both routine colonoscopy, higher risk surveillance, and in colitis are now accumulating. As NBI is now available commercially it may be appropriate to consider what its role might be in our routine clinical practice.

Theoretically, by increasing contrast for adenomas, NBI should help in adenoma detection, particularly for subtle, flat lesions that are otherwise hard to detect. There are now three published studies looking at this issue, two randomized studies in standard-risk patients, and one back-to-back study in high-risk patients having surveillance for hereditary nonpolyposis colorectal cancer (HNPCC) [10–12]. Neither randomized study showed a statistically significant increase in adenoma detection with NBI, with the single endoscopist, Rex et al. study, showing an almost identical number of adenomas detected with high-definition white light as with NBI, and one of the highest reported adenoma detection rates; nor did Rex see improved detection of flat lesions [10].

In the Adler et al. multi-endoscopist study, NBI appeared to increase detection at the start of the study compared with white light, but the white-light detection rates had improved by the end of the study to that of NBI, suggesting a possible ‘learning effect’ with NBI [11]. The initial white-light adenoma detection rates in this study were surprisingly low and it is unclear if the increase seen was an effect of NBI or feedback of initial adenoma detection rates. Both these studies used patients at relatively low baseline colorectal cancer and adenoma risk. By contrast, our back-to-back study in patients having surveillance for HNPCC, where even missing a small adenoma may lead to cancer, a second pass with NBI in the proximal colon significantly improved adenoma detection, beyond that expected by a second pass with white light [12]. The total number of adenomas detected after NBI was almost doubled, including a significantly higher proportion of flat adenomas, consistent with the idea that NBI highlights subtle lesions. How can these disparate results be explained? Colonoscopists behavior may be different in high-risk patients looking for small lesions, where by withdrawing slowly with the scope tip close to the mucosa and spending extra time washing, the benefits of NBI, which is darker than white light and significantly affected by poor bowel preparation, may be maximized. There may also be a learning curve with NBI for detection. In the HNPCC study, each endoscopist had experience of at least 100 colonoscopies with NBI at study start, whereas in the Adler study, colonoscopists had experience of only five training NBI colonoscopies [11,12]. Randomized data in abstract form suggest that other high-risk groups (cancer and polyp follow-up) may also benefit from NBI [13]. An alternative, but untested, possibility is that there are differences between NBI contrast for lesions provided by sequential red–green–blue monochrome video systems (Lucera Spectrum, Olympus), used in the back-to-back study, compared with color video chip-based systems (Excera II, Olympus) used in the randomized studies; however, the spectral bandwidths of the optical filters used are the same in each system. Conversely, the image stability for the color chip system may enhance white light detection. Currently therefore, NBI should only be considered for enhanced detection in high-risk groups, similar to suggestions for chromoendoscopy. Unlike chromoendoscopy, which is considered too time consuming for general clinical use, NBI is an instant push-button technique, and there are therefore minimal time implications for its use.

Two other detection issues deserve consideration. First, there has been considerable interest recently in the role of hyperplastic polyps as precursor lesions to sporadic colorectal carcinomas which have high level microsatellite instability (MSI-high), and which make up approximately 10% of all colorectal cancers [14]. Hyperplastic polyps (often large and proximally located) with additional histological architectural changes are termed sessile serrated adenomas (SSAs) and have a high rate of BRAF mutations, which are also seen in sporadic MSI-high cancers. Hyperplastic polyps are difficult to detect with white light, but appear pale with NBI and are more easily seen. In both the randomized Adler et al. study and in our back-to-back study, hyperplastic polyp detection was significantly increased with NBI [11,12]. Potentially, NBI may lead to more comprehensive MSI-high cancer prevention, and improved detection of cases of hyperplastic polyposis syndrome. The second issue is dysplasia detection in colitis. Although clearly NBI can detect dysplasia in colitis, a small, randomized, sequential endoscopy study by Dekker et al. suggests that both white light and NBI miss significant numbers of lesions [15,16]; however, this study used an older prototype NBI system, which gave a dark image, and potentially better results might be achieved with later generation systems. Anecdotally, we feel that the background inflammation in colitis negates the ability of NBI to improve contrast for hypervascular dysplastic lesions and feel that currently, chromoendoscopy remains the criterion standard for dysplasia detection at colonoscopic colitis surveillance, and is supported by a strong evidence base [17].

The clinical data for NBI lesion characterization are stronger, with five studies now published relating to NBI, with multiple endoscopists from multiple countries, both Eastern and Western hemispheres [2,8,9,18,19]. Interestingly, although NBI can provide pit pattern information, all published studies have focused on subjective measures of microvascular density, which appear simple to learn compared with the 200–300 lesion learning curve for pit pattern [20]. Only one small pilot study (from our group) has looked at NBI pit pattern that appears no better than microvascular measures, backed up by further data from a larger series of small (<10 mm) polyps now published in abstract form [8,21]. In fact, the combination of pit pattern data and microvascular measures may even increase sensitivity [21]. Nevertheless, the idea that a simple color-based system with a short learning curve to differentiate lesions could be very appealing to endoscopists. Summarizing the available studies suggests that NBI has a sensitivity of approximately 90–95% and a specificity of 80–85% for differentiating neoplastic from non-neoplastic polyps using microvascular measures. This level of accuracy is equivalent to that of expert chromoendoscopy using pit pattern, but with a rapid push-button system that does not require mucolytics or dye-spray catheters. In total, four of the five groups directly compared NBI with chromoendoscopy and found that NBI performed very similarly in terms of lesion characterization [2,8,18,19]. This level of accuracy may approach the level needed to abandon conventional histopathology for small or particularly diminutive lesions, where the histopathologist is only being asked to differentiate neoplastic from non-neoplastic lesions, with a minimal chance of an invasive lesion being encountered; however, this would represent a paradigm shift in approach to small polyps with a shift in responsibility for the determination of dysplasia in small polyps to endoscopists, who would also therefore determine surveillance intervals immediately postprocedure [22]. Patients might prefer this one-stop approach, and it could save cost as 90% of polyps removed at colonoscopy are less than 10 mm in size. It might also allow focusing of histological resource on higher risk lesions. In order to convince patients, pathologists and financiers (e.g., health maintenance organizations and other insurers) to embrace such an approach, several factors need to be considered. First, it would be helpful to perform a large, international, nonacademic-center based, multicenter trial to confirm that the findings seen in these early studies by enthusiasts can be reproduced in routine day-to-day clinical practice, and to determine the learning curve, which remains to be known. Second, that following a period of training, some form of accreditation process be required to show both the ability to distinguish neoplastic from non-neoplastic lesions, and ability to acquire diagnostic quality images. Finally, the ability to store high-quality digital still images of every polyp as a permanent record will be needed in case of dispute [22]. It is still unclear whether endoscopists will wish to take on this extra responsibility or how they might be rewarded for this, which might offset some of the costs of converting to NBI.

Assessing lesions for endoscopic resectability, primarily the presence or absence of submucosal invasion, ‘staging’, is becoming increasingly important as the endoscopic armamentarium for wide-field mucosal resection improves, and with endoscopic sub-mucosal dissection (ESD), the potential for en bloc resection of larger lesions now exists. Multiple methods have been used in the past to define irresectibility, including morphological features such as ulceration and fold convergence, chromoendoscopy showing Kudo type V pit pattern, and the nonlifting sign at submucosal injection. Some authors have also proposed the use of endoscopic ultrasound (EUS). Furthermore, for both chromoendoscopy and EUS, there is a significant learning curve. NBI can assess pit pattern, and we have reported the successful use of type V pit pattern seen with NBI in an 8 mm polyp to determine that it was a carcinoma in vivo with appropriate modification of the endoscopic approach [23]. However, using type V pit pattern with NBI may not be as accurate as chromoscopic pit pattern, and two groups have now reported on microvessel indicators of invasion [24,25]. The presence of thick and/or irregular microvessels predicts submucosal invasion with a high degree of accuracy, and in a multivariate analysis, irregular vessels were the strongest predictor of invasion [24,25]. Again these subjective vessel appearances seem relatively simple to learn, although formal learning curve data are lacking. Therefore, NBI using microvessel measures may divide polyps into threegroups: those without discernable microvessel pattern, non-neoplastic; those with discernable microvessels ‘meshed brown capillary vessels’, ‘strong vascular pattern intensity’, or ‘brown lesion hue’, neoplastic; and those with thick or irregular microvessels, submucosally invasive. Such a simple, and apparently accurate differentiation system, available at the push of a button, we feel represents a real advance for colonoscopy, and endoscopists are likely to employ NBI in a targeted way as they find and assess each lesion.

The role for lower GI NBI may be expanded in the future, particularly to other clinical areas where chromoendoscopy has been successful, such as predicting relapse in ulcerative colitis, chromopouchoscopy, and assessing the edges of lesions for remaining neoplastic tissue after resection. NBI-based lesion characterization is currently based on subjective microvessel measures. Potentially some of the ‘human factor’ might be removed from such decision making in the future with computer-aided diagnosis (CAD) using image analysis algorithms of still video frames, increasing accuracy. Endoscopic CAD could allow objective measures of vessel density, thickness and possibly irregularity, particularly if the distance from the mucosal surface was fixed by using a cap on the distal tip of the colonoscope. The expanding roles for NBI in the upper GI tract, in targeting biopsies in Barrett’s surveillance, detecting squamous esophageal carcinoma and staging early gastric cancer, may bring NBI availability to colonoscopists who initially have not felt the evidence justifies adoption.

In summary, we feel that NBI may have a role in adenoma detection at colonoscopy, but on the current evidence this should probably be restricted to high-risk cases, such as HNPCC, cancer follow-up, and hyperplastic polyposis screening and surveillance. The use of NBI for assessing lesions looks very promising, both to differentiate neoplastic from non-neoplastic lesions, and to assess potentially invasive lesions. Moving to NBI-based ‘endohistology’ as a form of optical biopsy for small polyps is appealing in time and cost, but represents a significant practice change, and must only be done with suitable accreditation and the consent of all interested parties to ensure safe clinical outcomes. Much research work remains to be done to fully assess the diagnostic potential of NBI, which may help colonoscopy retain its pre-eminence as the diagnostic colonic imaging modality of choice. The future for NBI colonoscopy looks bright – and blue.

Financial & competing interests disclosure

James E East and Brian P Saunders have both used equipment on loan from Olympus Keymed, UK and Olympus Medical Systems Corporation, Japan to conduct NBI-based research. The authors have no other relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript apart from those disclosed.

No writing assistance was utilized in the production of this manuscript.