Dr K. Ragunath, Wolfson Digestive Diseases Centre, Queens Medical Centre, University Hospital, Notingham NG7 2UH, UK. E-mail: K.Ragunath@nottingham.ac.uk
Background High resolution magnification endoscopy with narrow band imaging (NBI) may improve the detection of specialised intestinal metaplasia (SIM) and dysplasia in Barrett’s oesophagus.
Aims To describe the magnified endoscopic features with the use of NBI in Barrett’s oesophagus.
Methods Three hundred and forty-four areas from 50 patients with Barrett’s oesophagus were studied using high resolution magnification endoscopy (HRME) with NBI and targeted biopsies were obtained. The sensitivity, specificity, predictive values of the various patterns for the prediction of SIM and dysplasia were calculated.
Results The magnified endoscopic features of Barrett’s oesophagus with the use of NBI consist of microstructural/microvascular patterns. The yield of SIM according to the patterns was: (i) Regular microstructural pattern with tubular/linear/villous pattern 90.6% and with circular pattern 0%; and (ii) Absent microstructural pattern 98.9%. The sensitivity, specificity, positive and negative predictive values of the combination of regular microstructural pattern (tubular/villous/linear) and absent microstructural pattern to detect SIM were 100%, 78.8%, 93.5% and 100%, respectively. The sensitivity, specificity, positive and negative predictive values of the irregular microvascular/microstructural pattern for the prediction of high grade dysplasia were 90%, 100%, 99.2% and 100%, respectively.
Conclusion High resolution magnification endoscopy with NBI allows clear visualisation of microstructural and microvascular patterns within Barrett’s oesophagus, and allows targeted biopsy with a high yield of SIM and high grade dysplasia.
Oesophageal adenocarcinoma has the fastest rising incidence of all gastrointestinal cancers in the Western world especially in the UK.1 Barrett’s oesophagus is the only identifiable premalignant condition responsible for this increase.2 Barrett’s-associated oesophageal cancer occurs by way of the metaplasia-dysplasia-carcinoma sequence. The distribution of specialised intestinal metaplasia (SIM) and dysplasia within Barrett’s oesophagus is patchy and is not visible distinctly during normal white light endoscopy (WLE). Hence, surveillance endoscopy using the random biopsy technique is subject to sampling error.
Narrow band imaging (NBI) is a novel technique that enhances the diagnostic capability of endoscopes in characterising tissues by using a narrow-band width filter in a red-green-blue (RGB) sequential illumination system. NBI makes use of the optical phenomenon that the depth of light penetration into tissues is dependent on the wavelength; the shorter the wavelength, the more superficial the penetration. The blue light penetrates the tissue superficially, emphasizing the image of capillary vessels on surface mucosa.3, 4 Recent reports have indicated that NBI combined with high resolution magnification endoscopy may be helpful in identifying SIM and dysplasia in Barrett’s oesophagus and that NBI may lead to the same contrast enhancement capabilities as chromoendoscopy without using staining agents.5–7 The aim of our study was to describe the magnified endoscopic features with the use of NBI in Barrett’s oesophagus.
We included adult patients who were enrolled in our regular Barrett’s oesophagus surveillance programme, or were referred from other centres for workup of recently diagnosed dysplasia. The presence and extent of Barrett’s epithelium in patients with circumferential and non-circumferential segments was measured from the gastro-oesophageal junction according to the Prague C&M criteria.8, 9 The study was carried out in accordance with the Helsinki Declaration10 and all patients gave informed consent.
All examinations were performed with the same prototype NBI system (Olympus, Tokyo, Japan). This system is equipped with a prototype RGB sequential illumination light source (XCLV-260HP, Olympus) and video processor (XCV-260HP3P, Olympus). The light source contains one rotating RGB filter, and one NBI filter placed between the RGB filter and the light source. The NBI filter splits the white light into two specific narrow band in the ranges of blue (400–430 nm) and green (530–550 nm). As blue light penetrates only superficially, this setting allows for enhancement of the details of mucosal surface structures and optimal imaging of the epithelial microvasculature. The insertion of the NBI filter between the RGB filter and the xenon lamp is achieved by pressing a button on the, endoscope handle, allowing the endoscopist to alternate between WLE and NBI easily at any time during the procedure. Endoscopy was performed using the high resolution zoom gastroscope (GIF-Q240Z, Olympus).
Endoscopic examination-histologic examination
All patients were offered conscious sedation with intravenous midazolam (2–5 mg), or local pharyngeal anaesthesia with xylocaine spray. Prior to endoscopy a black hood (or cap) (MB-162 Olympus) was attached to the tip of the endoscope, to fix the distance between the mucosa and the endoscope to 2 mm, and therefore achieve maximum focus and magnification. Ten minutes prior to the endoscopy procedure, all patients were given to drink a mixture consisting of 50 mL of water with n-acetylcysteine and simethicone to help clear the mucus from the oesophageal surface. Following intubation of the oesophagus, additional 10–20 mL of this mixture was flushed through the instrument channel if there was any adherent mucous until the mucosa was clean. All endoscopies were performed by three endoscopists (GKA, KR, KY) with experience in magnification endoscopy. The oesophagus was first examined with WLE in the overview mode (without activating the zoom lens) and the endoscopist recorded the presence/length of hiatus hernia, length of Barrett’s segment according to the Prague criteria,8, 9 and any other abnormalities of note as defined by the macroscopic classification of early neoplasia of the digestive tract.11 Long segment’s Barrett’s oesophagus (LSBO) was defined as ≥3 cm of maximal columnar mucosa. Subsequently, NBI was activated and the whole mucosa was inspected in the overview mode for detection of any macroscopically visible abnormal lesions. If focal abnormalities were seen, these were examined under magnification. We then examined the rest of the mucosa at four quadrants every 1–2 cm intervals under NBI-magnification. To describe the magnified endoscopic patterns seen, all endoscopists followed the ‘hierarchial classification’ as proposed by the Amsterdam group.12 The endoscopic assessment of the pit patterns that were visualised was performed immediately during the procedure, therefore, the endoscopists were blinded to the histology findings. After a mucosal pattern was observed by the endoscopist, a biopsy forceps was then passed through the channel to the endoscope tip under magnification and the specific area, which had been just magnified, was then biopsied. Targeted biopsies were taken from areas with specific regular/irregular structural or micro-vascular patterns from the four quadrants at 1–2 cm intervals. The biopsies were taken first from the gastro-oesophageal junction proceeding proximally to the squamo-columnar junction allowing the blood to trickle down and prevent obscuring the views. Additionally the Barrett's segment was repeatedly washed to allow clear views.
All biopsies were taken using standard biopsy forceps (FB-24K-1, Olympus) and were immediately placed in 10% buffered formalin. Biopsies were processed and stained using standard methods, were subsequently evaluated by one pathologist with extensive experience in the evaluation of Barrett’s neoplasia, and reviewed by a second pathologist in cases of dysplasia. For purposes of this study, the pathologist was blinded to the endoscopic findings, but was aware of the clinical history of the patient. The histological outcome was classified into: no dysplasia, indefinite for dysplasia, low grade dysplasia (LGD), high grade dysplasia (HGD) and carcinoma, according to the revised Vienna classification.13
Biopsy confirmed SIM and dysplasia were used as the outcome measures. The sensitivity, specificity, predictive values of the various types of patterns for the prediction of intestinal metaplasia and dysplasia were calculated.
Three hundred and forty-four areas from 50 patients with Barrett’s oesophagus were studied (34 male, 16 female, mean age 62.1 years, range 32–86). Mean length of columnar mucosa was 3.0 cm (circumferential, range 0–13 cm) and 3.82 cm (maximal, range 1–13 cm). Twenty three patients (46%) had short segment Barrett’s oesophagus (SSBO) (mean length 1.4 cm), and twenty-seven patients had LSBO (mean length 5.5 cm). All patients were receiving antisecretory therapy with proton pump inhibitors and there was no evidence of erosive oesophagitis.
HRME plus NBI findings
As NBI can enhance microstructural and microvascular changes, the magnified endoscopic features of Barrett’s epithelium with the use of NBI were categorised according to microstructural and microvascular patterns:
(ii) Irregular microvascular pattern (IMVP) (tortuous microvessels with abnormal calibre)
Detection of SIM
The yield of SIM on target biopsies in per biopsy analysis according to the patterns was as follows: RMSP with tubular/linear/villous pattern 165/182 (90.6%), RMSP with circular pattern 0/50 (0%) and RMVP with AMSP 94/95 (98.9%). Sensitivity, specificity, positive and negative predictive values of the combination of the RMSP with regular tubular/villous/linear pattern and/or RMVP with AMSP for detecting SIM were 100% (98.1–100), 78.8% (68.3–86.6), 93.5% (89.7–95.9) and 100% (93.2–100), respectively. The sensitivity, specificity, positive and negative predictive values of the above patterns in patients with LSBO and SSBO were 100%, 87.5%, 97.6%, 100%, and 100%, 64%, 76.6%, and 100%, respectively. In per patient analysis, the sensitivity, specificity, positive and negative predictive values of the combination of RMSP and AMSP with RMVP for detecting SIM was 100% (88.8–100), 45.4% (18.1–75.4), 86.6% (72.5–94.4) and 100% (46.2–100), respectively.
Detection of dysplasia
Dysplasia was detected in 10 patients; four had LGD (three males, one female, mean age 60 years, mean length of Barrett’s 6 cm, one had SSBO), and six had HGD/early cancer (EC) (four males, two females, mean age 63.5 years, mean length of Barrett’s 4.3 cm, one had SSBO). Three patients with LGD had RMSP with regular tubular/linear/villous pattern, and one patient had a RMVP with AMSP. No endoscopically visible lesions were detected with conventional WLE, NBI or with Magnification endoscopy/NBI in any of the patients with LGD.
Four of six patients with HGD had IMVP. One patient showed IMSP. In all these patients, WLE showed a visible subtle lesion in the Barrett’s segment and magnification endoscopy with NBI revealed a clear demarcation line between the neoplastic and non-neoplastic mucosa. One patient with SSBO showed a RMSP with villous type pattern, and targeted biopsies revealed HGD. In this patient, there was no suspicious lesion seen with WLE or NBI. The sensitivity, specificity, positive and negative predictive values of the IMVP or IMSP for the prediction of HGD in per biopsy analysis was 90% (66.8–98.2), 100% (98.2–100), 99.2% (97.1–99.8) and 100% (78.1–100), respectively. The sensitivity, specificity, positive and negative predictive values of the IMVP or IMSP for the prediction of HGD in per patient analysis was 83.3% (36.4–99.1), 97.7% (86.4–99.8), 83.3% (36.4–99.1) and 97.7% (86.4–99.8), respectively.
Currently, conventional endoscopy with four-quadrant biopsies is the gold standard for the diagnosis and surveillance of patients with Barrett’s oesophagus. However, recently several reports have shown that HRME combined with chromoendoscopy can reveal mucosal patterns which correlate with SIM and HGD. In the study by Sharma et al. magnification chromoendoscopy with indigo carmine revealed specific patterns for SIM and HGD.14 Guelrud et al. combined HRME with instillation of acetic acid and produced a classification of mucosal surface pit patterns that predicted the histologic findings.15 Recently, we have also shown that high resolution magnification endoscopy (HRME) with instillation of acetic acid allows clear resolution of epithelial pit patterns within Barrett’s oesophagus, and allows targeted biopsy with a high yield of SIM and dysplasia, even in SSBO.16
A plethora of other new techniques have been applied to overcome the limitations of surveillance by random biopsy and optically detect SIM and early neoplasia, including light scattering spectroscopy, laser-induced fluorescence spectroscopy, optical coherence tomography, Raman spectroscopy and confocal endomicroscopy.17–20 However, most of these techniques require expensive instruments and are technically demanding. The NBI system we used in our study is composed of a sequential combination of an electronic endoscope and a light source equipped with one RGB and one narrow-band filter corresponding to green and blue. The use of this system in combination with a magnifying endoscope yields highly clear images of the capillaries on the mucosal surface. Hamamoto et al. reported that the visualisation of the oesophagogastric junction, blood vessels and Barrett’s segment was better with NBI than by conventional endoscopy. The relationship between the endoscopic and histopathologic diagnoses was more accurate by NBI endoscopy than by conventional endoscopy.21 Kara et al. used NBI with magnifying endoscopy to image and biopsy randomly selected areas in 63 patients with Barrett’s oesophagus. In their study, the magnified NBI images had a sensitivity of 94%, a specificity of 76%, a positive predictive value of 64% and a negative predictive value of 98% for HGD. In a similar study, Sharma et al. showed that the sensitivity, specificity and positive predictive value of the villous pattern for diagnosis of SIM were 93.5%, 86.7% and 94.7%, and for the diagnosis of HGD were 100%, 98.7% and 95.3%, respectively.23
In our series of patients with Barrett’s oesophagus, magnification endoscopy with NBI demonstrated specific microstructural and microvascular patterns. The RMVP with tubular/linear/villous pattern and the RMVP/AMSP had a sensitivity of 100% for the prediction of SIM. The lower specificity of 78.8% could be due to potential biopsy targeting problems or interpretative errors of patterns (including the learning curve that may be present in this preliminary study). In cases of LSBO, specificity was 87.5% while in SSBO only 64%, which may reflect the difficulty in obtaining targeted biopsies in an area which is very close to the gastric cardia.
Additionally, the IMVP or IMSP had sensitivity of 90% and specificity of 99% for the prediction of HGD. We have already described the microvascular architecture of the intramucosal carcinoma in Barrett’s oesophagus which is characterised by IMVP and a clear demarcation line between the neoplastic and non-neoplastic mucosa, showing that these findings are in common with gastric intramucosal carcinoma.24–26 In our study, in 5/6 cases with HGD/EC, we visualised an IMVP and a clear demarcation line between the neoplastic and non-neoplastic mucosa. One limitation of our study is that it was mainly a feasibility study aimed to describe the magnified endoscopic features in the Barrett’s epithelium. It was mainly focused on non-dysplastic Barrett’s oesophagus, with only six patients with Barrett’s-associated HGD included. Therefore, although it seems that the IMVP may discriminate between HGD/EC from LGD/non-neoplastic lesions, more studies with higher number of patients with Barrett’s-associated inconspicuous HGD are needed in order to describe the magnified endoscopic patterns in Barrett’s-associated HGD and further validate our findings. An additional limitation is that almost all lesions were seen with WLE, and this may account for the high specificity and sensitivity of the irregular patterns for HGD. However, we cannot conclude about the utility of NBI magnification endoscopy to detect subtle dysplastic lesions as all dysplastic lesions seen in our study were also detected with WLE. This is mainly a descriptive study and it cannot compare the above technique with the established four quadrant random biopsy technique for the surveillance of patients with Barrett’s oesophagus.
There was one case of HGD, where both microstructural and microvascular patterns were regular. In this particular case, the dysplastic glands were not superficially located but in deeper layers and this may explain why both microvascular and microstructural endoscopic patterns were regular.
Although our results show that there are specific microstructural and microvascular patterns associated with SIM and HGD, this was not observed in LGD which appeared similar to SIM pattern. Three cases with LGD had a RMSP with regular tubular/structural/linear pattern and one case had a RMVP with AMSP. This may be because the mucosal changes associated with LGD may not be significantly different than those observed in non-dysplastic areas (at the current magnification level) and also due to the high interobserver variability associated with the histological diagnosis of LGD.
In conclusion, HRME with NBI allows clear resolution of microstructural and microvascular pattern within Barrett’s oesophagus, and allows targeted biopsy with a high yield of SIM and HGD/EC, even in SSBO. However, adequately powered randomized studies are required in order to confirm our results, to test the inter- and intra-observer variability of this novel endoscopic technique and compare this technique with random four-quadrant biopsies for the surveillance of patients with Barrett’s oesophagus.
Declaration of personal interests: We would like to thank the Hellenic Gastroenterology Society and AstraZeneca, Greece, for providing a research scholarship to Dr Anagnostopoulos, University of Fukuoka, Japan, for providing an overseas sabbatical to Dr Yao, Digestive Disorders Foundation (CORE) and The Great Britain Sasakawa Foundation for providing travel and research grants to Dr Ragunath. The NBI endoscopy equipment was provided by Olympus-Keymed, UK. Declaration of funding interests: None.