This preliminary feasibility study assessed the utility of a novel narrow-band imaging (NBI) system (Olympus Exera III; 190 series) both as a detection and as a characterization tool in patients undergoing surveillance endoscopy for Barrett's esophagus (BE). Two hundred and twenty-one areas in 40 patients with BE were examined prospectively. The BE segment was initially evaluated with NBI overview as a ‘red flag’ technique. Abnormal areas identified with NBI overview were then further interrogated with NBI and a dual focus (DF) magnification system (NBI-DF) in order to aid characterization. Normal areas on NBI overview were also systematically assessed with NBI-DF systematically (four quadrants every 2 cm). A confidence system was utilized when each area was assessed with NBI-DF. All areas on NBI-DF were classified into three easily distinguishable mucosal patterns: (i) regular pits with regular microvasculature (no dysplasia); (ii) irregular pits with irregular microvasculature (early cancer/high-grade dysplasia [HGD]); and (iii) equivocal, where the endoscopist was not sure about the pattern (this could be areas with increased brownish discoloration on NBI overview and dilated vasculature but no change in caliber on NBI-DF [likely inflammation or low-grade dysplasia: LGD]). Corresponding biopsies of each area were then taken. The sensitivity (Sn), specificity (Sp), positive predictivevalue (PPV) and negative predictive value (NPV) of both modes (NBI overview and NBI-DF) were then compared with the final histopathological diagnosis. One hundred and eighty-three of 221 areas (82.8%) did not exhibit any dysplasia on final histopathological assessment. NBI overview and NBI-DF accurately called all these areas as non-dysplastic. The 38 areas that appeared suspicious on NBI overview were also further assessed with NBI-DF: seven of seven were accurately predicted as harboring no dysplasia; nine areas were predicted as irregular, of which four harbored early cancer, one HGD, three LGD and one inflammation on final histopathology assessment. Twenty-two areas were deemed to be equivocal (final histology: 18 LGD and four inflammation). The Sn, Sp, PPV and NPV for the prediction of dysplasia/early cancer using NBI overview and NBI-DF were thus 100%, 93.8%, 68.6%, 100% and 100%, 86.2%, 73.3%, 100%, respectively. If NBI-DF was used in addition to NBI overview, biopsies would have been avoided in 190 areas (86%). In addition, all early cancers and HGD could be accurately identified.
The incidence of esophageal adenocarcinoma in the West has quadrupled over the last four decades.[1-3] Barrett's esophagus (BE) is the only identifiable premalignant condition responsible for this increase. It is well known that outcomes for esophageal adenocarcinoma are better in patients with BE who are on a surveillance protocol. Present guidelines recommend random four-quadrant biopsies taken in every 2-cm segment of Barrett's segment. This approach has been frequently described as ‘hit and miss’ because areas of dysplasia or microscopic carcinoma can remain undetected. In recent years, with the advent of a plethora of new endoscopic mucosal imaging technologies, identification of dysplasia and cancer in patients with BE has improved. Although the increased yields are impressive, there has been some concern regarding uniformity in the methods used. The technical skill required to carry out these procedures can sometimes also be daunting. Recently, a novel promising technology, narrow band imaging (NBI), has come to the fore. This technology provides the endoscopist with a quick and simple way to visualize mucosal morphology. The addition of a magnification or zoom component has further allowed visualization of very minute details of the surface structural patterns allowing prediction of histology in real time. Preliminary descriptive studies done with NBI and optical magnification have shown a high correlation between NBI findings and histopathology.[9-13] The system, however, was not freely available commercially and the classification system used was difficult to reproduce, especially in the community. A novel endoscope with NBI capability with the addition of a push-button function termed the dual focus (DF) has been recently introduced. Magnification of up to 70× can be achieved by simply activating the DF button which can be programmed on the head of the endoscope. The aim of this preliminary study was to validate a simplified mucosal morphology classification in assessing this new system as a simple tool to detect and characterize lesions in BE.
The study was approved by the local institutional research and ethics committee. Patients known to have BE and undergoing surveillance endoscopy or referred from other centers for work-up of recently diagnosed dysplasia over a 1-year period between July 2011 and June 2012 were invited to participate in the study. BE was defined in accordance with the British Society of Gastroenterology guidelines as ‘an endoscopically apparent area above the oesophagogastric junction that is suggestive of Barrett's Esophagus and is supported by the finding of columnar lined oesophagus on histology’. All patients gave written informed consent. Patients were excluded if they had conditions that could preclude adequate sampling of the esophagus (coagulation disorders, anticoagulant therapy and esophageal varices), if they exhibited any endoscopic evidence of erosive esophagitis, or if they had an obvious esophageal cancer.
Histology was predicted in real time based on the mucosal morphological characteristics on NBI-DF as follows:
Regular pit/regular microvasculature (no dysplasia [ND]) (Fig. 1).
Regular/absent pits with dilated microvasculature but no change in the caliber of the microvessels (likely low-grade dysplasia [LGD]/inflammation) (Fig. 3).
This was a simplified classification. A confidence level was also given to each area which was predicted as high (a) or (b), or low (c).
All procedures were done with a prototype 190 series Exera III NBI system (Olympus Co., Tokyo, Japan) with DF capability. The NBI image has an increased intensity and is brightened by 150% while the DF function enables magnification of up to 70×. Both are push-button techniques that are simple to use. A transparent cap was utilized in all procedures.
After giving informed consent, all patients were offered local pharyngeal anesthesia with xylocaine spray followed by induction and maintenance of deep sedation with i.v. midazolam and propofol. As in all imaging studies, it is imperative to visualize the mucosa clearly; hence patients were also given a mucolytic agent, N-acetylcysteine, and a defoaming agent, simethicone, mixed with water (50 mL) to drink prior to the procedure. Following intubation of the esophagus, an additional 10–20 mL of the mucolytic agent was flushed to rid the surface of any adherent mucus. This was followed by flushing the surface with dilute adrenaline (1:20 000) (10–20 mL) to reduce interference with blood and improve mucosal visualization after biopsies were taken. To minimize movement artifacts and, if deemed necessary by the endoscopist, sedation was titrated further upwards. If the esophagus exhibited excessive peristaltic activity that interfered with the examination, an antispasmodic agent, hyoscine-N-butylbromide (Buscopan, 10–20 mg), was given i.v.
The esophagus was first examined with NBI in the overview mode (without activation of DF function). The endoscopist recorded the length of Barrett's segment according to the Prague C & M criteria and any macroscopically evident lesions. Only inconspicuous, flat areas or areas adjacent to macroscopically visible lesions on NBI in the Barrett's segment were further assessed by the NBI-DF mode. We ensured that the imaged view was the area biopsied by applying light suction pressure to the mucosa, thus enabling the cap to fix to the imaged area. Normal areas on NBI overview were then assessed with NBI-DF. This was done systematically every 2 cm in each quadrant of BE, starting from the gastroesophageal junction. After taking one targeted biopsy of the ‘normal’ area, we then focused on another quadrant and repeated the procedure (digital still image followed by targeted biopsy). The endoscopist classified the images obtained in real time as regular, irregular or equivocal. All normal and abnormal areas and prediction of histology on NBI overview and NBI-DF were recorded separately on a case record form. All images were stored as high-definition JPEG files (200–300 kb, 1280 × 1024 pixel array, 32-bit color). We were especially careful in visualization of the mucosa, and if blood from earlier biopsies obscured the views, adequate flushing with water was done until a clear visual field was obtained. We also alternated freely between the NBI overview and NBI-DF modes to ensure that we were consistently aware of the previous biopsy sites and the actual position of the scope in the Barrett's segment. All biopsies were taken using standard biopsy forceps (FB230K; Olympus Co., and placed in separate labeled pots filled with 10% buffered formalin. As we dealt with a large number of samples during each procedure, we ensured that each pot was clearly labeled and linked to the corresponding NBI image.
Biopsy specimens were processed with hematoxylin-eosin (HE) and Alcian blue stains. These were reviewed by an expert gastrointestinal pathologist who had extensive experience in gastrointestinal pathology for more than 10 years. The pathologist was blinded to the endoscopic findings. Dysplasia in BE was classified according to the Vienna classification and confirmed by a second experienced gastrointestinal pathologist. The NBI overview and NBI-DF prediction was then compared with the final histopathological diagnosis.
Forty patients comprising 31 men and nine women; mean age 62.9 years (age range 27–85 years) were recruited. The mean BE length according to the Prague C & M criteria was 4.35 cm. All patients were receiving treatment with proton pump inhibitors. Results were assessed on a per-area basis by comparing the prediction of each NBI overview and NBI-DF image with the final histopathological diagnosis. To further demonstrate clinical relevance, we felt it was necessary to also present the results on a per-patient basis, by taking the worst NBI-DF grade in each patient and correlating it with the worst histological diagnosis.
NBI overview assessment
A total of 221 areas were biopsied and assessed together with their corresponding histopathological diagnoses. Morphology of 183 areas was found to be regular on NBI overview, all of which were histologically proven to have no dysplasia. The other 38 areas were suspicious on NBI overview, of which 26 areas were abnormal on final histopathology (LGD, HGD or adenocarcinoma) but 12 demonstrating no dysplasia or inflammation only (Table 1). The sensitivity (Sn), specificity (Sp), positive predictive value (PPV) and negative predictive value (NPV) of NBI overview were thus 100%, 93.8%, 68.6% and 100%, respectively.
Table 1. Per-area analysis on NBI overview in the detection of dysplasia vs no dysplasia
All of the areas were also assessed with NBI-DF. In addition to the 183 areas (which were graded as regular on NBI overview), seven other areas were also identified as demonstrating a regular pattern (190 in total). These 190 areas corresponded to normal histology. Nine areas were graded as irregular in which histology revealed four adenocarcinomas, three LGD, one HGD and one inflammatory change. The endoscopist was unsure about the remaining 22 areas which were graded as equivocal (final histopathology revealed 18 LGD and four inflammatory change) (Table 2). Results with NBI-DF were thus more promising with the Sn, Sp, PPV and NPV 100%, 97.4%, 83.8% and 100%, respectively.
Table 2. Per-area analysis on NBI-DF in the differentiation of dysplasia vs no dysplasia
The worst histological lesion in any given BE segment was used for the per-patient analysis. Of the 40 patients enrolled, using NBI-DF, dysplasia/early cancer was predicted in the Barrett's segment in 15 patients. Four of these patients exhibited adenocarcinoma, one had HGD, six had LGD and four had inflammation. The Sn, Sp, PPV and NPV were 100%, 86.2%, 73.3% and 100%, respectively.
The present study looked at combining two very useful procedural paradigms (detection and characterization) in a straightforward, simple and practical manner using a single device with very promising early results. The new NBI technology in the overview mode could be the ideal red flag technique. In addition, further characterization of abnormal areas with NBI-DF can then be used to aid characterization. The combination of these techniques could potentially lead to a paradigm shift of how patients with BE are surveyed. Abnormal areas on overview mode are further evaluated with NBI-DF. Irregular areas characterized by NBI-DF can be resected, whereas areas that are equivocal, biopsied and normal, or non-suspicious are simply left behind.
There are some limitations in the present study that need to be addressed. The findings on NBI overview could have potentially biased the prediction of histology when NBI-DF was used. However, we designed the study to reflect how surveillance is done in day-to-day practice; hence, the systematic manner with which NBI overview was switched to NBI-DF. This was deliberately done to see if a one-stop approach could be feasible where NBI overview performs as a red-flag technique followed by NBI-DF which is used as a characterization tool. The accuracy of the area imaged and the corresponding biopsies was potentially an issue that was dealt with by the technology itself. The new system has a wider field of view and hence the biopsy forceps is always easily visualized and comes into view even after maximal magnification is applied.
Various studies have used NBI as a potential tool to predict histology. In a recent meta-analysis by Mannath et al., the pooled sensitivity and specificity of this technology was greater than 95%. Most of the studies analyzed, however, were carried out with endoscopes that were not commercially available. There were also multiple classification systems used to predict histology that appeared to be confusing. There have also been studies looking at using a combination of technologies to aid in detecting and characterizing dysplasia in BE in a single session; this has been described as multi-modal imaging. Endoscopic trimodal imaging which utilizes white light, NBI and autofluorescence imaging (AFI) initially showed tremendous potential where AFI was used as a red-flag technique and NBI as a characterization tool. However, the technology has fallen by the wayside given the reasonably high false-positive rates with AFI. Adding NBI with optical magnification did manage to lower the rate, but not convincingly. A recent multicenter study in the community has proven that the trimodal endoscope is still not ready for prime time. Another study looked at combining high-definition white light endoscopy (HD-WLE) as a screening tool and a probe-based confocal endomicroscope (pCLE) as a ‘point and shoot’ technique. pCLE combined with HD-WLE significantly improved the ability to detect neoplasia in BE patients compared with HD-WLE alone. One of the main drawbacks of all these studies was the expensive equipment needed. Moreover, some of these technologies have yet to cross the ‘tertiary referral center’ boundary into the community. The present study used a prototype endoscope that is now available commercially. The push-button technology for both modalities (NBI and NBI-DF) is not only quick, but enables the endoscopist to obtain crisp, clear images almost instantaneously.
In conclusion, this preliminary study demonstrated that if NBI-DF is used in addition to NBI overview, biopsies could have been avoided in 86% of areas imaged. In addition, all early cancers and HGD could be accurately identified. The strategy of using NBI overview as a red-flag technique and NBI-DF as an interrogation tool on any abnormal areas detected on NBI overview could have major repercussions for the general endoscopist. The tedious and often impractical presently recommended strategy of random four-quadrant 2-cm biopsies could potentially be abandoned if this combined approach is utilized.
This study was conducted with an Investigator Initiated Grant from Cancer Australia to the Primary Investigator, A/Prof. Rajvinder Singh.
Conflict of Interests
Authors declare no conflict of interests for this article.