Key points

Introduction

Barrett’s esophagus (BE) is defined by the presence of specialized intestinal metaplasia (SIM) of the esophagus and affects 1% to 2% of the general population. BE is associated with increased cellular proliferation and turnover that may result in progression to dysplasia, and is therefore a precursor lesion of esophageal adenocarcinoma (EAC). In a recent retrospective population-based cohort study, the presence of BE conferred a relative risk of EAC of 11.2 compared with that of the general population (95% confidence interval [CI], 8.8–14.4). Early detection results in improved patient outcomes, and therefore endoscopic surveillance of BE is recommended to detect dysplasia or cancer at an early stage and instigate therapy. Over the past decade, remarkable progress has been made in the endoscopic treatment of BE-related neoplasia, including endoscopic resection of focal lesions and ablative therapies.

Endoscopy is a key approach for detection and treatment of BE-related neoplasia. Random biopsies are susceptible to sampling variability and have been shown to miss up to 57% of neoplastic lesions, particularly flat neoplasias. Gastroenterology society guidelines and expert consensus statements recommend endoscopic surveillance using white-light endoscopy (WLE), with targeted biopsies of any endoscopically visible lesions, and random 4-quadrant biopsies of every 1 to 2 cm of the BE segment (Seattle protocol). Endoscopic techniques that might improve visualization of endoscopic features associated with neoplastic change would greatly enhance the diagnostic yield, efficacy, cost, and efficiency of current surveillance practices. This article describes a variety of “red flag” diagnostic techniques that have been developed with the goal of increasing the sensitivity for detecting lesions during a wide-field endoscopic examination.

Introduction

Barrett’s esophagus (BE) is defined by the presence of specialized intestinal metaplasia (SIM) of the esophagus and affects 1% to 2% of the general population. BE is associated with increased cellular proliferation and turnover that may result in progression to dysplasia, and is therefore a precursor lesion of esophageal adenocarcinoma (EAC). In a recent retrospective population-based cohort study, the presence of BE conferred a relative risk of EAC of 11.2 compared with that of the general population (95% confidence interval [CI], 8.8–14.4). Early detection results in improved patient outcomes, and therefore endoscopic surveillance of BE is recommended to detect dysplasia or cancer at an early stage and instigate therapy. Over the past decade, remarkable progress has been made in the endoscopic treatment of BE-related neoplasia, including endoscopic resection of focal lesions and ablative therapies.

Endoscopy is a key approach for detection and treatment of BE-related neoplasia. Random biopsies are susceptible to sampling variability and have been shown to miss up to 57% of neoplastic lesions, particularly flat neoplasias. Gastroenterology society guidelines and expert consensus statements recommend endoscopic surveillance using white-light endoscopy (WLE), with targeted biopsies of any endoscopically visible lesions, and random 4-quadrant biopsies of every 1 to 2 cm of the BE segment (Seattle protocol). Endoscopic techniques that might improve visualization of endoscopic features associated with neoplastic change would greatly enhance the diagnostic yield, efficacy, cost, and efficiency of current surveillance practices. This article describes a variety of “red flag” diagnostic techniques that have been developed with the goal of increasing the sensitivity for detecting lesions during a wide-field endoscopic examination.

Characteristics of ideal endoscopic enhancements for wide-field detection of BE-related neoplasia

In clinical practice, most patients with BE do not harbor dysplasia. These patients have a very low incidence of high-grade dysplasia (HGD) and EAC development, and thus most patients undergoing surveillance have multiple biopsy specimens showing no evidence of HGD/EAC. For populations with a low prevalence of disease, the most important metrics for a diagnostic test are the sensitivity and negative predictive value (NPV).

The American Society for Gastrointestinal Endoscopy (ASGE) convened a PIVI (Preservation and Incorporation of Valuable Endoscopic Innovations) initiative to develop a priori diagnostic and/or therapeutic thresholds for endoscopic technologies designed to resolve relevant clinical questions. The PIVI for BE provided thresholds for diagnosis of BE and associated neoplasia. The current 4-quadrant, 1- to 2-cm biopsy protocol has a reported sensitivity ranging from 28% to 85% for detecting HGD/EAC. In studies in which cost-effectiveness of a surveillance program has been demonstrated, sensitivity of 85% to 90% has been assumed. To eliminate the need for random mucosal biopsies during the endoscopic surveillance of patients with nondysplastic BE, an imaging technology with targeted biopsies should have a per-patient sensitivity of 90% or greater and an NPV of 98% or greater for detecting HGD or early EAC compared with the current standard protocol (WLE and targeted and random 4-quadrant biopsies every 2 cm). The specificity of biopsy protocols in patients with BE has ranged from 56% to 100%. Hence, the new imaging technology should have a specificity that is sufficiently high (80%) to allow a reduction in the number of biopsies (compared with random biopsies). Because abnormal imaging requires biopsies for confirmation, new imaging methods should not result in more biopsy specimens being obtained than would a random biopsy protocol.

The various “red flag” endoscopic imaging techniques are summarized in Table 1 .

High-definition WLE

Standard WLEs are the most commonly used method of endoscopic imaging, generating 300,000-pixel images. In contrast, modern high-definition endoscopes offer high pixel densities (600,000–1,000,000 pixels) and, when combined with high-definition monitors, provide marked improvement in image resolution ( Fig. 1 ). In a study by Kara and colleagues, high-definition WLE (HD-WLE) alone had a 79% sensitivity for detecting lesions with HGD. Gupta and colleagues performed HD-WLE in 112 patients, and 34% had HGD or EAC detected across 130 locations. Most importantly, longer inspection time was associated with a significant increase in the number of lesions detected; an average inspection time of greater than 1 minute per centimeter of BE enabled a higher lesion detection rate (54.2% vs 13.3%; P = .04), with a trend toward a higher detection rate of HGD/EAC (40.2% vs 6.7%; P = .06). Hence, most experts would recommend high-definition (>850,000 pixels) endoscopy over standard endoscopy in evaluating patients with BE. Standard-resolution endoscopes are not recommended, although scant scientific evidence exists for this recommendation.

Endoscopic endoscopy images of BE with a macroscopic Paris 2A neoplastic lesion seen with ( A ) standard-resolution WLE, ( B ) high-resolution WLE, and ( C ) high-resolution endoscopy with narrow band imaging.

Narrow band imaging

Narrow band imaging (NBI) was first described in 2004 (Olympus Corporation, Tokyo, Japan). Endoscopes equipped with NBI contain an additional filter that can be activated by the endoscopist. The filter narrows the bandwidths of the emitted blue (440–460 nm) and green light (540–560 nm), and the relative intensity of blue light is increased. The narrow band blue light displays superficial capillary networks, whereas green light displays the subepithelial vessels. A combination of the 2 images produces a high-definition image of the mucosal surface, allowing visualization of subtle mucosal irregularities and alterations in vascular patterns (see Fig. 1 ; Figs. 2 and 3 ).

High-resolution NBI of flat nondysplastic BE showing the contrast between squamous mucosa ( greenish ) and columnar Barrett’s mucosa ( brown-blue ) and regular mucosal pit pattern.

Irregular mucosal pattern and highlighted subtle nodularity in neoplastic BE (3–5 o’clock position) by high-resolution NBI.

In a small prospective study of 28 patients, NBI did not increase the number of cases diagnosed with HGD/EAC compared with high-resolution WLE. More areas of HGD were detected with NBI than with WLE (4 additional lesions in 3 patients); however, given the small size of the study, statistical significance was not reached. A prospective, blinded study by Sharma and colleagues in 2006 using a standardized classification of mucosal (ridge/villous, circular, and irregular/distorted) and vascular pattern (normal and abnormal) suggested that NBI had a high sensitivity and specificity for detecting HGD (100% and 99%). In a tandem crossover study, Wolfsen and colleagues showed that high-resolution NBI was superior to standard-resolution WLE in detecting dysplasia (57% vs 43%), with a sensitivity and specificity of 89% and 95%, respectively. Additionally, in the NBI group, fewer numbers of biopsies were required (mean, 4.7 vs 8.5; P <.001). However, the higher-resolution endoscopes used in the NBI group may have accounted for the higher detection rate.

Classification of mucosal and vascular patterns by NBI is key for the correct prediction of BE histology. Studies by Curvers and colleagues and by Herrero and colleagues have shown that NBI does not improve inter-observer agreement or accuracy over HD-WLE. In a study involving ex vivo examination of pedigreed NBI and WLE BE images, the interobserver agreement for NBI diagnosis ranged from 0.40 to 0.56 (moderate) and did not significantly differ between expert and nonexpert endoscopists. The overall yield for correctly identifying images of early neoplasia was 81% for high-resolution WLE, 72% for NBI, and 83% for high-resolution WLE with NBI, with no significant difference between experts and nonexperts. These kappa scores and accuracy rates are slightly lower than those reported for NBI in the colon (κ = 0.63; substantial agreement for prediction of polyp histology; accuracy rates 80%–85%). Widespread training will be needed for recognition of irregular NBI patterns in BE and a study of interobserver variability when used by community gastroenterologists to determine the incremental benefit of NBI over careful HD-WLE examination for routine imaging in BE.

Sharma and colleagues recently reported results of an international randomized controlled trial involving experts comparing NBI with HD-WLE, which showed that both techniques detected 92% of patients with intestinal metaplasia, but NBI required fewer biopsies per patient (3.6 vs 7.6; P <.0001). NBI also detected a higher proportion of areas with dysplasia (30% vs 21%; P = .01). All areas of HGD and cancer had an irregular mucosal or vascular pattern when examined with NBI and, importantly, no area of regular mucosal/vascular pattern harbored HGD or cancer, suggesting biopsies of these areas can be avoided. Not surprisingly, the accuracy of NBI for detecting low-grade dysplasia was limited. This trial showed that NBI-targeted biopsies are more efficient than random 4-quadrant biopsies.

Autofluorescence

Autofluorescence imaging (AFI) produces real-time pseudocolor images based on the detection of natural tissue fluorescence generated from endogenous tissue fluorophores, such as elastin, collagen, porphyrins, flavins, aromatic amino acids, and NADH. When exposed to short-wavelength light, fluorophores are excited and emit fluorescent light of longer wavelength (ie, autofluorescence). Normal and neoplastic tissue have different autofluorescence characteristics, enabling their differentiation.

During AFI, normal tissue is pseudo-colored as green, blood vessels as dark green, and dysplastic/neoplastic areas appear as magenta. Suspected neoplasia (AFI-positive lesion) is defined as any area that is different in color from the surrounding mucosa and that has a defined circumferential margin. The following characteristics of neoplastic tissue enable abnormal AFI: (1) increase in the nuclear-cytoplasmic ratio, (2) loss of collagen, and (3) neovascularization, inducing increased hemoglobin concentration, which absorbs autofluorescence light. Autofluorescence in BE is believed to result primarily from collagen in the stroma that is reabsorbed by hemoglobin.

Autofluorescent intensity can be low and earlier AFI systems failed to produce sufficient image quality for clinical use. However, the integration of AFI with high-resolution video endoscopes and a second, more sensitive, charge-coupled device that converts light into a digital image has significantly improved image quality.

In a feasibility study, Kara and colleagues examined 22 patients with HGD using both WLE and AFI. HGD was detected in 6 of these subjects with AFI alone. The sensitivity for detecting HGD was 91%; however, a high rate of false-positives as seen, yielding a specificity of only 43%. To overcome this inadequacy, subsequent studies combined NBI with AFI, hoping to improve the specificity of the technique. In a study of 20 patients with suspected HGD, Kara and colleagues showed that the addition of NBI to AFI reduced the false-positive rate from 40% to 10%. However, NBI also resulted in the misclassification of 8% to 17% of true-positive areas as nonneoplastic.

The combined use of WLE, NBI, and AFI resulted in the development of endoscopic trimodal imaging (ETMI), which incorporates the 3 platforms into 1 endoscope and is currently available in the United Kingdom and Asia. WLE and ETMI were compared in 87 patients with suspected HGD/EAC. Targeted detection of HGD/EAC was higher with ETMI than with WLE (65% vs 45%), but the overall detection rate for neoplasia was not statistically significant (84% vs 72%). Higher false-positive rates were also still noted with ETMI (71% vs 53%). Random 4-quadrant biopsies identified more areas of HGD/EAC than targeted biopsies (84% vs 64%). A subsequent randomized crossover study of 99 patients with low-grade intraepithelial neoplasia compared the use of ETMI versus WLE and found an improved targeted detection with ETMI (54% vs 34%), but no statistically significant increase in overall detection.

In summary, the major limitation of AFI seems to be the high false-positive rate, and data in the current literature have not yet supported its widespread use.