Confocal laser endomicroscopy (CLE) provides real-time microscopic imaging within the esophagus after administration of fluorescein. It permits immediate diagnosis of Barrett’s esophagus and associated intraepithelial neoplasia (dysplasia). Multiple trials have evaluated the accuracy of both CLE systems (endoscope-integrated CLE and probe-based CLE) in Barrett’s patients. CLE requires a thorough knowledge of mucosal pathology, but training sets based on a pattern recognition approach demonstrate that gastroenterologists can achieve good accuracy rates. Accuracy of predicting neoplasia is sufficient to allow for the use of optical biopsies instead of physical pinch biopsies according to the Seattle protocol in some but not all trials. A major advantage is the ability to use CLE for immediate targeted resection of suspicious lesions without the need to wait for histopathology results. CLE use for screening after circumferential radiofrequency ablation therapy has so far not met the clinical needs. In addition, many of the classical risk factors that are needed to determine if a local endoscopic resection is feasible and can be considered curative cannot (yet) be assessed by CLE. Molecular imaging using CLE aims at further optimization of lesion characterization but still mandates further clinical evaluation.
KeywordsEndomicroscopy, Barrett’s esophagus, intraepithelial neoplasia, radiofrequency ablation, endoscopic mucosal resection, molecular imaging
Barrett’s esophagus (BE) is associated with the risk of development of esophageal adenocarcinoma. Current surveillance according to the Seattle protocol includes white light endoscopy (WLE) with the collection of random four-quadrant biopsy specimens over every 1–2 cm of the columnar-lined esophagus. The aim of such surveillance is detection of neoplasia, ideally at an (endoscopically) curable stage. This state-of-the-art approach is labor intensive and prone to sampling error since only a minor part of the mucosal surface is undergoing microscopic analysis ex vivo. Despite high definition (HD) and virtual or spraying surface contrast enhancement to augment WLE, its ability to reliably detect premalignant lesions remains suboptimal. This is partly based on the fact that high-grade intraepithelial neoplasia (IN) intramucosal cancer in BE is found in a patchy pattern side by side with lower grades of dysplasia and metaplasia.
Endomicroscopy is a technique in which the mucosa is magnified by a confocal scanner. When combined with WLE this technique can be used to highlight areas that are suspicious for dysplastic epithelium. A fluorescent agent is injected intravenously prior to image acquisition and images are viewed at real time during endoscopy. This tool is best used when examining small areas of the esophageal mucosa. Endomicroscopy is fundamentally different from light microscopy: the use of the confocal technique allows microscopy even underneath the tissue surface in intact tissue without the need to physically shine light through thin tissue sections. Such optical sectioning permits magnification to about 1000-fold and reveals subtle structural details of the mucosa on a (sub-)cellular level. Since its first description over 10 years ago, many trials have covered confocal laser endomicroscopy (CLE) in BE. This indicates a clinical need to optimize detection of Barrett’s associated IN and reflects the low confidence of many gastroenterologists in untargeted quadrant biopsies to incidentally pick up preneoplastic lesions. Volumetric laser endomicroscopy (VLE) shows overlap in naming and indication but relies on optical coherence tomography, a technically different cross-sectional imaging technique that is covered in Chapter 8 : “Enhanced Imaging of the Esophagus: Optical Coherence Tomography.”
Confocal Endomicroscopy Devices
Two CLE systems are currently used in clinical practice, an endoscope-based system (eCLE; Pentax, Tokyo, Japan) and a probe-based system (pCLE; Mauna Kea Technologies, Paris, France) . Both are point techniques that achieve very high resolution of a small mucosal area.
In eCLE, that is not marketed at this point (early 2015) but still available in many endoscopy suites, a miniaturized confocal scanning device is integrated into the distal tip of a standard resolution endoscope. The tip of the scanner protrudes slightly and is visible in the endoscopic image at the seven o’clock position so that it can be placed onto the point of interest under endoscopic guidance. The free working channel can be used for labeling, targeted biopsies, or other interventions. The scanning mechanism relies on a single optical fiber working as a pinhole, mounted into a resonant magnetic tuning fork that scans the area at high resolution at two different speeds. Lateral resolution is 1024×1024 pixels (~0.7 µm) at a frame rate of approximately 0.6 s −1 or 512×1024 pixels at 1.2 s −1 . The depth of imaging can be actuated by the user at 7 µm steps from surface to about 200 µm.
In probe-based CLE (pCLE), the confocal probe is fitted through the working channel of any gastrointestinal endoscope. This carries with it the advantage of using it with different types of endoscopes, including state-of-the-art HD scopes. The probes are available with different diameters that even allow use within the bile duct or through an FNA needle during endoscopic ultrasonography (which are outside the focus of this review). The probes use a fiber bundle for laser light propagation and collection of fluorescence. The imaging plane is fixed for the different probe types and can be somewhat adapted by exerting different extent of pressure with the probe. Most trials have used the Gastroflex UHD CLE probe. Some authors use a short transparent cap on the endoscope to facilitate stabilization of the probe tip on a region of interest. Resolution of pCLE is lower than with eCLE, but image acquisition faster, providing microscopic video sequences at real time.
In both systems, the excitation wavelength is 488 nm (blue light), and emitted light is captured in the green range. Imaging relies on the application of a fluorescent agent. In most trials, fluorescein is injected intravenously at 2.5–5 mL of a 10% solution. Fluorescent contrast is available for tissue imaging after few seconds. Fluorescein is partly bound by plasma proteins and also extravasates into the tissue. As a result, the tissue structure becomes visible almost immediately with good resolution of the mucosal structures and the capillaries of the lamina propria. Nuclei are not discernible, but structural information of the mucosa is usually sufficient as a basis for a therapeutic decision (see later). Fluorescein is safe, and adverse events are rare . While some early trials have also used topical acriflavine (which visualizes nuclei) , this is largely abandoned due to a theoretical risk of harboring mutagenic effects of the nuclear staining. Cresyl violet results in an indirect visualization of nuclei but has only been evaluated in small trials .
In CLE, the resultant image on the screen is parallel to the tissue surface, ie, perpendicular to sectioning in conventional histopathology. Interpretation of the microscopic images is usually performed online in order to make use of the advantage of having a microscopic tissue analysis available during the endoscopic session. This requires a thorough knowledge of the mucosal histopathology by the endoscopist. For starting CLE in the endoscopy unit, it might be beneficial to ask a histopathologist into the room for the first CLE sessions, however, this is not mandatory. Usually, the endoscopist bases his microscopic diagnosis on a two-step decision: the first being the differentiation of normal versus abnormal (pattern recognition) and the second being appreciation of the fine suspicious alterations in abnormal tissue (detail description). This will be explained later for Barrett’s esophagus and associated neoplasia in more detail.
Confocal Laser Endomicroscopy in Barrett’s Esophagus
Confocal Laser Endomicroscopy Features of Barrett’s Esophagus
In BE, the squamous epithelium of the lower esophagus is replaced by metaplastic columnar-lined epithelium ( Fig. 9.1 ). One of the hallmarks of the metaplasia is the presence of goblet cells (although this has been questioned for a smaller subset of Barrett’s patients). Goblet cells are easily discernible by CLE as high columnar cells with dark mucin inclusions toward the luminal aspect. The differentiation of the glandular structure of BE against the squamous epithelium is usually very straightforward whereas delineation against the darker, cobbled-stone appearance of the cardiac epithelium may be more difficult.
In the first trial using CLE for BE , vessel and cellular aspects were described for the typical microscopic definition of BE: BE contains columnar-lined epithelium with goblet cells in the luminal superficial aspects of the mucosal layer. In deeper parts, villous-like glandular structures contain dark cylindrical epithelial cells. Capillaries within the Lamina propria are regular and visible in deeper optical sections. Often, a superficial double lining is visible that corresponds to the brush border of the epithelium. In BE-associated neoplasia, fluorescein leaks from superficial and deep irregular capillaries. Nests of malignant cells are darker probably due to a lower pH—fluorescence after fluorescein is pH dependent. Neoplastic cells may pile up to multilayer structures or are no longer contained by a normal basal membrane and infiltrate from the epithelium into the lamina propria.
For pCLE, criteria have been tested and validated for detection of BE-associated neoplasia : a saw-toothed epithelial surface, enlarged and pleomorphic cells, not equidistant glands unequal in size and shape, and goblet cells that are not easily identified, showed an overall accuracy in diagnosing BE-associated neoplasia of 81.5%.
Clinical Trials for Diagnosing Barrett’s Esophagus and Barrett’s Esophagus-Associated Neoplasia
Initial studies have aimed at the clinical feasibility of CLE in Barrett’s esophagus and the diagnostic accuracy in comparison to histopathology. These early trials defined the criteria to diagnose BE and BE-associated neoplasia that were then evaluated and used in follow-up trials. An early study using CLE in 63 patients with BE was published in 2006 . Here, based on the above-mentioned criteria eCLE was able to predict Barrett’s epithelium and Barrett’s associated neoplasia during endoscopy with a sensitivity of 98.1% and 92.9% and a specificity of 94.1% and 98.4%, respectively. Accuracy was 96.8% and 97.4%, respectively. Interobserver and intraobserver agreements for the prediction of histopathological diagnosis were substantial ( κ , 0.843 and 0.892, respectively). In a similar setting, eCLE was performed in 50 patients referred for known BE . Again, optical biopsies were performed in a circular fashion every 1–2 cm of the columnar-lined esophagus, corresponding to an “optical Seattle biopsy” protocol, and on a total of three visible lesions. With targeted histology as the gold standard, BE-associated neoplasia could be predicted with an accuracy of 98.1% and a substantial agreement between endomicroscopy and histology ( κ =0.76).
Early trials using pCLE demonstrated a high negative predictive value (NPV) of 99% for BE, however, showed a PPV of only 44% . Similar numbers were generated in a second trial with pCLE, where the NPV to rule out neoplasia was 95%, but PPV was only 18% . There has been some speculation about the reason for the initially observed differences in the numbers of pCLE and eCLE, and one possible explanation might be higher resolution of eCLE. In fact, one small case series has compared pCLE criteria and eCLE criteria, and found an interrater agreement of 0.17 and 0.68, respectively . Accordingly, the overall accuracy in detecting dysplasia was only 37% and 44.3%. However, this was an ex vivo study on only 13 specimens and thus does not provide a solid head-to-head comparison. Another useful explanation was that compromising on PPV rather than NPV makes sense in a clinical setting in order to be sure to not miss suspicious lesions. Accordingly, follow-up trials demonstrated higher sensitivity and specificity rates of 87.8% for pCLE . In a post hoc analysis in a meta-analysis, the type of CLE used did not make a significant difference as to sensitivity and specificity .
A following set of trials compared CLE optical biopsies in a Seattle fashion (with targeted real biopsies only on suspicious areas) with random physical biopsies according to the Seattle protocol, the current standard of care. This slightly different approach makes use of the fact that a larger mucosal area is covered by in vivo endomicroscopy by moving the scanner across the columnar-lined esophagus than by taking pinpoint real biopsies. In the first of these studies , the yield per biopsy was almost doubled by endomicroscopic targeting. In about two-thirds of patients, the need for random biopsies would have been abolished based on normal CLE findings. This study corroborated the point that gastroenterologists are able to reliably differentiate nonneoplastic from neoplastic Barrett’s epithelium and strengthened the argument that close observation with CLE would obviate the need for random biopsies in favor of CLE-targeted biopsies. In a multicenter follow-up trial, 192 BE patients were randomized to undergo screening with high-definition white light endoscopy (HD-WLE) and random biopsies versus HD-WLE followed by eCLE and targeted biopsies only . Diagnosis and proposed management were documented after HD-WLE in both groups, and after eCLE in the second group. In the eCLE group, a lower number of biopsies resulted in a significantly higher yield for BE-associated neoplasia (7% in the HD-WLE group vs 34% in HD-WLE eCLE group). In a per-biopsy analysis, the use of eCLE in addition to HD-WLE resulted in a 4.8-fold decrease in the number of biopsies during endoscopy. In a per-patient analysis, eCLE had a 2.7-fold higher diagnostic yield for neoplasia. With eCLE, 34% of patients had a correct change in management. Of 26 patients without lesions, 5 had inapparent high-grade IN in flat BE detected by eCLE and 21 did not require any biopsies.
The next set of studies established the use of CLE in addition to HD-WLE. In a trial comparing HD-WLE versus HD-WLE with pCLE in a total of 100 patients , HD-WLE alone did not find suspicious areas, but random biopsies according to the Seattle protocol picked up incident BE-associated neoplasia in 5 of 50 patients (10%, 1 patient with high-grade neoplasia). pCLE identified areas suspicious for neoplasia in 21 of 50 patients (42%), of which 14 cases were confirmed by histopathology, 2 with high grade. This was significantly higher than in the HD-WLE only group, with numbers for sensitivity, specificity, PPV and NPV of pCLE for neoplasia of 100%, 83%, 67%, and 100%, respectively. Similar results supporting the use of pCLE in addition to HD-WLE were found in an enriched population of 101 consecutive BE patients scheduled for surveillance or treatment of BE-associated neoplasia . In contrast, another study included 50 consecutive patients with known dysplastic Barrett’s esophagus, and a prediction of histology was performed with HD-WLE, followed by NBI, and finally by CLE, and correlated to biopsies targeted to this spot. A total of 91 biopsy spots harbored high-grade IN or intramucosal cancer. CLE did not add to the accuracy rates of 82.8% for HD-WLE or 81.4% for NBI. In these patients with a known high pretest probability, all mucosal carcinomas could be detected by targeted biopsies guided by HD-WLE or NBI alone.
Two meta-analyses have evaluated the value of CLE in BE patients, pooling pCLE and eCLE trials. A recent meta-analysis summarized results in BE patients up to 2013 in a total of 7 trials, including all prospective studies that compared the accuracy of CLE with standard four-quadrant biopsies (total of 345 patients and 3080 lesions) . On a per-lesion analysis, the pooled sensitivity for the detection of high-grade IN and early carcinoma in Barrett’s esophagus was 68%, the pooled specificity 88%. Respective numbers for sensitivity and specificity in a per-patient analysis were 86% and 83%. A second meta-analysis with slightly different eligibility criteria on 8 trials involving 709 patients and 4008 specimens showed comparable results in a per-lesion analysis (sensitivity 70%, specificity 91%) and per-patient analysis (sensitivity 89%, specificity 75%) for the detection of neoplasia. The authors of these meta-analyses conclude that CLE with targeted biopsies has good diagnostic accuracy for detecting high-grade IN and early carcinoma in BE. However, the data had been gathered from patients with a high overall prevalence of high-grade IN and early carcinoma, and may therefore not be generalizable to all screening patients. Thus abandoning the Seattle protocol at this stage cannot be recommended for the broad clinical practice, despite good results in expert centers.
A further set of trials used CLE for guidance of therapy or surveillance after therapy. It has been hypothesized that CLE could be an appropriate tool to confirm completeness of ablation and to rule out residual buried Barrett’s glands after radiofrequency ablation of BE. This is difficult to achieve with HD-WLE alone. A recent study assessed the usefulness of pCLE in addition to HD-WLE in a prospective, multicenter, randomized fashion . After endoscopic ablation, patients were followed up with HD-WLE or HD-WLE aided by pCLE, with treatment of residual Barrett’s mucosa guided by endoscopic or/and endomicroscopic findings. A total of 119 patients were included before termination of the trial at interim analysis. The addition of pCLE did not result in a higher proportion of patients with complete absence of BE at follow-up. However, only a low proportion of patients achieved complete eradication in both groups (15 of 57 patients (26%) for HD-WLE; 17/62, 27% with HD-WLE+pCLE) which might have influenced results. Another reason for this unanticipated result could be the limited depth of imaging especially in squamous epithelium, making transmission of blue laser light for CLE difficult. However, some neoplastic glands at least partially undermining squamous epithelium were diagnosed in two trials using eCLE , however, without prior ablation treatment, thus not representing true buried glands.
How long does it take for an experienced gastroenterologist and endoscopist to be confident in establishing a microscopic diagnosis based on CLE findings (rather than taking a biopsy and asking the histopathologist for a diagnosis)? A short learning curve after a structured teaching session and a set of surveilled teaching cases has been suggested by an abstract . In a prospective, double-blind review of pCLE images of 40 sites of BE and matched tissue specimens as a gold standard, the sensitivity for the diagnosis of neoplasia for 11 endoscopists (CLE specialists and nonspecialists) was 88%, with a specificity of 96% and a substantial diagnostic agreement ( κ , 0.72) that was even better when only pCLE specialists were evaluated ( κ , 0.83) . Recently, a thorough evaluation of the pCLE criteria was published . After development of the above-mentioned criteria, a validation set was used for the prediction of neoplasia in BE. Overall accuracy in predicting high-grade IN and cancer was high and interobserver agreement substantial. Accuracy and agreement between experienced and nonexperienced assessors were not different after initial structured teaching, suggesting a short learning curve. These pCLE criteria also served as a basis to establish the accuracy and interobserver agreement between endoscopists and histopathologists and among histopathologists . Among histopathologists, the accuracy for the diagnosis of all grades of neoplasia was 77.8% in 90 videos on BE and rose to 93.8% when pathologists had “high confidence” in their assessment of the videos, with substantial interobserver agreement. Comparable results were achieved by endoscopists in the same set of videos, suggesting that endoscopists can be trained to achieve a sound diagnosis in BE based on predefined criteria. This trial also underlines the need for high quality microscopic images to allow for a confident (and more accurate) diagnosis.