Optical coherence tomography in Barrett's esophagus: the road to clinical utility

High accuracy and biopsy concordance 

To improve on the current system of endoscopic surveillance—systematic, 4-quadrant biopsy every 1 to 2 cm, with standard white-light endoscopy—OCT will need to be sensitive and specific for dysplasia and subsquamous disease. In 2001, Poneros et al7 published the first study examining the accuracy of in vivo OCT in diagnosing Barrett's metaplasia and found a sensitivity of 97% and specificity of 92%. Although identification of intestinal metaplasia is useful, much current clinical decision-making in BE depends on the degree of dysplasia in the tissue. Subsequent work in the field has shown that the differences between OCT images of metaplasia and dysplasia are subtle. In 2005, Isenberg et al8 published a histological correlation study examining the accuracy of in vivo OCT in diagnosing dysplasia. They reported a sensitivity of 68%, specificity of 82%, and high interobserver variability. In 2006, Evans et al9 published a study further examining the OCT image characteristics of high-grade dysplasia and intramucosal adenocarcinoma. They created a 4-point scoring system based on histopathologic characteristics—a score of ≥2 had a sensitivity of 83% and specificity of 75% for high-grade dysplasia and/or intramucosal adenocarcinoma. Although the capacity for resolution has since improved,10 the sensitivity and specificity for dysplasia have not yet been studied in large numbers of participants at these higher resolutions.

Large-area image acquisition 

The current system of systematic, 4-quadrant biopsy surveillance is limited by sampling error. In order to improve on the current system, OCT will need to quickly and accurately survey the distal esophagus. OCT images can be obtained in a linear or radial manner. Linear scanning produces a linear image and is able to sample only a small area at one time. Early studies of OCT obtained linear images and were limited by sampling error. Radial scanning creates an image similar to that seen in radial endoscopic US and has the potential for assessment of larger areas and thus may be more amenable to use in endoscopic surveillance. Recent studies have attempted to improve the capacity of OCT to obtain radial images. In 2008, Suter et al11 conducted a pilot study by using OCT in combination with a balloon-centering catheter to assess the feasibility of radial imaging of the distal esophagus. Of the 10 participants completing the study, 2 had incomplete visualization of the squamocolumnar junction because of malpositioning of the system. Although this device could perform a comprehensive survey of the distal esophagus, the balloon compressed the mucosa of the esophagus and, according to the authors, distorted the images significantly. Additionally, blood, mucus, motion artifact, and variations in balloon contact affected image quality. Given these issues, the authors felt that the sensitivity and specificity of diagnostic criteria developed for linear sampling OCT systems were not applicable to this radial OCT balloon system, and the sensitivity and specificity of radial OCT using this system are unknown. The balloon-centering system was not compatible with concomitant biopsy nor was the technique able to register 3-dimensional data for later localization of a potential biopsy site, making a study correlating images to histology impossible with that system.

Fu et al12 conducted a similar pilot study by using OCT in combination with a balloon-centering catheter to assess the feasibility of large-area, comprehensive optical microscopy of the distal esophagus in swine. They reported that they were successful in imaging the distal esophagus. Unfortunately, the diagnostic capabilities, image quality, time to process the data, and the ability to perform biopsy concomitantly were not reported.

Quick image acquisition 

The images obtained via OCT-guided surveillance of the distal esophagus will need to be obtained in a timely manner compatible with running an endoscopy suite and the risks of patient sedation. The rate of image acquisition has significantly improved in the last 6 years, with preservation of resolution.13 The improvement in speed has made comprehensive OCT of the esophagus possible. The two recent studies noted earlier imaged the distal esophagus and added, on average, only 4 minutes to the procedure.11, 12 Although the images were obtained quickly, the interpretation of the images occurred after endoscopy and was described as “time consuming and laborious.”11

Easy and efficient interpretation with low interobserver variability 

For the procedure to be clinically feasible, interpretation of OCT images will need to be easy, reliable, and efficient. Given the potentially large volumes of data that may be produced, we will need to develop a method to select clinically relevant images for review. The interpretation of high-risk images will require either a gastroenterologist trained in histopathology or a pathologist trained in OCT and/or an accurate diagnostic computer program. Qi et al14 attempted to address this issue with the development of a computer-aided diagnostic algorithm by using OCT to screen for dysplasia. The investigators created an image dataset with 106 high-quality OCT image-biopsy pairs obtained during endoscopic surveillance in 13 patients. They used this dataset to create a computer-aided diagnosis algorithm for the diagnosis of dysplasia. This algorithm demonstrated a sensitivity of 82% and specificity of 74%. However, OCT surveillance of the esophagus would generate thousands of images of various qualities. The authors report that their computer-aided diagnostic algorithm required human assistance to choose regions of interest and, therefore, would be unsuitable as a stand-alone device in a surveillance protocol.

Cost effectiveness 

OCT systems are currently built by hand and adapted to the needs of a given research question. For wide-scale dissemination, a standardized, commercially available OCT system will be necessary, and it will come at a cost. What will that extra money buy? Will it eliminate the need for biopsy and histopathology, potentially paying for itself? Will it slow down the day in the procedure room, potentially costing money for something that is rapidly considered a standard part of endoscopy and not a separate line item? The economics are unclear.

If OCT solves the problems of accuracy and easy image interpretation, what is the likelihood that it will improve patient outcomes? OCT may locate many lesions of unknown significance. For example, Cobb et al2 address the issue of SSIM. We have not yet established the natural history of SSIM, much less developed guidelines for management. Should the patient with current SSIM and previous dysplastic BE have recurrent ablation? Should they have more intensive endoscopic surveillance? Esophagectomy? No one knows. It is vital to isolate SSIM to answer these important research questions, but if the lesion turns out to be relatively benign, it may not end up being clinically relevant to making this diagnosis. Treatment guidelines for BE in general are conservative, with frequent surveillance intervals and aggressive treatment for precancerous lesions. Much effort goes into preventing a relatively rare cancer in a large pool of patients. If OCT turns out to be superior to current methods in stratifying risk, perhaps the greatest impact on cost-effectiveness will be to tell us who we do not need to worry about.