Two recent publications focused on new technologies to detect Barrett's esophagus and early Barrett's neoplasia.
A clinical practice update from the American Gastroenterological Association, published on July 1 by Clinical Gastroenterology and Hepatology, reviewed new technology and innovation for surveillance and screening in patients with Barrett's esophagus. It presented best practice advice statements, which were developed from expert review of existing literature combined with discussion and expert opinion to provide practical advice for all gastroenterologists and endoscopists. The update does not include formal ratings of the quality of evidence or strength of the best practice advice statements.
Clinicians may consider screening with standard upper endoscopy in patients with at least three established risk factors for Barrett's esophagus and esophageal adenocarcinoma, according to the update. These risk factors include the presence of chronic gastroesophageal reflux disease and at least two of the following: age older than 50 years, male sex, non-Hispanic White race, smoking, obesity, and a family history of Barrett's esophagus or esophageal adenocarcinoma. Screening and surveillance endoscopic examination should be performed using high-definition white light endoscopy (WLE) and virtual chromoendoscopy, although nonendoscopic cell collection devices may also be considered as a screening option for Barrett's esophagus, according to the advice. Advanced imaging technologies may be used as adjunctive techniques to identify dysplasia. “The panel were supportive of the need to have improved imaging technologies to better identify areas of dysplasia and early cancer,” the update said. Panelists considered confocal or volumetric laser endomicroscopy (VLE) in this discussion. While they noted that VLE is not currently commercially available, it has introduced several new advances in regard to imaging in Barrett's esophagus, including laser marking and the interpretation of imaging using artificial intelligence (AI).
“The panelists felt strongly this was an important area where further innovation is needed but that the use of these techniques was not required for a high-quality exam and the data to date did not support its routine use,” the update said. “However, the panel felt these technologies were promising and carried potential benefits in select cases and currently might be best utilized in expert centers.”
The second paper, a systematic review and meta-analysis, provided additional evidence of AI's potential as a useful tool to detect early Barrett's neoplasia in patients with lesions of at least high-grade dysplasia. Researchers reviewed 12 studies with a total of 1,361 patients and 532,328 images that were acquired to train various AI models. In terms of the endoscopic imaging modality evaluated, 11 studies included WLE, three included narrow-band imaging, and three included virtual chromoendoscopy images. None of the studies included Barrett's esophagus lesions with low-grade dysplasia. The primary outcome was the detection of dysplasia and early cancer in Barrett's esophagus. Results were published on June 22 by Frontiers in Medicine.
Eleven studies of 191,278 endoscopic images or frames demonstrated that AI is very accurate at detecting early Barrett's esophagus neoplasia, with a pooled sensitivity, specificity, and diagnostic odds ratio of 90.3% (95% CI, 87.1% to 92.7%), 84.4% (95% CI, 80.2 to 87.9%), and 48.1% (95% CI, 28.4% to 81.5%); however, all primary outcome analyses had significant heterogeneity. Despite significant heterogeneity among the included studies, the area under the summary of receiver-operating characteristic curve was 0.94 (95% CI, 0.92 to 0.96).
Limitations of the meta-analysis include that most of the studies were retrospective in nature, the authors noted. They added that there were insufficient data to perform a subgroup analysis of the performance of AI versus endoscopists at detecting early neoplastic Barrett's esophagus.
“Our results support the need for more studies, including AI models to detect macroscopically visible low-grade dysplasia,” the study authors concluded. “In addition, well-designed prospective randomized controlled studies are needed to further explore if AI can indeed be effective both for experts and non-expert endoscopists.”