Big Data as the Universal Language for Barrett’s Esophagus

Excerpt

Eastern and Western gastroenterologists disagree on what endoscopic features define Barrett’s esophagus: salmon-colored mucosa in the West [1] and distal palisading vascular arborizations in the East [2]. British and American pathologists (and many others who embrace either the British or the American viewpoint) disagree on what type of columnar mucosa must be seen in a biopsy specimen from the lower esophagus to seal a diagnosis of Barrett’s mucosa [3, 4]. Eastern and Western pathologists—in spite of consensus meetings in Padua, Vienna, and other lovely venues—cannot decide what dysplasia is [5]. In addition to these geographic differences, locally codified and supported by venerable American, British, and Japanese professional associations [3, 6, 7], those supposedly united by common definitions are separated by an alarming degree of interobserver variability. In spite of Spechler’s partiality to pink [1], other American endoscopists may have very different perceptions of the color of a salmon. Three goblet cells at the squamocolumnar junction may be sufficient for a pathologist to declare Barrett’s esophagus and sentence the patient to a lifetime of surveillance and oncophobia, while a more prudent colleague would mention the very focal metaplasia without referring to the ‘B’ word. Since the 1980s, dysplasia in Barrett’s mucosa has been the object of innumerable studies based on the highly deceptive, but inexplicably revered, κ statistics, whose results can be cynically summarized as follows: there is good agreement on the absence of dysplasia and on the presence of high-grade dysplasia, but the categories ‘indefinite’ and ‘low-grade’ could just as well be decided by tossing dice [8, 9]. Yes, there is agreement on high-grade dysplasia, except that a Japanese and a German pathologist might call it cancer … [5, 10]. …