Background: It is difficult to decide which patients with reflux symptoms require endoscopy. The aim of this study was to develop a scoring system to predict esophageal findings at endoscopy.
Methods: A consecutive sample of 1011 adult patients scheduled for upper endoscopy were asked to complete a validated symptom questionnaire. The endoscopy reports were abstracted. Individual logistic regression models were developed to predict esophagitis, Barrett's esophagus (long and short segment) and esophageal stricture, including Schatzki's ring.
Results: Reflux esophagitis was independently associated with heartburn frequency (p<0.0001) but not severity or duration (p>0.05). Barrett's esophagus was associated with the duration of acid regurgitation (p<0.005) but not with frequency or severity (p>0.05). Strictures were associated with dysphagia severity (p<0.0001) and duration (p<0.0001) but not frequency (p>0.05). At a sensitivity of 80%, the models had a specificity of 49% for esophagitis, 57% for Barrett's esophagus, and 68% for strictures. At a specificity of 80%, the sensitivities were 51% for esophagitis, 62% for Barrett's esophagus and 71% for strictures.
Conclusions: Endoscopic findings were associated with distinct attributes of reflux symptoms. Symptoms are only modestly predictive of findings at endoscopy.