E1938. Revisiting CT Esophagography in the Emergency Setting: A Comparison of General and Expert Radiologists Diagnostic Accuracy
  1. Brad Evans; Maine Medical Center; University of Wisconsin
  2. Christina Cinelli; Maine Medical Center
  3. Sharon Siegel; Maine Medical Center
CT esophagography is now widely used to diagnose esophageal perforation and to assess extent of mediastinal contamination, often replacing fluoroscopic esophagography in the emergent setting. However, there is substantial variation in the interpretation of these studies, at least in part due to the relatively new technique and inexperience of some general radiologists, which has not been studied. This is not helped by the rarity of esophageal perforations. Our primary purpose is to compare the diagnostic accuracy of general and expert radiologist initial interpretations of chest CT and CT esophagography with fluoroscopy and final clinical diagnosis. The secondary purpose is to correlate CT findings to specific diagnoses to better understand the imaging spectrum of esophageal injuries.

Materials and Methods:
A retrospective study of patients presenting with suspected esophageal perforation from 2010 - 2020 was conducted. CT studies were reviewed by two experienced emergency radiologists who, independently and blinded to final diagnosis, identified key CT findings and categorized esophageal injuries as normal esophagus; abnormal esophagus without evidence of perforation; sealed/contained perforation; or free perforation. CT findings were correlated with final diagnosis. Initial general radiologist and expert consensus interpretations were compared with each other and with the gold standard fluoroscopic, endoscopic, surgical, and/or clinical diagnosis.

87 CT esophagograms and 69 chest CTs were performed for suspected esophageal perforation from 136 patients, with 46 confirmed perforations including sealed and free types. Most common CT features in esophageal perforation: esophageal wall thickening (98%), pneumomediastinum (91%), and mediastinal inflammation/fluid (83%). Both general and expert radiologists had high NPV for perforation compared to fluoroscopy when performed (94% and 100% respectively). CT esophagogram technique improved sensitivity for perforation in both the general group (76 to 79%) and the expert consensus group (86 to 90%) compared to chest CT without oral contrast. NPV improved from 86% in the general group to 95% in the expert group with fair interpretation agreement in chest CT (kappa 0.38) and moderate in CT esophagogram (kappa 0.42). Surprisingly, only fair agreement in the key finding of extra-luminal contrast (kappa 0.39), which was commonly confounded by streak artifact.

Both general and experienced radiologists had higher NPV for excluding esophageal perforation than fluoroscopy, both also seeing improved sensitivities reaching 90% using CT esophagogram technique over chest CT. General radiologists had lower accuracy diagnosing esophageal perforation on both chest CT and CT esophagogram compared to experienced emergency radiologists. We propose an imaging algorithm to evaluate suspected esophageal injuries as these share common CT features and recognition of these patterns can help categorize patients into important management pathways.