Gastroenterologists' interpretation of CTC: a pilot study demonstrating feasibility and similar accuracy compared with radiologists' interpretation.

Objectives: Computed tomography colonography (CTC) is an emerging colon cancer screening modality that has the potential to increase adherence to current screening recommendations. Traditionally, the interpretation of CTC has been limited to radiologists. As the technology of CTC has developed, three-dimensional endoluminal fly-through images have largely replaced two-dimensional CT images as the primary reading modality. Such a display is a realistic corollary to the endoscopic view obtained during colonoscopy. Our study sought to determine whether gastroenterologists could interpret the colonic display of CTC with an accuracy similar to that of trained radiologists.

Methods: Three board-certified gastroenterologists and four gastroenterology fellows in various stages of training interpreted a mean of 45 CTCs (range: 30-50) in which colonoscopy had also been performed. Before reading any cases, each reader underwent CTC interpretation training with an experienced CTC radiologist. After interpreting each CTC, the gastroenterologist had access to both the original radiology interpretation of the CTC and the corresponding colonoscopy results. Outcomes included accuracy of the gastroenterologists' interpretation, time required for CTC interpretation, evidence of learning, and the level of diagnostic agreement between gastroenterologists and radiologists.

Results: Gastroenterologist readers identified polyps >or=6 mm on CTC with a mean sensitivity and specificity of 83.5% (67-100%) and 78.8% (69-100%), respectively. Corresponding values for polyps >or=8 mm were 83.8% (68-100%) and 74% (30-93%), respectively, and those for polyps >or=10 mm were 87.8% (67-100%) and 85.2% (60-94%), respectively. Overall, 83% (5 of 6) of gastroenterologists achieved kappa scores >or=0.60, suggesting good agreement with radiologists; 66% achieved kappa>or=0.75. There was a direct relationship between diagnostic accuracy and level of gastroenterology training, with third-year fellows being nearly as accurate as the attendings. The average gastroenterologist CTC reading time was 18.4 min (range: 11.2-25.6).

Conclusions: The gastroenterologists in this study were able to read CTCs with an accuracy that approaches that of radiologists. The level of training affected the accuracy of CTC interpretation by the gastroenterologist. Average gastroenterologist CTC interpretation times in this study were similar to recommended colonoscopy times. Further studies are warranted to determine whether gastroenterologists are able to interpret CTCs independently in clinical practice.