Background & aims: Virtual colonoscopy using abdominal spiral computed tomography scanning allows total colonic evaluation with minimal invasiveness. Two-dimensional images and selective 3-dimensional images of the colon are used to detect colorectal lesions. This trial used conventional colonoscopy with segmental unblinding to determine the ability of virtual colonoscopy to identify patients with colorectal lesions who need conventional colonoscopy.
Methods: We studied 205 patients with virtual colonoscopy using oral iodinated contrast preceding conventional colonoscopy. Colonic lavage was achieved with an oral sodium phosphosoda preparation and colonic distention with a carbon dioxide electronic insufflator.
Results: The overall sensitivity and specificity of virtual colonoscopy in identifying patients with colorectal lesions was 61.8% and 70.7%, respectively. Virtual colonoscopy was more accurate in identifying patients with lesions >/=6 mm (sensitivity 84.4% and specificity 83.1%) and those with lesions >/=10 mm (sensitivity 90% and specificity 94.6%). The negative predictive value of virtual colonoscopy was 95% for a 6-mm cutoff size and 98.9% for a 10-mm cutoff. Using a 10-mm cutoff, virtual colonoscopy precludes the need for conventional colonoscopy in 86% of patients with a 1% false-negative rate (68% with a 3.4% false-negative rate when using a 6-mm cutoff).
Conclusions: Virtual colonoscopy has a high sensitivity and specificity for detecting patients with significant colorectal lesions. Its high negative predictive value may help reduce the number of negative screening colonoscopies. Further studies are needed to determine what lesion cutoff size is clinically acceptable and the appropriate interval time for repeat virtual colonoscopy when it detects lesions below this cutoff size.