Accuracy of routine nontargeted CT without colonography technique for the detection of large colorectal polyps and cancer.

Abstract

To assess the accuracy of routine contrast-enhanced computed tomography for the detection of large colorectal polyps and cancer. The study group consisted of 100 patients who underwent abdominal computed tomography and optical colonoscopy evaluation before any treatment or intervention. Invasive colorectal carcinoma and large polyps (>or=10mm) were found at colonoscopy in 29 and 16 patients, respectively. Computed tomography studies with oral and IV contrast were retrospectively reviewed as 5-mm thick sections in standard soft tissue windows by 3 readers. The readers scored each segment for polyp or cancer utilizing a 5-point scale for receiver operating characteristic analysis. By-patient sensitivity/specificity for cancer detection for each reader was 69.0%/88.7%, 65.5%/84.5%, and 82.8%/77.5%, respectively. Pooled sensitivity, specificity, and accuracy for cancer was 72.4%, 83.6%, and 80.3%, respectively. The empiric area under the receiver operating characteristic curves for the readers ranged from 0.775 to 0.857. By-patient sensitivity for large polyps was poor with a pooled sensitivity of 14.5%. Standard computed tomography not tailored to colorectal investigation was moderately effective for detecting invasive cancers, but insensitive for large polyps. Unsuspected cancer should be a part of the standard search pattern for routine computed tomography interpretation, but more advanced colonography techniques are necessary for detecting large polyps.