Prospective blinded evaluation of computed tomographic colonography for screen detection of colorectal polyps

This study used a low lesion prevalence population reflective of the screening setting to estimate the sensitivity and specificity of computerized tomographic (CT) colonography for detection of colorectal polyps. This prospective, blinded study comprised 703 asymptomatic persons at higher-than-average risk for colorectal cancer who underwent CT colonography followed by same-day colonoscopy. Two of 3 experienced readers interpreted each CT colonography examination. Overall lesion prevalence for adenomas ≥1 cm in diameter was 5%. Seventy percent of all lesions were proximal to the descending colon. With colonoscopy serving as the gold standard, CT colonography detected 34%, 32%, 73%, and 63% of the 59 polyps ≥1 cm for readers 1, 2, 3, and double-reading, respectively; and 35%, 29%, 57%, and 54% of the 94 polyps 5–9 mm for readers 1, 2, 3, and double-reading, respectively. Specificity for CT colonography ranged from 95% to 98% and 86% to 95% for >1 cm and 5–9-mm polyps, respectively. Interobserver variability was high for CT colonography with κ statistic values ranging from −0.67 to 0.89. In a low prevalence setting, polyp detection rates at CT colonography are well below those at colonoscopy. These rates are less than previous reports based largely on high lesion prevalence cohorts. High interobserver variability warrants further investigation but may be due to the low prevalence of polyps in this cohort and the high impact on total sensitivity of each missed polyp. Specificity, based on large numbers, is high and exhibits excellent agreement among observers.