Published studies have reported a wide range of sensitivities and specificities for computed tomographic (CT) colonography for polyp detection, generating controversy regarding its diagnostic accuracy.

A meta-analysis of published studies comparing the accuracies of CT colonography and colonoscopy for polyp detection was performed. The pooled per-patient sensitivities and specificities were calculated at various thresholds for polyp size. Summary receiver operating characteristic (sROC) curves were also constructed.

Thirty studies were included in the meta-analysis of CT colonography. The pooled per-patient sensitivity of CT colonography was higher for polyps greater than 10 mm (0.82, 95% confidence interval [CI], 0.76-0.88) compared with polyps 6 to 10 mm (0.63, 95% CI, 0.52-0.75) and polyps 0 to 5 mm (0.56, 95% CI, 0.42-0.70). Similarly, the exact area under the sROC curve (area ± standard error) was higher using a threshold greater than 10 mm (0.898 ± 0.063) compared with thresholds of greater than 5 mm and any size (0.884 ± 0.033 and 0.822 ± 0.059, respectively). There were no significant differences in the diagnostic characteristics of 2-dimensional versus 3-dimensional CT colonography. At a threshold greater than 5 mm, the exact area under the sROC curve was significantly higher for endoscopic colonoscopy compared with CT colonography (0.998 ± 0.006 vs 0.884 ± 0.033, P < .005).

CT colonography has a reasonable sensitivity and specificity for detecting large polyps but was less accurate than endoscopic colonoscopy for smaller polyps. Thus, CT colonography may not be a reasonable alternative in situations in which a small polyp may be clinically relevant.