Test Performance of CT Colonography: A Systematic Review of the Literature and Meta-Analysis

Abstract

Background: Computed Tomography Colonography (CTC) is being proposed as a colorectal cancer screening modality that may improve screening adherence rates. The purpose of this study was to systematically review the test performance of CTC. Methods: We searched PubMed, Medline, EMBASE and Cochrane Controlled Trials Register for articles published in English language between January 1975 and October 2004 with the following MeSH terms or text words: virtual colonoscopy, CT colonography, CT colography or CT pneumocolon. We included only prospective blinded studies of adults undergoing CT colonography after full bowel preparation, using colonoscopy or surgery as the gold standard. Further, these studies had to use state-of-the-art technology with at least a single-detector CT scanner, both supine and prone positioning, air or CO2 colon insufflation of the colon, CT collimation < 5 mm, and both 2D & 3D views during interpretation. The data extracted included the overall sensitivity and specificity, as well as sensitivity and specificity for polyps <6mm, 6-9 mm and > 9mm in size. When data was sufficiently homogenous, we calculated weighted (by sample size) sensitivities and specificities. Heterogeneity was examined using stratified analyses and meta-regression. Results: A total of 31 studies were included that provided data from 5731 patients. There was a wide range of CTC sensitivities on a per patient analysis (21%-96%) that prevented pooling, even for polyps > 9 mm (range: 48%-100%). This heterogeneity was not explained by year published, demographics or cancer risk, imaging software used or CT reconstruction interval. Of note, collimation intervals, type of scanner and mode of imaging used did have an impact on heterogeneity. Data for specificity was homogenous with CTC being 92% specific (95% CI: 0.91-0.94) for polyps 6-9mm and 96% (95% CI: 0.96-0.97) for polyps > 9 mm. Discussion: While CTC is highly specific, there is a broad range of sensitivities that is not clearly related to patient or scanner characteristics--though collimation, type of scanner and mode of imaging do explain some of the disparity. The substantial heterogeneity underlines the concerns that CTC is not yet ready for widespread application. Experts may have valuable insights into CTC technique, but this study defines some of the specific technical issues that have impacted optimal test performance and supports the assertion that further scrutiny of CT Colonography is required. Background: Computed Tomography Colonography (CTC) is being proposed as a colorectal cancer screening modality that may improve screening adherence rates. The purpose of this study was to systematically review the test performance of CTC. Methods: We searched PubMed, Medline, EMBASE and Cochrane Controlled Trials Register for articles published in English language between January 1975 and October 2004 with the following MeSH terms or text words: virtual colonoscopy, CT colonography, CT colography or CT pneumocolon. We included only prospective blinded studies of adults undergoing CT colonography after full bowel preparation, using colonoscopy or surgery as the gold standard. Further, these studies had to use state-of-the-art technology with at least a single-detector CT scanner, both supine and prone positioning, air or CO2 colon insufflation of the colon, CT collimation < 5 mm, and both 2D & 3D views during interpretation. The data extracted included the overall sensitivity and specificity, as well as sensitivity and specificity for polyps <6mm, 6-9 mm and > 9mm in size. When data was sufficiently homogenous, we calculated weighted (by sample size) sensitivities and specificities. Heterogeneity was examined using stratified analyses and meta-regression. Results: A total of 31 studies were included that provided data from 5731 patients. There was a wide range of CTC sensitivities on a per patient analysis (21%-96%) that prevented pooling, even for polyps > 9 mm (range: 48%-100%). This heterogeneity was not explained by year published, demographics or cancer risk, imaging software used or CT reconstruction interval. Of note, collimation intervals, type of scanner and mode of imaging used did have an impact on heterogeneity. Data for specificity was homogenous with CTC being 92% specific (95% CI: 0.91-0.94) for polyps 6-9mm and 96% (95% CI: 0.96-0.97) for polyps > 9 mm. Discussion: While CTC is highly specific, there is a broad range of sensitivities that is not clearly related to patient or scanner characteristics--though collimation, type of scanner and mode of imaging do explain some of the disparity. The substantial heterogeneity underlines the concerns that CTC is not yet ready for widespread application. Experts may have valuable insights into CTC technique, but this study defines some of the specific technical issues that have impacted optimal test performance and supports the assertion that further scrutiny of CT Colonography is required.