Background: Computed tomographic (CT) colonography, also called virtual colono scopy, is an evolving technology under evaluation as a new method of screening f or colorectal cancer. However, its performance as a test has varied widely acros s studies, and the reasons for these discrepancies are poorly defined. Purpose: To systematically review the test performance of CT colonography compared to col onoscopy or surgery and to assess variables that may affect test performance. Da ta Sources: The PubMed, MEDLINE, and EMBASE databases and the Cochrane Controlle d Trials Register were searched for Englishlanguage articles published between January 1975 and February 2005. Study Selection: Prospective studies of adults undergoing CT colonography after full bowel preparation, with colonoscopy or sur gery as the gold standard, were selected. Studies had to have used stateofth eart technology, including at least a singledetector CT scanner with supine an d prone positioning, insufflation of the colon with air or carbon dioxide, colli mation smaller than 5 mm, and both 2-dimensional and 3-dimensional views durin g scan interpretation. The evaluators of the colonogram had to be unaware of the findings from use of the gold standard test. Data Abstraction: Data on sensitiv ity and specificity overall and for detection of polyps less than 6 mm, 6 to 9 m m, and greater than 9 mm in size were abstracted. Sensitivities and specificitie s weighted by sample size were calculated, and heterogeneity was explored by usi ng stratified analyses and metaregression. Data Synthesis: 33 studies provided data on 6393 patients. The sensitivity of CT colonography was heterogeneous but improved as polyp size increased (48%[95%CI, 25%to 70%] for detection of p olyps< 6 mm, 70%[CI, 55%to 84%] for polyps 6 to 9 mm, and 85%[CI, 79%to 91%] for polyps >9 mm). Characteristics of the CT colonography scanner, including width of collimation, type of detector, a nd mode of imaging, explained some of this heterogeneity. In contrast, specifici ty was homogenous (92%[CI, 89%to 96%] for detection of polyps < 6 mm, 93% [CI , 91%to 95%] for polyps 6 to 9 mm, and 97%[CI, 96%to 97%] for polyps >9 mm) . Limitations: The studies differed widely, and the extractable variables explai ned only a small amount of the heterogeneity. In addition, only a few studies ex amined the newest CT colonography technology. Conclusions: Computed tomographic colonography is highly specific, but the range of reported sensitivities is wide . Patient or scanner characteristics do not fully account for this variability, but collimation, type of scanner, and mode of imaging explain some of the discre pancy. This heterogeneity raises concerns about consistency of performance and a bout technical variability. These issues must be resolved before CT colonography can be advocated for generalized screening for colorectal cancer. Background: Computed tomographic (CT) colonography, also called virtual colono scopy, is an evolving technology under evaluation as a new method of screening f or colorectal cancer. However, its performance as a test has varied covered acrosses studies, and the reasons for These discrepancies are poorly defined. Purpose: To systematically review the test performance of CT colonography compared to col onoscopy or surgery and to assess assess that that may affect test performance. Da ta Sources: The PubMed, MEDLINE, and EMBASE databases and the Cochrane Controlled d Trials Register were searched for English-language articles published between January 1975 and February 2005. Study Selection: Prospective studies of adults under CT colonography after full bowel preparation, with colonoscopy or sur gery as the gold standard, were selected. Studies had to have used stateofth eart technology, including at least one single detector CT scanner with supine an d prone positioning, insuffla tion of the colon with air or carbon dioxide, colli mation smaller than 5 mm, and both both 2-dimensional and 3-dimensional views durin g scan interpretation. The evaluators of the colonogram had to be unaware of the findings from use of the gold standard test. Data Abstraction: Data on sensitivity and specificity overall and for detection of polyps less than 6 mm, 6 to 9 mm, and greater than 9 mm in size were abstracted. Sensitivities and specificitie s weighted by sample size were calculated, and heterogeneity The sensitivity of CT colonography was heterogeneous but improved as polyp size increased (48% [95% CI, 25% to 70%]) was explored by usi ng stratified analyzes and meta regressions. Data Synthesis: 33 studies provided data on 6393 patients for detection of p olyps <6 mm, 70% [CI, 55% to 84%] for polyps 6 to 9 mm, and 85% [CI, 79% to 91%] for polyps> 9 mm). Characteristics of the CT colonography scanner, including width of collimation, type of detector, a nd mode of imaging, so so me of this93% [CI, 91% to 95%] for polyps 6 to 9 mm, and 97% [CI, 96% to 97%] for polyps> 9 mm). Limitations: The studies differed widely, and the extractable variables explai ned only a small amount of the heterogeneity. In addition, only a few studies ex amined the newest CT colonography technology. Conclusions: Computed tomographic colonography is highly specific, but the range of reported sensitivities is wide. Patient or scanner characteristics do not fully account for this variability, but collimation, type of scanner, and mode of imaging explain some of the discrepancies. This heterogeneity raises concerns about consistency of performance and a bout technical variability. These issues must be resolved before CT colonography can be advocated for generalized screening for colorectal cancer.