Twenty eight studies (3,674 patients; range 51 to 402) were included in the review. Nine studies were assessed as Level II diagnostic evidence, 15 as Level III-1 evidence and four as Level III-2 evidence. Regardless of the order of testing, all studies performed invasive coronary angiography independently of CT angiography. In all but two studies, CT angiography results were interpreted blind to the results of the reference standard. Nine recruited patients consecutively and there were two retrospective studies.
Patient level: The initial meta-analysis, which omitted studies that excluded equivocal test results, gave sensitivity estimates of 98.2% (95% CI 97.4 to 98.8%) and specificity estimates of 81.6% (95% CI 79.0 to 84.0%), 18 studies. Secondary analyses showed little difference in results.
Vessel level: The overall vessel level analyses, which omitted studies that excluded equivocal test results, showed similar diagnostic performance to that of the patient-level analysis. The sensitivity estimates were 94.9% (95% CI 93.9 to 95.8%) and specificity estimates were 89.5% (95% CI 88.8 to 90.2 %), 17 studies. Specificity was slightly higher in the left main artery 97.1% (95% CI 95.7 to 98.1%) and slightly lower in the left anterior descending artery 84.5% (95% CI 82.1 to 86.7%) compared with the other arteries.
Segment level: The overall segment level analysis gave sensitivity estimates of 91.3% (95% CI 90.2 to 92.2%) and specificity estimates of 94.0 (95% CI 93.7 to 94.2%), 17 studies.
Positive predictive values, negative predictive values and accuracy results were also reported.