Twenty-five studies (2,508 participants, range 30 to 444) were included in the review. Differential and partial verification biases and incorporation bias were avoided in all studies. Eighty per cent of studies reported blinded interpretation of the index test and reference standard.
The pooled sensitivity for per patient detection of significant coronary artery disease was 99% (95% CI 97% to 99%) and the pooled specificity was 89% (95% CI 84% to 92%) based on data from 24 studies. Use of prospectively ECG-triggered scanning was associated with a significant increase in specificity (p=0.01); no other covariate showed a significant effect.
The pooled sensitivity for per segment detection of significant coronary artery disease was 94% (95% CI 92% to 96%) and the pooled specificity was 97% (95% CI 96% to 98%), based on data from 24 studies. None of the covariates investigated showed a significant effect on sensitivity or specificity.
Similar sensitivity and specificity estimates were reported for assessment of specific arteries (left anterior descending, left circumflex, left main and right coronary artery).
Values were also reported for positive and negative likelihood ratios and diagnostic odds ratios.