Carotid artery stenosis: accuracy of noninvasive tests--individual patient data meta-analysis

To find clinically relevant estimates of the accuracy of noninvasive imaging-Doppler ultrasonography (US), computed tomographic (CT) angiography, magnetic resonance (MR) angiography, and contrast material-enhanced MR angiography-in diagnosing both severe and moderate symptomatic carotid artery stenosis; to ascertain the effect of prespecified clinical factors and clinical setting on diagnostic accuracy; and to estimate the probability of agreement between two noninvasive tests.

Original principal investigators obtained ethics approval for each data set. All data were anonymized. Individual patient data sets (IPDs) for noninvasive imaging tests were used to determine sensitivity, specificity, and agreement between the tests for symptomatic carotid artery stenosis; to compare ipsilateral with contralateral arteries; to compare IPDs with literature estimates; to compare routine audit and research data; and to determine the effect of age and sex on sensitivity and specificity.

Contrast-enhanced MR angiography was the most accurate (sensitivity, 0.85 [30 of 35]; 95% confidence interval [CI]: 0.69, 0.93; and specificity, 0.85 [67 of 78]; 95% CI: 0.76, 0.92) for 70%-99% symptomatic stenosis. Sensitivity for a 50%-69% stenosis was poor, although data were limited. Sensitivity and specificity were generally lower in the ipsilateral than in the contralateral artery. IPD estimates were lower than literature values. Results of comparison of research with audit-derived data were inconclusive. Neither age nor sex affected accuracy. Agreement was better between two Doppler US tests and between two contrast-enhanced MR angiographic tests than it was between Doppler US and contrast-enhanced MR angiography, except for a 70%-99% symptomatic stenosis.

Primary studies should distinguish ipsilateral from contralateral arteries and carefully describe the patients' characteristics and study environment. The literature overestimates noninvasive imaging accuracy. More data are needed to inform physicians in routine clinical practice.