The importance of noninvasive testing of the carotid arteries continues to increase with the recent reports of the Asymptomatic Carotid Atherosclerosis Study (ACAS) [1] and the North American Symptomatic Carotid Endarterectomy Trial (NASCET) [2]. The meta-analysis by Blakeley and colleagues [3] contrasting the sensitivity and specificity of noninvasive techniques provides information needed by clinicians in their evaluations of patient populations.

Unfortunately, the brief mention of "publication bias" as a shortcoming of this report may lead clinician-readers to underestimate this problem. Because laboratories finding a poor correlation between noninvasive evaluations and arteriography are unlikely to publish their results, the predictive value of noninvasive testing in these cases will be substantially overestimated.

As part of the quality-control efforts in the ACAS, each of the clinical centers was required to submit the results from 50 patients who had both noninvasive Doppler evaluation and arteriography. These data offered the unique opportunity to describe the distribution of the performance of noninvasive testing across a spectrum of ultrasound laboratories [4]. Despite recognized quality of the ACAS medical centers, the variation in the magnitude of association between noninvasive evaluation and arteriography between the centers was remarkable. Data from 9 of the 33 comparisons (27%) showed a strong relation (> 60% sensitivity at a 90% positive predictive value) between the noninvasive evaluation and arteriography. These excellent results support the results of Blakeley and coauthors. In 18 comparisons (55%), however, the relation was not as strong (< 60% sensitivity at a 90% positive predictive value), and 6 comparisons showed no statistical relation between the results of noninvasive evaluation and arteriography. As a result of these analyses, ACAS required that laboratories showing a weak relation between noninvasive evaluation and arteriography make substantial changes in evaluation methods or prohibited those centers from entering patients into ACAS on the basis of noninvasive results.

These results have direct implications for the meta-analysis by Blakeley and coworkers. If these 33 comparisons had been made independently, the data from the 9 comparisons showing a good relation would probably have been submitted and accepted for publication. In most laboratories showing a poor association, however, the results would probably not have been submitted and would be even less likely to be accepted for publication. Only because of the unique opportunity to assess the noninvasive-arteriographic relation across various medical centers were such diverse results publishable. We encourage readers to exercise great caution in generalizing published results to their local sites, and we encourage users of laboratories employing noninvasive techniques to insist on a high level of quality control when data from local sonographers and equipment are being used. Although noninvasive techniques can provide highly reliable and valid data, the need for local quality control is paramount.