We thank Drs. Napoli and Machan for their interest in our report on the diagnostic accuracy of 64-slice computed tomographic coronary angiography (CTA) in patients with intermediate pretest likelihood of coronary artery disease. They raise several issues regarding the analysis and interpretation of our data.

In our study, we compared the results of multislice CTA to those obtained during conventional coronary angiography at the patient, vessel, and segmental levels. We agree with the opinion of Drs. Napoli and Machan that the patient-based analysis may be most useful for physicians during routine clinical practice, because this information is most relevant for decisions regarding further diagnostic testing. Additional analyses were provided in our study to describe the relation between CTA and conventional coronary angiography at the vessel and segmental levels. Although less important than the patient-level data, the aim of these more detailed analyses was to describe the accuracy of CTA to quantify extent of disease (1-vessel vs 2- or 3-vessel disease) and its ability to correctly describe stenosis location. Indeed, these data may be valuable to guide diagnostic decision making. As Drs. Napoli and Machan point out, analysis at the vessel and segmental levels may be biased by intersubject correlation. Such a “clustering” problem may for instance occur when a positively scored segment results in more focused lesion detection in other vessels or segments in the same patient. We did not correct for this potential effect, because the purpose of our study was to determine the usefulness of CTA in a clinical setting, in which such corrections are not feasible.

In addition, Drs. Napoli and Machan comment on the small sample size of the population and the subsequent wide confidence intervals. Indeed, as we state in the discussion of our report, a limitation of our prospective study was the relatively small study population of patients with intermediate pretest likelihood for coronary artery disease. However, we considered it important to report on the diagnostic accuracy of CTA in this particular population; thus far, almost all previous diagnostic accuracy studies have been performed in populations with high pretest likelihood, while those with intermediate pretest likelihood are considered the target population for CTA. Importantly, we observed that also in this target population, the high negative predictive value of CTA was maintained. Accordingly, although we fully agree that larger studies are needed to confirm our findings, the initial data obtained in the study provide important insights regarding the implementation of CTA in this target population for noninvasive diagnostic imaging.