Coronary artery stenosis in high-risk patients: 64-section CT and coronary angiography--prospective study and analysis of discordance.

PURPOSE

To assess the diagnostic accuracy of multisection (64-section) computed tomography (CT) versus coronary angiography in detection of and assignment of grades for coronary artery stenoses in a high-risk population and to investigate causes for discordance between the two.

MATERIALS AND METHODS

The protocol was approved by the local ethics committee. Patients gave informed consent. The study included 114 patients (103 men, 11 women; mean age, 63 years +/- 8.2 [standard deviation]) with potential myocardial ischemia. Multisection CT images were interpreted independently by two radiologists with unequal experience in reading coronary CT angiograms. Diagnostic performance of 64-section CT in detection of stenoses of 50% or more was assessed per patient, per artery, and per segment. Interrater agreement was assessed by using the Cohen kappa coefficient. Agreement between 64-section CT and coronary angiography for assigning grades to stenoses was assessed by using Bland-Altman analysis.

RESULTS

Sixty-eight percent of patients had stenoses of 50% or more. Good interrater agreement was found, with kappa values of 0.77-0.85. For the most experienced radiologist, the sensitivity, specificity, positive likelihood ratio, and negative likelihood ratio were 73.4%, 95.0%, 14.7, and 0.28 per segment, 95.2%, 94.7%, 18.0, and 0.05 per artery, and 100%, 89.2%, 9.26, and zero per patient, respectively. Discordance between 64-section CT and coronary angiography was related to either under- or overestimation of the degree of stenosis, anatomic misclassification, and coronary artery segments that were not assessable at 64-section CT. Bland-Altman analysis showed poor agreement, especially for intermediate stenosis (mean bias, 1.3%; 95% limits of agreement: -27.3%, 29.9%).

CONCLUSION

Despite excellent sensitivity and negative likelihood ratios in a per-patient or per-vessel analysis, some coronary artery stenosis remained misdiagnosed with 64-section CT, resulting in limited sensitivity on a per-segment basis owing to anatomic discordance and failure to accurately quantify intermediate stenosis.