Coronary computed tomographic angiography in patients suspected of coronary artery disease: Impact of observer experience on diagnostic performance and interobserver reproducibility

Abstract

A high diagnostic performance of coronary computed tomographic angiography (CTA) in identifying coronary artery disease (CAD) has been shown in experienced high-volume centers. Whether this may be accomplished in centers with less CTA experience remains unknown. We determined the diagnostic performance and interobserver reproducibility of CTA in detecting significant CAD in a center with limited experience. In 209 patients, CTA was performed with 64-slice or dual-source CT technology, and analyses were performed independently by 2 inexperienced observers. Significant CAD by CTA was defined as >/=1 stenoses >/=50% or >/=1 nonevaluable segment, whereas significant CAD by invasive quantitative coronary angiography was defined as >/=1 stenoses >/=50%. We evaluated the influence of CAD pretest probability, Agatston score (AS), heart rate (HR), and observer experience on the diagnostic sensitivity, specificity, positive (PPV) and negative predictive values (NPV), interobserver reproducibility, and duration of CTA analysis. Per-patient (CAD prevalence, 35%) sensitivity was 88%-99%, specificity was 78%-82%, PPV was 68%-74%, and NPV was 92%-99%. Overall interobserver reproducibility was good (kappa = 0.65). A significant temporal improvement was observed in diagnostic specificity (observer A: 68%-89%, P = 0.007; observer B: 71%-89%, P = 0.02), and interobserver reproducibility (kappa = 0.35-0.89, P = 0.01) during the study period. Duration of analysis decreased during the study period and was positively associated with CAD pretest probability and AS. Suboptimal diagnostic performance and interobserver reproducibility must be anticipated during CTA implementation. A high diagnostic sensitivity, specificity, and interobserver reproducibility were achieved after a large number of studies performed with the state-of-the-art scanner technology.