Reproducibility of computed tomography angiography data analysis using semiautomated plaque quantification software:...

Reproducibility of the quantitative assessment of atherosclerosis by computed tomography coronary angiography (CTCA) is paramount for the design of longitudinal studies. The purpose of this study was to assess the inter- and intra-observer reproducibility using semiautomated CT plaque analysis software in symptomatic individuals. CTCA was performed in 10 symptomatic patients after percutaneous treatment of the culprit lesions and was repeated after 3 years. The plaque quantitative analysis was performed in untreated vessels with mild-to-moderate atherosclerosis and included geometrical and compositional characteristics using semiautomated CT plaque analysis software. A total of 945 matched cross-sections from 21 segments were analyzed independently by a second reviewer to assess inter-observer variability; the first observer repeated all the analyses after 3 months to assess intra-observer variability. The observer variability was also compared to the absolute plaque changes detected over time. Agreement was evaluated by Bland-Altman analysis and concordance correlation coefficient. Inter-observer relative differences for lumen, vessel, plaque area and plaque burden were 1.2, 0.6, 2.2, 1.6% respectively. Intra-observer relative differences for lumen, vessel, plaque area and plaque burden were 1.0, 0.4, 0.2, 0.4% respectively. For the average plaque attenuation values the inter- and intra-observer variability was 5 and 2% respectively. For the % low-attenuation-plaque the inter- and intra-observer variability was 16 and 6% respectively. The absolute intra-observer variability for the plaque burden was 1.30 ± 1.09%, while the temporal plaque burden difference was 3.55 ± 3.02 % (p = 0.001). The present study shows that the geometrical assessment of coronary atherosclerosis by CTCA is highly reproducible within and between observers using semiautomated quantification software and that serial plaque changes can be detected beyond observer variability. The compositional measurements are more variable between observers than geometrical measurements.