Accuracy of computed tomography angiography to identify thin-cap fibroatheroma detected by optical coherence tomography

Background: Thin-cap fibroatheroma (TCFA) is assumed to cause acute coronary syndromes. Objective: To compare the accuracy of different models for diagnosing TCFA using parameters derived by CT, validated against optical coherence tomography (OCT). Methods: One hundred twenty-nine plaques in 106 patients were analyzed using data acquired by 64 row CT with a reconstruction thickness of 0.67 mm and an increment of 0.33 mm. TCFA was defined by OCT as a plaque with lipid content in >= 2 quadrants and the thinnest part of the fibrous cap measuring <= 65 gm. The following parameters were obtained from CT: remodeling index (RI), proportion of low attenuation (LA) volume (<60 HU), minimum CT number and napkin-ring sign (NRS). We compared three models to predict TCFA: Model 1, RI > 1.1, minimum CT number <30 HU and NRS; Model 2, RI > 1.1, minimum CT number <30 HU or NRS; Model 3, regression model using RI, proportion of LA volume and NRS. Results: In OCT, 83 plaques fulfilled the criteria of TCFA. The area under the receiver operating characteristics curve significantly (p < 0.01) increased to 0.96 (95% confidence interval (CI), 0.92-1.0) in model 3 as compared to models 1 (0.74, 95% CI, 0.68-0.80) and 2 (0.72, 95% CI, 0.67-0.79). Diagnostic accuracy of model 3 (93%) was significantly higher than that of models 1 (67%, p < 0.001) and 2 (80%, p = 0.001). Sensitivity and specificity of model 3 was 94% and 91%, respectively. Conclusion: Diagnostic performance to identify TCFA by coronary CTA improves when RI and proportion of LA volume are used as continuous values rather than dichotomizing these parameters. (C) 2017 Society of Cardiovascular Computed Tomography. Published by Elsevier Inc. All rights reserved.