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 μm. 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.