Aims: To identify causes of misinterpretation in second generation, dual-source coronary computed tomography angiography (CCTA).Methods: A retrospective re-interpretation was performed on 100 consecutive CCTA studies, previously performed with a 2×128 slice dual-source CT. Results were compared with coronary angiography (CA). CCTA and CA images were interpreted by 2 independent readers. At CCTA vessel diameter, image quality, plaque characteristics and localization (bifurcation vs. non) were described for all segments. Finally, aortic contrast-to-noise ratio (CNR) and the total Agatston calcium score were quantified. Agreement between CCTA and CA was assessed with the Kappa statistic after categorizing the stenosis severity at significant (≥50%) and critical (≥70%) cut-offs, and independent predictors of disagreement were determined by multivariable logistic regression, including patient characteristics such as body mass index (BMI), heart rate (HR), age and gender.Results: Per-segment sensitivity and specificity at ≥50% and ≥70% stenosis was of 83-95%, and 73-97%, respectively. There was a substantial agreement between CCTA and CA (kappa-50%=0.78, SE=0.03; kappa-70%=0.72, SE=0.03). Worse motion-related quality score, smaller vessel diameter, calcification within the segment of interest and LAD location were independent predictors of disagreement at 50% stenosis. The same factors, excluded LAD location, in addition to bifurcation-location of the coronary lesion predicted misdiagnosis at 70% stenosis. HR per se and BMI did not predict disagreement.Conclusion: According to the literature a substantial agreement between CCTA and CA was found. However, discrepancies exist and are mainly related with motion-related degradation of image quality, specific vessel anatomy and plaque characteristics. Awareness of such potential limitations may help guiding interpretation of CCTA.