Objective To investigate the intra- and interobserver repeatability of coronary artery disease (CAD) diagnosis based on invasive coronary angiography (ICA) and CT coronary angiography (CTCA).Methods Two readers with comparable experience ( over 10 years) independently evaluated ICA results of 42 consecutive patients with a blind method. After 30 days,one of them reviewed the same patients again.Another two comparable-experience (over 10 years) readers evaluated the results of CTCA (prospectively ECG-triggering) from the same 42 patients in the same way.The inter-reader and intra-reader repeatability of ICA and CTCA were analyzed by performing Kappa test and calculating the percentage of the segments with agreement on stenotic degree.Using ICA as reference,the accuracy of CTCA in diagnosing CAD was studied by comparing the area under ROC. Results The Kappa between readers for ICA and CTCA were 0.91 and 0.81.Intra-reader Kappa were 0.92 and 0.83 respectively (x2 =509.4 and 432.5,all P ＜0.01 ).The percentage of the segments with agreement between readers on the degree of stenosis were 80.8％ (494/611) in ICA and 75.2％ (469/624) in CTCA ( x2 =2.75,P =0.10),and within the same reader,86.9％ (531/611)in ICA and 81.9％ (511/624) in CTCA(x2 =3.76,P =0.053).With≥ 50％narrowing as a CAD diagnosis criterion,the agreement rates for two readers were 96.6％ (590/611 ) in ICA and 94.4％ (589/624) in CTCA( x2 =3.36,P =0.07),and for the same reader,97.4％ (595/611) in ICA,95.4％ (595/624) in CTCA ( x2 =3.62,P =0.06).Using ICA as reference,two readers of CTCA results achieved a sensitivity and specificity of 84.9％ (530/624)and 98.1％ (612/624).The area under ROC was 0.94 (95％ CI 0.91-0.97).Conclusions Both ICA and CTCA demonstrate good repeatability in diagnosing CAD.The repeatability of ICA is superior to that of CTCA.A certain discrepancy exists in two readings from the same reader or two readers.