Inter-scan variability of coronary artery calcium scoring assessed on 64-multidetector computed tomography vs. dual-source computed tomography: a head-to-head comparison.

AIMS Coronary artery calcium (CAC) scoring has emerged as a tool for risk stratification and potentially for monitoring response to risk factor modification. Therefore, repeat measurements should provide robust results and low inter-scanner variability for allowing meaningful comparison. The purpose of this study was to investigate inter-scanner variability of CAC for Agatston, volume, and mass scores by head-to-head comparison using two different cardiac computed tomography scanners: 64-detector multislice CT (MSCT) and 64-slice dual-source CT (DSCT). METHODS AND RESULTS Thirty patients underwent CAC measurements on both 64-MSCT (GE LightSpeed XT scanner: 120 kV, 70 mAs, 2.5 mm slices) and 64-DSCT (Siemens Somatom Definition: 120 kV, 80 mAs, 3 mm slices) within <100 days (0-97). Retrospective intra-scan comparison revealed an excellent correlation. The excellent intra-scan (inter-observer) agreement was documented by narrow limits of agreement and a correlation coefficient of variation (COV) of r ≥ 0.99 (P < 0.001) for all CAC scores with a low COV for both scanners (64-MSCT/64-DSCT), i.e. Agatston (2.0/2.1%), mass (3.0/2.0%), and volume (4.7/3.9%). Inter-scanner comparison revealed larger Bland-Altman (BA) limits of agreement, despite high correlation (r ≥ 0.97) for all scores, with COV at 15.1, 21.6, and 44.9% for Agatston, mass, and volume scores. The largest BA limits were observed for volume scores (-1552.8 to 574.2), which was massively improved (-241.0 to 300.4, COV 11.5%) after reanalysing the 64-DSCT scans (Siemens) with GE software/workstation (while Siemens software/workstation does not allow cross-vendor analysis). Phantom measurements confirmed overestimation of volume scores by 'syngo Ca-Scoring' (Siemens) software which should therefore be reviewed (vendor has been notified). CONCLUSION Intra- and inter-scan agreement of CAC measurement in a given data set is excellent. Inter-scanner variability is reasonable, particularly for Agatston units in the clinically most relevant range <1000. The use of different software solutions has a greater influence particularly on volume scores than the use of different scanner types.