Introduction 1 2 3 4 5 7 8 9 10 11 We set out to study inter-scan reproducibility of coronary calcium measurements from MDCT images and to evaluate whether reproducibility is affected by different measurement protocols, slice thickness, selected cardiovascular risk factors and technical variables. Materials and methods 12 2 TM Coronary imaging and calcium measurements  The amount of calcium in the coronary arteries was assessed with a Multi Detector-Row CT (MDCT) scanner (Mx 8000 IDT 16, Philips Medical Systems, Best, The Netherlands). Subjects were positioned within the gantry of the MDCT scanner in supine position. During a single breath hold, images of the heart, from the level of the tracheal bifurcation to below the base of the heart, were acquired using prospective ECG triggering at 50–80% of the RR-interval, depending on the heart rate. Scan parameters were 16 × 1.5 mm collimation, 205 mm field of view (FOV), 0.42 s rotation time, 0.28 s scan time per table position, 120 kVp and 40–70 mAs (patient weight <70 kg: 40 mAs; 70–90 kg: 55 mAs; >90 kg: 70 mAs). Scan duration was approximately 10 s, depending on heart rate and patient size. We had the participant get up from the table and lay down again since in studies on change in CAC over one year it is not realistic to assume exactly the same position of the participant at both occasions. Therefore our patients sat up and consequently moved slightly between scans to mimic two separate scan runs. From the acquired raw data, the whole volume was reconstructed with an intermediate reconstruction algorithm in non-overlapping data sets of 1.5 mm and 3 mm thick sections. Quantification of coronary calcium was performed on a separate workstation with software for calcium scoring (Heartbeat-CS, EBW, Philips Medical Systems, Best, The Netherlands). All regions with a density over 130 Hounsfield units were identified as potential calcifications. 2 2 13 14 2 Data analysis The mean and standard deviations (SD) of coronary calcium were calculated for all scoring methods separately. Because of the skewed distribution of scores, medians were also computed. The Intra-class correlation coefficient was estimated for between scans data and for 1.5 and 3.0 mm slices thicknesses separately. The mean difference in score between scans was calculated as well as the absolute and relative differences. To distinguish between random differences or systematic difference, information on mean and absolute differences is needed. One may assume a priori a non-differential misclassification in the calcium scores, but one has to show that with the results. When the chance of the 2nd result being higher or lower is equal, one would expect a mean difference of zero, with some standard deviation. The absolute difference will not be zero since all differences are ‘absolutised’, but it is expected that at least the mean difference is much less than the absolute difference. If however the chance of a higher or lower value in the 2nd scan is not equal, the mean difference will be plus or minus a certain value. In addition, the absolute difference will have a value close to that of the mean difference. Therefore we need both parameters. 15 16 P- Results 1 Table 1 N  Mean Std. deviation Age (year) 67.3 5.2 2 26.3 3.9 WHR 0.84 0.06 SBP (mmHg) 133.9 18.9 DBP (mmHg) 71.7 9.1 Total cholesterol (mmol/l) 6.09 0.86 LDL cholesterol (mmol/l) 4.31 0.97 HDL cholesterol (mmol/l) 1.51 0.36 Triglycerides (mmol/l) 1.28 0.62 Glucose (mmol/l)  4.05 0.69 Heart rate (beat/minute) 72 11 Current smoking (%)* 11  Former smoking (%) 43  Previous CVD (%) 1  Family history of CAD in either parents (%) 10  BMI = Body Mass Index; CAD = Coronary Artery Diseases; DBP = Diastolic Blood Pressure; LDL = Low Density Lipoprotein; HDL = High Density Lipoprotein; SBP = Systolic Blood Pressure; WHR = Waist to Hip Ratio * Percentages have been rounded 2 2 Table 2 Characteristics of different coronary calcium scoring methods; effect of slice thickness on inter-scan reproducibility  Mass 1st Scan Mass 2nd Scan  Volume 1st Scan  Volume 2nd Scan  Agatston 1st Scan  Agatston 2nd Scan  Slice thickness 1.5 mm Mean 32.21 31.88 154.52 149.40 170.33 163.63 Median 6.15 6.05 39.97 36.52 31.85 32.00 Agreement (k) Rumberger categories 0.97 0.89 0.87 Agreement (k) Quartiles 0.84 0.81 0.88 Mean difference 0.3 5.1 6.7 Absolute difference 4.0 22.3 24.3 Relative difference (%) 12.4 14.6 14.5 ICCC* 0.99 0.99 0.98 Slice thickness 3.0 mm Mean 25.57 25.45 131.45 126.98 140.06 135.82 Median 4.00 3.65 30.30 21.90 20.30 18.00 Agreement (k) Rumberger categories 0.92 0.83 0.73 Agreement (k) Quartiles 0.84 0.84 0.84 Mean difference 0.1 4.4 4.2 Absolute difference 3.5 18.7 21.3 Relative difference (%) 13.7 14.7 15.4 ICCC* 0.99 0.98 0.98 * Intra-class correlation coefficient 3 4 1 2 Table 3 Relationship between cardiovascular risk factors and inter-scan mean difference of coronary calcium scoring methods by MDCT (Slice thickness 1.5 mm)  Inter-scan mean difference CCS methods Mass Volume Agatston Biological variables r P r P r P 2 0.04 0.73 0.03 0.74 0.02 0.80 Age (year) 0.18 0.10 0.31 0.00 0.28 0.01 Smoking(Categorical) −0.00 0.98 0.04 0.71 0.07 0.49 WHR −0.03 0.73 0.08 0.48 0.13 0.24 SBP (mmHg) 0.10 0.37 0.16 0.14 0.24 0.03 DBP (mmHg) 0.16 0.14 0.05 0.61 0.11 0.34 Cholesterol (mmol/l) −0.27 0.05 −0.12 0.40 −0.20 0.17 LDL (mmol/l) −0.18 0.10 −0.19 0.09 −0.09 0.40 HDL (mmol/l) −0.04 0.72 −0.16 0.14 −0.11 0.34 Triglyceride (mmol/l) −0.02 0.85 0.13 0.24 0.11 0.34 Glucose (mmol/l) 0.16 0.24 −0.00 0.98 0.00 0.98 Mean heart rate −0.03 0.77 −0.03 0.73 −0.02 0.81 Technical variables Mean breathing artifact 0.01 0.88 −0.03 0.78 −0.02 0.87 Mean SD of noise 0.13 0.26 0.08 0.49 0.07 0.52 Coronary calcium Mean mass score 0.00 0.98     Mean volume score   0.03 0.75   Mean Agatston score     0.02 0.86 Mean log mass score 0.00 0.99     Mean log volume score   0.03 0.76   Mean log Agatston score     0.02 0.85 BMI = Body Mass Index; DBP = Diastolic Blood Pressure; LDL = Low Density Lipoprotein; HDL = High Density Lipoprotein; r = spearman correlation coefficient; SBP = Systolic Blood Pressure; WHR = Waist to Hip Ratio Table 4 Relationship between cardiovascular risk factors and inter-scan absolute and relative difference of coronary calcium scoring methods by MDCT (Slice thickness 1.5 mm)  Inter-scan relative difference CCS methods Mass Volume Agatston Biological variables r P r P r P 2 0.07 0.53 0.08 0.46 0.09 0.43 Age (year) 0.21 0.06 0.24 0.03 0.l17 0.12 Smoking(Categorical) −0.03 0.73 −0.07 0.51 -0.14 0.20 WHR 0.07 0.55 0.05 0.66 0.05 0.66 SBP (mmHg) 0.06 0.57 0.04 0.68 0.11 0.32 DBP (mmHg) 0.32 0.004 0.31 0.005 0.33 0.003 Cholesterol (mmol/l) 0.13 0.37 0.10 0.50 0.00 1.00 LDL (mmol/l) −0.14 0.21 −0.17 0.12 -0.18 0.11 HDL (mmol/l) 0.07 0.52 0.04 0.67 0.06 0.57 Triglyceride (mmol/l) 0.03 0.78 0.07 0.49 0.00 0.99 Glucose (mmol/l) 0.23 0.09 0.26 0.05 0.24 0.08 Mean heart rate −0.01 0.91 0.01 0.93 0.00 0.97 Technical variables Mean breathing artifact 0.10 0.44 0.09 0.49 0.15 0.23 Mean SD of noise 0.19 0.09 0.19 0.09 0.18 0.11 Coronary calcium Mean mass score 0.29 0.009     Mean volume score   0.33 0.003   Mean Agatston score     0.38 0.001 Mean log mass score 0.29 0.010     Mean log volume score   0.33 0.003   Mean log Agatston score     0.37 0.001  Inter-scan absolute difference Technical variables Mean breathing artifact 0.12 0.32 0.12 0.33 0.15 0.22 Mean SD of noise 0.20 0.08 0.19 0.09 0.15 0.17 Coronary calcium Mean mass score 0.86 <0.001     Mean volume score   0.84 <0.001   Mean Agatston score     0.89 <0.001 Mean log mass score 0.86 <0.001     Mean log volume score   0.83 <0.001   Mean log Agatston score     0.89 <0.001 BMI = Body Mass Index; DBP = Diastolic Blood Pressure; LDL = Low Density Lipoprotein; HDL = High Density Lipoprotein; r = spearman correlation coefficient; SBP = Systolic Blood Pressure; WHR = Waist to Hip Ratio Fig. 1 Relation between mean calcium score and inter-scan difference in mean calcium scores (Bland-Altman plots)  Fig. 2 Relation between mean calcium score and inter-scan absolute difference  Discussion With respect to ranking of subjects, the inter-scan reproducibility of coronary calcium measurements by MDCT using Agatston, volume and mass scoring algorithms is excellent. The inter-scan reproducibility showed no major differences between scoring methods. The slice thickness did not affect reproducibility, nor did heart rate and technical parameters. Measurement error was related to increased coronary artery calcification, although our findings suggest that the error in the measurements is a random phenomenon. 17 18 19 20 17 The implications of our main findings depend on the research question that is asked in studies using CAC measurements. When the interest is using CAC measurements for prognostic studies our results for kappa and ICCC show that ranking of subjects is adequate based on one CT scan. So the need for duplicate CAC scan is absent. The fact that measurement error increases with increasing CAC values, is in prognostic studies not of major importance since the categorization of individuals seems adequate. When the interest is in etiologic studies using CAC as outcome parameter, our findings show that risk factor relations will be validly estimated since none of the risk factors relates to measurement error. When the interest is in using CAC as risk factor for future events (assessment of relative risks), it is most likely that in analyses with CAC as continues variable the magnitude of association of high CAC levels with events reflects an underestimation of the true magnitude. The direction of the relation will not change since based on our results measurement error is random, leading to random misclassification of the exposure variable. When the interest is in diagnostic value of CAC measurements, which is usually done in categories of CAC, again the relations will be valid given our high kappa coefficients. Although our study was performed in healthy postmenopausal women, we expect that the finding will also be applicable for men. Our findings are important in the light of the wider availability of MDCT in countries compared to EBCT. One reason for that is lower equipment cost. Other advantages of MDCT over EBCT have been suggested to be less quantum noise, thinner section thickness, and simultaneous acquisition of four sections (with 16-slice or with 64-slice ), which is reported to reduce misregistration artifact. In conclusion, our findings demonstrate that coronary calcium measurements by MDCT are highly reproducible and are not affected by scoring protocols, slice thicknesses and technical factors.