Purpose: To compare the measurement variability for coronary artery calcium (CAC) measurements using mineral mass compared with a modified Agatston score (AS) or volume score (VS) with multi-detector CT (MDCT) scanning, and to estimate the potential impact of these methods on the design of CAC progression studies.
Materials and methods: We studied 162 consecutive subjects (83 women, 79 men, mean age 51 +/- 11 years) from a general Caucasian community-based cohort (Framingham Heart Study) with duplicate runs of prospective electrocardiographically-triggered MDCT scanning. Each scan was independently evaluated for the presence of CAC by four experienced observers who determined a "modified" AS, VS and mineral mass.
Results: Of the 162 subjects, CAC was detected in both scans in 69 (42%) and no CAC was detected in either scan in 72 (45%). Calcium scores were low in the 21/162 subjects (12%) for whom CAC was present in one but not the other scan (modified AS < 20 in 20/21 subjects, mean AS 4.6 +/- 1.9). For all three quantification algorithms, the inter- and intraobserver correlation were excellent (r > 0.96). However, the mean interscan variability was significantly different between mineral mass, modified AS, and VS (coefficient of variation 26 +/- 19%, 41 +/- 28% and 34 +/- 25%, respectively; p < 0.04), with significantly smaller mean differences in pair-wise comparisons for mineral mass compared with modified AS (p < 0.002) or with VS (p < 0.03). The amount of CAC but not heart rate was an independent predictor of interscan variability (r = -0.638, -0.614 and -0.577 for AS, VS, and mineral mass, respectively; all p < 0.0001). The decreased interscan variability of mineral mass would allow a sample size reduction of 5.5% compared with modified AS for observational studies of CAC progression and for randomized clinical trials.
Conclusion: There is significantly reduced interscan variability of CAC measurements with mineral mass compared with the modified AS or VS. However, the measurement variability of all quantification methods is predicted by the amount of CAC and is inversely correlated to the extent of partial volume artifacts. Moreover, the improvement of measurement reproducibility leads to a modest reduction in sample size for observational epidemiological studies or randomized clinical trials to assess the progression of CAC.