Evaluating changes in coronary artery calcium: An analytic method that accounts for interscan variability

OBJECTIVE. Coronary calcium measured by electron beam CT is associated with coronary disease and can be used to predict coronary disease events. Studies investigating changes in coronary calcium need to address interscan variability as it relates to the overall coronary calcium score in defining progression or regression of coronary calcium over time. SUBJECTS AND METHODS. Electron beam CT was performed on 1,074 participants. Coronary calcium volume scores were repeated 5 min apart. Interscan variability was examined using Bland-Altman plots and homogeneity tests. Transformations of the Box-Cox family (including power, roots, and logarithm) were applied to calcium volume scores. The transformation that stabilized the variation in calcium volume scores was applied to progression of calcium volume scores in 109 subjects with diabetes. RESULTS. The variability in calcium volume score increased as the level of coronary calcium increased (rho = 0.67, p < 0.001 for the relation between the absolute difference and the mean value of calcium volume scores). This heterogeneity was removed using the square root transformation of the calcium volume score (rho = 0.09, p < 0. 15 for the relation between the absolute difference in the square root of the calcium volume score and the mean square root of the calcium volume score). This transformation was applied to calcium volume scores taken a mean of 2.7 years apart in 109 subjects with diabetes. A significant change in calcium volume score was defined as a difference between the square root-transformed to calcium volume scores greater than or equal to 2.5 mm(3) (> 99th percentile of interscan variability). Significant progression was observed in 10% of the subjects. The square root of the calcium volume score corrected for the bias in progression of calcium volume because of the level of coronary calcium. CONCLUSION. Using the square root of the calcium volume score stabilized interscan variability across the range of coronary calcium. Defining change in coronary calcium as greater than or equal to 2.5 mm3 of the difference in the square root-transformed calcium volume scores provided an estimate that was unbiased with respect to baseline coronary calcium. This analytic technique may facilitate investigations of the relevance of changes in coronary calcium to clinical outcomes and the use of changes in coronary calcium as a measure of the therapeutic impact on subclinical disease in clinical trials.