The following CT features were learn more analyzed for the common and internal carotid arteries at baseline and follow-up: lumen volume, wall volume, volume of calcium, volume of fibrous tissue, volume of lipid, number of lipid clusters, largest size of lipid clusters, location of largest lipid clusters, number of calcium clusters, largest size of calcium clusters, and location of largest calcium clusters.

The locations of the largest lipid and calcium clusters were described as a percent of the carotid wall thickness. For example, 0% indicates that the center of the cluster is immediately adjacent to the inner contour of the carotid artery, and 100% indicates that the center of the cluster is immediately adjacent to the outer contour of the carotid artery. CT features were measured and recorded separately for the following three segments of the carotid arteries: the 3 cm of the common carotid artery (CCA) immediately proximal to the carotid bifurcation, the 3 cm of the internal carotid artery (ICA) immediately distal to the carotid bifurcation, and both of these segments considered together (BIF). The software automatically Inhibitor Library register the carotid contours as determined on the baseline and the 1-year follow-up CTA studies (Supp Fig 2), and measure changes over 1 year in terms of lumen volume, wall volume, volume

of calcium, and volume of lipid. Baseline values of carotid imaging features and clinical variables were assessed for their Celecoxib ability to significantly predict changes in these imaging features over 1 year. Our outcome variables were as follows: change in lumen volume, change in wall volume, change in volume of calcium, and change in volume of lipid. Our predictor variables were as follows: baseline lumen volume, wall volume, volume of calcium, volume of fibrous tissue, volume of lipid, largest size of lipid clusters, location of largest lipid clusters, number of calcium clusters, size of calcium clusters, and location of largest calcium clusters, in addition to the following

clinical variables: age, gender, baseline BMI, current smoking status, hypertension, diabetes, baseline significant coronary artery disease, statin use. Time between baseline and follow-up exams was considered as a possible confounder. For each outcome feature, we looked at the change in its value over 1 year’s time. Using a mixed regression model with random effects, we looked for significant effects that the baseline values of carotid imaging features along with the clinical variables had on this change. We first did this in a univariate analysis using a threshold of .30 for significance. This lenient threshold was selected to avoid ruling out negative confounders for the subsequent multivariate analysis. See an example of this analysis for the change in volume of lipid over 1 year in Table 2.