Figure 4. Distribution of plaque types by disease course and age.

(A) The bar plot shows the actual plaque distribution in relation to clinical course. The percentages shown for each subgroup or bar are calculated among all plaques for those subjects within the subgroup. Bar width is related to subgroup size. On multinomial modeling plaque type distribution differs significantly (p < 0.0001) among disease courses.

(B) Graph B provides estimates of the percentage of plaques by type expressed as a function of the patient’s age at death using multinomial modeling accounting for clustering of plaques within patient. Percentages across the curves add to 100 percent at a given age. Shaded regions represent 95% cluster bootstrap confidence intervals. Early active and late active plaques are pooled as active plaques. Significant differences are observed (p < 0.0001) based on multinomial modeling.

(C) In graph C the estimated proportion of patients with at least one of each plaque type as a function of age is depicted. Estimates are based on separate logistic regression models at the patient-level. Percentages across the curves add to more than 100 percent since subjects could have at least one of several types of plaques. Early active and late active plaques are pooled as active plaques.