Plasma proteomics and coronary artery calcium score: synergistic, concordant and contrasting predictions of cardiovascular outcomes in The Multi-Ethnic Study of Atherosclerosis

Figures

• 

  Figure 1:. Cox proportional hazard (CoxPH) ratios (HR) by qualifying event type. The CoxPH survival was modeled as event status (“event” or “no event”) to follow-up days (event follow-up days or censoring if no event) distributed by a center scaled metric, Prot-CVR or CAC score. HRs are reported as the exponentiated model coefficients [exp(coeff)], with 95% confidence intervals. HR = 1 indicates no predicted change in risk for the "event" group relative to the “no event” group over the follow-up period, HR >1 indicates increased predicted risk, and HR <1 indicates decreased predicted risk. HRs for Prot-CVR and CAC were compared for each event type using a Z-statistic calculated from the difference in log(HR) divided by the standard error of the difference. Reported p-values were corrected for multiple testing using False Discovery Rate (FDR) adjustment.

• 

  Figure 2:. Kaplan-Meier 4-year event-free probability (with 95% CI bands) of all participants stratified by (A) Prot-CVR risk bins, and (B) CAC risk bins. The number of participants diagnosed with a CV or mortality event/total number of participants by risk bin is included in the in-line legend. The median time (years) to CV event across all groups is indicated with a vertical dashed line (2.73 years).

• 

  Figure 3:. CoxPH models evaluate the performance of one or more scoring metric (Prot-CVR, CAC) with respect to predicting combined cardiovascular events. Models include all participants at exam 5 that are part of the Prot-CVR intended use population (individuals with stable CVD, type 2 diabetes, chronic kidney disease, a history of cancer, or > 65 years of age). A: A CoxPH survival model for the composite CV and mortality outcome distributed by center scaled Prot-CVR and adjusted in an add-one manner for clinical covariates (between horizontal dashed lines) known to impact CV risk [diabetes status, sex, high-density lipoprotein cholesterol (mg/dL), systolic blood pressure (mmHg), smoking status, total cholesterol (mg/dL)], for additional clinical covariates significant in a full combined model [ethnicity, antihypertensive medication use, and body mass index (BMI)], and finally for center scaled CAC. B: 4-year AUC and C-Index metrics for simplified models. 95% confidence intervals reflect 2-sided confidence interval around derived 4-year AUC and C-Index metrics. Bold font indicates a significant (p ≤0.05) increase in performance of the simplified combined models (Prot-CVR + CAC and Prot-CVR…

• 

  Figure 4:. Prot-CVR scores for participants with a CAC score of 0 at Exam 1 for CV and mortality composite outcome events at 4 years and over the full follow-up period, shown as a notched box plot with outliers. Differences between mean Prot-CVR scores in the “event” versus "no event" group were assessed using a one-sided t-test with an alternative hypothesis that the mean “event” score would greater than the “no event” score. Group means between “event” and “no event” groups that are significantly different from each other (p-value <0.01) are indicated with “*”. Non-CV mortality rate was 2.2% over 4 years and 23.4% over the full follow-up period.

• 

  Figure 5:. Sensitivity to change in risk as measured by Prot-CVR and CAC from baseline to follow-up (Exam 5) in advance of an upcoming CV event. Change in risk from baseline to exam 5 in participants with paired samples that are part of the Prot-CVR intended use population (individuals with stable CVD, type 2 diabetes, chronic kidney disease, a history of cancer, or over 65 years old) that went on to have an event (CV and mortality composite outcome: MI, Stroke/TIA, HF hospitalization or all-cause death). Error bars represent 95% confidence intervals around the mean score. The left panels represent Prot-CVR predictions, and the right panels are CAC scores. The top panels are for all participants, while the bottom panels represent participants with non-zero CAC scores at exam 1 and exam 5. Significance testing for within-group change from baseline to follow-up and between-group difference was calculated using two-sided, two-sample t-tests (alpha = 0.05), *** <0.001, ** <0.01, * <0.05.