Figure 1.

Average RMSEs across simulations for both scenarios using different models

Average RMSEs of simulated datasets were calculated using different models in scenario 1 (A) and scenario 2 (B), respectively, with different EPV. The models examined include full model, BE with significance levels α = 0.05 and α = 0.5 (BE 0.05 and BE 0.5), ridge, lasso, and elastic net. Scenario 1 considers patients’ clinical variables relevant to CAD and blood-based biomarkers as predictors. Predicted event probabilities were computed at time points 0.08, 0.17, and 0.25 years, respectively. In scenario 2, information on 55 genetic variants is also considered besides the predictors used in scenario 1, while predicted event probabilities were computed at time points 1, 2.5, and 5 years, respectively. BE, backward elimination; RMSE, root mean square error; EPV, events per variable.