Extended Data Fig. 8: Receiver Operating Characteristic (ROC) Curve analysis.

Panel A shows data for the sample with data on Pace of Aging and blood-chemistry measures of biological age (n = 7,537). Panel B shows data for the sample with data on Pace of Aging and epigenetic clocks (n = 2,848). ROC curves are generated by graphing sensitivity against 1-specificity for each value of a prediction metric. A predictor that generates no improvement in classification relative to random chance generates a diagonal line (sensitivity=1-specificty). ROC curves for predictors can be summarized by the area between the ROC curve and the diagonal, referred to as area under the curve (AUC). ROC Curves are drawn for model-based predictions of the outcomes including the aging measure indicated in the legend and a set of covariates (age, sex, race, education level, and smoking history). For reference, predictions from a model including only the covariates is also shown (‘base model’).