Figure 2.

Ability to estimate Aβ PET status from plasma biomarkers. A logistic regression model was fit to predict Aβ PET status from age, plasma Aβ42/Aβ40 and APOE ε4 status. The effect of each variable in the model is shown for all individuals in subplot A (note that the upper line in the far-left subplot shows the effect of positive APOE status), the distribution of predicted amyloid PET risk (i.e. the kernel density estimate based on the probability density function) derived from the logistic regression model is visualized in subplot B, and a ROC curve showing results of the fitted model is visualized in subplot C. Note that plasma Aβ42/Aβ40 values have been negated to improve interpretability so that higher levels are associated with worsening risk for Aβ PET positivity.