Plasma biomarkers of Alzheimer’s disease improve prediction of cognitive decline in cognitively unimpaired elderly populations

Fig. 1. Relationship between plasma biomarkers and change in PACC. This figure shows the longitudinal PACC trajectory expected for a biomarker-negative or biomarker-positive CU individual with average age, average education, and female sex. β-coefficients at the top left of each panel are presented in terms of “points/year per standard deviation change in biomarker value” and are derived from linear mixed effects models with longitudinal PACC as outcome and age, sex, education, plus each plasma biomarker included separately from each other as predictors. All statistical tests were two-sided with no adjustment for multiple comparisons. Shaded areas represent 95% confidence intervals of the regression lines.

Fig. 2. Relationship between plasma biomarkers and conversion to AD. This figure shows the risk of AD dementia overtime expected for a biomarker-negative or biomarker-positive CU individual with average age, average education, and female sex. Hazard ratios at the top left of each panel are presented in terms of “increased risk of converting to AD dementia per standard deviation change in biomarker value” and are derived from Cox regression models with conversion to AD dementia as outcome and age, sex, education, plus each plasma biomarker included separately from each other. All statistical tests were two-sided with no adjustment for multiple comparisons. Shaded areas represent 95% confidence intervals of the regression lines.

Fig. 3. Power enrichment analysis of plasma biomarkers in a theoretical clinical trial. This figure shows the reduction in sample size resulting from using plasma biomarkers for inclusion enrichment in theoretical clinical trials aimed at slowing decline in PACC or reducing risk of AD dementia in a CU population. Sample sizes were estimated for a trial enriched using pre-defined cutoffs for each biomarker as inclusion threshold. Confidence intervals were derived by calculating the 0.025% and 95.75% percentile values from 1000 bootstrapped trials over n = 435 individuals. Source data is available in Supplementary Table 8.

Fig. 4. Effect of inclusion threshold variability on biomarker effectiveness. This figure shows the effect that varying the inclusion threshold by 20% in either direction of the pre-defined cutoffs has on the reduction in sample size in a theoretical clinical trial using biomarkers for inclusion enrichment. The reduction in sample size is compared to an unenriched trial with PACC or conversion to AD dementia as primary outcome. For datapoints in the figure, “circles” represent cutoffs for which the required sample size was significantly reduced, while “crosses” represent cutoffs for which required sample size was not significantly reduced. Shaded areas represent 95% confidence intervals of the loess regression lines fit over the entire data range.