n = 80 | Model | Odds ratio (p-value) |
|---|
AICc | AUC [95% CI] | p-tau217 | NfL | GFAP | Aβ42/Aβ40 |
|---|
Model 1 (best model) | 76.2 | 0.840 [0.748, 0.933] | 3.11 (0.0002) | | | |
Model 2 | 76.5 | 0.843 [0.754, 0.931] | 2.78 (0.0009) | 1.54 (0.16) | | |
Model 3 | 76.9 | 0.836 [0.745, 0.927] | 2.86 (0.0008) | | 1.44 (0.22) | |
Model 4 | 77.9 | 0.846 [0.761, 0.931] | 2.65 (0.0022) | 1.44 (0.26) | 1.31 (0.37) | |
Model 5 | 78.3 | 0.841 [0.748, 0.934] | 3.04 (0.0003) | | | 0.90 (0.75) |
- Results from logistic regression models discriminating MCI patients who progressed to AD dementia within three years (n = 21) vs. those who did not (n = 59) in the subsample with all plasma biomarkers. Models are ordered based on AICc (lower values representing better model fit) and odds ratio (p-value) of each variable included in the corresponding models are reported. Odds ratio values represent the “increased risk” of converting to AD dementia for each increase in standard deviation of the plasma biomarker value. Note that a difference in AICc greater than 2 between models would imply a better fit for the model with the lowest AICc
- Abbreviations: Aβ beta-amyloid, AICc corrected Akaike information criteria, AUC area under the curve, GFAP glial fibrillary acidic protein, NfL neurofilament light, p-tau217 phosphorylated tau 217
