Table 3.
Binary logistic regression model to predict the odds of conversion to CDR above zero based on baseline biomarkers
| Odds of conversion from CDR = 0 to CDR > 0 | ||||
|---|---|---|---|---|
| Univariable | Multivariable | |||
| p-value | OR (95% CI) | p-value § | OR b (95% CI) | |
| Age (years) | 0.996 | 1 (0.951–1.051) | - | - |
| Sex (male versus female) | 0.911 | 0.963 (0.496–1.87) | - | - |
| Ln ICA calcium score | 0.210 | 0.891 (0.743–1.068) | - | - |
| Ln ICA calcium volume | 0.265 | 0.899 (0.845–1.084) | - | - |
| WMH voluma (1000 mm3) | 0.340 | 1.01 (0.989–1.032) | - | - |
| Total hippocampal volumea (100 mm3) | 0.006 | 0.940 (0.9–0.982) | 0.071 | 0.958 (0.915–1.004) |
| AD cortical signature thickness (mm) | 0.125 | 0.110 (0.007–1.837) | - | - |
| Centiloida (5 unit) | 0.003 | 1.08 (1.027–1.136) | 0.012 | 1.070 (1.015–1.127) |
Abbreviations: ICA, internal carotid artery, WMH volume, white matter hyperintensities volume, Centiloid, measure of global amyloid disposition based on conversion of PIB or AV45 PET SUVRs to a standardized scale, AD cortical signature thickness, cortical thickness in signature regions affected in Alzheimer disease (see Dincer et al. 2020), Ln ICA calcium score/volume, natural log transformation of internal carotid artery Agatston calcium score/volume, OR (95% CI), odds ratio and 95% confidence interval
aOdds ratios are demonstrated for 1000-mm3 increments in WMH volume, 100-mm3 increments in total hippocampal volume, and 5-unit increments in Centiloid scale
bSignificant variables from the univariable model were entered into a stepwise multivariable model where a probability of 0.3 was set to enter the variables into the model and a probability of 0.1 to remove variables from the model
§ p-values were corrected for multiple comparisons through estimation of the false discovery rate via the Benjamini and Hochberg method. Variables with a p-value of 0.05 were considered statistically significant in both regression models (denoted in bold)