Table 3.
Best prediction models for annualized change rates including both sGFAP and sNfL.
| Annualized whole brain volume change rate Adj. R2 = 0.12, F = 3.93, p = 0.006 | ||
|---|---|---|
| Fixed effects | Std. β | p Value |
| Baseline age | −0.12 | 0.25 |
| Female sex | 0.11 | 0.29 |
| Ln NfL at 12M | −0.27 | 0.01 |
| Normalized WBF | −0.28 | <0.01 |
| Annualized ventricular volume change rateAdj. R2 = 0.18, F = 5.61, p = 0.006 | ||
| Fixed effects | Std. β | p Value |
| Baseline age | 0.05 | 0.64 |
| Female sex | −0.16 | 0.10 |
| Ln NfL at 12M | 0.26 | 0.02 |
| Ln GFAP at 3M | 0.79 | <0.01 |
| Annualized thalamic volume change rateAdj. R2 = 0.07, F = 3.00, p = 0.04 | ||
| Fixed effects | Std. β | p Value |
| Baseline age | −0.09 | 0.42 |
| Female sex | 0.16 | 0.14 |
| Ln NfL at 12M | −0.25 | 0.03 |
Multivariate linear regression with forward selection procedure (cutoff p value < 0.05) was used to establish the best prediction model for the different MRI volume changes. sGFAP and sNfL levels were natural log transformed. Singificant p-values have been emboldened. sGFAP: serum glial fibrillary acidic protein; sNfL: serum neurofilament light; Ln: natural log transformed; 3M: 3 months after baseline; 12M: 12 months after baseline; WBF: whole brain parenchymal fraction of intracranial volume.