Table 4.

Regression models to predict cognitive decline and cognition at baseline.

	Final models		Bootstrapped validation final models
	B	P	B mean	95% CImean	P median*
Cognition at baseline (MEG model; $R_{\mathrm{adj}}^2=10\%$)
Leaf fraction alpha1	13.99	0.004*	13.82	4.36–22.84	0.004*
Sex	0.30	0.029*	0.30	0.04–0.57	0.027*
Age	–0.01	0.184	–0.008	–0.019 to 0.004	0.185
Education middle vs. low	0.25	0.141	0.25	–0.11 to 0.60	0.145
Education high vs. low	0.37	0.025*	0.37	0.04 to 0.68	0.025*
Cognition at baseline (MEG and MRI model; $R_{\mathrm{adj}}^2=34\%$)
Leaf fraction alpha1	8.79	0.041*	8.56	−1.36 to 18.08	0.043*
Cortical GM volume (L)	4.83	0.007*	4.85	1.42–8.12	0.006*
Deep GM volume (ml)	0.04	0.009*	0.04	0.008–0.06	0.008*
Sex	0.15	0.204	0.16	−0.07 to 0.37	0.183
Age	0.01	0.064	0.01	−0.002 to 0.02	0.062
Education middle vs. low	0.31	0.037*	0.30	−0.02 to 0.63	0.038*
Education high vs. low	0.37	0.010*	0.36	0.05 to 0.69	0.010*
Longitudinal cognitive decline (MEG model; $R_{\mathrm{adj}}^2=15\%$)
Leaf fraction delta band	3.37	0.001*	3.40	0.94 to 5.93	0.001*
Diameter beta band	0.07	0.003*	0.07	0.03 to 0.10	0.003*
Cognition at baseline	–0.02	0.148	–0.02	–0.05 to 0.01	0.158
Sex	0.03	0.264	0.03	–0.03 to 0.08	0.237
Age	–0.003	0.011*	–0.002	–0.004 to 0.001	0.009*
Education middle vs. low	0.04	0.190	0.04	–0.01 to 0.09	0.194
Education high vs. low	0.06	0.031*	0.06	0.01–0.11	0.030*
Longitudinal cognitive decline (MEG and MRI model; $R_{\mathrm{adj}}^2=21\%$)
Leaf fraction delta	3.35	0.001*	3.41	1.13–5.74	0.001*
Diameter beta	0.07	0.003*	0.07	0.03–0.10	0.003*
Cortical GM volume (L)	0.70	0.006*	0.69	0.24–1.16	0.006*
Cognition at baseline	–0.04	0.009*	–0.04	–0.07 to 0.007	0.010*
Sex	0.02	0.363	0.02	–0.03 to 0.07	0.291
Age	–0.001	0.641	–0.001	–0.003 to 0.002	0.454
Education middle vs. low	0.05	0.076	0.05	0.003 to 0.10	0.079
Education high vs. low	0.07	0.010*	0.07	0.02 to 0.12	0.010*

CI: confidence interval; MEG: magnetoencephalography; MRI: magnetic resonance imaging; GM: gray matter.

Covariates are presented in italics.

^*Due to a skewed distribution of the p values across the 10,000 bootstrap samples, the median p value is noted.