Figure 3.

SEM model for the whole group analyses (n = 400) for the dopaminergic system. The respective included brain areas are shown estimating latent factors ‘areas high in DA’ and ‘areas low in DA’. Light grey arrows indicate added intercorrelations between brain regions based on the top three modification indices for model fit improvement. CSF biomarkers p-tau and Aß42/40 are inversely; ‘areas high in DA’ and ‘areas low in DA’ are positively correlated. p-tau is not significantly linked to ‘areas high in DA’ and ‘areas low in DA’, while Aß42/40 shows a positive link to ‘areas high in DA’, but not to ‘areas low in DA’. ‘Areas high in DA’ were positively linked to memory, language and executive function, while ‘areas low in DA’ were positively linked to working memory. For an overview of regressions across neurotransmitters, see Tables 2 and 3. Not shown here for representational clarity: CSF biomarkers, latent factors and cognitive variables were controlled for additional AD risk factors (see Supplementary Table 3 for estimates of relationships). Latent factors are shown in ellipses, and observed variables are shown in squares. Numbers given in ellipses and squares indicate R² of explained variance in the respective dependent variable. Only significant links are shown. CTX, cortex. ***P ≤ 0.001. **P ≤ 0.01. *P ≤ 0.05.