Characteristics of participant in the CABLE study

A total of 1,078 individuals from the CABLE study were included in the cross-sectional study. Demographical and clinical characteristics of our participants were described in Table 2. The total individuals had a mean age of 62.58 years [SD 10.06], a female proportion of 44.16%, as well as an APOE ε4 carrier proportion of 15.90%. The mean ANU-ADRI score was 4.65[SD 9.00]. According the total ANU-ADRI scores, they were grouped into low risk group (n = 110), intermediate risk group (n = 793), and high risk group (n = 175). Compared with low and intermedium risk groups, participants from the high risk group tended to be older and less-educated.

Table 2 Characteristics of participants across ANU-ADRI categoriesIntergroup differences in CSF AD biomarkers and cognition

The significant differences were found in cognition (both MMSE and MoCA scores) among the three ANU-ADRI risk groups. To be specific, high risk group had worse cognitive performance compared to the low risk group (MMSE: P < 0.001; MoCA: P < 0.001) and intermedium (MMSE: P < 0.001; MoCA: P < 0.001) risk group (Fig. 1A-B). As for CSF AD biomarkers, the high risk group had higher levels of CSF AD biomarkers level compared with the low (t-tau: P < 0.001; p-tau: P < 0.001; t-tau/Aβ42: P = 0.002) and intermedium (t-tau: P < 0.001; p-tau: P < 0.001; t-tau/Aβ42: P = 0.002) risk groups (Fig. 1C-E). However, no significant intergroup differences in CSF Aβ42 and p-tau/Aβ42 ratio were found (Supplementary Fig. 1 and Fig. 1).

Fig. 1

Differences in CSF biomarkers and cognition between the three ANU-ADRI categories. Differences in cognition (A-B) and CSF tau-related biomarkers (C-F) were examined by the analysis of variance. ANU-ADRI, Australian National University Alzheimer Disease Risk Index; CSF, cerebrospinal fluid; Aβ, amyloid beta; P-tau, phosphorylated tau; T-tau, total tau; MMSE, Mini-Mental State Examination; MoCA, Montreal Cognitive Assessment

Association of ANU-ADRI scores with cognition and several biomarkers of AD pathology

The MLR models were conducted to evaluate the relationships of ANU-ADRI scores with cognitive scores and several biomarkers of AD pathology (Fig. 2). After adjusting for APOE ɛ4, significant negative correlations of ANU-ADRI scores with MMSE (β = -0.264, P < 0.001, Fig. 2A) and MoCA (β = -0.393, P < 0.001, Fig. 2B) scores were observed. Higher ANU-ADRI scores were associated with increased CSF t-tau (β = 0.236, P < 0.001, Fig. 2C), p-tau (β = 0.183, P < 0.001, Fig. 2D), and t-tau/Aβ42 ratio (β = 0.094, P = 0.005, Fig. 2E). However, ANU-ADRI scores were not associated with Aβ42 (Supplementary Table 1). Sensitivity analysis was conducted by additionally adjusting for the comorbidities of coronary heart disease, stroke, as well as hypertension, which yielded similar findings (Supplementary Table 2). The interaction analyses demonstrated that associations between ANU-ADRI scores and cognition were not affected by APOE ε4 status (Supplementary Table 3).

Fig. 2

Associations between ANU-ADRI scores and CSF tau-related biomarkers and cognition. Multiple linear regression models were used to examine the associations between the ANU-ADRI scores and cognition (A-B) and CSF tau-related biomarkers (C-F) adjusting for APOE ɛ4 allele statuses. ANU-ADRI, Australian National University Alzheimer Disease Risk Index; CSF, cerebrospinal fluid; Aβ, amyloid beta; P-tau, phosphorylated tau; T-tau, total tau; MMSE, Mini-Mental State Examination; MoCA, Montreal Cognitive Assessment

CSF AD biomarkers mediates the association of ANU-ADRI scores with cognition

Our mediation analyses revealed that the correlation of ANU-ADRI scores with MMSE and MoCA scores were partially mediated by CSF t-tau (MMSE score: mediation proportion = 10.50%; MoCA score: mediation proportion = 6.60%, Fig. 3A-B). And p-tau also mediated the relationship of ANU-ADRI scores with cognition (MMSE score: mediation proportion = 8.01%; MoCA score: mediation proportion = 4.45%, Fig. 3C-D).

Fig. 3

CSF p-tau and t-tau mediated association between ANU-ADRI scores and cognition. Models of mediation for ANU-ADRI scores, cognition (MMSE and MoCA), and CSF biomarkers (p-tau and t-tau), with ANU-ADRI scores as independent variable and CSF biomarkers as mediator and cognition as dependent variable. Each model path was adjusted for APOE ε4 status. a is the effect of the independent variable on mediators; b is the effect of mediators on dependent variables after controlling the influence of independent variables; c is the total effect of independent variables on dependent variables; c’ is the direct effect; IE is the indirect effect. ANU-ADRI, Australian National University Alzheimer Disease Risk Index; CSF, cerebrospinal fluid; P-tau, phosphorylated tau; T-tau, total tau; MMSE, Mini-Mental State Examination; MoCA, Montreal Cognitive Assessment

Serial mediation between ANU-ADRI scores and cognition

CSF sTREM2, an inflammatory-related protein, showed an association with CSF p-tau level, suggesting a possible link between inflammation and tau pathology [24]. Given the mediation effects of CSF AD biomarkers observed in our mediation analyses, we thus further examined whether CSF AD biomarkers and sTREM2 mediated the effects of ANU-ADRI scores on cognition using the SEM. Therefore, three mediation pathway analyses were conducted: (1) ANU-ADRI → t-tau/p-tau → sTREM2 → MMSE/MoCA; (2) ANU-ADRI → t-tau/p-tau → MMSE/MoCA; (3) ANU-ADRI → sTREM2 → MMSE/MoCA. The model fit of the serial mediation was reported in supplementary Table 7. The results suggested that ANU-ADRI scores had a significant negative influence on MMSE/MoCA in the first model (Fig. 4 A, β = -0.287, P < 0.001; B, β = -0.398, P < 0.001; C, β = -0.287, P < 0.001; D, β = -0.398, P < 0.001). ANU-ADRI scores were positively corelated with CSF t-tau/p-tau (Fig. 4 A, β = 0.250, P < 0.001; B, β = 0.249, P < 0.001; C, β = 0.191, P < 0.001; D, β = 0.195, P < 0.001), CSF t-tau/p-tau was positively corelated with sTREM2 (Fig. 4 A, β = 0.374, P < 0.001; B, β = 0.393, P < 0.001; C, β = 0.333, P < 0.001; D, β = 0.346, P < 0.001), but sTREM2 was not significantly associated with MMSE/MoCA (Fig. 4 A, β = -0.031, P = 0.397; B, β = -0.037, P = 0.307; C, β = -0.027, P = 0.454; D, β = -0.037, P = 0.309). The indirect pathway by which ANU-ADRI scores affects MMSE/MoCA via CSF t-tau/p-tau and sTREM2 was not significant (Fig. 4 A, β 1 = -0.003, P = 0.408; B, β 1 = -0.004, P = 0.324; C, β 1 = -0.002, P = 0.468; D, β 1 = -0.002, P = 0.333). In the second mediation model, the association of ANU-ADRI scores with MMSE/MoCA was significantly mediated by t-tau/p-tau (Fig. 4 A, β 2 = -0.031, P < 0.001; B, β 2 = -0.027, P = 0.002 C, β 2 = -0.023, P = 0.001; D, β 2 = -0.018, P = 0.009). Specifically, ANU-ADRI scores were positively corelated with CSF t-tau/p-tau (Fig. 4 A, β = 0.243, P < 0.001; B, β = 0.240, P < 0.001; C, β = 0.191, P < 0.001; D, β = 0.194, P < 0.001) and CSF t-tau/p-tau was negatively corelated with MMSE/MoCA (Fig. 4 A, β = -0.126, P < 0.001; B, β = -0.111, P = 0.001; C, β = -0.122, P < 0.001; D, β = -0.094, P = 0.003). However, mediation analyses demonstrated that sTREM2 was not a significant mediator for the correlation between ANU-ADRI scores and MMSE/MoCA in the third mediation model, (Fig. 4 A, β 3 = -0.004, P = 0.452; B, β 3 = -0.005, P = 0.377; C, β 3 = -0.004, P = 0.449; D, β 3 = -0.005, P = 0.369). ANU-ADRI scores were positively corelated with sTREM2 (Fig. 4 A, β = 0.119, P = 0.002; B, β = 0.132, P = 0.001; C, β = 0.119, P = 0.002; D, β = 0.132, P = 0.001), but sTREM2 was not significantly associated with MMSE/MoCA (Fig. 4 A, β = -0.029, P = 0.414; B, β = -0.036, P = 0.331; C, β = -0.029, P = 0.410; D, β = -0.036, P = 0.327).

Fig. 4

Mediation analysis. A Three mediation pathways were conducted between ANU-ADRI scores and MMSE: (1) ANU-ADRI → t-tau → sTREM2 → MMSE; (2) ANU-ADRI → t-tau → MMSE; (3) ANU-ADRI → sTREM2 → MMSE. The serial mediation pathway via t-tau and sTREM2 was not significant (β 1 = -0.003, P = 0.408). sTREM2 was not significant mediator for this association (β 3 = -0.004, P = 0.452), but t-tau was significant mediator for this association (β 2 = -0.031, P < 0.001). B Three mediation pathways were conducted between ANU-ADRI scores and MoCA: (1) ANU-ADRI → t-tau → sTREM2 → MoCA; (2) ANU-ADRI → t-tau → MoCA; (3) ANU-ADRI → sTREM2 → MoCA. The serial mediation pathway via t-tau and sTREM2 was not significant (β 1 = -0.004, P = 0.324). sTREM2 was not significant mediator for this association (β 3 = -0.005, P = 0.377), but t-tau was significant mediator for this association (β 2 = -0.027, P = 0.002). C Three mediation pathways were conducted between ANU-ADRI scores and MMSE: (1) ANU-ADRI → p-tau → sTREM2 → MMSE; (2) ANU-ADRI → p-tau → MMSE; (3) ANU-ADRI → sTREM2 → MMSE. The serial mediation pathway via p-tau and sTREM2 was not significant (β 1 = -0.002, P = 0.468). sTREM2 was not significant mediator for this association (β 3 = -0.004, P = 0.449), but p-tau was significant mediator for this association (β 2 = -0.023, P = 0.001). D Three mediation pathways were tested between ANU-ADRI scores and MoCA: (1) ANU-ADRI → p-tau → sTREM2 → MoCA; (2) ANU-ADRI → p-tau → MoCA; (3) ANU-ADRI → sTREM2 → MoCA. The serial mediation pathway via p-tau and sTREM2 was not significant (β 1 = -0.002, P = 0.333). sTREM2 was not significant mediator for this association (β 3 = -0.005, P = 0.369), but p-tau was significant mediator for this association (β 2 = -0.018, P = 0.009). These three pathways are presented using green, blue, red lines. All mediation paths are adjusted for APOE ɛ4 allele status. The β coefficients in each path and P-values for mediation effects were calculated by a bootstrap test with 10,000 resampling iteration. The dotted line indicates that the indirect effect is not significant (P ≥ 0.05), the solid line indicates that the indirect is significant (P < 0.05). *P < 0.05, **P < 0.01 and ***P < 0.001