Demographic characteristics

Group demographic and clinical characteristics are reported in Table 1. Participants with aMCI due to AD were significantly older (p = .001), carried the APOE ε4 allele at significantly higher rates (p = .009), and scored significantly worse on BNT-30 (p = .017) compared to participants with MCI due to FTLD. On the other hand, individuals with aMCI due to AD scored significantly better on P-VF (p < .001) and S-VF-A (p = .008) compared to individuals with MCI due to FTLD. Participants with AD dementia were significantly older (p = .015), carried the APOE ε4 allele at significantly higher rates (p < .001), and scored significantly worse on MMSE (p = .022), ROCF-R (p = .006), ROCF-C (p = .014), LOG-I (p = .025), and LOG-D (p < .001) compared to FTLD dementia participants. Furthermore, individuals with AD dementia scored significantly better on P-VF (p = .001).

Table 1 Demographic features and cognitive performance of patients with aMCI due to AD, MCI due to FTLD, AD dementia and FTLD dementia

Ng concentrations were significantly higher in patients with AD dementia compared to FTLD dementia (p = .035), and in aMCI due to AD compared to MCI due to FTLD (p < .001) (see Table 1). Additionally, participants with aMCI due to AD had significantly lower levels of Aβ1−42 (p < .001) and Aβ1−42/1−40 (p < .001), and higher levels of p-tau 181 (p < .001) and t-tau (p < .001) compared to participants with MCI due to FTLD, and participants with AD dementia had significantly lower levels of Aβ1−42 (p < .001), Aβ1−42/1−40 (p < .001) and NfL (p = .013), and higher levels of p-tau 181 (p = .036) compared to participants with FTLD dementia.

Participants with SCD had significantly more years of education than participants with AD dementia (p < .001) and FTLD dementia (p = .016), carried the APOE ε4 allele at significantly lower rates than participants with aMCI due to AD (p = .042) and AD dementia (p = .003), had significantly higher MMSE, LOG-I, LOG-D, ROCF-R, F-DigitSpan-SC, B-DigitSpan-SC, P-VF, S-VF-A scores compared to all other subgroups (p < .001), significantly higher CDT and ROCF-C scores compared to aMCI due to AD, AD dementia, and FTLD dementia (p < .001), and significantly lower TMT-A, TMT-B, and BNT-30 scores than all other subgroups (p < .001). Ng concentrations were significantly lower in participants with SCD compared to those with aMCI due to AD (p = .001) and AD dementia (p < .001). Participants with SCD had significantly higher levels of Aβ1−42 and Aβ1−42/1−40 compared to those with aMCI due to AD and AD dementia (p < .001), significantly lower levels of p-tau 181 compared to patients with aMCI due to AD (p < .001), AD dementia (p < .001), and FTLD dementia (p = .005), t-tau compared to those with aMCI due to AD (p < .001), AD dementia (p < .001), and FTLD dementia (p < .001), and NfL compared to FTLD dementia (p = .009).

Ng levels across diagnostic groups

In ANCOVA adjusted for age, sex, and years of education the main effect of diagnosis across the five diagnostic subgroups on Ng levels was significant (F[4, 263] = 8.19, p < .001). After performing pairwise comparisons, patients with aMCI due to AD had significantly higher Ng concentrations than patients with MCI due to FTLD (F[1, 134] = 15.16, p < .001, η2 = 0.08, AUC = 0.66, 95% CI: 0.62–0.84, with a sensitivity of 0.68 and specificity of 0.72) and those with AD dementia had significantly higher Ng concentrations than those with FTLD dementia (F[1, 96] = 4.60, p = .029, η2 = 0.06, AUC = 0.64, 95% CI: 0.49–0.77, with a sensitivity of 0.73 and specificity of 0.52). Ng levels were also higher in patients with AD dementia than in patients with aMCI due to AD although the result only approached statistical significance (F[1, 176] = 2.68, p = .067, η2 = 0.02, AUC = 0.58, 95% CI: 0.48–0.67, with a sensitivity of 0.25 and specificity of 0.92). Ng concentrations were also significantly higher in patients with FTLD dementia compared to those with MCI due to FTLD (F [1, 54] = 4.35, p = .034, η2 = 0.09, AUC = 0.67, 95% CI: 0.49–0.81, with a sensitivity of 0.72 and specificity of 0.57) (Fig. 1).

Fig. 1

CSF Ng concentrations in subjective cognitive decline (SCD), aMCI due to AD, AD dementia, MCI due to FTLD and FTLD dementia patients. Violin plots showing distribution, box plots, and significant differences assessed by ANCOVA adjusted for age, sex, and education of CSF Ng concentrations. For visual purposes, CSF Ng values are presented raw, not log transformed. *p < .05, **p < .01, ***p < .001

Ng levels were significantly lower in patients with SCD than in patients with aMCI due to AD (F[1, 142] = 10.72, p = .001, η2 = 0.07, AUC = 0.65, 95% CI: 0.59–0.78, with a sensitivity of 0.62 and specificity of 0.76) or those with AD dementia (F[1, 100] = 20.90, p < .001, η2 = 0.19, AUC = 0.75, 95% CI: 0.65–0.84, with a sensitivity of 0.67 and specificity of 0.79). Ng levels were also lower in patients with SCD than in patients with FTLD dementia although the result only approached statistical significance (F [1, 62] = 2.27, p = .051, η2 = 0.07, AUC = 0.65, 95% CI: 0.43–0.73, with a sensitivity of 0.45 and specificity of 0.82). Ng levels did not differ between patients with SCD and MCI due to FTLD (F [1, 58] = 1.02, p = .491, η2 = 0.01, AUC = 0.56, 95% CI: 0.42–0.73, with a sensitivity of 0.52 and specificity of 0.70).

Additionally, when we stratified the AD sample by APOE ε4 carrier status, Ng levels were significantly higher in aMCI due to AD APOE ε4 carriers compared to non-carriers (F[1, 109] = 8.22, p = .005, η2 = 0.07, AUC = 0.65, 95% CI: 0.57–0.78, with a sensitivity of 0.59 and specificity of 0.69) but not in AD dementia APOE ε4 carriers compared to non-carriers (F [1, 67] = 2.20, p = .144, η2 = 0.03, AUC = 0.60, 95% CI: 0.49–0.79, with a sensitivity of 0.58 and specificity of 0.33).

In addition, we analyzed the Aβ1−42/Ng ratio too. In ANCOVA adjusted for age, sex, and years of education the main effect of diagnosis across the five diagnostic subgroups on Aβ1−42/Ng levels was significant (F[4, 263]=, p < .001). After performing pairwise comparisons, patients with aMCI due to AD had significantly lower Aβ1−42/Ng ratio compared to patients with MCI due to FTLD (F[1, 134] = 41.26, p < .001, η2 = 0.23, AUC = 0.78, 95% CI: 0.82–0.94, with a sensitivity of 0.95 and specificity of 0.82), and patients with AD dementia had significantly lower levels of Aβ1−42/Ng compared to FTLD dementia too (F[1, 96] = 62.69, p < .001, η2 = 0.44, AUC = 0.90, 95% CI: 0.84–0.99, with a sensitivity of 0.89 and specificity of 0.87). Patients with AD dementia had lower levels of Aβ1−42/Ng compared to patients aMCI due to AD (F[1, 176] = 7.58, p = .007, η2 = 0.04, AUC = 0.61, 95% CI: 0.50–0.68, with a sensitivity of 0.26 and specificity of 0.91). The Aβ1−42/Ng ratio differentiated between patients with MCI due to FTLD and FTLD dementia too (F [1, 54] = 4.35, p = .045, η2 = 0.10, AUC = 0.68, 95% CI: 0.39–0.76, with a sensitivity of 0.95 and specificity of 0.26).

Participants with SCD had significantly higher levels of Aβ1−42/Ng ratio compared to patients with aMCI due to AD (F[1, 142] = 23.95, p < .001, η2 = 0.14, AUC = 0.72, 95% CI: 0.71–0.88, with a sensitivity of 0.87 and specificity of 0.69) and AD dementia (F[1, 100] = 71.90, p < .001, η2 = 0.40, AUC = 0.88, 95% CI: 0.82–0.96, with a sensitivity of 0.87 and specificity of 0.82). Participants with SCD had only marginally lower levels of Aβ1−42/Ng ratio compared to patients with MCI due to FTLD (F [1, 58] = 3.75, p = .060, η2 = 0.06, AUC = 0.64, 95% CI: 0.48–0.79, with a sensitivity of 0.43 and specificity of 0.95). There were no statistically significant differences between Aβ1−42/Ng ratio in participants with SCD and patients with FTLD dementia (F [1, 62] = 0.00, p = .99, η2 = 0.00, AUC = 0.50, 95% CI: 0.38–0.72, with a sensitivity of 0.47 and specificity of 0.74).

CSF Ng levels and cognitive performance

The results for the association between CSF Ng levels and performance across cognitive domains are presented in Table 2. Using the entire AD sample, linear regression models adjusted for age, sex, and years of education indicated that Ng levels were negatively, albeit non-significantly, associated with memory scores overall (β=-0.25, p = .154), as well as in analyses with aMCI due to AD (β=-0.30, p = .147) and positively, albeit non-significantly, with AD dementia (β = 0.20, p = .165) (Fig. 2). No significant results emerged for the association between Ng levels and scores on other cognitive domains overall or in diagnosis-specific analyses.

Table 2 Association between Ng levels and cognitive domains by diagnosisFig. 2

Memory z-scores and Ng levels in aMCI due to AD and AD dementia patients. Memory data are presented as z-scores and Ng data are presented as log-transformed values

Ng levels were again negatively but non-significantly associated with memory scores overall (β=-0.16, p = .563) and with MCI due to FTLD (β=-0.10, p = .753), and positively but non-significantly with FTLD dementia (β = 0.48, p = .173) (Fig. 3). As in AD-related analyses, results for Ng levels and other cognitive domains also did not yield significant results both overall and in analyses conducted by diagnostic group (see Table 2).

Fig. 3

Memory z-scores and Ng levels in MCI due to FTLD and FTLD dementia patients. Memory data are presented as z-scores and Ng data are presented as log-transformed values

Finally, Ng levels were negatively but non-significantly associated with memory scores in participants with SCD (β=-0.78, p = .187). As in AD- and FTLD-related analyses, results for Ng levels and other cognitive domains also did not yield significant results in analyses using the SCD subgroup.

Additionally, after adjusting for age, sex, and years of education, Pearson’s correlation analysis revealed that Ng levels did not correlate with MMSE scores in patients with AD (r=-.14, p = .086) or in patients with FTLD (r=-.17, p = .273).

CSF Ng levels and cognitive performance in relation to APOE ε4 carrier status in aMCI due to AD and AD dementia

In linear regression models adjusted for age, sex, and years of education with memory scores as dependent variable and an interaction between Ng levels and APOE ε4 carrier status as an independent variable, we found that the association between Ng levels and memory scores did not statistically differ between AD APOE ε4 carriers and APOE ε4 non-carriers (β=-0.32, p = .358) (Fig. 4).

Fig. 4

Memory z-scores and Ng levels in APOE ε4 carriers and non-carriers with aMCI due to AD and AD dementia. Memory data are presented as z-scores and Ng data are presented as log-transformed values

In addition, the association between Ng levels and scores of other cognitive domains did not statistically differ between APOE ε4 carriers and APOE ε4 non-carriers, including language (β = 0.46, p = .181), attention and working memory (β = 0.01, p = .980), executive function (β = 0.35, p = .335), and visuospatial function (β=-0.11, p = .772).

Correlations of CSF Ng with AD biomarkers

Using the entire AD sample, Ng correlated positively with t-tau (r = .77, p < .001) and p-tau 181 (r = .79, p < .001), and negatively with Aβ1−42/1−40 (r=-.43, p < .001). Ng levels did not correlate with Aβ1−42 (r = .11, p = .151) or serum NfL (r = .02, p = .765). Additionally, in patients with aMCI due to AD, Ng correlated positively with t-tau (r = .77, p < .001) and p-tau 181 (r = .77, p < .001), and negatively with Aβ1−42/1−40 (r=-.52, p < .001). Ng levels did not correlate with Aβ1−42 (r = .12, p = .228) and serum NfL (r = .15, p = .135). In patients with AD dementia, Ng correlated positively with t-tau (r = .85, p < .001) and p-tau 181 (r = .83, p < .001), and negatively with Aβ1−42/1−40 (r=-.53, p < .001) and serum NfL (r=-.28, p = .033) too. Ng levels did not correlate with Aβ1−42 (r = .16, p = .185).

Using the entire FTLD sample, Ng positively correlated with t-tau (r = .82, p < .001) and p-tau 181 (r = .78, p < .001), and negatively with Aβ1−42/1−40 (r=-.51, p < .001). Ng did not correlate with Aβ1−42 (r = .16, p = .241) and serum NfL (r=-.20, p = .161). In subgroup analysis, in patients with MCI due to FTLD, Ng correlated positively with t-tau (r = .90, p < .001) and p-tau 181 (r = .83, p < .001), and negatively with Aβ1−42 (r=-.50, p = .010). Ng levels did not correlate with Aβ1−42/1−40 (r=-.22, p = .301) or serum NfL (r=-.07, p = .727). In patients with FTLD dementia, Ng positively correlated with t-tau (r = .82, p < .001) and p-tau 181 (r = .84, p < .001), and negatively with Aβ1−42/1−40 (r=-.64, p < .001) and serum NfL (r=-.44, p = .018). Ng levels did not correlate with Aβ1−42 (r=-.05, p = .818).

Additionally, using the SCD sample, Ng positively correlated with t-tau (r = .76, p < .001), p-tau 181 (r = .66, p < .001), and Aβ1−42 (r = .68, p < .001). Ng did not correlate with Aβ1−42/1−40 (r = .10, p = .590) and serum NfL (r=-.17, p = .348).