Population characteristics
Demographics and characteristics are illustrated in Table 1. A total of 320 participants were enrolled in the study including 158 patients with AD, 32 with FTD, 30 with DLB, 59 patients with NND and 41 individuals with SCD. Overall, 220 patients (68.7%) were diagnosed with a cognitive decline due to a neurodegenerative disease at various severity (AD, FTD, or DLB). With regards to clinical syndromes, 41 subjects (12.8%) were classified as having SCD, 131 individuals (40.9%) had mild cognitive impairment (MCI), and 148 (46.3%) had major cognitive impairment (dementia). Compared with other groups, SCD subjects were younger and had higher education level. The sex ratio differed between the groups and patients with AD had more frequently major cognitive impairment than MCI. Body mass index differed in between diagnoses, with AD having a lower BMI than other groups.
Plasma NfL levels results were positively correlated with age (r = 0.46, P < 0.001), glomerular filtration rate (r = -0.34, P < 0.001), BMI (r = -0.24, P < 0.001), but no associations were observed with sex (P = 0.32) or APOE ε4 carriership (P = 0.24). In subgroup analysis plasma NfL was associated with sex only in SCD (P = 0.025), but not in any other subgroup. ApoE carriership was associated with plasma NfL only in LBD subgroup, with more elevated levels beeing associated with non-carriers P = 0.04. Regarding BMI, significant and negative correlations were found in AD and FTD subgroups (respectively r = -0.2, P = 0.01 ; r=-0.38, P = 0.03). Glomerular filtration rate was significantly and negatively correlated with plasma NfL across all diagnosis subgroups (P < 0.04 in all subgroups), except for FTD.
Association of neuropsychological and imaging data with plasma NfL levels across the whole population
All cognitive z-scores, except for attention and visuospatial abilities, were associated with plasma NfL levels across the overall population, even after adjusting for age and education level. Specifically, language (β = -1.72 (-2.49 ; -0.95), P < 0.001), praxis (β = -2.02 (-2.91 ; -1.13), P < 0.001), and memory (β = -1.48 (-2.38 ; -0.59), P < 0.001) displayed the strongest associations with plasma NfL levels when compared with executive functions (β = -0.81 (-1.30 ; -0.37), P = 0.001).
Except for praxis (β = 5.18 (1.50 ; 8.85), P = 0.006), age was not associated with any cognitive z-score, whereas education level was strongly associated with all cognitive domains (P < 0.001 for each cognitive function). BMI and GFR were not associated with any cognitive function, except with visuospatial functions (β = -1.9 (-3.76 ; -0.08), P = 0.04) and (β = -1.52 (-2.97 ; -0.05), P = 0.04).
Considering neuroimaging scores, plasma NfL levels were associated with medial temporal lobe atrophy (β = 1.31 (0.73 ; 1.90), P < 0.001) and white matter hyperintensities (β = 0.63 (0.11 ; 1.15), P = 0.02) in the global population. Age was also associated with medial temporal lobe atrophy (β = 2.75 (0.36 ; 5.14), P = 0.02) and white matter hyperintensities (β = 4.46 (2.35 ; 6.60), P < 0.001). Detailed results are reported in Table 2A.
Association of neuropsychological tests’ results and imaging scores with plasma NfL levels across diagnostic groups
Global cognition was associated with plasma NfL levels in all but SCD diagnosis groups (AD, β = -1.13 (-1.67 ; -0.59), P < 0.001; LBD β = -0.89 (-1.52 ; -0.25), P < 0.001; FTD β = -0.87 (-1.43 ; -0.33), P = 0.002; NND β = -0.80 (-1.38 ; -0.22), P = 0.007), with a stronger association found in the AD group.
Concerning memory, an association was only found with plasma NfL levels in the AD group (β = -1.29 (-2.17 ; -0.42), P = 0.004), but not in other diagnosis groups.
Plasma NfL levels were associated with language in all diagnosis groups (AD, β = -1.56 (-2.37 ; -0.74), P < 0.001; LBD β = -1.28 (-2.22; -0.33), P < 0.001: FTD β = -1.52 (-2.35 ; -0.69), P < 0.001; NND β = -1.30 (-2.17 ; -0.43), P = 0.003; SCD β = -1.06 (-2.08 ; -0.004), P = 0.0042). The strongest associations were found in the AD and FTD groups.
Praxis was also associated with plasma NfL levels in all but the SCD group (AD, β = -1.81 (-2.74 ; -0.87), P < 0.001; LBD, β = -1.60 (-2.70 ; -0.50), P = 0.004; FTD, β = -1.79 (-2.73 ; -0.84), P < 0.001; NND, β = -1.63 (-2.64 ; -0.63), P = 0.001; SCD, β = -1.12 (-2.29 ; 0.63), P = 0.06). The strongest associations were found in the AD and FTD groups.
Executive functions were associated with plasma NfL levels in AD (β = -0.71 (-1.21 ; -0.211), P = 0.005), LBD (β = -0.58 (-1.16 ; -0.002), P = 0.05), and tend to be associated with plasma NfL levels in FTD (β = -0.49 (-1.01 ; 0.02), P = 0.06).
DLB was the only group where plasma NfL and attention were associated (β = -0.81 (-1.55 ; -0.08), P = 0.03). No association was found between visuospatial functions and plasma NfL levels in diagnosis groups.
BMI, but not GFR, is also associated with visuospatial functions in this model (β = -2.01, P = 0.03). Education level displayed an association with all cognitive functions (P < 0.001).
Concerning neuroimaging scores, medial temporal lobe atrophy, measured with Scheltens’ visual scale, showed an association with plasma NfL levels across all diagnostic groups (AD, β = 1.16 (0.55 ; 1.77), P < 0.001; FTD, β = 1.35 (0.73 ; 1.96), P < 0.001; LBD, β = 1.04 (0.32 ; 1.75) P = 0.005; NND, β = 0.91(0.29 ; 1.54), P = 0.004; SCD, β = 0.8 (0.04 ; 1.57), P = 0.039). Detailed results are described in Table 2B and Fig. 1. Associations between plasma NfL levels and white matter hyperintensities were found in the AD group (β = 0.62 (0.08 ; 1.17), P = 0.026) and the NND group (β = 0.78 (0.22 ; 1.35), P = 0.007).
Association of plasma NfL with cognitive function across diagnosis subgroups
Plasma NfL association with global cognition and specific cognitive domains by diagnosis subgroups, including AD, FTD, LBD, NND and SCD. Standardized estimates with the corresponding 95% CI were plotted for the association of baseline plasma NfL levels with baseline cognitive domain scores for each diagnosis group. Linear regressions were adjusted on age, sex and education level.
Association of plasma NfL levels with cognitive decline across the global population
239 patients had available follow-up MMSE data and were included in a longitudinal analysis. Patients with no follow up examinations (n = 81) were not included in the longitudinal analysis. Their characteristics did not differ significantly with the included patients. See supplementary Table 2. The mean follow-up duration was 29.7 ± 15.2 months, with a mean of 3.7 ± 1.4 MMSE time points by patient. 127 had AD (53.1%), 21 had DLB (8.8%), 21 had FTD (8.8%), 40 had NND (16.7%), and 30 had SCD (12.6%). Considering neurodegeneration, 169 patients had a neurodegenerative disorder (70.7%), while 70 had any neurodegenerative disease (both SCD and NND groups, 29.3%).
In the global population, mixed model cross-sectional analysis revealed an association of plasma NfL levels (β = -5.27 (-7.60 ; -2.94), P < 0.001) and education level (β = 11.27 (8.19 ; 14.36), P < 0.001) with MMSE score (Table 3; Fig. 2A). Age, GFR and BMI were not associated with MMSE. In longitudinal analysis, the change in MMSE scores over time was inversely associated with plasma NfL levels at baseline (β =-0.15 (-0.26 ; -0.04), P = 0.006), but not with age, education level, BMI, GFR or baseline MMSE.(Table 3; Fig. 2B).
Association of plasma NfL and confounding variables with MMSE over time in the whole cohort, neurodegenerative conditions group and non-neurodegenerative conditions group
(A) Factors associated with MMSE in cross-sectional analysis in the whole cohort. (B) Factors associated with MMSE change over time in the whole cohort (C) Factors associated with baseline MMSE in the neurodegenerative condition group. (D) Factors associated with MMSE decline in the neurodegenerative condition group. (E) Factors associated with baseline MMSE in the non-degenerative condition group. (F) Factors associated with MMSE decline in the non-degenerative group.
Standardized estimates with the corresponding 95% CI were plotted for the association of baseline plasma NfL levels and confounding variables with MMSE scores at baseline (cross-sectional) and during follow-up (longitudinal) in the different groups.
All models represented are linear mixed models with fixed effects on plasma NfL, age, education level, body mass index and glomerular filtration rate. Random intercepts and slopes were included. All variables were log-transformed prior to analysis.
Association of plasma NfL levels with cognitive decline in subgroups analysis
In the non-neurodegenerative condition group, no association was found between plasma NfL levels, age, BMI or GFR and MMSE score cross-sectionally. Similarly, no association was found in longitudinal analyses. (Table 4; Fig. 2E and F)
In the neurodegenerative condition group, cross-sectional analysis revealed an association of plasma NfL levels (β =-5.66 (-8.87 ; -2.46), P = 0.001), age (β = 13.86 (1.93 ; 25.79), P = 0.001) and education level (β = 12.45 (8.74 ; 16.15), P < 0.001) with MMSE score. GFR and BMI were not associated with MMSE.
In the longitudinal analysis, plasma NfL levels at baseline were associated with MMSE change over time (β =-0.21 (-0.37 ; -0.06), P = 0.07), as well as baseline MMSE (β =-0.17 (-0.31 ; -0.03), P = 0.17). Results are reported Tables 4and Fig. 2C and D. Figure 3 illustrates the predicted MMSE change over time according to linear mixed model coefficients. For each group, we computed the MMSE evolution according to linear mixed model coefficients, with neurodegenerative and non-neurodegenerative conditions groups mean age, education level, as well as plasma NfL 10th, 50th and 90th quantiles values of the whole population.
Predicted MMSE trajectories at hypothesized 10th, 50th and 90th deciles plasma NfL levels in the non-neurodegenerative and degenerative subgroups
Predicted MMSE trajectories for hypothesized individuals with plasma NfL levels at the 10th, 50th and 90th centiles, respectively with a non-neurodegenerative (blue curves) and a degenerative (red curves) disorder, computed using mixed model coefficients. Age, sex, education level, glomerular filtration rate and body mass index used for analysis are, respectively, the mean of the neurodegenerative group and non-neurodegenerative groups.
The stratified lines represent the estimated slope across time for each baseline plasma NfL levels (Q10 = 9.27 pg/mL; Q50 = 19.18 pg/mL; Q90 = 35.48 pg/mL).
Add Comment