Plasma biomarkers of amyloid, tau, axonal, and neuroinflammation pathologies in dementia with Lewy bodies

Abstract

Background: Increasing evidence supports the use of plasma biomarkers of amyloid, tau, neurodegeneration, and neuroinflammation for diagnosis of dementia. However, their performance for positive and differential diagnosis of dementia with Lewy bodies (DLB) in clinical settings is still uncertain.

Methods: We conducted a retrospective biomarker study in two tertiary memory centers, Paris Lariboisière and CM2RR Strasbourg, France, enrolling patients with DLB (n = 104), Alzheimer's disease (AD, n = 76), and neurological controls (NC, n = 27). Measured biomarkers included plasma Aβ40/Aβ42 ratio, p-tau181, NfL, and GFAP using SIMOA and plasma YKL-40 and sTREM2 using ELISA. DLB patients with available CSF analysis (n = 90) were stratified according to their CSF Aβ profile.

Results: DLB patients displayed modified plasma Aβ ratio, p-tau181, and GFAP levels compared with NC and modified plasma Aβ ratio, p-tau181, GFAP, NfL, and sTREM2 levels compared with AD patients. Plasma p-tau181 best differentiated DLB from AD patients (ROC analysis, area under the curve [AUC] = 0.80) and NC (AUC = 0.78), and combining biomarkers did not improve diagnosis performance. Plasma p-tau181 was the best standalone biomarker to differentiate amyloid-positive from amyloid-negative DLB cases (AUC = 0.75) and was associated with cognitive status in the DLB group. Combining plasma Aβ ratio, p-tau181 and NfL increased performance to identify amyloid copathology (AUC = 0.79). Principal component analysis identified different segregation patterns of biomarkers in the DLB and AD groups.

Conclusions: Amyloid, tau, neurodegeneration and neuroinflammation plasma biomarkers are modified in DLB, albeit with moderate diagnosis performance. Plasma p-tau181 can contribute to identify Aβ copathology in DLB.

Keywords: Alzheimer’s disease; Amyloid pathology; Dementia with Lewy bodies; Plasma biomarkers.

Fig. 1

Plasma biomarkers levels across diagnosis groups. Plasma biomarkers levels across diagnosis groups including a, Aβ ratio; b, p-tau181; c, NfL; d, GFAP; e, sTREM2; and f, YKL-40. P-values were obtained through one-way ANCOVA followed by post hoc Tukey’s test, adjusting for multiple comparisons. Significant differences (P < 0.05) are reported. The effect size was determined using Cohen’s d. Boxplots display the median, IQR, and value for all participants

Fig. 2

Plasma biomarkers performance to identify DLB. ROC analysis: a, to compare single biomarkers performance to discriminate between DLB patients and NC; b, to compare biomarkers combination to discriminate between DLB and NC; c, to compare single biomarkers performance to discriminate between DLB and AD patients; d, to compare biomarkers combination to discriminate between DLB and AD patients. ROC analysis results are presented as AUC (95% CI). Combinations of biomarkers were selected through binary logistic regression with backward stepwise elimination, including age and sex as constant variables. ^a model including p-tau181 outperformed all other models (∂AIC > 4), ^b no significant difference in the model’s fit with the All markers model (∂AIC < 4).

Fig. 3

Plasma biomarkers levels in relation to amyloid pathology in DLB patients. Plasma biomarkers levels across amyloid-negative (A-) DLB and amyloid-positive (A+) DLB patients including a, Aβ ratio; b, p-tau181; c, NfL; d, GFAP; e, sTREM2; f, YKL-40; and across A-T- DLB and A + T + DLB patients including: g, Aβ ratio; h, p-tau181; i, NfL; j, GFAP; k, sTREM2; l, YKL-40. For biomarker levels comparison, P-values were obtained through one-way ANCOVA adjusting for multiple comparisons. Significant differences (P < 0.05) are reported in bold. The effect size was determined using η2. Boxplots display the median, IQR, and value for all participants

Fig. 4

Plasma biomarkers performance for identification of amyloid copathology in DLB patients. ROC analysis: a, to compare single biomarkers performance to discriminate between A- and A + DLB patients; b, to compare biomarkers combination to discriminate between A- and A + DLB patients; c, to compare single biomarkers performance to discriminate between A-T- and A + T + DLB patients; d, to compare biomarkers combination to discriminate between A-T- and A + T + DLB patients. ROC analysis results are presented as AUC (95% CI). Combinations of biomarkers were selected through binary logistic regression with backward stepwise elimination, including age and sex as constant variables. ^athe model including p-tau181 outperformed all other models (∂AIC > 4). ^bthe model associating plasma p-tau181, GFAP, and NfL was equivalent to the All markers models (∂AIC < 4).^cthe model including p-tau181 outperformed the All markers model (∂AIC > 4), with the best trade-off between parsimony and performance

Fig. 5

Principal component analysis of biomarker data in DLB and AD patients. a, Principal component analysis in DLB patients (n = 103). Component 1 associating plasma Aβ ratio, p-tau181, and GFAP explained 19% of the variance of the biomarkers data. Component 2 associating neuroinflammation sTREM2 and YKL-40 and axonal damage NfL makers explained 40% of variance. b, Principal component analysis in AD patients (n = 76). Component 1 associating plasma Aβ ratio, p-tau181, and GFAP explained 20% of the variance of the biomarkers data. Component 2 associating neuroinflammation sTREM2 and YKL-40 and axonal damage NfL markers explained 36% of variance