Cognitive effects of Lewy body pathology in clinically unimpaired individuals

Abstract

α-Synuclein aggregates constitute the pathology of Lewy body (LB) disease. Little is known about the effects of LB pathology in preclinical (presymptomatic) individuals, either as isolated pathology or coexisting with Alzheimer's disease (AD) pathology (β-amyloid (Aβ) and tau). We examined the effects of LB pathology using a cerebrospinal fluid α-synuclein-seed amplification assay in 1,182 cognitively and neurologically unimpaired participants from the BioFINDER study: 8% were LB positive, 26% Aβ positive (13% of those were LB positive) and 16% tau positive. LB positivity occurred more often in the presence of Aβ positivity but not tau positivity. LB pathology had independently negative effects on cross-sectional and longitudinal global cognition and memory and on longitudinal attention/executive function. Tau had cognitive effects of a similar magnitude, but these were less pronounced for Aβ. Participants with both LB and AD (Aβ and tau) pathology exhibited faster cognitive decline than those with only LB or AD pathology. LB, but not AD, pathology was associated with reduced sense of smell. Only LB-positive participants progressed to clinical LB disease over 10 years. These results are important for individualized prognosis, recruitment and choice of outcome measures in preclinical LB disease trials, but also for the design of early AD trials because >10% of individuals with preclinical AD have coexisting LB pathology.

Fig. 1. Prevalence of Aβ, tau and LB pathologies.

a, Prevalence of Aβ (A), tau (T) and LB positivity. b, Prevalence of A/T/LB groups. c, Prevalence of AD/LB groups. d–f, Proportions of groups a–c, respectively, with increasing age, where d shows A, LB and T positivity, e combinations of A/T/LB positivity/negativity and f combinations of AD/LB positivity/negativity. Note that AD positivity refers to being both A⁺ and T⁺ while LB positivity refers to being α-syn SAA⁺. d, Using age as independent variable and pathology as dependent in logistic regression models, age had an OR of 1.066 (95% CI 1.036–1.099) for LB, 1.067 (95% CI 1.048–1.087) for A and 1.071 (95% CI 1.048–1.095) for T pathology.

Fig. 2. Comparisons between AD/LB groups and independent effects of LB, Aβ and tau pathologies on cross-sectional clinical outcomes.

a–j, Significant effects (two-sided) were examined with linear regression models using either two AD/LB groups (a–e) or all three pathologies binarized (f–j) in the same model (to examine independent effects) while adjusting for age, sex and education (motor function was not adjusted for education). a,f, Global cognition. b,g, Memory. c,h, Attention/executive function. d,i, Smell. e,j, Motor function. Outcomes were z-scored cognitive tests (a–c,f–h), smell identification test (d,i) and an informant-based motor questionnaire (e,j). a–e, Boxes show interquartile range, horizontal lines are medians and whiskers were plotted using the Tukey method. f–j, Dot/center denotes estimate of the pathology and error bars 95% CI. Red indicates significant association between pathology and worse performance. In total, 941 participants were AD^–/LB^–, 74 AD^–/LB⁺, 147 AD⁺/LB^– and 20 AD⁺/LB⁺; 94 were LB⁺, 304 Aβ⁺ and 195 tau⁺. Extended Data Fig. 2a,b shows the effect on motor function using the UPDRS-III scale (no significant effect of LB pathology). Statistical analyses with corrections for multiple comparisons are shown in Supplementary Fig. 1 (all effects of LB pathology were significant following correction). The effect of LB on clinical outcomes with/without adjustment for Aβ and tau is shown in Extended Data Table 1. Missing data shown in Supplementary Table 2. h, When restricting the analysis of attention/executive function to participants with available SMDT data (n = 854) the results were consistent, showing a significant effect for Aβ (P = 0.01) but not for tau and LB. *P < 0.05, **P < 0.01, ***P < 0.001 (two-sided).

Fig. 3. Independent effect of AD/LB groups and LB, Aβ and tau pathologies on longitudinal cognitive performance.

a–c, Significant effects (two-sided) were examined with LME models focusing on the interaction of AD/LB group × time, adjusted for age, sex and education. a,d, Longitudinal global cognition. b,e, Longitudinal memory. c,f, Longitudinal attention/executive function. d–f, Interaction time × all three pathologies (binarized) was used in the same model to examine the independent effects of each pathology on cognitive progression while adjusting for age, sex and education. Outcomes were z-scored cognitive tests. d–f, Red indicates significant association between pathology and worse cognitive decline. The effect of LB on clinical outcomes with/without adjusting for Aβ and tau is shown in Extended Data Table 3. a–c, Estimated marginal means and 95% CI of means obtained from LME models by AD/LB group. d–f, Dot/center indicates the interaction estimate of time × pathology; error bars 95% CI. In total, 941 participants were AD^–/LB^–, 74 AD^–/LB⁺, 147 AD⁺/LB^– and 20 AD⁺/LB⁺; 94 were LB⁺, 304 Aβ⁺ and 195 tau⁺. Statistical analyses with corrections for multiple comparisons are shown in Supplementary Fig. 3 (all significant differences/associations were still significant after correction). Missing data shown in Supplementary Table 2. *P < 0.05, **P < 0.01, ***P < 0.001 (two-sided).

Fig. 4. Survival curves for progression to PD or DLB stratified by LB status at baseline.

The event was met when a participant fulfilled the clinical criteria for PD or DLB (alternatively, prodromal DLB). Lines show point estimates of survival curves and shaded areas 95% CI. Vertical lines indicate time points of censoring. The table below shows the number of participants at each time point that had not yet progressed to PD/DLB. No participants who were LB^– at baseline progressed to PD/DLB. See Extended Data Table 4 for specifications of progression to a clinical diagnosis based on AD/LB positivity. P value derived from the log-rank test and indicates that the survival curves (that is, time to PD/DLB) of LB⁺ and LB^– participants are significantly different.

Extended Data Fig. 1. ROC analysis for discriminating LB positive versus LB negative participants using the Smell test.

At the highest Youden index (−0.9 z-score on the smell test), the sensitivity was 63% and the specificity 86% with an overall percent correctly classified of 85%. Of those with available Smell test score (n = 398), 371 were LB- and 27 LB+.

Extended Data Fig. 2. Comparison between AD/LB groups and independent effects of LB, Aβ and tau on motor function (UPDRS-III).

The analyses were performed using linear regression models using two AD/LB groups (A) or all three pathologies binarized (B) in the same model (to examine independent effects), adjusted for age, sex, and UPDRS rater (three raters in total). The outcome was z-scored UPDRS-III values. Boxes in a show interquartile range, the horizontal lines are medians and the whiskers were plotted using the Tukey method. b shows the estimate of the pathology with 95% confidence interval. The statistical analyses with corrections for multiple comparisons are shown in Supplementary Fig. 2. 660 participants had UPDRS data, of whom 510 AD-/LB-, 48 AD-/LB+, 90 AD+/LB-, and 12 AD+/LB+. 60 were LB+, 209 Aβ+, and 122 tau+. Since UPDRS-III was not available for all participants, we performed a power analysis to examine what effect size the study was powered to detect (that is, effect of LB pathology on UPDRS in panel B). Using a power calculation for linear regression models, the smallest detectable effect size (f2) was 0.012, at 80% power and alpha=0.05. This suggests that our study was powered to detect even a small effect size of LB pathology on motor function. After applying multiple comparison correction with the FDR method (at α = 0.05), the AD+/LB+ group did not perform significantly different compared with AD-/LB- (Supplementary Fig. 2). * p < 0.05 (two-sided).