Association of TDP-43 proteinopathy, cerebral amyloid angiopathy, and Lewy bodies with cognitive impairment in individuals with or without Alzheimer's disease neuropathology

Abstract

Alzheimer's disease patients typically present with multiple co-morbid neuropathologies at autopsy, but the impact of these pathologies on cognitive impairment during life is poorly understood. In this study, we developed cognitive trajectories for patients with common co-pathologies in the presence and absence of Alzheimer's disease neuropathology. Cognitive trajectories were modelled in a Bayesian hierarchical regression framework to estimate the effects of each neuropathology on cognitive decline as assessed by the mini-mental state examination and the clinical dementia rating scale sum of boxes scores. We show that both TDP-43 proteinopathy and cerebral amyloid angiopathy associate with cognitive impairment of similar magnitude to that associated with Alzheimer's disease neuropathology. Within our study population, 63% of individuals given the 'gold-standard' neuropathological diagnosis of Alzheimer's disease in fact possessed either TDP-43 proteinopathy or cerebral amyloid angiopathy of sufficient severity to independently explain the majority of their cognitive impairment. This suggests that many individuals diagnosed with Alzheimer's disease may actually suffer from a mixed dementia, and therapeutics targeting only Alzheimer's disease-related processes may have severely limited efficacy in these co-morbid populations.

Figure 1

Venn diagram of the prevalence of CAA, Lewy bodies, and TDP-43 proteinopathy in the presence of different severities of Alzheimer's disease neuropathological change. ADNC Alzheimer’s disease neuropathological change, CAA cerebral amyloid angiopathy. Percentages represent the proportion of ADNC− (left) or ADNC + (right) individuals with the indicated co-pathological signature.

Figure 2

Trajectories of cognitive decline in individuals with or without Alzheimer’s disease neuropathological change and other neuropathological changes at autopsy. Scores have been predicted using a hierarchical model with random intercepts and slopes for the following values of the covariates—72.5 years age at death, APOE ε4 positive, Male, 15.5 years of education, positive for the presence of some infarcts or lacunes, with remaining two co-pathologies positive. Solid and dashed curves are the expected cognitive scores. Shaded regions represent 95% Bayesian credible intervals. MMSE mini mental state examination (low score = worse cognition). CDR-SB clinical dementia rating scale sum of boxes (high score = worse cognition). ADNC Alzheimer’s disease neuropathological change. CAA cerebral amyloid angiopathy. LB lewy bodies. In the models with three-way interactions our reference category was ADNC− and absence of a specific co-morbid pathology (similar to above, one model each for TDP-43, CAA and Lewy Bodies). We found that in ADNC− individuals, having TDP-43 proteinopathy (β_MMSE = − 1.19, 95% BCI (− 1.71, − 0.69); β_CDR-SB = 0.70, 95% BCI (0.35, 1.06)) or CAA (β_MMSE = -0.89, 95% BCI (− 1.67, − 0.05); β_CDR-SB = 0.50, 95% BCI (− 0.02, 1.04)) was associated with a steeper rate of cognitive decline compared to the corresponding reference category, while the presence of Lewy bodies (β_MMSE = − 0.12, 95% BCI (− 0.62, 0.40); β_CDR-SB = 0.11, 95% BCI (− 0.24, 0.45)) was not associated in the same manner. The association between the presence TDP-43 proteinopathy or CAA with cognitive decline in ADNC− individuals was of similar magnitude to the association between being ADNC + in the absence of TDP-43 proteinopathy (β_MMSE = − 1.28, 95% BCI (− 1.59, − 0.96); β_CDR-SB = 0.78, 95% BCI (0.56, 0.99)) or CAA (β_MMSE = − 1.07, 95% BCI (− 1.40, − 0.75); β_CDR-SB = 0.70, 95% BCI (0.48, 0.92)). The estimated proportion of ADNC + individuals that also have substantial TDP-43 proteinopathy, CAA, or both is 0.63, 95% CI (0.59, 0.68). Bayesian beta-regression models were produced as sensitivity analyses corresponding to each previous model. In almost all cases, the directionality and strength of association (measured by the width of 95% BCIs and inclusion of the null value zero) of our results were confirmed in the beta-regression model (see Supplementary Table S1 and Supplementary Table S2 online).