The Association Between Neuropathological Lesions and Body Mass Index Is Independent of Cognitive Abilities

Abstract

Background:

The association of moderate and severe dementia with low body mass index (BMI) is well described, but weight decline seems to also occur in individuals with preclinical neuropathologies. Considering that up to one-fifth of individuals with normal cognition meet the criteria for a dementia-related neuropathological diagnosis, autopsy studies are key to detecting preclinical neurodegenerative and cerebrovascular diseases that could be underlying weight changes.

Objective:

We investigated the association between dementia-related brain lesions and BMI and evaluated whether the cognitive function was a mediator of this association.

Methods:

In 1,170 participants, sociodemographic data, clinical history, and cognitive post-mortem evaluation were assessed with an informant. Neuropathological evaluation was performed in all cases. Linear regression models were used to investigate the association between neuropathological lesions (exposure variable) and BMI (outcome) adjusted for demographic, clinical, and cognitive variables in the whole sample, and in only those with normal cognition. Corrections for multiple comparisons were performed. In addition, a mediation analysis was performed to investigate the direct and indirect effects of cognitive abilities on the association between neuropathology and BMI.

Results:

Individuals with lower BMI had a higher burden of neuropathological lesions and poorer cognitive abilities. Only neurofibrillary tangles (NFT) and neuropathological comorbidity were associated with low BMI, while other neurodegenerative and cerebrovascular lesions were not. NFT were indirectly associated with BMI through cognitive abilities, and also directly, even in participants with normal cognition.

Conclusions:

Neurofibrillary tangles were directly associated with low BMI even in individuals with preclinical Alzheimer’s disease.

Keywords

Alzheimer’s disease autopsy body mass index dementia mediation neuropathology

INTRODUCTION

Midlife obesity has been consistently related to a higher risk of cognitive impairment, mainly when adiposity measurements were performed at least 10 years before the cognitive evaluation.1 - 16 In 10,000 participants of the Whitehall II study, 17 obesity at the age of 50 years was a risk factor for dementia incidence after 28 years of follow-up. Higher body mass index (BMI) was associated with accelerated rates of brain atrophy evaluated by magnetic resonance imaging (MRI), particularly in temporal areas.18 - 19 Obesity has been related to a higher risk of Alzheimer’s disease (AD), vascular dementia (VaD), and all-cause dementia in prior studies.20 - 22 Neuropathological studies are in agreement with these clinical findings; higher midlife BMI has been associated with greater severity of amyloid deposition. 14 Midlife BMI may also affect the age-at-onset of AD as each one-unit increase in BMI was related to anticipation of seven months in AD initial symptoms. 14

However, BMI in late life showed a null or protective association with dementia risk.17,23,24 , 17,23,24 Although initially referred to as the obesity paradox, 25 a growing body of evidence suggests the possibility of reverse causation, since early neuropsychiatric symptoms before the onset of cognitive symptoms may lead to weight loss.26,27, 26,27 Indeed, a recent meta-analysis highlighted that the inverse association of high BMI and dementia in older individuals has been only found in studies with short follow-up periods (<9 years). 28 Conversely, studies of late-life BMI with longer follow-ups (10–18 years) showed the harmful consequences of being overweight and obese on increased dementia risk.3,28,29 , 3,28,29 As seen in younger cohorts, obesity at late-life was also associated with brain atrophy.17,30, 17,30 Moreover, in an autopsy study from the National Alzheimer’s Coordinating Center Neuropathology Data Set, 31 late-life increases in BMI over time predicted greater odds of AD pathology.

Although the association of moderate and severe dementia with low BMI has been widely demonstrated,2,17,21,26,32 , 2,17,21,26,32 evidence of this association in older adults with mild dementia or preclinical stages is scarce. 32 The negative impact of dementia over body weight is well documented, 26 and seems to be driven by the presence of cognitive symptoms. However, there is evidence that weight loss can also occur in individuals with normal cognition,26,33, 26,33 which suggests an effect of neuropathological lesions on body weight before clinical dementia. Weight loss is one of the most common manifestations of dementia and may precede its clinical onset by many years.16,33, 16,33 Changes in appetite and neuropsychiatric symptoms before cognitive impairment, such as apathy and depression, may lead to weight loss during the preclinical phase. 32 Since up to 20% of individuals with normal cognition meet the criteria for one or more neuropathological diagnoses,34 - 36 neuropathological studies are important to define the neurodegenerative and cerebrovascular diseases that could be underlying weight changes during the entire dementia course. Therefore, we aimed to investigate the association between dementia-related brain lesions and BMI in participants with different severity of cognitive symptoms, and whether the cognitive abilities were a mediator of this association. We hypothesized that dementia-related neuropathological lesions will be directly associated with BMI, particularly for AD pathology, independent of cognitive symptoms.

METHODS

Participants

Brain samples were part of the Biobank for Aging Studies (BAS) collection, which is an ongoing population-based study that started in 2004. In Brazil, an autopsy exam is a legal requirement for individuals with non-traumatic deaths, whose cause is unclear (unassisted or in which the chain of events that lead to death were not clear from the available data). The Sao Paulo Autopsy Service is the only morgue in the metropolitan area of Sao Paulo, which is the Brazilian most populous city. The deceased’s next-of-kin signed an informed consent after agreeing to donate the deceased’s brain. BAS procedures follow international and Brazilian human research regulations and were approved by local and federal research committees. BAS procedures have been described elsewhere. 37

Inclusion criteria were an age of 50 years or older at the time of death and the presence of a knowledgeable informant, defined as a family member who had at least weekly contact with the deceased in the six months before death and was able to provide clinical and functional information. Brain tissue was obtained within 24 h after death. Samples were excluded when the brain tissue was incompatible with neuropathological analyses (e.g., cerebrospinal fluid pH < 6.5 or major acute brain lesions including hemorrhages). For this study, we additionally excluded cases with missing data for study variables. Of the 1,441 participants from the BAS, 1,170 met the eligibility criteria for this study. Participants were excluded due to age at death less than 50 years old (n = 31), lack of a knowledgeable informant (n = 132), and missing data for covariates (n = 108) (Supplementary Table 1).

The present study was approved by the local research committee (protocol 52182421.2.0000.0068).

Postmortem clinical evaluation

Interviews with knowledgeable informants were performed by trained gerontologists through a semi-structured interview, in a private room, while autopsy procedures were taking place, within 24 h after death. Sociodemographic variables included age, sex, race, and education. Race was reported by the informant and categorized into: White, Black, Brown (mixed of Black and White), Asian, or other race. Education was measured as the number of years of formal education. Hypertension, diabetes mellitus, and dyslipidemia were defined by a previous medical diagnosis. Smoking and alcohol consumption were classified as never, current, or past users, while physical activity was defined as present if the participant engaged in physical activity for at least 150 min per week. Cancer was defined by a previous medical diagnosis or the information on cancer from the full-body autopsy report.

The cognitive abilities were assessed using the informant section of the Clinical Dementia Rating (CDR), 38 and family members were interviewed concerning the deceased’s performance at three months before death to avoid the frequent interference of mental status changes close to death. The CDR is a scale used to stage dementia presence and severity by assessing six domains (memory, orientation, judgment and problem-solving, community affairs, home and hobbies, and personal care). The CDR originates two metrics, the global result (referred to as CDR) and the CDR sum of boxes (CDR-SOB). The CDR is based on the classification of the individual six domains and participant’s cognitive abilities were classified into five categories: normal cognition (CDR 0); questionable dementia (CDR 0.5); mild dementia (CDR 1); moderate dementia (CDR 2); or severe dementia (CDR 3). The CDR sum of boxes (CDR-SOB) results from the sum of the patient’s performance in the individual cognitive domains (ranging from 0 to 18, with increasing severity). The CDR has been previously validated for the evaluation of cognitive and functional abilities by informants in postmortem settings. 39

Upon autopsy, the deceased’s weight and height were measured, without clothes, in the supine position, using an electronic scale and a stadiometer, respectively. BMI was calculated by dividing the weight in kilos by the square of the height in meters.

Neuropathological evaluation

One brain hemisphere was fixed in 4% buffered paraformaldehyde, and selected brain areas of interest from the other hemisphere were frozen at –80°C. The samples from the following areas were embedded in paraffin: middle frontal gyrus, middle and superior temporal gyri, angular gyrus, superior frontal and anterior cingulate gyrus, visual cortex, hippocampal formation at the level of the lateral geniculate body, amygdala, basal ganglia at the level of the anterior commissure, thalamus, midbrain, pons, medulla oblongata, and cerebellum. Blocks were sectioned into 5-μm-thickness sections. All sections were stained with hematoxylin and eosin. Immunohistochemistry with antibodies against amyloid-β (4G8, 1 : 10 000; Signet Pathology Systems, Dedham, Massachusetts), phosphorylated tau (PHF-1, 1 : 2 000), Transactive response DNA-binding Protein-43 (TDP-43) (1 : 500 Proteintech, Chicago, Illinois), and α-synuclein (EqQV-1, 1 : 10.000) were performed in selected sections. Internationally accepted neuropathological criteria were used to diagnose and stage brain pathologies. Diagnoses were made by trained pathologists blinded to clinical status.

AD pathology was scored using the Braak and Braak staging system for neurofibrillary tangles (Braak NFT), 40 into Stages 0 to VI, and the Consortium to Establish a Registry for AD (CERAD) criteria for neuritic plaques, 41 measured as “None”, “Sparse”, “Moderate” or “Frequent”. Neuropathological diagnosis of AD was ascertained for individuals with Braak stage III or above, and with a CERAD score of moderate or frequent.

The evaluation of cerebrovascular lesions was performed both macroscopically, by naked eye examination, and microscopically, using hematoxylin and eosin-stained slides in 13 sampled areas (middle frontal gyrus, middle and superior temporal gyri, angular gyrus, superior frontal and anterior cingulate gyrus, visual cortex, hippocampal formation at the level of the lateral geniculate body, amygdala, basal ganglia at the level of the anterior commissure, thalamus, midbrain, pons, medulla oblongata, and cerebellum). The presence of hyaline arteriolosclerosis (HA) was defined by the degree of vessel thickness, localization, and extension, with diagnosis requiring at least moderate microvascular changes in three or more cortical regions. Additionally, lacunae and large infarcts were registered by topography, stage, size, and number. The infarct variable was considered present in participants with either one large chronic infarct (>1 cm) or three lacunae (<1 cm) in any of the following strategic areas: thalamus, frontocingular cortex, basal forebrain, caudate, medial temporal area, or angular gyrus.

Siderocalcinosis, the development of calcium and iron encrustations in vessel’s middle layers, was ascertained in the basal ganglia, and labeled as present/absent. Hippocampal sclerosis, characterized by pyramidal cell loss and gliosis in CA1 and subiculum of the hippocampal formation, was searched for and also scored as present/absent. Moreover, cerebral amyloid angiopathy (CAA) was analysed, through amyloid-β immunostaining, and evaluated regarding localization (meningeal, gray matter, and/or white matter) and severity, plus the presence of capillary amyloid deposition. CAA was considered present when found in at least three different cortical areas.

Argyrophilic grain disease (AGD) was defined by the presence of phosphorylated tau-positive grains and pretangles in the hippocampus (notably in the hippocampal sectors CA1 and CA2), and oligodendrocytes with coiled bodies in the hippocampal/temporal white matter. 42 Lewy bodies were staged following Braak classification for Parkinson’s disease (Braak PD), 43 and the Lewy body pathology diagnosis was defined by Braak PD stage ≥III. TDP-43 pathology was evaluated with immunohistochemistry using specific antibodies, primarily in the hippocampal formation and amygdala.

The Neuropathological Comorbidity (NPC) score was previously developed considering the neuropathologic lesions (Braak NFT, hippocampal sclerosis, lacunar infarcts, hyaline arteriolosclerosis, siderocalcinosis, and Braak PD) that were associated with dementia in another study with BAS sample. 34 The NPC is a weighted score, where each lesion receives a score based on the effect size of its association with cognitive impairment (Supplementary Table 2). The NPC ranges from 0 to 17 points (with higher values indicating higher neuropathologic comorbidity). Except for NPC, which is a discrete variable, neuropathological markers were measured as categorical variables (Supplementary Table 3). Despite being ordinal variables, Braak NFT and CERAD were treated as continuous variables in our analyses, as corroborated by the literature. 44

Statistical analysis

We used cross sectional data to compare participants classified into four BMI categories (<18.5 kg/m²; 18.5–24.9 kg/m²; 25–29.9 kg/m²; and ≥30 kg/m²) 45 regarding sociodemographic, clinical, and neuropathologic variables using the chi-square and one-way ANOVA tests for categorical and quantitative variables, respectively. Linear regression models were used to investigate the associations of BMI (continuous dependent variable) and neuropathological lesions (independent variable) in separate models, and with all lesions in the same model, to investigate their independent associations with the outcome. Multicollinearity of neuropathological lesions in the same model was assessed by the Variance Inflation Factor (VIF) method. We also investigated the association between the NPC score and BMI using linear regression models. Afterwards, the analysis was repeated removing the Braak NFT score from the NPC score, to assess whether the association was mainly driven by NFTs. The unadjusted linear regression models included only one neuropathologic lesion at a time. Model 1 included each neuropathologic lesion and demographic variables (age, sex, education, and race). We added clinical comorbidities and lifestyle variables (hypertension, diabetes mellitus, dyslipidemia, physical activity, smoking, and alcohol use) to Model 2. We further added cancer in Model 3, while Model 4 included all the previous variables plus the CDR-SOB. The same regression models, except Model 4, were applied in a subanalysis considering only individuals with normal cognition (CDR = 0). All models were corrected for multiple comparisons using the Bonferroni correction. Individuals with and without cognitive abilities impairment were also compared regarding BMI, and sociodemographic and clinical variables (Supplementary Table 4 and Supplementary Figure 1).

The direct and indirect effects of each neuropathological lesion on BMI were also evaluated, to evaluate the role of possible mediation by cognitive abilities. The exposure/outcome (neuropathologic lesions and BMI) and mediator/outcome (CDR-SOB and BMI) associations were evaluated using the aforementioned adjusted linear regression models. Another linear regression model was used to assess the effect of neuropathology over CDR-SOB (exposure/mediator). The direct and indirect effects of each neuropathologic lesion (through CDR-SOB) on BMI were calculated in a mediation analysis by both the difference method, a more traditional approach, and the product method, which was recently described. 46 The high degree of agreement between these two approaches endorsed the robustness of our findings and the fitness of the adopted regression models’ methodology to the variables used. 47 The alpha level was set at 0.05 level in two-sided tests. All analyses were performed through codes written on the Jupyter Notebook platform (v.6.0.3) in Python language (v.3.8.5).

RESULTS

The mean age of the sample was 74.2±11.8 years old, 50% of the individuals were women, and 66% were White. The mean education was 4.4±3.9 years. The demographic characteristics observed in the study participants were similar to those who died in the city of São Paulo in the same period. 34 Comorbidities were prevalent, as 65% had hypertension, 28% diabetes, and 13% cancer (Table 1), and participants had on average 1.7 clinical comorbidities. Regarding cognition, 67% had normal cognition (CDR = 0). On the other hand, only 41% did not have any neuropathological diagnosis. The mean value of CDR-SOB was 3.4±6.2 (Median value = 0; Interquartile range = 0; 3.5). The most frequent neuropathological diagnosis was AD (13.8%), followed by VaD (9.8%), and the combination of both (6.2%) (Table 2). The mean BMI was 23.2±5.1 kg/m². Participants with obesity were younger and had a higher education level than those with normal BMI. They also had a more significant burden of metabolic and cardiovascular diseases, while the diagnosis of cancer was more frequent in participants with low BMI (Table 1). In addition, individuals with low BMI were more likely to have poor cognitive abilities and more often met the neuropathological criteria for AD and mixed AD/VaD diagnoses. Further, they had a higher burden of brain lesions, both on the neuropathologic comorbidity score and when AD pathology (CERAD and Braak NFT), and AGD were assessed independently (Table 2).

Table 1

Characteristics of the sample according to body mass index categories (n = 1,170)

	All	BMI <18.5 kg/m²	BMI 18.5–24.9 kg/m²	BMI 25–29.9 kg/m²	BMI ≥30 kg/m²	p ^*
	N=1,170	N=164	N=663	N=237	N=106
Sociodemographics
Age (y), mean (SD)	74.2 (11.8)	76.7 (12.0)	75.1 (11.6)	71.8 (11.0)	69.7 (12.5)	<0.001
Female gender, %	50.5	61.0	46.8	48.9	61.3	0.001
Race, %						0.166
White	66.3	54.9	68.8	66.2	68.9
Black	10.9	15.2	10.1	10.1	10.4
Brown	20.4	26.2	19.3	20.7	17.9
Asian	2.0	3.1	1.8	1.7	1.9
Missing data	0.4	0.6	0.0	1.3	0.9
Education (y), mean (SD)	4.5 (3.9)	3.9 (3.5)	4.3 (3.8)	5.1 (4.1)	5.3 (4.1)	<0.001
Comorbidities/Lifestyle
Hypertension, %	65.2	50.0	63.1	75.5	78.3	<0.001
Diabetes Mellitus, %	28.2	16.5	27.6	30.8	44.3	<0.001
Dyslipidemia, %	10.3	3.1	8.6	16.0	18.9	<0.001
Stroke, %	14.0	11.6	15.7	12.2	11.3	0.308
Coronary Artery Disease, %	19.8	12.2	19.0	24.9	25.5	0.007
Heart Failure, %	17.1	11.0	16.6	14.3	35.8	<0.001
Cancer, %	12.6	20.1	13.0	8.4	7.5	0.002
Smoking, %						0.182
Current User	22.5	26.8	23.2	21.9	12.3
Past User	29.6	27.4	29.3	29.1	35.8
Never	47.9	45.8	47.5	49.0	51.9
Alcohol, %						0.390
Current User	23.3	20.1	24.3	23.6	21.7
Past User	18.9	24.4	19.0	16.5	15.1
Never	57.8	55.5	56.7	59.9	63.2
Physically inactive, %	59.9	73.2	60.5	52.7	51.9	<0.001
Cognition
CDR, %						<0.001
0	66.9	51.8	65.6	75.9	78.3
0.5	9.4	7.9	9.6	10.1	8.5
1	5.4	6.7	5.9	3.4	4.7
2	5.0	5.5	6.2	3.4	0.9
3	13.2	28.1	12.7	7.2	7.6
CDR-SOB, mean (SD)	3.5 (6.2)	6.3 (7.8)	4.3 (3.8)	2.1 (4.9)	1.8 (4.6)	<0.001

Table 2

Description of neuropathological lesions and diagnoses in the total sample and stratified by body mass index categories (n = 1,170)

All	BMI <18.5 kg/m²	BMI 18.5–24.9 kg/m²	BMI 25–29.9 kg/m²	BMI ≥30 kg/m²	p ^*
	N = 1,170	N = 164	N = 663	N = 237	N = 106
Neuropathological Lesions
CERAD NP, %						0.001
None	61.2	57.3	58.7	63.7	77.3
Sparse	13.8	11.0	14.6	14.3	12.3
Moderate	13.5	12.2	15.4	12.7	5.7
Frequent	11.5	19.5	11.3	9.3	4.7
Braak NFT, %						<0.001
Stages 0–II	61.6	47.0	60.8	68.4	73.6
Stages III–IV	26.8	28.6	27.7	24.0	24.5
Stages V–VI	11.6	24.4	11.5	7.6	1.9
Braak PD, %						0.243
Stages 0–III	93.2	89.6	93.5	93.7	95.3
Stages IV–VI	6.8	10.4	6.5	6.3	4.7
Hippocampal sclerosis, %	2.6	4.3	3.2	1.3	0.0	0.163
Lacunar infarcts, %	10.6	15.2	10.6	9.3	6.6	0.114
Hyaline arteriolosclerosis, %	23.8	26.2	25.3	19.8	19.8	0.224
Siderocalcinosis, %	19.2	21.3	20.4	16.9	14.2	0.305
Cerebral amyloid angiopathy, %	6.8	11.6	6.3	6.3	3.8	0.052
Argyrophilic grain disease, %	22.4	32.9	20.4	23.6	16.0	0.002
TDP-43, %	5.6	7.9	5.6	3.4	6.6	0.248
Neuropathological Comorbidity Score, mean (SD)	2.2 (3.0)	3.4 (3.6)	2.3 (3.1)	1.8 (2.5)	1.2 (1.8)	<0.001
Neuropathological Diagnoses
Normal, %	40.9	22.6	41.0	43.9	62.3	<0.001
AD, %	13.8	15.8	15.2	12.2	4.7	0.022
VaD, %	9.8	9.7	10.4	9.7	6.6	0.682
AD/VaD, %	6.2	11.6	6.2	4.2	2.8	0.008
Mixed dementia Not-AD/VaD, %	6.2	10.4	5.6	5.1	5.7	0.114
Lewy Body/Parkinson’s Disease, %	3.5	4.9	3.5	3.0	2.8	0.736
Other/Not defined, %	19.6	25.0	18.1	21.9	15.1	0.103

^*p-values for the comparisons among BMI categories. Chi-square test was used for categorical variables, and one-way ANOVA test was used for quantitative variables. BMI, body mass index; CERAD, Consortium to Establish a Registry for Alzheimer’s Disease for neuritic plaques; Braak NFT, Braak staging for neurofibrillary tangles; Braak PD, Braak staging for Parkinson’s disease and Lewy body pathology; TDP-43, Transactive DNA-binding Protein 43; AD, Alzheimer’s disease; VaD, vascular dementia; AD/VaD, mixed dementia of Alzheimer’s disease and vascular dementia.

In unadjusted analysis, CERAD NP, Braak NFT, HA, and NPC score had a significant association with BMI. However, when the models were adjusted for demographic and clinical variables, most neuropathologic lesions were not associated with BMI. Lower BMI was only associated with higher Braak NFT (β=–0.060, 95% CI=–0.088; –0.032, p < 0.001) and NPC score (β=–0.084, 95% CI=–0.119; –0.049, p < 0.001) (Table 3). When we removed the Braak NFT score from the NPC score, a significant association with BMI was still seen in the unadjusted regression model. However, no significant association was observed in the subsequent adjusted models (Supplementary Table 5). When we included all neuropathologic lesions in the same model, Braak NFT was consistently associated with lower BMI, even after adjustments (β=–0.061, 95% CI=–0.099; –0.023, p = 0.001). The variation in BMI was mainly due to clinical comorbidities and neuropathological lesions (Adjusted R² = 0.048), being only minorly explained by demographic variables. Additionally, there was no multicollinearity between neuropathological variables in the regression model with all neuropathological lesions (Supplementary Table 6).When we analysed data only from individuals with normal cognition (CDR = 0), the association between Braak NFT and BMI was significant, but the same was not seen with NPC score (Table 4). Furthermore, no associations were found for the NPC score without the Braak NFT component in the models considering only individuals with CDR = 0 (Supplementary Table 5).We found a direct effect of Braak NFT (β=–0.048, 95% CI=–0.078; –0.018, p = 0.002) and NPC score (β=–0.067, 95% CI=–0.107; –0.027, p = 0.001) on BMI, besides an indirect effect mediated by cognitive abilities (CDR-SOB) (Fig. 1 and Supplementary Table 7). Braak NFT and NPC scores were also associated with CDR-SOB (Supplementary Table 7). However, the direct association was at least four times greater than that of the association mediated by cognitive abilities, for both neuropathological variables.

Fig. 1

Mediation analysis for the association between the Neuropathological Comorbidity (NPC) Score and Braak staging for neurofibrillary tangles (Braak NFT) with Body Mass Index (BMI), considering a direct and indirect effect mediated by cognitive abilities, evaluated by Clinical Dementia Rating sum of boxes (CDR-SOB). (A) Findings for the whole sample; and (B) Findings when considering only individuals with normal cognition.

Table 3

Association between neuropathological lesions and body mass index (n = 1,170)

	Unadjusted			Model 1			Model 2			Model 3			Model 4
	β^*	CI (95%)	p	β^*	CI (95%)	P	β^*	CI (95%)	p	β^*	CI (95%)	p	β^*	CI (95%)	p
CERAD NP	–0.043	–0.062; –0.024	<0.001	–0.023	–0.044; –0.002	0.031	–0.021	–0.041; –0.001	0.043	–0.022	–0.042; –0.002	0.035	–0.009	–0.031; 0.012	0.386
Braak NFT	–0.090	–0.113; –0.066	<0.001	–0.058	–0.087; –0.029	<0.001	–0.057	–0.084; –0.029	<0.001	–0.060	–0.088; –0.032	<0.001	–0.048	–0.078; –0.018	0.002
Braak PD	–0.044	–0.076; –0.012	0.007	–0.029	–0.061; 0.003	0.072	–0.022	–0.053; 0.009	0.158	–0.023	–0.054; 0.008	0.143	–0.013	–0.044; 0.018	0.422
Hippocampal sclerosis	–0.059	–0.101; –0.016	0.007	–0.063	–0.107; –0.020	0.004	–0.055	–0.097; –0.013	0.010	–0.056	–0.098; –0.015	0.008	–0.048	–0.089; –0.006	0.024
Lacunar infarcts	–0.023	–0.046; –0.001	0.039	–0.013	–0.035; 0.009	0.259	–0.009	–0.030; 0.013	0.420	–0.010	–0.032; 0.011	0.339	–0.004	–0.025; 0.018	0.740
Hyaline arteriolosclerosis	–0.024	–0.040; –0.008	0.004	–0.016	–0.032; 0.000	0.047	–0.017	–0.032; –0.001	0.037	–0.018	–0.034; –0.003	0.021	–0.015	–0.030; 0.001	0.062
Siderocalcinosis	–0.021	–0.039; –0.004	0.017	–0.013	–0.030; 0.004	0.135	–0.010	–0.027; 0.006	0.228	–0.011	–0.027; 0.006	0.197	–0.009	–0.026; 0.007	0.264
Cerebral amyloid angiopathy	–0.023	–0.050; 0.004	0.096	–0.011	–0.038; 0.016	0.413	–0.009	–0.035; 0.017	0.482	–0.012	–0.037; 0.014	0.378	–0.006	–0.032; 0.020	0.655
Argyrophilic grain disease	–0.020	–0.036; –0.003	0.018	–0.015	–0.031; 0.002	0.088	–0.013	–0.029; 0.003	0.113	–0.011	–0.027; 0.005	0.161	–0.014	–0.030; 0.002	0.082
TDP-43	–0.020	–0.050; 0.010	0.196	–0.016	–0.046; 0.014	0.302	–0.007	–0.036; 0.021	0.611	–0.007	–0.036; 0.021	0.618	–0.003	–0.032; 0.026	0.832
Neuropathological Comorbidity Score	–0.121	–0.154; –0.087	<0.001	–0.086	–0.121; –0.050	<0.001	–0.079	–0.114; –0.044	<0.001	–0.084	–0.119; –0.049	<0.001	–0.067	–0.107; –0.027	0.001

Model 1: Linear regression model adjusted for age, sex, race, and education. Model 2: Linear regression model adjusted for age, sex, race, education, hypertension, diabetes, dyslipidemia, physical activity, smoking, and alcohol use. Model 3: Linear regression model adjusted for age, sex, race, education, hypertension, diabetes, dyslipidemia, physical activity, smoking, alcohol use, and cancer. Model 4: Linear regression model adjusted for age, sex, race, education, hypertension, diabetes, dyslipidemia, physical activity, smoking, alcohol use, cancer, and Clinical Dementia Rating sum of boxes. CERAD NP, Consortium to Establish a Registry for Alzheimer’s Disease for neuritic plaques; Braak NFT, Braak staging for neurofibrillary tangles; Braak PD, Braak staging for Parkinson’s disease and Lewy body pathology; TDP-43, Transactive DNA-binding Protein 43.

Table 4

Association between neuropathological lesions and body mass index among individuals with normal cognition (n = 783)

Unadjusted			Model 1			Model 2			Model 3
β^*	CI (95%)	p	β^*	CI (95%)	p	β^*	CI (95%)	p	β^*	CI (95%)	p
CERAD NP	–0.040	–0.070; –0.011	0.007	–0.011	–0.042; 0.021	0.516	–0.014	–0.044; 0.016	0.361	–0.011	–0.041; 0.018	0.453
Braak NFT	–0.095	–0.132; –0.059	<0.001	–0.089	–0.137; –0.042	<0.001	–0.075	–0.118; –0.032	0.001	–0.074	–0.116; –0.031	0.001
Braak PD	–0.052	–0.102; –0.002	0.043	–0.019	–0.070; 0.033	0.479	–0.026	–0.074; 0.023	0.303	–0.027	–0.075; 0.022	0.280
Hippocampal sclerosis	–0.043	–0.127; 0.042	0.324	–0.024	–0.108; 0.060	0.575	–0.035	–0.116; 0.045	0.387	–0.043	–0.123; 0.036	0.287
Lacunar infarcts	–0.013	–0.048; 0.023	0.483	–0.008	–0.043; 0.027	0.661	–0.007	–0.041; 0.027	0.682	–0.009	–0.042; 0.024	0.600
Hyaline arteriolosclerosis	–0.018	–0.039; 0.003	0.099	–0.007	–0.028; 0.014	0.507	–0.016	–0.036; 0.005	0.137	–0.014	–0.035; 0.006	0.161
Siderocalcinosis	–0.012	–0.035; 0.012	0.329	–0.004	–0.027; 0.019	0.730	0.003	–0.019; 0.025	0.816	0.003	–0.019; 0.024	0.825
Cerebral amyloid angiopathy	–0.026	–0.067; 0.014	0.205	–0.015	–0.055; 0.026	0.475	–0.023	–0.062; 0.016	0.245	–0.025	–0.063; 0.013	0.198
Argyrophilic grain disease	–0.032	–0.052; –0.012	0.002	–0.025	–0.045; –0.004	0.018	–0.020	–0.039; 0.000	0.053	–0.021	–0.040; –0.001	0.037
TDP-43	–0.009	–0.053; 0.035	0.688	0.007	–0.036; 0.050	0.749	0.015	–0.026; 0.057	0.469	0.018	–0.023; 0.059	0.400
Neuropathological Comorbidity Score	–0.118	–0.184; –0.052	<0.001	–0.067	–0.135; 0.002	0.056	–0.074	–0.141; –0.008	0.029	–0.076	–0.142; –0.011	0.023

Model 1: Linear regression model adjusted for age, sex, race, and education. Model 2: Linear regression model adjusted for age, sex, race, education, hypertension, diabetes, dyslipidemia, physical activity, smoking, and alcohol use. Model 3: Linear regression model adjusted for age, sex, race, education, hypertension, diabetes, dyslipidemia, physical activity, smoking, alcohol use, and cancer. CERAD NP: Consortium to Establish a Registry for Alzheimer’s Disease for neuritic plaques; Braak NFT: Braak staging for neurofibrillary tangles; Braak PD: Braak staging for Parkinson’s disease and Lewy body pathology; TDP-43: Transactive DNA-binding Protein 43.

DISCUSSION

In this study, individuals with lower BMI were more likely to have a higher burden of neuropathological lesions and poorer cognitive and functional abilities. NFT burden and NPC were associated with lower BMI in adjusted models applied to the whole sample. Furthermore, when we performed a mediation analysis, only a small part of the association was mediated by cognitive abilities, while the direct effect of neuropathology (mainly NFT and NPC) on BMI accounted for 80% of the total association. In participants with normal cognition, we found an association of BMI with Braak NFT, reinforcing the importance of these lesions on BMI before the onset of cognitive impairment.

Prior findings on the association between dementia and BMI have already suggested that neuropathological changes and cognitive impairment may lead to weight loss.26,27, 26,27 A higher risk of dementia was found in underweight older individuals from the Cardiovascular Health Cognition Study.16,26, 16,26 In addition, a decrease in BMI during late life might be an early sign that can be present before the cognitive symptoms. 48 Although BMI decline is more pronounced in individuals already exhibiting cognitive symptoms, 26 weight loss might also occur in cognitively normal individuals with preclinical neuropathological processes.26,33,49,50 , 26,33,49,50 These findings are in line with longitudinal studies of BMI trajectories along the dementia course, which showed a downward trend in weight before the clinical onset. 51 This finding had been seen in participants as young as those in their late 50 s. 26 In our study, weight loss seemed to precede cognitive symptoms, once the association between neuropathological lesions and lower BMI was ascertained in the subset with normal cognition as well. Therefore, neurodegenerative diseases can be related to underlying weight changes before the onset of cognitive symptoms. Once a greater understanding of the mechanisms that guide weight variation is established, weight loss might be useful as a possible early marker of neuropathological processes, thus favoring earlier investigation, diagnosis and, eventually, early interventions for preclinical phases.

Except for AGD, all neuropathological lesions observed in this study have already been independently associated with cognitive impairment and dementia. 34 A higher neuropathological burden, reflected by higher NPC scores, was associated with higher CDR-SOB scores. 34 Moreover, NFTs and NPC were related to lower BMI in adjusted analyses. As NFTs were the neuropathologic lesion that contributed most to the NPC score, 34 it is plausible that the deposition of these lesions was the main driver of the weight loss in our sample. This finding is reinforced by the fact that, when we removed the Braak NFT variable from the NPC score, the association between the NPC score and BMI lost significance in the adjusted models. Also, no significant association was observed when the same analysis was repeated in the subset of individuals with normal cognitive abilities. In other words, when the Braak NFT was removed from the NPC score, the association between NPC and BMI was only mediated by CDR-SOB (indirect effect). Further, AD pathology has been associated with decreased late-life BMI.51,52, 51,52 Indeed, in individuals with normal cognition, lower late-life BMI predicted greater odds of AD pathology. 31 On the other hand, previous studies could not demonstrate the same association for patients with dementia.31,51, 31,51 Nonetheless, our findings confirm the association between lower BMI by the time of death and NFTs, regardless of CDR-SOB score.

Brain structures where AD pathology is commonly found (e.g. olfactory bulbs, hippocampus, and amygdala) are hypothesized to be major determinants of weight control. Those areas have been linked to weight/energy regulation mechanisms,51,53, 51,53 neuropsychiatric symptoms, such as depression and apathy, 51 food perception (through olfaction and taste), 54 and the drive to eat.51,55, 51,55 In addition, Braak staging is based on the stereotypic spread of NFT in AD from transentorhinal regions to hippocampal areas, and then to limbic areas, such as the amygdala. 40 Hence, is plausible that Braak NFT staging has a consistent association with weight loss mechanisms and with BMI, as we found in our study. Despite subjects with AD pathology having a persistently lower baseline BMI during follow-up compared to individuals without AD pathology, BMI did not change in 1,421 patients from the National Alzheimer’s Coordinating Centers. 51 Therefore, while individuals with AD pathology consistently had lower body weight over time, they did not have a faster decline in BMI. However, an alternative hypothesis could explain the BMI stability in that study. Individuals with dementia probably receive more frequent healthcare attention, including closer monitoring of weight and diet, and they are more often exposed to medications that induce weight gain (e.g., antipsychotics).51,56, 51,56

Our study has several advantages. We performed a large autopsy study, which is not common when investigating the association between BMI and cognition in older populations. Our study is population-based, which contributed to reducing selection bias and potential underrepresentation of minority groups, increasing the study’s external validity. Among the unique characteristics of our sample, we have a large number of individuals with normal cognition and different socioeconomic profiles with less education and admixed race compared to previous studies.31,49,51 , 31,49,51 The neuropathological evaluation was extensive, including the evaluation of less commonly assessed abnormalities, as TDP-43 and argyrophilic grain disease. Moreover, we adjusted the analyses for several potential confounders, including cancer information from the full-body autopsy report (considering its potential effect on weight loss unrelated to dementia). 32 A formal mediation analysis, which was possible because of our large sample, was also performed to address the participation of possible mediators. However, our study must be interpreted considering several limitations. As we did not follow participants during life, clinical variables and cognition were obtained retrospectively, through an interview with a proxy. The cognitive post-mortem evaluation has been validated and showed good accuracy for dementia diagnosis in a previous validation study. 39 Additionally, given the study design, we could not measure BMI before death at multiple time points, which precluded us from documenting the weight trajectories before death. Moreover, a postmortem measure of BMI may reflect an acute weight loss related to the cause of death. Even considering these limitations, postmortem measures of BMI were associated with dementia and atherosclerosis in previous studies.32,57, 32,57 Another study limitation is that we used the CERAD score to classify neuritic plaques since the more recent classification proposed by Thal et al. 58 is not available for our entire collection. Also, data on atherosclerosis of the circle of Willis and APOE status were not available for most individuals in our sample, so they were not included in the analyses. Even though we have a large sample, some pathologies were infrequent (e.g., HS had a frequency of 2.6%), which could explain the absence of association between these lesions and BMI. Furthermore, although mediation analysis only played a complementary role in our study, the adequate interpretation of this analysis depends on assumptions that are subject to errors, such as model misspecification (linear association between BMI and cognition or non-additive effects of neuropathology and cognition on BMI), absence of a potential interaction between the neuropathology and cognitive function, and small different results from the difference and product approaches.

In conclusion, we investigated the association of neuropathological lesions and BMI in a large and heterogeneous sample of participants. In concordance with previous studies, NFTs were associated with lower BMI. Our findings also support the reverse causation theory, since the association between neuropathology and BMI was present even in individuals with normal cognitive abilities. Longitudinal clinicopathological studies are needed to confirm our findings and deepen the understanding of the possible bidirectional associations between BMI and neuropathology over time in older individuals.

AUTHOR CONTRIBUTIONS

Raul dos Reis Ururahy (Conceptualization; Data curation; Formal analysis; Writing – original draft; Writing – review & editing); Marina Scott do Val (Data curation; Formal analysis; Software; Writing – original draft; Writing – review & editing); Aline Maria Macagnan Ciciliati (Methodology; Writing – original draft); Renata Elaine Paraizo Leite (Methodology; Supervision); Vitor Ribeiro Paes (Data curation; Investigation; Methodology; Writing – review & editing); Roberta Diehl Rodrigues (Methodology; Supervision); Lea Tenenholz Grinberg (Data curation; Methodology; Supervision; Writing – review & editing); Carlos Augusto Pasqualucci (Data curation; Investigation; Methodology); Wilson Jacob Filho (Methodology; Supervision); Claudia Kimie Suemoto (Conceptualization; Data curation; Formal analysis; Funding acquisition; Methodology; Project administration; Supervision; Validation; Writing – original draft; Writing – review & editing).

Footnotes

ACKNOWLEDGMENTS

The authors have no acknowledgments to report.

FUNDING

Alzheimer’s Association Research Grant (AARG-20-678884).

CONFLICT OF INTEREST

The authors have no conflict of interest to report.

DATA AVAILABILITY

The data that support the findings of this study are available upon request after analysis from the BAS executive board.

The supplementary material is available in the electronic version of this article: .

References

Kivipelto

, Ngandu

, Fratiglioni

, et al. Obesity and vascular risk factors at midlife and the risk of dementia and Alzheimer disease. Arch Neurol 2005; 62: 1556–1560.

Whitmer

, Gustafson

, Barrett-Connor

, et al. Central obesity and increased risk of dementia more than three decades later. Neurology 2008; 71: 1057–1064.

Gustafson

, Rothenberg

, Blennow

, et al. An 18-year follow-up of overweight and risk of Alzheimer disease. Arch Intern Med 2003; 163: 1524–1528.

Hayden

, Zandi

, Lyketsos

, et al. Vascular risk factors for incident Alzheimer disease and vascular dementia: the Cache County study. Alzheimer Dis Assoc Disord 2006; 20: 93–100.

Rosengren

, Skoog

, Gustafson

, et al. Body mass index, other cardiovascular risk factors, and hospitalization for dementia. Arch Intern Med 2005; 165: 321–326.

Whitmer

, Gunderson

, Barrett-Connor

, et al. Obesity in middle age and future risk of dementia: a 27 year longitudinal population based study. BMJ 2005; 330: 1360.

Chiang

, Yip

, Wu

, et al. Midlife risk factors for subtypes of dementia: a nested case-control study in Taiwan. Am J Geriatr Psychiatry 2007; 15: 762–771.

Whitmer

, Gunderson

, Quesenberry

Jr , et al. Body mass index in midlife and risk of Alzheimer disease and vascular dementia. Curr Alzheimer Res 2007; 4: 103–109.

Dahl

, Hassing

, Fransson

, et al. Body mass index across midlife and cognitive change in late life. Int J Obes (Lond) 2013; 37: 296–302.

10.

Dahl

, Hassing

, Fransson

, et al. Being overweight in midlife is associated with lower cognitive ability and steeper cognitive decline in late life. J Gerontol A Biol Sci Med Sci 2010; 65: 57–62.

11.

, Atti

, Gatz

, et al. Midlife overweight and obesity increase late-life dementia risk: a population-based twin study. Neurology 2011; 76: 1568–1574.

12.

Hassing

, Dahl

, Thorvaldsson

, et al. Overweight in midlife and risk of dementia: a 40-year follow-up study. Int J Obes (Lond) 2009; 33: 893–898.

13.

Gorospe

and Dave

. The risk of dementia with increased body mass index. Age Ageing 2007; 36: 23–29.

14.

Chuang

, An

, Bilgel

, et al. Midlife adiposity predicts earlier onset of Alzheimer’s dementia, neuropathology and presymptomatic cerebral amyloid accumulation. Mol Psychiatry 2016; 21: 910–915.

15.

Debette

, Beiser

, Hoffmann

, et al. Visceral fat is associated with lower brain volume in healthy middle-aged adults. Ann Neurol 2010; 68: 136–144.

16.

Fitzpatrick

, Kuller

, Lopez

, et al. Midlife and late-life obesity and the risk of dementia: cardiovascular health study. Arch Neurol 2009; 66: 336–342.

17.

Singh-Manoux

, Dugravot

, Shipley

, et al. Obesity trajectories and risk of dementia: 28 years of follow-up in the Whitehall II Study. Alzheimers Dement 2018; 14: 178–186.

18.

Enzinger

, Fazekas

, Matthews

, et al. Risk factors for progression of brain atrophy in aging: six-year follow-up of normal subjects. Neurology 2005; 64: 1704–1711.

19.

Gustafson

, Lissner

, Bengtsson

, et al. A 24-year follow-up of body mass index and cerebral atrophy. Neurology 2004; 63: 1876–1881.

20.

Beydoun

, Beydoun

and Wang

. Obesity and central obesity as risk factors for incident dementia and its subtypes: a systematic review and meta-analysis. Obes Rev 2008; 9: 204–218.

21.

Anstey

, Cherbuin

, Budge

, et al. Body mass index in midlife and late-life as a risk factor for dementia: a meta-analysis of prospective studies. Obes Rev 2011; 12: e426–e437.

22.

Emmerzaal

, Kiliaan

and Gustafson

. – a decade of body mass index, Alzheimer’s disease, and dementia. J Alzheimers Dis 2015; 43: 739–755.

23.

Dahl

, Löppönen

, Isoaho

, et al. Overweight and obesity in old age are not associated with greater dementia risk. J Am Geriatr Soc 2008; 56: 2261–2266.

24.

Power

, Alfonso

, Flicker

, et al. Body adiposity in later life and the incidence of dementia: the health in men study. PLoS One 2011; 6: e17902.

25.

Oreopoulos

, Kalantar-Zadeh

, Sharma

, et al. The obesity paradox in the elderly: potential mechanisms and clinical implications. Clin Geriatr Med 2009; 25: 643–659, viii.

26.

Suemoto

, Gilsanz

, Mayeda

, et al. Body mass index and cognitive function: the potential for reverse causation. Int J Obes (Lond) 2015; 39: 1383–1389.

27.

Rabin

, Shirzadi

, Swardfager

, et al. Amyloid-beta burden predicts prospective decline in body mass index in clinically normal adults. Neurobiol Aging 2020; 93: 124–130.

28.

Danat

, Clifford

, Partridge

, et al. Impacts of overweight and obesity in older age on the risk of dementia: a systematic literature review and a meta-analysis. J Alzheimers Dis 2019; 70(s1): S87–S99.

29.

Danat

and Chen

. Do older adults obesity and overweight prevent dementia risk? – A new population-based cohort study from China. Alzheimer’s Disease, Dementia Care Research and Awareness Conference. Boston, Massachusetts, USA. August 31– September 01, 2018. Available from: https://alzheimers.global-summit.com/scientificprogram.

30.

, Raji

, Becker

, et al. Cardiovascular Health Study; ADNI. Obesity is linked with lower brain volume in 700 AD and MCI patients. Neurobiol Aging 2010; 31: 1326–1339.

31.

Besser

, Alosco

, Ramirez Gomez

, et al. Late-life vascular risk factors and Alzheimer disease neuropathology in individuals with normal cognition. J Neuropathol Exp Neurol 2016; 75: 955–962.

32.

Ciciliati

AMM

, Adriazola

, Souza Farias-Itao

, et al. Severe dementia predicts weight loss by the time of death. Front Neurol 2021; 12: 610302.

33.

Knopman

, Edland

, Cha

, et al. Incident dementia in women is preceded by weight loss by at least a decade. Neurology 2007; 69: 739–746.

34.

Suemoto

, Ferretti-Rebustini

, Rodriguez

, et al. Neuropathological diagnoses and clinical correlates in older adults in Brazil: A cross-sectional study. PLoS Med 2017; 14: e1002267.

35.

Neuropathology Group; Medical Research Council Cognitive Function and Aging Study. Pathological correlates of late-onset dementia in a multicentre, community-based population in England and Wales. Lancet 2001; 357: 169–175.

36.

Polvikoski

, Sulkava

, Myllykangas

, et al. Prevalence of Alzheimer’s disease in very elderly people: a prospective neuropathological study. Neurology 2001; 56: 1690–1696.

37.

Grinberg

, Ferretti

, Farfel

, et al. Brain bank of the Brazilian aging brain study group – a milestone reached and more than 1,600 collected brains. Cell Tissue Bank 2007; 8: 151–162.

38.

Morris

. The Clinical Dementia Rating (CDR): current version and scoring rules. Neurology 1993; 43: 2412–2414.

39.

Ferretti

REL

, Damin

, Brucki

SMD

, et al. Post-mortem diagnosis of dementia by informant interview. Dement Neuropsychol 2010; 4: 138–144.

40.

Braak

and Braak

. Neuropathological stageing of Alzheimer-related changes. Acta Neuropathol 1991; 82: 239–259.

41.

Mirra

, Heyman

, McKeel

, et al. The Consortium to Establish a Registry for Alzheimer’s Disease (CERAD). Part II. Standardization of the neuropathologic assessment of Alzheimer’s disease. Neurology 1991; 41: 479–486.

42.

Saito

, Ruberu

, Sawabe

, et al. Staging of argyrophilic grains: an age-associated tauopathy. J Neuropathol Exp Neurol 2004; 63: 911–918.

43.

Jellinger

. A critical reappraisal of current staging of Lewy-related pathology in human brain. Acta Neuropathol 2008; 116: 1–16.

44.

Robitzsch

. Why ordinal variables can (almost) always be treated as continuous variables: clarifying assumptions of robust continuous and ordinal factor analysis estimation methods. Front Educ 2020; 5: 589965.

45.

WHO Physical Status: The Use and Interpretation of Anthropometry: Report of a World Health Organization (WHO) Expert Committee. Geneva, Switzerland: World Health Organization, 1995.

46.

Hayes

and Rockwood

. Regression-based statistical mediation and moderation analysis in clinical research: Observations, recommendations, and implementation. Behav Res Ther 2017; 98: 39–57.

47.

VanderWeele

. Mediation analysis: A practitioner’s guide. Annu Rev Public Health 2016; 37: 17–32.

48.

Besser

, Gill

, Monsell

, et al. Body mass index, weight change, and clinical progression in mild cognitive impairment and Alzheimer disease. Alzheimer Dis Assoc Disord 2014; 28: 36–43.

49.

Stewart

, Masaki

, Xue

, et al. A 32-year prospective study of change in body weight and incident dementia: the Honolulu-Asia Aging Study. Arch Neurol 2005; 62: 55–60.

50.

Johnson

, Wilkins

and Morris

. Accelerated weight loss may precede diagnosis in Alzheimer disease. Arch Neurol 2006; 63: 1312–1317.

51.

Alosco

, Duskin

, Besser

, et al. Modeling the relationships among late-life body mass index, cerebrovascular disease, and Alzheimer’s disease neuropathology in an autopsy sample of 1,421 subjects from the National Alzheimer’s Coordinating Center Data Set. J Alzheimers Dis 2017; 57: 953–968.

52.

Buchman

, Schneider

, Wilson

, et al. Body mass index in older persons is associated with Alzheimer disease pathology. Neurology 2006; 67: 1949–1954.

53.

Cenquizca

and Swanson

. Analysis of direct hippocampal cortical field CA1 axonal projections to diencephalon in the rat. J Comp Neurol 2006; 497: 101–114.

54.

Albers

, Gilmore

, Kaye

, et al. At the interface of sensory and motor dysfunctions and Alzheimer’s disease. Alzheimers Dement 2015; 11: 70–98.

55.

Guyonnet

, Nourhashemi

, Reyes-Ortega

, et al. La perte de poids chez les sujets présentant une démence de type Alzheimer [Weight loss in patients with Alzheimer-type dementia]. Rev Med Interne 1997; 18: 776–785.

56.

, Scarmeas

, Cosentino

, et al. Change in body mass index before and after Alzheimer’s disease onset. Curr Alzheimer Res 2014; 11: 349–356.

57.

Nishizawa

, Suemoto

, Farias-Itao

, et al. Morphometric measurements of systemic atherosclerosis and visceral fat: Evidence from an autopsy study. PLoS One 2017; 12: e0186630.

58.

Thal

, Rüb

, Orantes

, et al. Phases of A beta-deposition in the human brain and its relevance for the development of AD. Neurology 2002; 58: 1791–1800.