Sex differences in dementia pathology and cognitive performance in a population‐based, ethnically diverse Brazilian autopsy study

Abstract

INTRODUCTION

Sex differences in dementia‐related neuropathology are understudied in diverse populations.

METHODS

We analyzed sex differences in neuropathological and cognitive data from the Brazilian Biobank for Aging Studies. Cognitive performance was evaluated with the Clinical Dementia Rating–Sum of Boxes (CDR‐SOB). Linear and logistic regression models were conducted, including interaction terms for age, race, and education.

RESULTS

In 2229 participants (50.7% female, mean age ± SD 75.4 ± 12.3 years, 62.1% White), female sex was associated with a higher odds of AD pathology (Braak: odds ratio [OR] = 1.42, 95% confidence interval [CI]  = 1.17–1.72; Consortium to Establish a Registry for Alzheimer's Disease (CERAD): OR = 1.58, 95% CI = 1.27–1.98), and trans‐activation response (TAR) DNA‐binding protein 43 (TDP‐43) (OR = 1.77, IC 95% = 1.22–2.59), and lower odds of Lewy body disease (OR = 0.69, 95% CI = 0.51–0.95). Female participant had worse cognitive performance (β = 1.56, 95% CI = 1.03–2.10). Sex modified associations of Braak, CERAD, TDP‐43, and cerebral amyloid angiopathy with cognition. Age also interacted with sex and pathology on CDR‐SOB.

DISCUSSION

Sex differences in the associations between sex and neuropathology suggest the need for sex‐informed dementia research.

Highlights

• Female sex was associated with higher odds of Alzheimer's disease (AD) and trans‐activation response (TAR) DNA‐binding protein 43 (TDP‐43) pathology, and lower odds of Lewy body disease.

• Female participants had worse cognitive performance, and sex modified the association of AD pathology, TDP‐43, and cerebral amyloid angiopathy with cognition, with more pronounced differences in older participants.

• Education and race did not modify the association between sex, neuropathology, and cognition.

1. BACKGROUND

Epidemiological studies showed that Alzheimer's disease (AD) is two to three times more frequent in female than in male individuals. ¹ However, previous studies about sex differences related to other forms of dementia, such as vascular dementia (VaD) and Lewy body disease (LBD), showed inconsistent results. ² , ³ Although some clinical studies reported a higher frequency of LBD in men, a community‐based study found a slight predominance of LBD in women and no sex differences in Parkinson's disease (PD) or PD dementia. ⁴ , ⁵

Notably, the increased risk of dementia is concentrated disproportionately in low and middle‐income countries (LMICs), where two‐thirds of people with dementia live, suggesting that socioeconomic factors may contribute to an increased dementia prevalence in LMICs. ⁶ , ⁷ Indeed, sex disparities in dementia risk in LMICs may be further influenced by socioeconomic inequalities, such as differences in access to education, high‐complexity occupations, and healthcare. ⁷ Despite the high dementia burden, studies examining sex disparities in dementia are scarce in LMICs. ⁷ , ⁸ Moreover, the few neuropathological studies that have explored sex‐related differences in dementia have been conducted primarily in high‐income countries, particularly in the United States. ⁹ , ¹⁰ Most studies on this topic used community‐based or convenience samples, with a main focus on AD pathology, and included mostly White individuals age 65 or older, with high levels of education. ⁹ , ¹⁰ In a community‐based sample of 1453 older participants, women had higher levels of AD pathology, particularly tau pathology. ¹⁰ In addition, compared to men, women were more likely to have more severe arteriolosclerosis and less likely to have gross infarcts. ¹⁰ Population‐based autopsy studies in LMICs are essential to overcoming these limitations, as they enable comprehensive assessment of neuropathology in diverse populations. ¹¹ In this context, our study aimed to investigate the associations of sex with dementia‐related neuropathologies, as well as whether these associations with cognitive performance differ by sex, using a large and ethnically diverse Brazilian sample with low education.

2. METHODS

2.1. Participants

We analyzed data collected from 2004 to 2026 from participants of the Biobank for Aging Studies (BAS) at the University of São Paulo Medical School, Brazil. An autopsy is mandatory in Brazil for individuals with a natural death of unknown cause without suspicion of a traumatic or violent cause. A detailed description of the BAS procedures can be found elsewhere. ¹² The BAS sample includes individuals 18 years or older at death. The next of kin (NOK) provided informed consent for brain collection and provided clinical information and had at least weekly contact with the deceased in the 6 months preceding death. ¹³ Exclusion criteria for the BAS were a post‐mortem interval greater than 24 h, incomplete or inconsistent clinical information provided by the NOK, or brain tissue unsuitable for neuropathological examination, such as cerebrospinal fluid (CSF) pH below 6.5. ¹⁴ For this study, we further restricted the analytical sample to participants 50 years of age or older, considering that dementia‐related neuropathology is more frequent in older individuals. We also excluded participants with missing data for study variables. The final analytical sample comprised 2229 participants after applying the eligibility criteria. For race‐interaction and race‐stratified analyses, we excluded Asian participants (n = 50) from the main sample due to small sample size, leaving 2179 participants. For sensitivity analyses with the semiquantitative classification for hyaline arteriolosclerosis (HA) and cerebral amyloid angiopathy (CAA), we excluded participants with missing data for these variables (HA: n = 731; CAA: n = 4), leaving an analytical sample of 1494 participants (Figure S1). This study was approved by the local ethics committee (Approval Number: 63942322.3.0000.0068) and conducted in accordance with the ethical principles outlined in the Declaration of Helsinki.

RESEARCH IN CONTEXT

1. Systematic review: Epidemiological and clinical studies indicate that Alzheimer's disease (AD) is two to three times more prevalent in women. The burden of dementia is disproportionately higher in low‐ and middle‐income countries (LMICs), where socioeconomic inequalities may be associated with sex disparities in dementia risk. Research on neuropathology in these regions is limited, and the association between sex and neuropathological burden of dementia in diverse populations remains poorly understood.

2. Interpretation: Using data from a large and diverse sample of Brazilian participants with low educational levels, we found that sex modified the associations between neuropathology and cognition. Specifically, women showed stronger associations between AD pathology and trans‐activation response (TAR) DNA‐binding protein 43 (TDP‐43) pathology and cognitive impairment compared to men, and these associations were more pronounced among older participants. Neither education nor race modified these associations.

3. Future directions: Sex differences in the associations between neuropathology and cognitive outcomes were observed in diverse, low‐educated populations in LMICs. Future longitudinal and mechanistic studies integrating sex hormone data, genetic profiling, and domain‐specific neuropsychological testing are warranted to elucidate sex‐specific pathways and inform targeted dementia prevention strategies.

2.2. Neuropathological assessment

The neuropathological evaluation was performed according to internationally accepted pathological staging and diagnosis criteria (Figure S2). ¹⁵ , ¹⁶ , ¹⁷ , ¹⁸ , ¹⁹ AD pathology was assessed by immunostaining with anti‐β‐amyloid antibodies (4G8, 1:10,000; BioLegend #800701) and classified using the Consortium to Establish a Registry for Alzheimer's Disease (CERAD) criteria for neuritic plaques. ¹⁶ Phospho‐tau was identified with specific antibodies (AT8, 1:400; Invitrogen MN1020) and staged using Braak and Braak scores. ¹⁵ We categorized the CERAD staging also into three groups (none/sparse, moderate, and frequent) and the Braak staging system into three groups (0–II, III–IV, and V–VI). AD diagnosis was ascertained for individuals with Braak Stage III or above, and a CERAD neuritic plaque density of moderate or frequent. ²⁰

For LBD pathology, we used alpha‐synuclein (LB509, 1:500; Sigma‐Aldrich MABN82) immunostaining and classified according to the Braak staging for Parkinson's disease, categorizing LBD diagnosis when scored III or higher. ¹⁸ Trans‐activation response (TAR) DNA‐binding protein 43 (TDP‐43) was analyzed semi‐quantitatively using a TDP‐43 antibody (1:500; BioLegend #829901) and classified as present when stages were I or higher, according to the limbic‐predominant age‐related TDP‐43 encephalopathy neuropathologic changes (LATE‐NC) diagnosis consensus recommendation. ¹⁹

Cerebrovascular lesions were assessed macroscopically during the sampling of fixed brain regions from both hemispheres, cerebellum, and brainstem, and were analyzed semi‐quantitatively using hematoxylin and eosin (H&E) staining in 15 areas: middle frontal gyrus, middle and superior temporal gyri, angular gyrus, superior frontal and anterior cingulate gyri, 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. Large and lacunar infarcts were documented based on their location, size, and number. VaD diagnosis was considered present if there was one large infarct (≥1 cm) or lacunar infarcts (<1 cm) detected in three or more cortical regions or any strategic area for cognition (i.e., thalamus, frontal‐cingulate cortex, basal forebrain, caudate, medial temporal area, or angular gyrus). ¹⁷ We evaluated HA using H&E staining, assessing the location, extent, and severity of vascular changes. HA was considered present if moderate or severe lesions were found in at least three cortical regions. CAA was assessed semi‐quantitatively using anti‐amyloid beta (Aβ) immunostaining to evaluate the severity of amyloid deposition in vessels. CAA was considered present if moderate or severe lesions were detected in at least three different cortical regions. ¹⁷ As sensitivity analyses, HA and CAA were also considered as ordinal variables (none, mild, moderate, and severe).

Rare neuropathological diagnoses (e.g., LATE‐NC, corticobasal degeneration, progressive supranuclear palsy, chronic traumatic encephalopathy) were classified as “Other.” Mixed diagnoses involving AD and VaD were classified as AD+VaD; AD and LBD as AD+LBD; and VaD and LBD as VaD+LBD. Cases with AD combined with two or three additional pathologies were classified as AD+2 pathologies and AD+3 pathologies, respectively.

2.3. Clinical assessment

A trained gerontologist applied a semi‐structured interview to the NOK. Sociodemographic variables were ascertained during structured interviews and included age, years of education, and race (classified as White, Black, and Asian). Clinical information included previous medical diagnoses of hypertension, diabetes, and dyslipidemia. Smoking status was categorized as current, former, or never smoker, and alcohol use was classified as non‐use/social use, current, or former alcohol abuse. ¹² , ¹³

We used the Clinical Dementia Rating (CDR) scale to assess cognitive and functional abilities. ²¹ The CDR assesses six domains: (i) memory, (ii) orientation, (iii) judgment and problem‐solving, (iv) community affairs, (v) home and hobbies tasks, and (vi) personal care. Each domain is scored from zero to three based on the presence and severity. The severity of clinical dementia was classified based on an algorithm into five groups: no cognitive impairment (CDR = 0), questionable dementia (CDR = 0.5), mild (CDR = 1), moderate (CDR = 2), and severe dementia (CDR = 3). We computed the CDR–Sum of Boxes (CDR‐SOB) by summing up the scores of the six domains. CDR‐SOB ranged from 0 to 18, where higher scores indicate worse cognitive performance. ²²

2.4. Statistical analysis

Descriptive statistics were presented as mean and standard deviation (SD) for normally distributed continuous variables, and as median and interquartile range (IQR) for variables with non‐normal distributions. Categorical variables were described as relative frequencies. We compared sociodemographic and clinical characteristics and neuropathological lesions and diagnoses according to sex using the Mann–Whitney or t‐test for continuous variables, and the chi‐square test or Fisher's exact test for categorical variables.

Logistic ordinal regressions were performed to examine the association between sex and AD pathology ordinal variables (Braak and CERAD stages) and CDR global classification (normal cognition, questionable, mild, moderate, and severe dementia). In addition, we used binary logistic regression to investigate the associations of sex with LBD pathology, TDP‐43, lacunar infarcts, HA, and CAA. As sensitivity analyses, HA and CAA were modeled as ordinal variables (none, mild, moderate, and severe) using ordinal logistic regression, in a subset of participants with complete data for these two variables (n = 1494).

Furthermore, we conducted linear regression models to examine the association between sex and CDR‐SOB scores. Regression models were adjusted for age, education, and race (Model 1), and further adjusted for hypertension, dyslipidemia, diabetes, smoking, and alcohol use (Model 2). Because co‐pathologies were common, particularly in the very old, ²³ we investigated the independent association between each neuropathology and CDR‐SOB stratified by sex, by adding all neuropathologies as predictors in a multivariable linear model adjusted for the confounders described above in Model 2.

Furthermore, because social determinants of health may influence the association of sex with neuropathology and dementia, we further investigated whether age, race, and education could act as effect modifiers in the association between neuropathology, sex, and cognitive performance. ¹⁰ , ²⁴ , ²⁵ To investigate age as an effect modifier in the associations between neuropathology and cognition, participants were dichotomized at the sample median age into younger (<76 years) and older groups (≥76 years), and the triple interaction among neuropathological lesions, sex, and age on CDR‐SOB was assessed. In addition, participants were categorized into two educational groups (less than 4 years and 4 or more years of formal education). This cutoff categorization was used because 4 years was the median education of the sample and is related to the first stage of primary education in Brazil. ²⁶ We then assessed the triple interaction among neuropathologies, sex, and education level on cognitive performance by CDR‐SOB. Similarly, we evaluated effect modification by race, categorizing race into White and Black individuals (excluding Asian participants due to small number) and including a triple interaction among neuropathological lesions, sex, and race on CDR‐SOB. Triple interaction models were adjusted for age, education, race, hypertension, dyslipidemia, diabetes, smoking, and alcohol use, excluding race or education, as appropriate (for the triple interaction with binary age, the models were adjusted for continuous age). To improve the interpretability of the results, predicted CDR‐SOB scores were reported according to the interactions between each neuropathology and sex. Given the large number of multiple comparisons, p‐values were corrected using the false discovery rate (FDR), ²⁷ except for the interaction analyses due to their exploratory nature. Statistical analyses were performed using R version 4.0.3.

3. RESULTS

We included data from 2229 participants. Our sample comprised 1098 men (49%) and 1131 women (51%), and the mean (SD) age was 75.4 (12.3) years. The mean education was 4.96 (4.18) years. Regarding race distribution, 62% were White, 36% Black, and 2% Asian. Compared to male participants, female participants were older, less educated, had a higher frequency of hypertension, and consumed less tobacco and alcohol (Table 1). In addition, women had a higher frequency of dementia and higher CDR‐SOB scores compared to men (Table 1). In univariate analyses, Braak stages, CERAD stages, TDP‐43, HA, and CAA were more frequent in women compared to men (Figure 1A and Table S1). Furthermore, AD, AD+VaD, AD+LBD, and AD+2 pathologies were more frequent in women compared to men (Figure 1B). The absence of a neuropathological diagnosis (normal) was more frequent in men; however, men also had a higher frequency of VaD and LBD pathology compared to women (Figure 1B).

TABLE 1.

Sociodemographic and clinical characteristics of the sample (n = 2229).

	All	Male	Female
Variable	(n = 2229)	(n = 1098)	(n = 1,131)	p‐value
Age, years, mean (SD) a	75.4 (12.3)	72.7 (11.7)	78.1 (12.4)	< 0.001
Race, n (%) b				0.225
White	1385 (62.1)	664 (60.5)	721 (63.7)
Black	794 (35.6)	406 (37.0)	388 (34.3)
Asian	50 (2.24)	28 (2.55)	22 (1.95)
Education, years, mean (SD) a	4.96 (4.18)	5.59 (4.21)	4.34 (4.05)	<0.001
Hypertension, n (%) b	1443 (64.7)	659 (60.0)	784 (69.3)	<0.001
Diabetes, n (%) b	693 (31.1)	341 (31.1)	352 (31.1)	1.000
Dyslipidemia, n (%) b	348 (15.6)	165 (15.0)	183 (16.2)	0.489
Smoking, n (%) b				<0.001
Never	1148 (51.5)	424 (38.6)	724 (64.0)
Current	446 (20.0)	277 (25.2)	169 (14.9)
Former	635 (28.5)	397 (36.2)	238 (21.0)
Alcohol use, n (%) b				<0.001
Non‐use/social use	1,627 (73.0)	622 (56.6)	1005 (88.9)
Abuse	160 (7.18)	140 (12.8)	20 (1.77)
Former abuse	442 (19.8)	336 (30.6)	106 (9.37)
Clinical Dementia Rating (CDR), n (%) b , c				<0.001
No cognitive impairment	1463 (65.6)	803 (73.1)	660 (58.4)
Questionable dementia	210 (9.42)	103 (9.38)	107 (9.46)
Mild dementia	103 (4.62)	46 (4.19)	57 (5.04)
Moderate dementia	126 (5.65)	50 (4.55)	76 (6.72)
Severe dementia	327 (14.7)	96 (8.74)	231 (20.4)
CDR‐SOB, mean (SD) a	3.73 (6.35)	2.54 (5.29)	4.88 (7.05)	<0.001

Abbreviations: CDR‐SOB, Clinical Dementia Rating—Sum of Boxes; SD, standard deviation.

^a Student's t‐test.

^b Chi‐square test.

^c CDR categories: No cognitive impairment, CDR = 0; Questionable dementia, CDR = 0.5; Mild dementia, CDR = 1; Moderate dementia, CDR = 2; Severe dementia, CDR = 3.

FIGURE 1.

Comparisons of (A) neuropathological lesions and (B) neuropathological diagnoses between male (n = 1098) and female (n = 1131) participants. Statistically significant differences were marked with *. Comparisons were conducted using chi‐square tests, except for AD+3 pathologies, AD+other, and VaD+LBD, for which Fisher's exact tests were used due to unmet chi‐square assumptions (i.e., expected cell counts less than 5). Braak stages were classified as present if Braak ≥III and CERAD stages were classified as present if CERAD ≥moderate. AD, Alzheimer's disease; CAA, cerebral amyloid angiopathy; CERAD, Consortium to Establish a Registry for Alzheimer´s Disease; LBD, Lewy body disease; TDP‐43, trans‐activation response (TAR) DNA‐binding protein 43; VaD, vascular dementia.

In fully adjusted analyses, female sex was associated with higher Braak stages (odds ratio [OR] = 1.42, 95% confidence interval [CI] = 1.17–1.72) and CERAD stages (OR = 1.58, 95% CI = 1.27–1.98), as well as higher odds of TDP‐43 (OR = 1.77, 95% CI = 1.22–2.59). On the other hand, female sex was associated with lower odds of LBD (OR = 0.69, 95% CI = 0.51–0.95). Sex was not associated with infarcts, HA, or CAA (Table 2). HA and CAA remained not associated with sex when considered as ordinal variables in 1494 participants (Table S2).

TABLE 2.

Association between sex and neuropathological lesions (n = 2229).

Neuropathological lesion	Unadjusted OR (95% CI)	p‐value	Model 1 a OR (95% CI)	p‐value	Model 2 b OR (95% CI)	p‐value	Ranking	Threshold	Significant c
Braak stages d	1.91 (1.63–2.24)	<0.001	1.33 (1.11–1.59)	0.002	1.42 (1.17–1.72)	<0.001	2	0.014	Yes
CERAD stages d	2.24 (1.85–2.72)	<0.001	1.62 (1.32–1.99)	<0.001	1.58 (1.27–1.98)	<0.001	1	0.007	Yes
Lewy body disease e	1.06 (0.80–1.39)	0.698	0.75 (0.56–1.01)	0.056	0.69 (0.51–0.95)	0.022	4	0.029	Yes
TDP‐43 e	2.56 (1.85–3.59)	<0.001	1.69 (1.20–2.42)	0.003	1.77 (1.22–2.59)	0.003	3	0.021	Yes
Lacunar infarcts e	0.85 (0.63–1.13)	0.268	0.72 (0.53–0.97)	0.032	0.74 (0.54–1.03)	0.077	7	0.050	No
Hyaline arteriolosclerosis e	1.46 (1.22–1.75)	<0.001	1.15 (0.94–1.40)	0.176	1.22 (0.98–1.51)	0.071	6	0.043	No
CAA e	1.78 (1.43–2.23)	<0.001	1.25 (0.98–1.59)	0.068	1.29 (0.99–1.67)	0.056	5	0.036	No

Reference: Male participants (n = 1098).

Abbreviations: CAA, cerebral amyloid angiopathy; CERAD, Consortium to Establish a Registry for Alzheimer´s Disease; CI, confidence interval; OR, odds ratio; TDP‐43: trans‐activation response (TAR) DNA‐binding protein 43.

^a Model 1 was adjusted for age, education, and race.

^b Model 2 was adjusted for age, education, race, hypertension, dyslipidemia, diabetes, smoking, and alcohol use.

^c Significance adjustment for multiple comparisons using the Benjamini–Hochberg false discovery rate method.

^d Ordinal logistic regression.

^e Binary logistic regression.

Regarding cognition and clinical dementia diagnosis, female sex was associated with higher CDR‐SOB scores (β = 1.56, 95% CI = 1.03–2.10), meaning worse cognitive performance, and with higher odds of dementia severity according to CDR global classification (OR = 1.66, 95% CI = 1.36–2.04) compared to male sex. In the analyses assessing the independent association between neuropathological lesions and CDR‐SOB stratified by sex, Braak stages V–VI were independently associated with higher CDR‐SOB in both men (β = 5.11, 95%CI = 3.71–6.51) and women (β = 2.83, 95% CI = 1.37–4.29), and so was frequent CERAD stage (men: β = 1.93, 95% CI = 0.54–3.32; women: β = 3.95, 95% CI = 2.53–5.36), whereas moderate CERAD stage was only associated with higher CDR‐SOB in women (β = 2.23, 95% CI = 1.23–3.23). LBD was also independently associated with CDR‐SOB in men (β = 2.13, 95% CI = 1.20–3.06) and women (β = 1.64, 95% CI = 0.54–2.73), and so were TDP‐43 (men: β = 2.20, 95% CI = 0.89–3.52; women: β = 3.85, 95% CI = 2.73–4.96), lacunar infarcts (men: β = 1.87, 95% CI = 0.94–2.80; women: β = 2.25, 95% CI = 1.06–3.44), and HA (men: β = 1.24, 95% CI = 0.59–1.90; women: β = 1.24, 95% CI 0.51–1.98). On the other hand, CAA was independently associated with higher CDR‐SOB only in women (β = 1.46, 95% CI = 0.58–2.34) (Table S3).

The interaction analyses between neuropathological lesions and sex on CDR‐SOB scores showed significant interactions for Braak (p = 0.005), CERAD (p < 0.001), TDP‐43 (p < 0.001), and CAA (p = 0.002) (Figure 2; Table S4). Accordingly, predicted CDR‐SOB scores by sex showed that women had higher predicted scores than men for the same level of neuropathology regarding Braak stages III–IV (men: CDR‐SOB = 3.07, 95% CI = 2.35–3.79; women: CDR‐SOB = 4.84, 95% CI = 4.17–5.52), CERAD stages moderate (men: CDR‐SOB = 4.94, 95% CI = 3.90–5.98; women: CDR‐SOB = 7.76, 95% CI = 6.91–8.60) and frequent (men: CDR‐SOB = 8.58, 95% CI = 7.28–9.87; women: CDR‐SOB = 11.37, 95% CI = 10.46–12.29), TDP‐43 (men: CDR‐SOB = 6.97, 95% CI = 5.38–8.56; women: CDR‐SOB = 11.53, 95% CI = 10.47–12.59), and CAA (men: CDR‐SOB = 5.97, 95% CI = 4.92–7.01; women: CDR‐SOB = 9.04, 95% CI = 8.19–9.88) (Table S4).

FIGURE 2.

Predicted CDR‐SOB, considering interaction terms between neuropathological lesions and sex. Male participants are shown in black and female participants in red. (A) Braak stages; (B) CERAD stages; (C) Lewy body disease; (D) TDP‐43 pathology; (E) lacunar infarcts; (F) hyaline arteriolosclerosis; and (G) CAA. Linear regression models were adjusted for age, education, race, hypertension, dyslipidemia, diabetes, smoking, and alcohol use. The p‐values were for the interaction between each neuropathological lesion and sex. CAA, cerebral amyloid angiopathy; CDR‐SOB, Clinical Dementia Rating–Sum of Boxes; CERAD, Consortium to Establish a Registry for Alzheimer's Disease; TDP‐43, trans‐activation response (TAR) DNA‐binding protein 43.

The triple interactions between neuropathologies, sex, and age on CDR‐SOB scores were significant for Braak stages (p = 0.015), CERAD stages (p < 0.001), HA (p = 0.041), and CAA (p = 0.035) (Table S5). Among older participants (≥76 years), women had higher predicted CDR‐SOB scores than men in Braak stages III–IV (men: CDR‐SOB = 4.19, 95% CI = 3.39–4.99; women: CDR‐SOB = 6.49, 95% CI = 5.79–7.19), CERAD stage moderate (men: CDR‐SOB = 6.07, 95% CI = 4.83–7.31; women: CDR‐SOB = 9.51, 95% CI = 8.64–10.37), in the presence of moderate/severe HA (men: CDR‐SOB = 7.28, 95% CI = 6.24–8.31; women: CDR‐SOB = 9.25, 95% CI = 8.47–10.03), and in the presence of moderate/severe CAA (men: CDR‐SOB = 8.06, 95% CI = 6.86–9.26; women: CDR‐SOB = 10.78, 95% CI = 9.92–11.64). Among younger participants (<76 years), similar findings were found for severe CERAD stages and CAA only (Table S5). None of the triple interactions between neuropathologies, race, and sex on CDR‐SOB scores were significant (Table S6); nor were the triple interactions between neuropathologies, education, and sex on CDR‐SOB scores (Table S7).

4. DISCUSSION

In this large and diverse sample with low educational attainment, women were older and had lower levels of education than men. In addition, women had a higher burden of AD (i.e., higher accumulation of neuritic plaques and neurofibrillary tangles) and TDP‐43 pathologies, and lower odds of LBD pathology compared to men, whereas no sex differences were observed between sexes for cerebrovascular lesions (i.e., lacunar infarcts, HA, or CAA) in adjusted models. Furthermore, female participants had worse cognitive performance and higher odds of more severe dementia than male participants. Sex modified the associations between neuropathologies and cognition; women had poorer cognitive performance compared to men in the presence of the same levels of AD pathology, including Braak and CERAD staging, as well as TDP‐43 and CAA. These sex differences were more pronounced among older participants. Education and race did not modify the interactions of sex and pathologies on cognitive outcomes.

The higher prevalence of severe dementia among women in our sample is consistent with the epidemiological data, including from Brazil, which show a greater burden of dementia in women, particularly in the context of older age and lower educational attainment. ⁸ , ²⁸ Of interest, these factors were also more prominent among female participants in our sample. The higher prevalence of AD pathology in women is well‐established in the literature. ²⁹ Female participants in our sample exhibited higher odds of AD pathology, as well as poorer cognitive performance than male participants at the same level of AD pathology. Similarly, previous studies have found that the accumulation of AD pathology, including neurofibrillary tangles and amyloid plaques, was more pronounced in women and correlates with more severe clinical symptoms. ¹⁰ , ³⁰ A longitudinal neuroimaging study revealed that amyloid accumulation was associated with higher levels of phosphorylated tau 181 (p‐tau181) in the CSF in women compared to men. ³¹ Furthermore, women with Aβ positivity presented with faster neurofibrillary tangle accumulation compared with their male counterparts. ³¹ Recently, our group has shown that female sex was associated with Aβ burden, particularly in White female participants. ³² However, the association between apolipoprotein E (APOE) ε4 and Aβ burden was not modified by sex, but by race and African ancestry. The current study expanded our previous study by further investigating the associations of sex with other pathologies, beyond AD pathology. ³²

Biological factors proposed to contribute to AD pathology vulnerability include sex‐specific genetic and epigenetic factors, X‐chromosome effects on tau burden, hormonal transitions such as menopause, telomere shortening, and increased cellular senescence. ¹ , ³³ Several mechanisms, including differences in life expectancy, variations in brain structures influenced by sexual hormones throughout life, and exposure to diverse psychosocial factors, may further contribute to the stronger clinical symptoms and AD pathology in females in our study. ²⁹

We observed higher TDP‐43 odds in women and a stronger association of TDP‐43 with poor cognitive performance in women than in men. These findings suggest that neuropathological and clinical manifestations of TDP‐43 may differ between women and men and are consistent with recent literature on LATE and sex differences in clinical presentation, even after controlling for age, which is an important confounder, since LATE predominantly affects older adults. ³⁴ Consistent with our findings on the association of female sex with lower odds of LBD, prior studies have reported a higher LBD prevalence in men. ⁴ , ³⁵

Our study showed a sex effect modification of the association between CAA and cognitive performance. When CAA was present at moderate or severe levels, CDR‐SOB scores were higher in female than male participants. The literature on sex differences in the frequency of CAA is conflicting, and the investigation of sex as a modifier on the association between CAA and cognitive performance is lacking. In a study with 6120 participants from the National Alzheimer's Coordinating Center (NACC), CAA frequencies were similar between male and female participants. ³⁶ On the other hand, CAA frequency was higher in men in another smaller autopsy study, when the models were adjusted for AD pathology. ³⁷

The triple interaction between age, sex, and neuropathologies is novel and yielded interesting results. In adults 76 years of age or older, female participants had worse CDR‐SOB scores than male participants at the same level of AD pathology, HA, and CAA, whereas the sex differences in cognitive expression of pathology were more pronounced for CERAD and CAA in younger participants. Exploratory analyses of triple interactions of neuropathologies and sex with education and race on cognitive performance were not significant, suggesting that the described interactions between neuropathologies and sex were not further explained by social determinants of health, such as race and education. Although these findings may reflect true null effects, it is also possible that limited statistical power for the triple interactions analyses has limited our ability to detect real interactions. Moreover, the absence of similar investigations of effect modifiers of the association between sex and neuropathologies precluded comparisons between our findings and other studies.

Our study has several strengths. We analyzed clinical and neuropathological data from a large sample of individuals with less educational attainment (a mean of 5 years of formal education) and diverse racial backgrounds. In addition, the BAS is a population‐based autopsy study, which allows for the collection of a large number of brains from individuals with normal cognition. ¹⁴ However, our findings should be interpreted considering certain limitations. We did not follow participants during their lifetime, and clinical variables were assessed postmortem through interviews with informants. The cross‐sectional nature of this observational study also limited causal inference conclusions. To enhance the reliability of the study data, we included only participants who had at least weekly contact with the informant and excluded individuals if the informant provided conflicting information during the clinical interview. In addition, we acknowledge the limitations of not evaluating clinical data related to sex, such as menopause, pregnancies, hormone replacement therapy, and the analysis of sex hormones, which could be important modifiers of our results. ²⁹ Moreover, the semiquantitative classifications for HA and CAA were available for a limited number of participants, and future studies should evaluate the association between sex and more discriminative variables for vascular lesions. Although our study is population based and the sociodemographic data of the study sample are similar to the data of the people who died in the city of São Paulo (mean age ± SD: 73 ± 12, 50%), selection bias may have occurred. ¹⁴ Finally, we performed several statistical analyses that can increase the chances of false‐positive associations. Although we corrected the results for multiple comparisons, Type 1 error is still possible.

Our findings contribute to the growing evidence that sex influences the expression of neuropathology and cognition. Female participants demonstrated a higher burden of AD and TDP‐43 pathologies, a lower burden of LBD, and poorer cognitive outcomes compared to male participants. At the same level of pathology, female participants presented poorer cognitive scores than male participants for AD, TDP‐43, and CAA pathologies. Future longitudinal and mechanistic studies, integrating sex hormones, genetic profiling, and domain‐specific neuropsychological testing, are needed to clarify these sex‐specific pathways and inform tailored approaches to dementia prevention and treatment.

CONFLICT OF INTEREST STATEMENT

The authors declare no conflicts of interest. Author disclosures are available in the Supporting Information.

CONSENT STATEMENT

A close family member authorized the brain donation and signed an informed consent document for study participation.

Supporting information

Supporting Information