Association Between Body Composition Patterns, Cardiovascular Disease, and Risk of Neurodegenerative Disease in the UK Biobank

Abstract

Background and Objectives

Accumulating evidence connects diverse components of body composition (e.g., fat, muscle, and bone) to neurodegenerative disease risk, yet their interplay remains underexplored. This study examines the associations between patterns of body composition and the risk of neurodegenerative diseases, exploring the mediating role of cardiovascular diseases (CVDs).

Methods

This retrospective analysis used data from the UK Biobank, a prospective community-based cohort study. We included participants free of neurodegenerative diseases and with requisite body composition measurements at recruitment, who were followed from 5 years after recruitment until April 1, 2023, to identify incident neurodegenerative diseases. We assessed the associations between different components and major patterns of body composition (identified by principal component analysis) with the risk of neurodegenerative diseases, using multivariable Cox models. Analyses were stratified by disease susceptibility, indexed by polygenetic risk scores for Alzheimer and Parkinson diseases, APOE genotype, and family history of neurodegenerative diseases. Furthermore, we performed mediation analysis to estimate the contribution of CVDs to these associations. In addition, in a subcohort of 40,790 participants, we examined the relationship between body composition patterns and brain aging biomarkers (i.e., brain atrophy and cerebral small vessel disease).

Results

Among 412,691 participants (mean age 56.0 years, 55.1% female), 8,224 new cases of neurodegenerative diseases were identified over an average follow-up of 9.1 years. Patterns identified as “fat-to-lean mass,” “muscle strength,” “bone density,” and “leg-dominant fat distribution” were associated with a lower rate of neurodegenerative diseases (hazard ratio [HR] = 0.74–0.94) while “central obesity” and “arm-dominant fat distribution” patterns were associated with a higher rate (HR = 1.13–1.18). Stratification analysis yielded comparable risk estimates across different susceptibility groups. Notably, 10.7%–35.3% of the observed associations were mediated by CVDs, particularly cerebrovascular diseases. The subcohort analysis of brain aging biomarkers corroborated the findings for “central obesity,” “muscle strength,” and “arm-dominant fat distribution” patterns.

Discussion

Our analyses demonstrated robust associations of body composition patterns featured by “central obesity,” “muscle strength,” and “arm-dominant fat distribution” with both neurodegenerative diseases and brain aging, which were partially mediated by CVDs. These findings underscore the potential of improving body composition and early CVD management in mitigating risk of neurodegenerative diseases.

Introduction

Neurodegenerative diseases, including Alzheimer disease (AD) and Parkinson disease (PD), currently affect over 60 million people and rank as the seventh leading cause of death worldwide.^1,2 As populations age, this situation is expected to worsen. Unfortunately, disease-modifying treatments remain scarce for these diseases. Hence, to alleviate disease burden, it is crucial to identify modifiable risk factors to facilitate the development of preventive measures. In addition, it is important to consider an individual's genetic predisposition when investigating these modifiable risk factors, aiming for more precise and tailored preventive measures.

Cardiovascular diseases (CVDs), such as coronary heart disease and cerebrovascular diseases, have been previously suggested as risk factors of neurodegenerative diseases.³ However, in some studies, obesity, a risk factor of CVDs, was linked to a lower risk of all-cause dementia and PD (i.e., obesity paradox phenomenon).^4,5 One possible explanation for this paradox is unintentional weight loss during the preclinical stage of neurodegenerative diseases, leading to a reverse causation effect. Moreover, this paradox may stem from defining obesity solely based on body mass index (BMI). BMI not only faces categorization issues due to its foundation on data from nondiverse populations but also fails to account for variations in body composition (e.g., BMI fails to distinguish muscle from fat, misclassifying muscular individuals as having high adiposity). Diverse body components may play distinct roles in the development of neurodegenerative diseases. For instance, emerging evidence suggests a protective effect of improved muscle function against cognitive decline and Alzheimer dementia^6,7 while abdominal fat deposition seems detrimental to neurodegenerative diseases.^8,9 Nevertheless, findings on fat mass (FM), lean mass (LM), and bone density remain inconsistent.^7,10-12 These discrepancies could arise from the complex interplay among diverse body components, driven by the intricate cross-talks between fat, muscle, and bone tissues.^13,14 However, to date, few studies have simultaneously considered diverse body components when exploring their health consequences. A comprehensive understanding of their respective contributions to health could provide insights into preventing illnesses, for example, neurodegenerative diseases, by targeting body composition beyond the conventional focus on overall weight control.

Leveraging enriched phenotypic information and high-quality genetic data from the community-based UK Biobank cohort, we aimed to comprehensively examine the associations between diverse body components, focusing on identifying patterns of body composition to address their complex interplay, and risk of neurodegenerative diseases. We also explored whether such associations would differ for individuals with different susceptibilities to neurodegenerative diseases. Finally, our study sought to elucidate the potential mediating role of CVDs in the studied associations.

Methods

Study Population and Design

The UK Biobank enrolled 502,507 participants aged 40 to 70 years across the United Kingdom during 2006–2010.^15,16 In brief, at recruitment, information on sociodemographic, lifestyle, medical, and genetic factors was collected from all participants through touchscreen questionnaires, physical measures, and sample assays. Information on health-related outcomes was obtained through periodic linkages with multiple national data sets, including primary care, hospital inpatient, and death registry records.

In this study, we excluded participants who withdrew their information (n = 139), had prevalent neurodegenerative diseases at recruitment (n = 559), or had missing (n = 22,786) or extreme (i.e., exceeding ±5 standard derivations, n = 5,229) values in requisite body composition measurements. To mitigate potential confounding by preexisting conditions, both diagnosed and potentially undiagnosed, we further excluded participants identified with severe somatic diseases (as per the Charlson Comorbidity Index,¹⁷ n = 59,256) and those classified as underweight (BMI < 18.5 kg/m², n = 1,847), leaving 412,691 eligible participants for analyses (eFigure 1). To mitigate reverse causation, we started the follow-up for all participants from 5 years after recruitment (i.e., 5-year lag time) until a diagnosis of neurodegenerative diseases, loss to follow-up, death, or April 1, 2023, whichever occurred first.

Standard Protocol Approvals, Registrations, and Patient Consents

The UK Biobank has full ethical approval from the NHS National Research Ethics Service (16/NW/0274), and informed consent was obtained before data collection from each participant. This study (the study protocol can be found in the eSAP) was also approved by the biomedical research ethics committee of West China Hospital (reference number: 2019-1171).

Measurements of Body Composition

At recruitment, a range of body composition measurements were performed using well-calibrated instruments by trained staff following a standard protocol. Specifically, height, waist, and hip circumference were measured manually. Whole-body and regional (i.e., arm, leg, and trunk) FM and LM, as well as weight, were evaluated by bioimpedance using the Tanita BC-418 MA body composition analyzer. Grip strength of both hands was measured using the Jamar J00105 hydraulic hand dynamometer, with instructions for maximal effort. Bone density assessment was conducted on calcaneus using the Norland McCue Contact Ultrasound Bone Analyzer. These measurements and their derivations, including a total of 28 items, were categorized into 8 groups based on the dimensions of body composition they assessed (eTable 1).

Ascertainment of Neurodegenerative Diseases

The main outcome of this study was the incidence of any neurodegenerative disease. Primary or vascular neurodegenerative diseases, as well as specific neurodegenerative diseases (i.e., PD, AD, and all-cause dementia), were considered separately as secondary outcomes. Cases were identified based on hospital inpatient (primary or secondary hospital diagnosis) or death registry (underlying or contributory cause of death) records, according to the corresponding International Classification of Diseases coding system (ICD-10) codes (eTable 2). The accuracy of diagnosing neurodegenerative diseases using UK Biobank inpatient and death registry data has been validated, with a positive predictive value of 84.5% for all-cause dementia and 84.0% for PD against detailed clinical evaluations.¹⁸

Furthermore, we used neuroimaging data derived from brain magnetic resonance imaging (MRI) as a supplement outcome (i.e., biomarkers for brain aging). In this analysis, we similarly excluded 42 participants with neurodegenerative diseases before attending the MRI examination in 2014,¹⁹ leaving 40,790 participants in this subanalysis (eFigure 1). We assessed brain atrophy biomarkers using white matter, grey matter, and hippocampal volumes and evaluated cerebral small vessel disease (cSVD)-related biomarkers with white matter hyperintensity volume, fractional anisotropy, and mean diffusivity²⁰ (details about these measurements given in eMethods).

Susceptibility to Neurodegenerative Diseases

We assessed the susceptibility to neurodegenerative diseases by polygenetic risk score (PRS) for AD, PRS for PD, APOE genotype, and family history of neurodegenerative disease, among 403,733 participants with both phenotypic and genetic data (eFigure 1). The PRS quantifies common genetic load, and APOE is a well-established genetic risk factor of AD.²¹ Family history encapsulates both genetic and shared environmental influences (details in eMethods). Our validation results demonstrated that the PRS for both AD and PD, generated by UK Biobank,²² showed strong associations with a clinical diagnosis of neurodegenerative diseases (odds ratio and 95% CI per 1-unit increase in the PRS for AD: 1.51 [1.48–1.53] and for PD: 1.15 [1.13–1.17]).

Covariates and Mediators

Information on sociodemographic factors (e.g., age, sex, ethnicity, educational level, and income), lifestyle factors (smoking and drinking status, diet, and physical activity), and family history was collected through questionnaires at recruitment. Cognitive function was evaluated by a brief computerized cognitive test battery at recruitment. We used reaction time as an indicator of cognition, with a higher value indicating poorer overall cognition (good cognition: reaction time ≤ median level and poor cognition: reaction time > median level). Ectopic fat deposition, presented by abdominal fat ratio (abdominal fat volume divided by the sum of abdominal fat volume and thigh muscle volume) and muscle fat infiltration (fat fraction in thigh muscle volume), was assessed by abdominal MRI examinations in approximately 60,000 participants in 2014. More information about covariates is given in eMethods.

In this study, we considered the occurrence of CVDs (i.e., coronary heart diseases, heart failure, or cerebrovascular diseases) after recruitment but before the onset of neurodegenerative diseases as a potential mediator of the studied associations. This was ascertained by diagnoses from hospital inpatient records according to the corresponding ICD-10 codes (eTable 2).

Statistical Analysis

Identification of Patterns of Body Composition

To capture the complex interplay among body components, we first mapped the correlations for 28 components (eFigure 2) and then used principal component analysis (PCA) to identify patterns in body composition by sex, considering the inherent sex-based differences. PCA is a dimensionality reduction technique that reorganizes variables with high correlation into a new set of low-related or unrelated variables, known as principal components (PCs). Consequently, we identified 7 significant PCs (i.e., with eigenvalue >1.0), representing major patterns of body composition (eFigure 3). The identified patterns showed considerable similarity between male and female sex and collectively explained over 90% of variance in both sexes. Hence, to ensure consistency in PCA algorithms, we applied male-derived loading coefficients—which determine the influence of each variable on the PCs—to both sexes in subsequent analyses. We labeled each pattern based on its contributing variables with top loading coefficients, resulting in patterns named “fat-to-lean mass” (indicating FM relative to LM), “lean mass,” “central obesity,” “muscle strength,” “bone density,” “leg-dominant fat distribution” (highlighting FM primarily distributed in the legs, as opposed to the trunk), and “arm-dominant fat distribution” (characterized by FM mainly in the arms, with LM predominantly in the legs rather than in the arms and trunk). The exposure level to each identified pattern, represented as PC scores, was calculated for all study participants, through integrating all included variables weighted by their corresponding loading coefficients.

Phenotypic Association of Body Composition With Risk of Neurodegenerative Diseases

We assessed the associations between the identified patterns of body composition, as well as individual components of body composition, and incident neurodegenerative diseases. First, we treated the exposures as continuous variables and used restricted cubic splines with knots at the 5th, 35th, 65th, and 95th percentiles using the R package “rms.” This approach allowed us to flexibly model the shape (linear or nonlinear) of the associations between body composition and neurodegenerative disease for men and women separately. We tested the potential nonlinearity by using the likelihood-ratio test comparing models with only a linear term against models with linear and cubic terms. Subsequently, exposure levels were categorized into low, moderate, and high groups based on sex-specific tertile distribution (low: <first tertile; moderate: first-second tertile; and high: >second tertile) because of the observed nonlinear nature of most associations. Cox models with attained age as the timescale were then used to estimate hazard ratios (HRs) and 95% CIs in relation to the moderate and high levels of exposure, using the low level as the reference. Models were adjusted for sex; annual household income; Townsend deprivation index; educational level; smoking and drinking status; physical activity; fruit or vegetable intake; history of hypertension, diabetes, or dyslipidemia; baseline cognition function; family history of neurodegenerative diseases; height; and weight. In addition, for brain aging biomarkers as continuous variables, we used multivariable linear regression models to examine the association with identified patterns of body composition, adjusting for aforementioned confounders, age at imaging assessment, and intracranial volumes.

Modification Role of Disease Susceptibility on the Phenotypic Associations

To explore whether the studied associations might be modified by disease susceptibility, we performed stratified analyses by the PRS of AD (high risk: >second tertile or low risk: ≤first tertile), or PRS of PD (high or low risk), APOE genotype (APOE ε4 carriers or noncarriers), and family history of neurodegenerative diseases (yes or no).

Mediating Effect of CVDs on the Phenotypic Associations

Mediation analysis was used to assess the potential contribution of CVDs on the observed phenotypic associations. Initially, a simple-mediator model was used to evaluate the individual contribution of each CVD. Subsequently, a multimediator model was used to assess the combined mediating effect of all CVDs. These analyses were conducted with the R package “CMAverse,”²³ using a regression-based approach with Cox models adjusted for the aforementioned covariates. Parametric bootstrapping (n = 400 times) was used to calculate 95% CIs and p-values.

Subgroup and Sensitivity Analyses

To disentangle whether the studied associations differed between individuals with normal weight and overweight/obese individuals, we performed separate analyses for individuals with BMI ≤25 kg/m² or BMI >25 kg/m². Considering that cognitive decline may be a prodromal phase of neurodegenerative disease, we further stratified our analysis by cognition function at baseline (good or poor). Stratification analyses by age at baseline (≤60 years or >60 years) were also performed.

In the sensitivity analyses, we used alternative cutoff points to define high or low genetic susceptibility, using median (> or < median) or quartiles (>3rd or <1st quartile) of PRS for AD and PD. We applied a 10-year lag time to further minimize the probability of reverse causation. We also excluded individuals with self-reported weight loss or gain within the year before recruitment because their body composition measurements might not accurately represent long-term exposure. Moreover, to address potential influence of ethnicity on body composition, we restricted our analyses to encompass solely individuals of European descent. Finally, we used Fine-Grey models to account for the competing risk of death.

The R (version 4.0.2) and Python (version 3.8) packages were used for the analyses. Statistical significance was set at p-value <0.05, with false discovery rate adjustments for the total number of comparisons across all interconnected analyses (e.g., body composition components or patterns).

Data Availability

Data from the UK Biobank (ukbiobank.ac.uk/) are available to all researchers on making an application.

Results

Among the 412,691 participants of this study, the mean age at recruitment was 56.0 years and 55.1% (n = 227,349) were female. During a mean follow-up of 9.1 years, 8,224 cases of incident neurodegenerative diseases were identified, of which 6,274 and 1,194 were attributed to primary and vascular causes, respectively. The specific neurodegenerative diseases analyzed included 2,427 cases of PD, 2,933 AD cases, and 6,076 all-cause dementia cases. The baseline characteristics of the study participants, by level of exposure to the identified body composition patterns (i.e., low, moderate, or high), are summarized in eTable 3. Opposite to the pattern characterized by “muscle strength,” individuals with high level of exposure to the other 6 patterns generally tended to have a high BMI (all p-values < 0.001). Participants exhibiting high levels in the “fat-to-lean mass,” “lean mass,” “central obesity,” and “arm-dominant fat distribution” patterns tended to have a high abdomen fat ratio and muscle fat infiltration while the opposite tendency was observed for the “muscle strength” and “leg-dominant fat distribution” patterns.

Phenotypic Associations Between Identified Patterns or Individual Components of Body Composition and Subsequent Neurodegenerative Diseases

Analyses of the 7 identified patterns of body composition, as continuous (Figure 1) or categorized (Figure 2A) variables, found that 6 of the 7 patterns showed significant associations, with largely comparable estimates between men and women. We observed largely linear-shaped relationships for the “muscle strength” and “arm-dominant fat distribution” patterns while the other associations were mostly nonlinear (U, L, or J shapes, Figure 1). Compared with low exposure level, higher exposure level to “fat-to-lean mass” (HR for moderate level: 0.92 [95% CI 0.86–0.99]), “muscle strength” (HR for moderate and high levels: 0.81 [0.77–0.85] and 0.74 [0.69–0.79]), “bone density” (HR for high level: 0.94 [0.89–0.99]), and “leg-dominant fat distribution” (HR for moderate level: 0.94 [0.89–0.99]) patterns was associated with a lower rate of neurodegenerative disease, whereas higher exposure level to “central obesity” (HR for high level: 1.13 [1.06–1.21]) and “arm-dominant fat distribution” (HR for high level: 1.18 [1.10–1.26]) patterns was associated with a higher rate of neurodegenerative diseases. For subtypes (eFigure 4) and specific neurodegenerative diseases (eFigure 5), the findings were generally consistent, except for the “lean mass” pattern, which was associated with an elevated rate of vascular neurodegenerative diseases but a reduced risk of AD.

Figure 1. Associations Between Identified Patterns of Body Composition and Rate of Incident Neurodegenerative Diseases, With Separate Analyses for Male (Blue) and Female (Red).

Exposure levels were categorized into low, moderate (md), and high groups based on sex-specific tertile distribution (low: <first tertile; moderate: first-second tertile; and high: >second tertile). HRs are indicated by solid lines, and 95% CIs are indicated by shaded areas. The vertical dashed line represents the sex-specific tertile level of each pattern of body composition. Restricted cubic splines were constructed with 4 knots located at the 5th, 35th, 65th, and 95th percentiles of each pattern. We tested the associations between identified patterns of body composition and risk of incident neurodegenerative diseases by treating exposures as continuous variables, using Cox models with attained age as the timescale after adjusting for annual household income; Townsend deprivation index; educational level; smoking and drinking status; physical activity; fruit and vegetable intake; history of hypertension, diabetes, or dyslipidemia; baseline cognition function; family history of neurodegenerative diseases; height; and weight (results denoted as “P overall”). We tested the potential nonlinearity by using the likelihood-ratio test comparing models with only a linear term against models with linear and cubic terms (results denoted as “P nonlinearity”). HR = hazard ratio.

Figure 2. (A) Associations Between Identified Patterns of Body Composition and Rate of Incident Neurodegenerative Diseases and (B) Mediating Effect of Cardiovascular Diseases on the Observed Associations.

IR was calculated as numbers of cases per 1000 person-years. HRs and 95% CIs were derived from Cox models with attained age as the timescale after adjusting for sex; annual household income; Townsend deprivation index; educational level; smoking and drinking status; physical activity; fruit and vegetable intake; history of hypertension, diabetes, or dyslipidemia; baseline cognition function; family history of neurodegenerative diseases; height; and weight. CeVD = cerebrovascular diseases; CHD = coronary heart diseases; HF = heart failure; HR = hazard ratio; IR = incidence rate; NS = not statistically significant.

The associations between individual components of body composition and neurodegenerative diseases showed partial similarity to the results of body composition patterns (Figure 3, eFigure 6, and eTable 4). For instance, we observed that most central obesity indicators were positively associated with an elevated rate of neurodegenerative disease (high vs low, HR for waist-to-hip ratio = 1.10 [1.04–1.17]) and all muscle strength-related indicators were inversely related with the neurodegenerative disease rate (high vs low, HR for grip strength = 0.73 [0.68–0.78]). Positive associations were also noted for arm/whole FM (high vs low, 1.19 [1.12–1.26]) and leg/whole LM (high vs low, 1.10 [1.04–1.17]). In addition, we noted that arm/whole LM was related to a reduced rate of neurodegenerative diseases (high vs low, 0.81 [0.76–0.86]).

Figure 3. Associations Between Individual Items of Body Composition and Rate of Incident Neurodegenerative Diseases.

Color of the circle represents the value of hazard ratios, derived from Cox models with attained age as the timescale after adjusting for sex; annual household income; Townsend deprivation index; educational level; smoking and drinking status; physical activity; fruit and vegetable intake; history of hypertension, diabetes, or dyslipidemia; baseline cognition function; family history of neurodegenerative diseases; height; and weight. The inner circle represents the hazard ratios of the moderate groups, and the outer circle represents the hazard ratios of the high groups, using the low groups as the reference. *Statistically significant (i.e., p-value < 0.05 using the false discovery rate for adjustment of multiple testing). BMD = bone mineral density; BSI = body shape index; FM = fat mass; HC = hip circumference; LM = lean mass; WC = waist circumference; WTR = waist-to-hip ratio; WTHR = waist-to-height ratio; WWI = waist-to-weight index.

Phenotypic Associations Between Identified Patterns of Body Composition and Brain Aging Biomarkers

Individuals with subsequent neurodegenerative diseases showed poorer brain aging biomarkers than those without (eTable 5). We observed that the “central obesity” and “arm-dominant fat distribution” patterns were positively linked to poorer brain atrophy and cSVD-related biomarkers while the “muscle strength” pattern was negatively correlated with these biomarkers (Figure 4 and eFigure 7), corroborating the main analysis. However, the “fat-to-lean mass” pattern was shown to be positively associated with poorer brain aging profiles, contrasting the main analysis.

Figure 4. Associations Between Identified Patterns of Body Composition and Brain Aging Biomarkers.

The darkness of color corresponds to the magnitude of regression coefficients, with green denoting inverse associations and red denoting positive associations. Only those results that reached statistical significance (i.e., p-value <0.05 using the false discovery rate for adjustment of multiple testing) are marked with color. These regression coefficients and 95% CIs were derived from linear regression models after adjusting for age at imaging assessment; sex; annual household income; Townsend deprivation index; educational level; smoking and drinking status; physical activity; fruit and vegetable intake; history of hypertension, diabetes, or dyslipidemia; baseline cognition function; family history of neurodegenerative diseases; height; weight; and intracranial volume. †We reoriented WM, GM, hippocampus volumes, and FA values so that increases in these values reflected more severe damage, consistent with WMH and MD. FA = fractional anisotropy; GM = grey matter; MD = mean diffusivity; WM = white matter; WMH = white matter hyperintensity.

Modification Role of Disease Susceptibility on the Observed Associations

The associations between body composition patterns and neurodegenerative diseases were largely comparable for individuals with low or high disease susceptibility, index by PRS of AD and PD (Figure 5), APOE genotype (eFigure 8), or family history of neurodegenerative diseases (eFigure 9). For the “bone density” pattern, we only noted significant association among people with low disease susceptibility. These results remained largely consistent when applying alternative definitions to define low or high genetic risk (eFigure 10-11).

Figure 5. Associations Between Identified Patterns of Body Composition and Rate of Incident Neurodegenerative Diseases Among Individuals With Different Disease Susceptibilities.

High genetic risk of AD (or PD) was defined as having a polygenetic risk score (PRS) of AD (or PD) above the second tertile of the entire study sample. Low genetic risk of AD (or PD) was defined as having a PRS of AD (or PD) ≤first tertile. IR was calculated as numbers of cases per 1000 person-years. HRs and 95% CIs were derived from Cox models with attained age as the timescale after adjusting for sex; annual household income; Townsend deprivation index; educational level; smoking and drinking status; physical activity; fruit and vegetable intake; history of hypertension, diabetes, or dyslipidemia; baseline cognition function; family history of neurodegenerative diseases; height; and weight. AD = Alzheimer disease; HR = hazard ratio; IR = incidence rate; PD = Parkinson disease.

Mediating Effect of CVDs in the Phenotypic Associations

In the analysis of combined mediating effect (Figure 2B and eTable 6 and 7), CVDs significantly mediated the observed associations for “central obesity” (35.3%), “muscle strength” (10.7%), “bone density” (21.2%), “leg-dominant fat distribution” (22.1%), and “arm-dominant fat distribution” (14.3%) patterns but did not exhibit a significant mediating effect for the “fat-to-lean mass” pattern. Among individual CVDs, cerebrovascular diseases consistently contributed the highest proportion of mediation (8.9%–28.9%).

Subgroup and Sensitivity Analyses

The stratified analyses by age at baseline (eFigure 12), BMI categories (eFigure 13), and cognition function (eFigure 14) retained largely comparable results. The sensitivity analyses did not modify these results to any meaningful extent (eFigure 15–18).

Discussion

Our findings from this community-based cohort study of 412,691 UK Biobank participants with a 9-year follow-up confirm the substantial influence of body composition on the risk of neurodegenerative diseases in middle-aged individuals. It is important to note that through applying PCA to extract major body composition patterns using information from highly interrelated body measurements and leveraging disease diagnosis and brain imaging data to identify both clinical diagnosis of neurodegenerative disease and brain aging biomarkers, we consistently found increased risks of both outcomes in association with “central obesity” and “arm-dominant fat distribution” patterns and reduced risks with the “muscle strength” pattern. Furthermore, these associations persisted across individuals with different susceptibilities to neurodegenerative disease and were significantly mediated through CVDs (10.7%–35.3%), particularly cerebrovascular diseases. Our findings highlight the potential for improvement in body composition and early interventions in CVDs as a target in mitigating the future risk of neurodegenerative diseases.

Our findings were largely in line with previous studies that have revealed associations between diverse body components and risk of neurodegenerative diseases, although most of them rarely considered the interrelationships among various components of body composition and focused solely on clinically diagnosed neurodegenerative diseases as outcomes. The robust association of “central obesity” and “muscle strength” patterns with both neurodegenerative diseases and MRI-based brain aging biomarkers aligned with previous evidence from Mendelian randomization⁹ and observational studies.^6,7,24,25 The inverse associations of “bone density”¹² (high vs low) and “fat-to-lean mass”²⁶ (moderate vs low) patterns with the clinical onset of neurodegenerative diseases partly collaborated with previous studies yet showed inconsistency or even contradiction with the subclinical outcomes of brain aging biomarkers. Such discrepancies suggest potential reverse causation in clinical outcomes, even after incorporating a 5-year or 10-year lag time. Supporting this, evidence shows that bone density's effect on dementia decreased over time.¹² In addition, the null results of the “lean mass” pattern, together with previous inconsistent findings,^7,10,11 suggest that muscle quality, compared with muscle quantity, might play a more important role in the development of neurodegenerative diseases. Finally, our study expanded existing knowledge by demonstrating that the aforementioned associations were largely independent of disease susceptibility.

Our study contributes novel insights by identifying significant associations between “leg-dominant fat distribution” and “arm-dominant fat distribution” patterns with neurodegeneration risk, emphasizing the pivotal role of fat and muscle distribution in predicting and potentially mitigating neurodegenerative diseases, beyond traditional weight metrics. The potential protective effect of “leg-dominant fat distribution,” particularly among men, aligns with studies linking gynoid fat distribution (i.e., “pear-shaped” body) to reduced dementia risk.²⁷ The preferential fat storage around the hips and upper thigh areas has been linked to health benefits, mainly because it could guard against ectopic fat deposition within vital organs (e.g., muscles), enhance insulin sensitivity, and reduce inflammation.²⁸ Conversely, the “arm-dominant fat distribution” pattern was found to potentially exacerbate neurodegeneration risk, partly supported by literature indicating adverse effects of higher FM and lower LM in arms.²⁶ This pattern's correlation with increased abdominal fat ratio and muscle fat infiltration indicates that this kind of accumulation of leg LM may be compromised by excessive adiposity, reflecting a passive loading effect. This sharply contrasts with beneficial LM in the arms, which is typically enhanced through physical activity. These findings highlighted that targeted interventions to modulate body composition—reducing trunk and arm adiposity while promoting healthy muscle development—may be more effective for neurodegeneration protection than general weight control. For instance, lifestyle modifications, such as engaging in resistance training, reducing sedentary behavior, and adhering to a balanced diet, can effectively help in reducing central fat and enhancing muscle strength, which may potentially offer greater neuroprotective benefits than weight-focused antiobesity medications. Our study underscores the need for further research to explore these targeted strategies and elucidate their underlying biological mechanisms.

The underlying mechanisms linking body composition and neurodegeneration have not been fully investigated. Previous evidence suggests that vascular endothelial dysfunction could be a crucial factor in the pathogenesis of both vascular and primary neurodegenerative diseases.^29,30 Our study has demonstrated that individuals with high exposure levels to patterns such as “central obesity” and “arm-dominant fat distribution” and low exposure levels to the “muscle strength” pattern tend to exhibit a higher abdomen fat ratio and increased muscle fat infiltration. Ectopic fat deposition has been linked to dysfunction of adipose and muscle tissues, impairment in insulin signaling, and the release of proinflammatory cytokines,^31,32 ultimately leading to vascular endothelial dysfunction, which is also a key contributor to CVDs. Our exploratory mediation analysis validated previous observations, highlighting the substantial involvement of CVDs, particularly cerebrovascular disease, in the association between these body composition patterns and neurodegenerative diseases. Our findings indicate, therefore, that early management of CVDs after exposure to adverse body composition may be a pathway to mitigate future neurodegeneration.

Strengths of this study include the large sample size, long follow-up duration, assessment of major body composition patterns to account for complex interplay between different body components, coverage of clinical and subclinical outcomes for neurodegeneration, consideration of effect modification by disease susceptibility, and examination of CVDs as a potential mediating factor to unveil underlying mechanisms. Our study also has some limitations. First, body components measured by bioimpedance are based on estimation on electrical impedance, which may introduce greater variability compared with direct measurements such as dual-energy X-ray absorptiometry (DXA). However, the bioimpedance used in UK Biobank was proven to have good agreement with DXA,³³ while also offering practical advantages such as ease of use and wide applicability. Second, although we used a 5-year and 10-year lag time to minimize reverse causation, it is important to acknowledge that the preclinical phase of neurodegenerative disease may last 10–20 years or longer, introducing the possibility of residual reverse causation. Nevertheless, our complementary investigation of MRI-based brain aging biomarkers consistently confirmed the associations with “central obesity,” “muscle strength,” and “arm-dominant fat distribution” patterns. Third, despite the high diagnostic accuracy for all-cause dementia, the potential for subtype misclassification requires further validation through imaging or additional biomarkers. Fourth, because the UK Biobank participants are predominantly of European (primarily Anglo-Saxon) ancestry and not directly representative of the general UK population (e.g., generally healthier, especially those with brain MRI imaging assessments),³⁴ generalization of our results to other ethnicities or populations might be a concern. For instance, the variation in body composition between ethnicities—with Asian populations often having higher visceral adiposity for a given BMI compared with their European counterparts³⁵—may increase their predisposition to CVDs and influence the applicability of our neurodegenerative risk findings. Further research involving diverse ethnic cohorts is essential to confirm our findings.

In summary, our study consistently observed a positive association of body composition patterns featured with “central obesity” and “arm-dominant fat distribution” while a negative association of body composition pattern featured with “muscle strength,” with both clinical neurodegenerative disease onset and brain aging. These associations were mostly independent of disease susceptibility and partly mediated by CVDs. Our results underscore the feasibility of improving body composition, including mitigating CVD burden at an early stage, as a strategy to reduce future neurodegenerative disease risk.

Glossary

AD

Alzheimer's disease

BMI

body mass index

cSVD

cerebral small vessel disease

CVD

cardiovascular disease

DXA

dual-energy X-ray absorptiometry

FM

fat mass

HR

hazard ratio

LM

lean mass

PCA

principal component analysis

PCs

principal components

PD

Parkinson's disease

PRS

polygenetic risk score

Appendix. Authors

Name	Location	Contribution
Shishi Xu, MD, PhD	West China Hospital of Sichuan University, Chengdu, China	Drafting/revision of the manuscript for content, including medical writing for content; study concept or design; analysis or interpretation of data
Shu Wen, MD, PhD	West China Hospital of Sichuan University, Chengdu, China	Analysis or interpretation of data
Yao Yang, MSc	West China Hospital of Sichuan University, Chengdu, China	Analysis or interpretation of data
Junhui He, MSc	West China Hospital of Sichuan University, Chengdu, China	Analysis or interpretation of data
Huazhen Yang, MSc	West China Hospital of Sichuan University, Chengdu, China	Major role in the acquisition of data
Yuanyuan Qu, MSc	West China Hospital of Sichuan University, Chengdu, China	Major role in the acquisition of data
Yu Zeng, MSc	West China Hospital of Sichuan University, Chengdu, China	Major role in the acquisition of data
Jianwei Zhu, MD, PhD	West China Hospital of Sichuan University, Chengdu, China	Analysis or interpretation of data
Fang Fang, PhD	Karolinska Institutet, Solna, Sweden	Drafting/revision of the manuscript for content, including medical writing for content; analysis or interpretation of data
Huan Song, MD, PhD	West China Hospital of Sichuan University, Chengdu, China	Drafting/revision of the manuscript for content, including medical writing for content; study concept or design; analysis or interpretation of data

Study Funding

This research was conducted using the UK Biobank Resource under Application 54803; it used data assets made available by National Safe Haven as part of the Data and Connectivity National Core Study, led by Health Data Research UK in partnership with the Office for National Statistics and funded by UK Research and Innovation (grant ref: MC_PC_20029 and MC_PC_20058). This work was also supported by 1.3.5 Project for Disciplines of Excellence, West China Hospital, Sichuan University (grant no. ZYYC21005 to HS), Key Research and Development Project of Sichuan Provincial Science and Technology Department (grant no. 2023YFS0258 to SX), and the Swedish Research Council (Joint Program on Neurodegenerative Disease, 2021-00696 to FF).

Disclosure

The authors report no relevant disclosures. Go to Neurology.org/N for full disclosures.