Associations of body habitus and its changes with incident dementia in older adults

This study examined the associations of body mass index (BMI) and waist circumference (WC), as well as their short‐ and long‐term changes over time, with incident dementia in older individuals.


INTRODUCTION
Dementia is one of the leading causes of disability and death, currently affecting over 55 million individuals. 1 With a growing aging population, the number of individuals with dementia is projected to be over 130 million in 2050. 2 This will impose burdens on individuals, the healthcare system and society.
Obesity in midlife is considered a risk factor for dementia, 3 but the relationship between body habitus and dementia becomes more complex in later life.Although some studies have shown that being overweight and obese is not associated with dementia risk, 4,5 a meta-analysis found that a higher body mass index (BMI) was associated with decreased dementia risk. 6However, individuals in the studies had a broad age range (e.g., ≥18 years), with only a few specifically focused on older individuals.Moreover, some studies suggested that waist circumference (WC) is also a marker of adiposity, as transitions in body composition (e.g., fat/lean mass) may be not characterized by weight change. 6,7Meanwhile, being underweight in later life has been associated more consistently with increased dementia risk, 6 but the extent to which this reflects prodromal dementia, 6,8 and whether weight loss or gain is a risk factor, remains inconclusive.
Using data from a large cohort of older communitydwelling adults aged predominantly 70+ years, this study aimed to examine the associations of (1) BMI and WC in older age; (2) the variability and change of these measures across 2 years; (3) long-term change in BMI (from 18 years to 70+ years), with the risk of incident dementia; and (4) potential sex differences in these associations.

Overview
Details of methods used in the current study are presented below.In brief, participants were generally healthy older individuals aged 65-98 years at enrolment.Using Cox models, three analyses examining the

Key points
• An underweight body mass index (BMI) and substantial decrease in BMI >5% over 2 years in older age were associated with increased dementia risk, but these associations were attenuated over time, suggesting potential reverse causality.• An overweight or obese BMI, as well as an elevated waist circumference (WC) in later life, were associated with lower dementia risk, but a two-year increase in BMI >5% showed an association with greater risk of dementia.• Compared to participants with a normal-weight BMI at both age 18 and 70+, those with obesity at both time-points showed an association with greater risk of dementia (especially in women), while a lower risk was shown in those with an overweight BMI at either or both time-points, and in those being non-obese in early adulthood but developed obesity at 70+ years.

Why does this paper matter?
These points provide insights into the complex relationships between body habitus and dementia risk across the lifecourse.First, the findings indicate that being underweight and losing weight may be signs of prodromal dementia, highlighting the importance of weight monitoring for older individuals.Second, maintaining body weight at a stable level in later life showed protective effects against dementia, suggesting that weight management via diet and exercise may be beneficial.Third, the impact of body habitus on later-life cognitive health may be established even in earlier life.Therefore, management of body weight and shape from an early age throughout the lifetime may provide long-term cognitive benefits.
prospective association with incident dementia were conducted for (1) BMI and WC at baseline, (2) the 2-year variability and change of BMI and WC, and (3) long-term BMI change from the age of 18 to 70 years.Timeon-study was used as the time axis.

Study participants
The participants of this study were enrolled in the ASPREE (ASPirin in Reducing Events in the Elderly) clinical trial and its follow-up observational study-ASPREE-eXTension.

Anthropometric measures at baseline
According to a pre-defined protocol, 12 anthropometric information including height, weight and WC was objectively measured at baseline.The weight and WC were remeasured at annual follow-ups.Height at baseline was used for calculations of BMI at all time-points.Height and weight were measured using a stadiometer where possible, and if not, a non-stretch measuring tape and an electronic scale.WC was measured with arms down on bare skin, head up and normal breathing.The categories of BMI were defined as 12 underweight (<20.0 kg/m 2 ), normal weight (20.0-24.9kg/m 2 ), overweight (25-29.9kg/m 2 ), and obese (30+ kg/m 2 ), which closely match the recommendations from the World Health Organization. 13For WC, three subgroups were defined according to recommendations from the Australian Government Department of Health and Aged Care 14 : low (<94 cm for men, <80 cm for women), elevated (94-101.9cm for men, 80-87.9cm for women), and highly elevated (102+ cm for men, 88+ cm for women).

Anthropometric changes across 2 years
The absolute change and variability (intra-individual standard deviation) in BMI and WC across 2 years (baseline, year 1 and year 2) were calculated.For variability measures, tertiles were created.[17]

BMI changes from 18 to 70+ years
In the ALSOP sub-cohort, participants self-reported body weight at the time of the survey and recalled their weight at the age of 18 years. 11Five patterns of long-term BMI change from 18 to 70+ years were defined, with the exclusion of participants being underweight at baseline due to small sample size 17 : long-term normal weight (normal BMI at both times), long-term overweight (overweight at either or both times, but neither obese), obese to non-obese (obese at 18 but not obese at 70+ years), non-obese to obese (not obese at 18 but obese at 70+ years), and long-term obese (obese at both times).

Dementia ascertainment
All participants underwent regular assessment at baseline and annually during follow-up in global cognition, verbal fluency, episodic memory and psychomotor speed.Participants meeting any of the following criteria (triggers) were identified as suspected dementia cases: (1) a score <78/100 or substantial decline (>10.15 points from age-and education-adjusted predicted score) on 3MS (which was used for cognitive assessment at baseline and annually during follow-up); (2) self-reported cognitive concerns; (3) a clinical diagnosis of dementia noted on the medical records; (4) prescription of cholinesterase inhibitors.Subsequently, a series of cognitive and functional evaluations were conducted after the initial dementia triggers, including Alzheimer's Disease Assessment Scale-Cognitive subscale, Color Trails, Lurian overlapping figures and the Alzheimer's Disease Cooperative Instrumental Study Activities of Daily Living scale.Additional documents, such as hospital records, medical reports and the results of laboratory tests and brain imaging, were also collected if available.An international committee of neurologists/geriatricians with expertise in dementia reviewed all materials and confirmed dementia according to the DSM-IV criteria.Adjudication of dementia cases required evidence of memory impairment and at least one of these cognitive deficits: executive dysfunction, agnosia, aphasia, or apraxia.These needed to be severe enough to impact social/occupational functioning, with substantial functional decline. 18

Statistical analysis
Proportional hazards Cox regression models were used to examine the association between anthropometry and incident dementia.The time of enrolment was used as the 'entry time' for the analyses of baseline BMI/WC and long-term BMI change.For models including variability and change of BMI/WC, participants diagnosed with dementia, died, or lost to follow-up within the first 2 years after enrolment were excluded, making the year 2 visit the 'entry time'.Days between the entry time and the date of dementia diagnosis (for dementia cases) or the last follow-up visit (for dementia-free participants) were used as the time axis.
We performed stratified analysis by sex and the status of Apolipoprotein E (ApoE) ε4 allele carrier.To examine reverse causation, analyses were conducted by dividing the follow-up into two intervals and assessing if the associations are consistent across intervals. 19All analyses were performed using Stata version 16.0 (Stata Corp., College Station, Texas, USA).The statistical significance was set at a 2-tailed p-value<0.05.

Ethics and informed consent
The ethics review board at each participating institution approved the ASPREE, ASPREE-XT, and the ALSOP study.Written informed consent was obtained from each participant.

RESULTS
The participants included in the analyses are described in Figure 1.A total of 118,621 participants (116,443 personyears) with complete data for all baseline variables included in the analysis.All dementia cases between baseline and year 2 were additionally excluded for analyses of two-year change/variability.The participant characteristics are summarized in Table 1 and Supplementary Table 1.Compared to those without dementia during follow-up, dementia cases were older, with a higher proportion of men, and were more likely to live alone at home, have depressive symptoms, or have never had alcohol.
There is some evidence of reverse causality in these associations, given we observed attenuation in some associations in the second interval (Supplementary Table 3).This was notable for underweight individuals, with the effect size decreasing from 98% increased dementia risk over the first 4 years ( p = 0.006) to 33% ( p = 0.30) after year 4.This was also shown in the HRs of obese BMI, dropping from 0.64 ( p = 0.003) to 0.82 ( p = 0.13).In contrast, attenuation of the associations was not observed for overweight BMI and elevated WC.
Figure 3 and Supplementary Table 5 show the results of long-term BMI change.Compared to those who maintained normal weight from 18 to 70 years, participants who were overweight at both times and those who were non-obese at 18 years but obese in later life, were less likely to develop dementia (overweight: HR:0.80, 95% CI:0.64-0.99,p = 0.04; non-obese to obesity, HR:0.70, 95% CI:0.51-0.95,p = 0.02).In contrast, being obese at both times was associated with the highest dementia risk (HR:2.27,95% CI:1.22-4.24,p = 0.01).

DISCUSSION
In this large cohort of older community-dwelling individuals, an overweight BMI and elevated WC at baseline were associated with decreased risk of dementia, whereas weight gain over the first 2 years after baseline showed an association with increased dementia risk.While obesity exclusively in old age did not appear to be a risk factor, individuals who reported obesity in both early and Categories of waist circumference were defined as: for men, low <94 cm; elevated 94 to <102 cm; highly elevated ≥102 cm; for women, low <80 cm; elevated, 80 to <88 cm; highly elevated ≥88 cm.
later life had increased dementia risk.Being underweight and losing weight in later life were associated with a higher risk of dementia.This finding, however, may be attributed to reverse causation, due to the attenuation of these associations over time.The associations of elevated WC and long-term obesity with dementia were stronger in women, while the associations of being underweight and weight change (either direction) were stronger in ApoE ε4 carriers.Consistent with our findings, weight loss in older age has been documented as a risk factor for dementia in prior literature. 6,15,161][22] These conditions may result in a decrease in body weight and meanwhile, adversely affect cognitive function.][25] There are also theories suggesting that the associations between being underweight or undergoing weight loss, and lower dementia risk are due to reverse causality. 15,16,26In line with this speculation are the attenuated associations over time, among the participants who were underweight at baseline and those whose BMI or WC decreased.The early neuropathological changes during the prodromal phase of dementia may influence body weight before the manifestation of cognitive symptoms.For example, those with preclinical neurodegeneration may experience difficulties in cooking and food F I G U R E 2 Adjusted associations of baseline BMI and WC with incident dementia (n = 18,621).BMI, body mass index; CI, confidence interval; HR, hazard ratio; WC, waist circumference.(1) Categories of BMI were defined as underweight: BMI < 20 kg/m 2 ; normal: 20-24.9kg/m 2 ; overweight: 25-29.9kg/m 2 ; obese: 30 kg/m 2 or above.(2) Categories of waist circumference were defined as: for men, low <94 cm; elevated 94 to <102 cm; highly elevated ≥102 cm; for women, low <80 cm; elevated, 80 to <88 cm; highly elevated ≥88 cm.(3) The models adjusted for the following covariates at baseline: age (continuous), sex (men, women), ethnicity (Australian white; American white; Hispanic/Latino; Black; other), education (<12 years; 12-15 years; ≥16 years), living situation (at home alone; at home with someone or in a residential home), smoking status (never; former; current), alcohol intake (never; former; current low risk; current high risk), hypertension (yes; no), diabetes (yes; no), dyslipidemia (yes; no), depression (yes; no), pulse pressure (continuous) and triglycerides (continuous).The models adjusted for the following covariates at baseline: age (continuous), sex (men; women), ethnicity (Australian white; American white; Hispanic/Latino; Black; other), education (<12 years; 12-15 years; ≥16 years), living situation (at home alone; at home with someone or in a residential home), smoking status (never; former; current), alcohol intake (never; former; current low risk; current high risk), hypertension (yes; no), diabetes (yes; no), dyslipidemia (yes; no), depression (yes; no), pulse pressure (continuous), triglycerides F I G U R E 3 Adjusted associations of long-term changes in BMI from early to later life with incident dementia (n = 11,033).BMI, body mass index; CI, confidence interval; HR, hazard ratio.(1) Early to later life BMI categories were defined as 1) normal weight at both times, overweight at either or both times (but neither obese), non-obese to obese (not obese at age 18 and obese at age 70+ years), obese to nonobese (obese at age 18 and not obese at age 70+ years), and obese at both times.(2) The models adjusted for the following covariates at baseline: age (continuous), sex (men; women), ethnicity (Australian white; American white; Hispanic/Latino; Black; other), education (<12 years; 12-15 years; ≥16 years), living situation (at home alone; at home with someone or in a residential home), smoking status (never; former; current), alcohol intake (never; former; current low risk; current high risk), hypertension (yes; no), diabetes (yes; no), dyslipidemia (yes; no), depression (yes; no), pulse pressure (continuous) and triglycerides (continuous).
purchasing, impairments in oral, olfactory and motor function, as well as loss of appetite, [27][28][29][30][31] resulting in poorer dietary patterns and lower calorie intake.These can be linked to weight loss as early signs of dementia onset.
Being overweight and even obese in later life appeared to be a protective factor, while long-term obesity was associated with an increased risk. 6These findings, which may appear counterintuitive, highlight the complex relationship between body habitus and dementia.Excess weight has been proposed to alter the cerebral structure and energy metabolism, with links to brain atrophy, ischemia and hypoperfusion. 32Obesity also increases the risk of many cardiometabolic conditions, such as cardiovascular diseases, which are themselves associated with dementia. 3,33As such, obesity in midlife is recognized as a risk factor for dementia, 3 and our research suggests that excess weight in earlier adulthood also appears to increase risk, even decades later for individuals who have survived into older age.In contrast, being overweight appears to be protective in both earlier and later life.One explanation is that individuals with excess body weight may have high blood levels of leptin and adiponectin, which are neuroprotective hormones produced predominantly by adipose tissue. 34Another explanation is superior nutritional status and physical function, which provides resilience against illness.This is supported by findings from ASPREE-XT 35 and elsewhere 36 of lower mortality risk in overweight individuals. 35,368][39] However, with further increased adipose tissue beyond a tipping point, these benefits may be outweighed by cardiovascular risks and reduced muscle. 40,41This is consolidated by our findings that protective effects of obesity against dementia disappeared after the initial 4 years, while the benefits of overweight BMI remained consistent over time.The relationship between BMI and dementia in older age is complex.Higher late-life BMI has been associated with a lower risk of all-cause dementia, but a higher risk of vascular dementia, 6 which indicates both cognitive benefits and cardiovascular risks.One study found such an inverse association between late-life BMI and dementia only in those who are genetically prone to higher BMI. 42nother study observed that the association between obesity and cognitive function was modified by APOE status, as well as the stage and subtype of dementia. 43These findings suggest a complex role of excess weight in cognitive aging with an interplay between genes and environment.Further, obese participants fulfilling the eligibility criteria at enrolment were likely much healthier than obese individuals who did not eligibility criteria, and whose disease status may have prevented their enrolment.
Our results demonstrating that weight gain is associated with dementia add to the few studies that investigated this. 15,16,44Apart from the cardiovascular risk, weight gain contributes to homeostatic imbalance and cellular dysfunction via certain pathways (e.g., insulin resistance), which leads to neuronal damage. 16,45,46However, the impact of elevated BMI on dementia has been observed to flip across the life course being detrimental in earlier life but protective after the age of 70. 6,47,48herefore, weight gain may be neuroprotective when it is gradual, slow, and not excessive with the timeframe lengthened to decades.In contrast, it may be a risk factor when occurring in later life, suggesting that keeping a stable normal or modestly elevated weight may be beneficial for brain aging.Despite the lack of evidence for reverse causation, rapid weight gain may also be a sign of dementia onset, resulting from behavioral changes.For example, individuals at this stage may reduce their physical activity due to physical limitations or changes in motivation. 49Also, altered eating habits characterized by overeating and a preference for strong flavors may occur as dementia prodromes, 50 possibly due to impaired sensory function.Meanwhile, emotional disturbances such as apathy, depression and anxiety, which are common in prodromal dementia, 51 may exacerbate these conditions, with increased food intake for self-comfort.
We found that both higher BMI and WC were associated with reduced dementia risk, but the association with WC was no longer significant when adjusted for BMI.A meta-analysis found lower dementia risk in those with a larger WC, 6 yet another study pooling over five million individuals across a wide age range (36-84 years) showed opposite results with the adjustment of BMI. 52These discrepancies suggest that the neuroprotective effects of WC may be largely driven by BMI.Since excess body mass may be caused by fat accumulated in various regions, it is crucial to consider fat distribution for adiposity-related risks.Central adiposity has been suggested as an indicator of increased visceral fat and a risk factor for cognitive impairment, possibly by increasing the level of proinflammatory cytokines. 7,53,54his study adds to the knowledge regarding the association between body habitus and dementia in older age.First, the anthropometric assessments followed a stringent protocol, with stadiometers cross-checked for accuracy and trained staff monitored annually for compliance with the protocol.This minimized any potential measurement error.Second, we assessed the change in anthropometric measures over time in variability and absolute difference, allowing an investigation into both the extent and pattern of weight change.Second, reverse causality was tested by stratifying the observational period into two intervals.Third, we analyzed both earlyand later-life BMI, to explore the potential impact of weight at these different time-points.These findings will provide new information for the management of body habitus for cognitive aging.
Limitations include the relatively healthy population because of the inclusion criteria of the clinical trial.As a result, they may be more resilient to health risks conferred by adiposity and thus, findings may not be entirely generalizable to a wider population.Second, we cannot differentiate between intentional and unintentional weight change, which may be attributable to different underlying health conditions and/or behaviors. 55,56herefore, further studies should consider whether weight change is driven by behaviors or illness.Third, BMI at the age of 18 years was calculated using the body height objectively measured at age 70+ years.As body height generally decreases with age, using a laterlife body height overestimates earlier-life BMI, and thus some participants may have been misclassified to a higher BMI category at age 18 years.Therefore, the results of lifetime obesity in our study should be interpreted cautiously.Fourth, the two-year span for the short-term analysis may not adequately capture the longitudinal patterns of BMI and WC in later life, limiting the temporal resolution.In the long-term analysis, the lack of intermediate measurement between the age of 18 and 70+ years may lead to an oversimplification of the variation and fluctuation of body weight across the life course.Fifth, the prodromal phase of dementia may last for 10-20 years, so a long period of lag time has been suggested for longitudinal analysis to fully account for the reverse causation in dementia research.However, our study, even with a four-year time, showed some evidence of reverse causation.Similarly, another prospective study examining social/cognitive activities and dementia risk in 1.3 million women observed a decrement in the association for some activities, across three 5-year intervals. 19hese findings suggest that certain weight loss may occur at a relatively late stage of prodromal dementia.
Sixth, there may be potential multiple comparisons arising from the extensive subgroups in the analyses.However, most findings showed p-values substantially below the conventional threshold, indicating a high level of statistical significance.As such, these results should be evaluated from both a statistical and clinical perspective, ensuring an unbiased understanding of their implications.

CONCLUSIONS
In this cohort of older adults, an overweight or obese BMI, or a higher WC in old age was associated with reduced dementia risk.In contrast, an increased risk of dementia was observed in those who experienced rapid BMI increase in old age, and those persistently obese in both early and later adulthood.The underlying mechanisms remain unclear (Supplementary Figure 1), necessitating further research.Notably, being underweight and weight loss also showed an association with higher dementia risk.However, there is evidence of reverse causation, since their associations were attenuated over time.Therefore, weight loss may be not a causal factor of cognitive decline, but an early reflection of the behavioral and health changes that may occur before dementia is diagnosed.
continuous), BMI (underweight, normal, overweight, obese; only for BMI variability and BMI change ), WC (low, elevated, highly elevated; only for WC variability and WC change ).b BMI variability and WC variability were defined for each individual, respectively, as the standard deviation of BMI and WC measured at baseline, year 1 and year 2. c BMI change and WC change were defined as the percentage of change from baseline to year 2, categorized as stable (change between À5% and 5%), increase (change >5%) and decrease (change < À5%).
Characteristics of the participants included in the analysis according to dementia status, over follow-up (n = 18,621).
T A B L E 1 *p-values are based on Pearson's chi-squared test or Fisher's exact test.a T A B L E 2 Associations of change of BMI and WC with incid4ent dementia (n = 15,887).