Sex Differences in the Association Between Body Mass Index and Dementia Risk in Community-Dwelling Japanese People Aged 40–74 Years

Study design and participants

The present study is a cohort study with an 8-year follow-up. The baseline study included 14,364 middle-aged and elderly individuals, aged 40–74 years, living in the Murakami region (Murakami city, Sekikawa village, and Awashimaura village) of Niigata Prefecture, Japan. Among the baseline participants, the following were excluded: 23 who had long-term care insurance at baseline, 515 with missing baseline data regarding education level, lifestyle factors, and/or BMI, and 24 with outlier BMI values. The final analyzed population consisted of 13,802 participants. Informed consent was obtained from all participants. The study protocol was approved by the Ethics Committee of Niigata University (Nos. 452, 1324, and 2018-0417).

Baseline survey

In the baseline survey conducted in 2011–2013, a self-administered questionnaire requested information on body size (height, weight, and waist circumference), sociodemographic characteristics (sex, age, marital status, education level, and occupation), lifestyle, and disease history. We verified self-reported height, weight, and waist circumference by comparing them with those measured directly in sub-samples. Spearman’s correlation coefficients between self-reported and measured data were 0.974 for men’s height, 0.975 for women’s height, 0.972 for men’s weight, and 0.973 for women’s weight (N = 1,752 for men, N = 2,259 for women, p < 0.0001 for all coefficients) [15]. Differences in mean values (self-reported minus measured) in men and women were 0.4 (SD, 1.6) and 0.4 cm (SD, 1.4) for height, 0.1 (SD, 2.3) and 0.2 kg (SD, 1.8) for weight, and –0.1 (SD, 0.9) and 0.0 kg/cm² (SD, 0.8) for BMI, respectively, after excluding extreme values exceeding±3SD. Spearman’s correlation coefficients between self-reported and measured waist circumference at the umbilical level in the standing position were 0.830 for men and 0.769 for women (N = 1,545 for men, N = 1,898 for women, p < 0.0001 for all coefficients). BMI was calculated as weight (kg) divided by height squared (m²). Marital status was classified as married, never married, and divorced, separated, or bereaved. Education level was categorized as junior high school, high school, junior or vocational college, and university or higher. Occupation was classified as office work, sales, and service; professional or management; manual (security; farming, forestry, or fishery; transportation; and labor services); and no job or others. Total physical activity levels were assessed with the metabolic equivalents (MET) score (MET-h/d), which was calculated by multiplying the time spent on activities per day by its MET intensity. Smoking habit was classified as non-smoker, past smoker, 1–20 cigarettes/day, and≥20 cigarettes/day. Alcohol consumption was classified as non- or rare-drinkers, 1–149, 150–299, 300–449, and≥500 g ethanol/week. History of heart disease (myocardial infarction, heart failure, atrial fibrillation, angina, and other), stroke, and diabetes was obtained from the self-reported questionnaire. The study protocol of this cohort study was previously described [15].

Follow-up survey

Dementia cases were identified using the long-term social care insurance (LTCI) database in Japan for frail individuals and individuals aged≥40 years [16]. Under the LTCI system, an individual’s eligibility is assessed according to the Opinion Paper of a physician in charge [17], and the levels of disabling dementia are evaluated by a physician and classified into six ranks (0, I, II, III, IV, V) ranging from no dementia (0) to individuals with severe dementia-related behavioral disturbance and cognitive impairment requiring medical treatment (V); individuals with Rank II (moderate dementia-related behavioral disturbance and cognitive impairment with slight dependence) or higher are considered to have disabling dementia [18]. The specificity of this dementia determination, using criteria defined by the International Psychogeriatric Association in collaboration with the World Health Organization as a gold standard [19], was reported to be high (94–97%) [20]. Person-years of observation were determined using data on moving out and death from residency registration and death registration records in conformity with the Basic Residential Registry Law and Family Registry Law.

Statistical methods

Mean values and standard deviations (SDs) were calculated for continuous variables. Values exceeding 3 SDs of BMI were excluded as outliers. BMI was categorized as underweight (<18.5 kg/m²), normal (18.5–24.9 kg/m²), overweight (25–29.9 kg/m²), and obese (≥30.0 kg/m²) according to the WHO criteria (WHO 2000) [4]. Since a U-shaped association between BMI and dementia risk is assumed in the present population, we divided normal weight into three levels: low-normal, mid-normal, and high-normal. BMI categories were ultimately divided into six groups, as follows: <18.5 (underweight), 18.5–20.6 (low-normal), 20.7–22.6 (mid-normal), 22.7–24.9 (high-normal), 25.0–29.9 (overweight), and≥30.0 (obese).

Cox proportional hazards models with years of follow-up as the time variable were used to calculate hazard ratios (HRs) and p for trend values for dementia according to BMI group by sex and age group. We conducted a similar analysis (sensitivity analysis) by using waist circumference as an alternative outcome. Multivariate analysis was conducted by adjusting for covariates of age, marital status (dummy variable), education, occupation (dummy variable), total physical activity, smoking, drinking (coded as 1, none/rarely or 450 g ethanol/week; and 0, other categories), coffee consumption, and disease history, where alcohol consumption was not linearly associated with dementia risk [21] and thus was converted into a dichotomous variable.

Associations between BMI and dementia risk were analyzed similarly in the cohort, in which participants diagnosed with dementia within the first four years of follow-up were excluded, because having preclinical dementia may affect BMI. When comparing HRs among BMI groups, we considered the group with the lowest HR as the reference group. However, the BMI group with the lowest HR (“obese” group) had a relatively small number of male participants. Therefore, we set the 2nd lowest BMI group (“overweight” group) as the alternative reference. SAS statistical software (release 9.4, SAS Institute Inc., Cary, NC, USA) was used for statistical analysis. p < 0.05 was considered statistically significant.

Main findings

The present 8-year cohort study, targeting community-dwelling people aged 40–74 years, demonstrated a dose-dependent inverse association between BMI and dementia risk in men and a U-shaped association in women. This sex difference was characterized by differences in the risk of dementia in overweight/obese individuals relative to non-overweight/non-obese individuals.

Interpretation of findings

Although obesity is traditionally considered a potential risk factor for dementia [22], the impact of being obese/overweight on dementia risk differs by life stage, i.e., high BMI in mid-life is associated with a high risk of dementia, and high BMI in late-life is associated with a lower or no risk [8–11]. Our findings, which covered mid-life through early late-life, may have been more reflective of mid-life characteristics. In this context, no association was observed between overweight/obese men and dementia risk, whereas overweight/obese women were found to be at a higher risk of dementia than the reference (high-normal BMI group). There are several potential explanations for this sex difference. First, obesity during mid-life is a well-established risk factor for the subsequent onset of dementia [9,23,24, 9,23,24]. Because women are more prone to dementia than men, the increased risk may have been observed specifically in women. Second, ethnicity may play a role. For example, while the meta-analysis by Lee et al. [7] did not find a sex difference in the manner of association between obesity/overweight and dementia risk, cohort studies included in the analysis mainly targeted Caucasian populations (22 of 23 studies), with only one study targeting an Asian population. The percent body fat of East Asians, including Japanese people, is reportedly higher than that of Caucasians, despite the BMI of Asians being lower [25], and a comparative study on adult body composition showed that Japanese men and women had significantly lower percent fat mass than their Germen counterparts in the range between underweight and normal [26]. In addition, percent fat mass in middle-aged and older Japanese people is higher in women than in men [27]. These findings, i.e., that mid-life adiposity is an important risk factor for dementia, may explain at least in part the high risk of dementia in overweight/obese women but not in men [24,28, 24,28]. Finally, cardiometabolic risk factors may be related to the sex differences observed in overweight/obese individuals with dementia. Although women have a lower cardiometabolic risk, they seem to have a higher risk of diabetic complications than men, including, for example, cardiovascular disease and depression, which are also risk factors for Alzheimer-type dementia [29]. In addition, microvascular disease is a greater contributor to cardiovascular disease in women than in men [29].

Jacob et al. [14] reported that being overweight was associated slightly, but significantly, with a lower risk of dementia in women, but not in men, in a German clinical setting. This discrepancy with our present findings might be explained by the older population of that study (10 years older than our study population), as being overweight/obese in late-life is potentially protective against dementia [12,23, 12,23].

The mechanisms linking overweight/obesity with cognitive decline/dementia have been reported by some studies. For instance, obesity is associated with neuroinflammation by the systemic and chronic presence of proinflammatory cytokines [30], which potentially increases the risk of Alzheimer-type dementia. Additionally, low levels of adiponectin in obese women, especially during the menopausal transition, could adversely affect the brain and cause dementia [31,32, 31,32]. Meanwhile, others have reported that abdominal obesity may be key, independently of BMI [33] or inflammation markers [28]. Mechanisms underlying the association between adiposity and dementia risk should be further investigated.

Low BMI was associated with a high dementia risk in both men and women in the present study. This finding is consistent with several meta-analyses [7,9,34, 7,9,34] and recently published studies targeting East Asian populations [5,6, 5,6]. Although low BMI may be a risk factor for dementia, reverse causation has recently been suggested in some studies [9,35, 9,35].

We also found that a high waist circumference is associated with a high risk of dementia in women, which was not so pronounced in men. This finding is consistent with the association between BMI and dementia risk. Waist circumference is an index of central obesity, which may be another independent risk factor that can explain the sex differences in the observed associations.

Strengths and limitations

The strength of the present study is that it is the first to examine the association between BMI and dementia by sex in an East Asian population. However, the present study also has some limitations worth nothing. First, dementia cases were not directly diagnosed in clinical settings, but instead identified using the LTCI database. This may have resulted in some misclassification. Second, the type of dementia was not specified in the present study, preventing a clearer understanding of the mechanism of association. However, two-thirds of dementia cases have been reported to be of the Alzheimer’s type in Japan [36], suggesting that our findings may apply to individuals with Alzheimer’s disease.

Conclusions

We compared dementia risk between sexes in a Japanese population aged 40–74 years and found sex differences in the association between BMI and dementia risk, with a dose-dependent inverse association in men and a U-shaped association in women. This suggests an increased risk of dementia due to being overweight/obese in women but not in men, and that sex differences in adiposity may underlie the association. Further cohort studies in other populations will be needed to confirm these findings.