Abstract
Background
Body mass index (BMI) affects cognitive health throughout life. BMI variability has emerged as a promising indicator for tracking longitudinal changes in BMI. The association of BMI variability with cognitive impairment remains unclear.
Objective
This study investigated the relationship between BMI variability and cognitive outcomes across age groups defined by a 50-year threshold.
Methods
The current study enrolled 1474 participants (> 18 years) from the China Suboptimal Health Cohort Study (COCAS). Various metrics, including the coefficient of variation (CV), standard deviation (SD), variability independent of the mean (VIM), and average real variability (ARV), were used to measure BMI variability. The main clinical outcome was the incidence of cognitive impairment, which was diagnosed via the Mini-Mental State Examination (MMSE). We stratified the subjects into two groups on the basis of age (> 50 years and ≤ 50 years), and multivariable logistic regression analysis was performed to evaluate the association between BMI variability and cognitive changes in these groups. We further used a latent growth curve model (LGCM) to explore the BMI trajectories of these two groups.
Results
Among those age > 50 years at baseline, stratified analysis revealed that the odds ratio (OR) comparing participants in the highest tertile with those in the lowest tertile of CV was 0.45 (95% confidence interval [CI], 0.30 to 0.96; P for trend = 0.002). Consistent results were observed when the BMI variability indices SD, VIM, and ARV were calculated. There was no association between BMI variability and cognitive performance among those age ≤50 years. LGCM showed a decline in BMI over time among those age > 50 years, whereas an increase in BMI was noted among those age ≤50 years.
Conclusions
In the older population, higher BMI variability was inversely associated with cognitive impairment risk, which may be partly explained by healthy weight loss patterns in this population.
Supplementary Information
The online version contains supplementary material available at 10.1186/s12889-025-24243-5.
Keywords: Body mass index, Cognition, Variability, BMI trajectory
Introduction
Cognitive impairment is an intermediate state between normal brain aging and dementia. Individuals with cognitive impairment have a heightened risk of progressing to dementia, with clinical progression rates ranging from 10 to 15% annually [1]. Recent data indicates that the global prevalence of cognitive impairment ranges approximately from 12 to 18% among individuals aged over 60 years, this number will continue to increase in the global aging process [2]. Hence, it becomes imperative to identify and manage risk factors associated with cognitive decline.
Enhancing brain health through scientific weight management has garnered much attention. In specific subgroups of individuals, being overweight or mildly obese was associated with a reduced risk of cognitive impairment, a phenomenon known as the “obesity paradox“ [3]. Although the association between BMI and cognition is not entirely consistent, the obesity paradox is more likely to occur when the study population is older and the study follow-up is shorter, possibly due to reverse causation [4–7].
Recently, there has been growing interest in the intrasubject variability in BMI. Unlike static BMI and BMI change, BMI variability measures the fluctuations in body weight over a given period. To some extent, this indicator addresses the gap left by previous longitudinal studies. Until now, research on the relationship between BMI variability and cognitive impairment has been limited. Several clinical trials have indicated that elevated BMI variability is linked to increased amyloid-beta peptide deposition and accelerated cognitive decline, regardless of gender and BMI at baseline. However, most of these studies have focused primarily on the older population (mean age 60–70 years), leaving the role of age itself insufficiently clarified [8–12]. This study investigated the association between BMI variability and the incidence of cognitive impairment stratified by age. Given existing evidence suggesting a greater propensity for weight loss in older adults, we hypothesized that (1) the association between BMI variability and cognitive impairment differs across age groups, and that (2) such differences may be attributable to distinct BMI trajectories across age strata.
Methods
Study design and participants
Our participants were derived from the COCAS in 2015 and were followed for four years until 2019. This community-based study aims to explore how suboptimal health status contributes to the incidence of chronic disease. Longitudinal cognitive change was a subgroup of this study. This cohort was established in July 2013 and will continue follow-up until December 2024. Residents over 18 years who agreed to be involved in future follow-up investigations were included. Detailed information about this cohort has been described in prior publications [13]. Following the declaration of Helsinki, this study was approved by the Ethical Committees of the Staff Hospital of Jidong Oilfield of Chinese National Petroleum and it was registered in the Chinese Clinical Trial Registry on October 21, 2020 with the identifier ChiCTR2000039204. All participants or their legal guardians gave written informed consent to the procedure.
Measures of BMI variability
The height and weight of the participants were measured while in a relaxed standing position. BMI was calculated as weight in kilograms divided by height in square meters. Four indices of BMI variability, as previously described, were computed, including the standard deviation in BMI (BMI-SD), coefficient of variation in BMI (BMI-CV), variability independent of the mean in BMI (BMI-VIM), and average real variability in BMI (BMI-ARV). The BMI-CV was calculated as SD/mean × 100 (%). The BMI-VIM was calculated as 100 × SD/meanbeta, where beta is the regression coefficient based on a natural logarithm of SD on the natural logarithm of the mean. The BMI-ARV was calculated as the average absolute difference among successive measurements [14]. For each participant, all of these indices were derived from four BMI measurements. The CV was used as the primary measure of variability.
Cognitive assessment
All participants underwent a face-to-face interview and completed the Chinese version of the MMSE during the later evaluation. The MMSE is a widely recognized tool for evaluating global cognitive status across various domains, including orientation, registration, attention, calculation, recall, and language. The MMSE scores ranged from 0 to 30 points, with higher scores indicating better cognitive performance. Cognitive impairment was defined based on the education-specific cutoffs of MMSE score: <18 for illiterates, < 21 for primary school graduates, and < 25 for junior school graduates or above [15].
Covariates
Blood samples of participants were collected from the antecubital vein in the morning under fasting conditions. Fresh samples underwent various biochemical analyses using an auto-analyzer (Hitachi 747; Hitachi, Tokyo, Japan). Serum creatinine (CR), cholesterol (TC), triglycerides (TG), high-density lipoproteins (HDL-C), low-density lipoproteins (LDL-C), and uric acid (UR) levels were assessed annually during the follow-up for all participants.
Additional covariates were collected through a set of combined self-administered questionnaires (refer to Supplementary material). Demographic variables included age, gender (male or female), and education level (middle school or below, high school or above). Health behavior variables included smoking status (current or not current), alcohol consumption (current or not current), and exercise frequency (never or more than once per week). Health status was evaluated using the following variables: hypertension, diabetes, cancer, cardiovascular events, and cerebrovascular events. Hypertension was defined as any current use of antihypertensive drugs, a diagnosis of hypertension during a healthcare examination, or a self-reported history. Diabetes was defined as any current use of glucose-lowering drugs, fasting glucose level ≥ 7.0 mmol/L, or a self-reported history. Cardiovascular and cerebrovascular events included stroke, coronary heart disease, atrial flutter/fibrillation, and heart failure.
Statistical analysis
All analyses were conducted separately for groups stratified by age. We divided the population into two groups using age 50 as the boundary, with individuals aged exactly 50 years included in the younger group. This cut-off was chosen based on previous studies indicating that health-related risk factors tend to accumulate more rapidly from this period [16, 17]. Baseline characteristics were presented as means ± standard deviation for normally distributed continuous variables, median (lower quartile-upper quartile) for skewed continuous variables, and frequency (percentage, %) for categorical variables. ANOVA was used for normally distributed data and the Kruskal-Wallis test was used for non-normally distributed data. Fisher’s exact test was used for categorical variables. We further divided the participants into three groups based on BMI variability tertiles and treated intraindividual variability as categorical variables in subsequent analyses. The association between BMI variability and the risk of cognitive impairment was determined with multivariable logistic regression models. Covariates with clinical relevance or a univariate association with the outcome were included in the multivariale logistic regression model. Variables for inclusion were carefully chosen, given the number of events available, to ensure parsimony of the final model. Model 1 adjusted for no covariates and Model 2 adjusted for age, gender, mean LDL-C, hypertension, diabetes, exercise frequency, smoking status, and alcohol consumption. Collinearity diagnostics were performed for all covariates, and the final included covariates had variance inflation factor values less than 5. Education was not included since most participants had completed high school education or above. Additionally, tests for linear trends were conducted by introducing the median value of each category of BMI variability as a continuous variable in Model 2 [18]. We further used LGCM to describe the BMI trajectories in two age-stratified groups. LGCM summarizes these trajectories into intercepts and slopes, representing the average BMI on the starting point and the average rates of change, respectively [19]. We assumed all trajectory shapes were cubic, quadratic, or linear. We assessed the fit of various models using the Root Mean Square Error of Approximation (RMSEA), the Comparative Fit Index (CFI), the Tucker-Lewis Index (TLI), the Akaike Information Criteria (AIC), and the Bayesian Information Criteria (BIC). A model was considered a good fit if it had an RMSEA less than 0.06, CFI and TLI greater than 0.95, and the lowest AIC and BIC. We also used Normed Chi-square (χ2/df) to determine whether the model is overfitting, the ideal value was 1–3, and the value below 1 suggested overfitting of the model [20]. We used SAS version 9.4 to manage the data, SPSS statistics 25 to analyze the data, and Mplus version 8.7 to complete the trajectory model. All p-values were two-tailed, and statistical significance was defined as a p-value less than 0.05.
Results
Study participants and characteristics
From 2015 to 2019, 6656 adults without cognitive impairment at baseline were initially included. However, we excluded 3136 participants due to missing education and MMSE data, 2009 participants due to missing BMI data, and 37 participants due to cancer, cardiovascular disease, or cerebrovascular disease. Ultimately, the current study enrolled 1474 participants, with 538 participants (36.5%) over the age of 50 (refer to Fig. 1). We presented comparison between participants included and excluded (refer to Supplementary Table 1). There were differences in all BMI variability indices and education between participants included and excluded.
Fig. 1.
Flow chart of the current study
Table 1 summarizes participant characteristics stratified by BMI-CV and age. By the end of the follow-up period, 146 participants (9.9%) had developed cognitive impairment, with 115 participants over fifty showing cognitive impairment compared to only 31 participants under fifty. In the younger group, participants with higher BMI-CV had a higher proportion of females, smokers, and drinkers, higher average HDL-C, and lower average WC, TG, UR, and CR. In the older group, participants with higher BMI-CV had a lower WC and proportion of males. We also stratified participants based on tertiles of other BMI variability indices and age. The baseline characteristics across these groups are summarized in Supplementary Tables 2–4. While some overlapping patterns were observed, there were also indicator-specific differences in group characteristics.
Table 1.
Baseline characteristics according to age and BMI-CV status
| Age ≤ 50 years (N = 936) | Age > 50 years (N = 538) | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| Lowest Tertile (N = 312) | Middle Tertile (N = 312) |
Highest Tertile (N = 312) |
P value | Lowest Tertile (N = 179) | Middle Tertile (N = 181) |
Highest Tertile (N = 178) |
P value | ||
| Age (years) | 36.10 ± 7.31 | 36.12 ± 7.79 | 35.04 ± 7.52 | 0.177 | 60.38 ± 5.41 | 60.07 ± 4.96 | 59.45 ± 5.29 | 0.239 | |
| Gender (n,%) | < 0.001a, b,c | 0.025a | |||||||
| Male | 207(66.3) | 173(55.4) | 131(42.00) | 93(52.0) | 91(50.2) | 69(38.8) | |||
| Female | 105(33.7) | 139(44.6) | 181(58.0) | 86(48.0) | 90(49.8) | 109(61.2)) | |||
| BMI (kg/m2) | 24.97 ± 3.74 | 24.11 ± 3.42 | 24.55 ± 3.47 | 0.011b | 25.39 ± 2.96 | 25.05 ± 2.96 | 24.79 ± 3.16 | 0.173 | |
| WC (cm) | 86.83 ± 10.77 | 83.63 ± 9.98 | 84.14 ± 10.76 | 0.001a, b | 90.21 ± 9.07 | 88.44 ± 8.69 | 87.80 ± 9.87 | 0.035a | |
| BMI variability | |||||||||
| SD | 0.32(0.24–0.38) | 0.55(0.47–0.62) | 0.93(0.77–1.28) | < 0.001a, b,c | 0.29(0.22–0.36) | 0.53(0.47–0.59) | 0.86(0.73–1.13) | < 0.001a, b,c | |
| CV | 0.01(0.01–0.01) | 0.02(0.02–0.03) | 0.04(0.03–0.05) | < 0.001a, b,c | 0.01(0.01–0.01) | 0.02(0.02–0.02) | 0.04(0.03–0.04) | < 0.001a, b,c | |
| ARV | 0.30(0.23–0.40) | 0.55(0.40–0.69) | 0.90(0.68–1.18) | < 0.001a, b,c | 0.28(0.20–0.35) | 0.50(0.40–0.63) | 0.80(0.58–1.01) | < 0.001a, b,c | |
| VIM | 0.37(0.29–0.44) | 0.67(0.59–0.75) | 1.10(0.94–1.48) | < 0.001a, b,c | 0.34(0.24–0.41) | 0.60(0.54–0.68) | 1.03(0.88–1.27) | < 0.001a, b,c | |
| Education (n,%) | 0.607 | 0.848 | |||||||
| Middle school or below | 18(5.7) | 15(4.8) | 21(6.7) | 11(6.1) | 14(7.7) | 12(6.7) | |||
| High school or above | 294(94.3) | 297(95.2) | 291(93.3) | 168(93.9) | 167(92.3) | 166(93.3) | |||
| Exercise (n,%) | 0.301 | 0.876 | |||||||
| Yes | 164(52.6) | 177(56.7) | 158(50.6) | 120(67.0) | 121(66.9) | 123(69.1) | |||
| No | 148(47.4) | 135(43.3) | 154(49.4) | 59(33.0) | 60(33.1) | 55(30.9) | |||
| Smoking status (n,%) | 0.013a, b | 0.140 | |||||||
| Yes | 89(28.5) | 76(24.4) | 58(18.6) | 30(16.8) | 41(22.7) | 22(12.4) | |||
| No | 223(71.5) | 236(75.6) | 254(81.4) | 149(83.2) | 140(77.3) | 156(87.6) | |||
| Alcohol consumption (n,%) | 0.001a | 0.111 | |||||||
| Yes | 142(45.5) | 117(37.5) | 97(31.1) | 61(34.1) | 50(27.6) | 43(24.2) | |||
| No | 170(54.5) | 195(62.5) | 215(68.9) | 118(65.9) | 131(72.4) | 135(75.8) | |||
| Hypertension (n,%) | 0.176 | 0.506 | |||||||
| Yes | 83(26.6) | 64(20.5) | 69(22.1) | 71(39.2) | 82(45.3) | 72(40.4) | |||
| No | 229(73.4) | 248(79.5) | 243(77.9) | 108(60.8) | 99(54.7) | 106(59.6) | |||
| Diabetes (n,%) | 0.747 | 0.897 | |||||||
| Yes | 10(3.2) | 11(3.5) | 14(4.4) | 29(16.2) | 30(16.6) | 32(18.0) | |||
| No | 302(96.8) | 301(96.5) | 298(95.6) | 150(83.8) | 151(83.4) | 146(82.0) | |||
| TC (mg/dL) | 4.74(4.19–5.34) | 4.84(4.33–5.42) | 4.75(4.17–5.34) | 0.241 | 5.31(4.56–5.92) | 5.23(4.68–5.80) | 5.21(4.48–5.89) | 0.931 | |
| TG (mmol/L) | 1.49(0.94–2.41) | 1.45(0.95–2.09) | 1.26(0.94–1.85) | 0.006a | 1.49(1.10–2.15) | 1.46(1.07–2.09) | 1.41(0.95–2.09) | 0.267 | |
| HDL-C (mmol/L) | 1.09(0.96–1.32) | 1.15(1.00-1.31) | 1.17(1.01–1.33) | 0.028a | 1.19(1.05–1.33) | 1.17(1.06–1.36) | 1.21(1.05–1.36) | 0.574 | |
| LDL-C (mmol/L) | 2.03(1.46–2.45) | 2.06(1.57–2.53) | 2.03(1.58–2.40) | 0.510 | 2.44(1.91–2.88) | 2.35(1.91–2.83) | 2.31(1.82–2.78) | 0.519 | |
| UR (umol/L) | 317.00(267.00-392.00) | 311.00(252.00-357.50) | 292.00(234.75–350.50) | 0.001a | 297.00(251.00-344.00) | 292.00(243.50–346.00) | 289.50(246.50-334.25) | 0.396 | |
| CR (µmol/L) | 100.00(85.00-113.00) | 97.00(85.00-111.00) | 94.00(82.00-109.25) | 0.030a | 95.00(84.00-107.00) | 97.50(85.25-108.75) | 94.00(82.75–107.00) | 0.423 | |
aParticipants with the highest tertile were significantly different from the lowest tertile (p < 0.05); bParticipants with middle tertile were significantly different from the lowest tertile(p < 0.05); cParticipants with the highest tertile were significantly different from the middle tertile(p < 0.05)
BMI: body mass index; WC: waist circumference; CV: coefficient of variation; SD: standard deviation; VIM: variability independent of the mean; ARV: average real variability; TC: total cholesterol; TG: triglycerides; HDL-C: high-density lipoprotein-cholesterol; LDL-C: low-density lipoprotein cholesterol; UR: uric acid; CR: creatinine
Association of BMI variability with cognitive outcomes
As shown in Tables 2 and 3, the effect of BMI variability on the incidence of cognitive impairment varied by age. Among participants over 50 years old, the highest tertile of BMI-CV was significantly inversely associated with the risk of cognitive impairment compared to the lowest tertile in the unadjusted model (OR 0.44, 95% CI 0.26–0.75, p for trend = 0.002). This inverse association remained statistically significant after adjusting for confounders (OR 0.45, 95% CI 0.26–0.77, p for trend = 0.004). Additionally, for a per-SD increase in BMI-CV, the fully adjusted OR for incident cognitive impairment was 0.60 (95%CI, 0.38–0.96). Similar results were observed when participants were stratified by BMI-SD, BMI-ARV, or BMI-VIM. However, among those under 50, BMI variability showed positive associations with the risk of cognitive impairment, but these results did not reach statistical significance.
Table 2.
Association of BMI variability with the incidence of cognitive impairment in the group aged > 50
| Variable | Lowest Tertile | Middle Tertile | Highest Tertile | Per-SD increase | P for trend |
|---|---|---|---|---|---|
| SD | |||||
| Total | 187 | 196 | 155 | ||
| Event | 47 | 46 | 22 | ||
| Model 1 | Reference | 0.91(0.57–1.46) | 0.49(0.28–0.86) | 0.56(0.33–0.94) | 0.014 |
| Model 2 | Reference | 0.91(0.56–1.47) | 0.50(0.28–0.89) | 0.57(0.33–0.99) | 0.019 |
| CV | |||||
| Total | 179 | 181 | 178 | ||
| Event | 50 | 39 | 26 | ||
| Model 1 | Reference | 0.71(0.44–1.15) | 0.44(0.26–0.75) | 0.58(0.38–0.91) | 0.002 |
| Model 2 | Reference | 0.68(0.41–1.12) | 0.45(0.26–0.77) | 0.60(0.38–0.96) | 0.004 |
| ARV | |||||
| Total | 217 | 168 | 153 | ||
| Event | 57 | 33 | 25 | ||
| Model 1 | Reference | 0.69(0.42–1.12) | 0.55(0.32–0.93) | 0.65(0.43–0.98) | 0.021 |
| Model 2 | Reference | 0.63(0.38–1.04) | 0.58(0.34–0.99) | 0.66(0.43–1.02) | 0.037 |
| VIM | |||||
| Total | 178 | 182 | 178 | ||
| Event | 48 | 40 | 27 | ||
| Model 1 | Reference | 0.76(0.47–1.24) | 0.48(0.29–0.82) | 0.60(0.40–0.90) | 0.013 |
| Model 2 | Reference | 0.74(0.45–1.21) | 0.49(0.28–0.84) | 0.61(0.40–0.95) | 0.010 |
Model 1: unadjusted; Model 2: adjusted for age, gender, mean LDL-C, hypertension, diabetes, exercise, smoking, and drinking status. CV: coefficient of variation; SD: standard deviation; VIM: variability independent of the mean; ARV: average real variability
Bold data are data with statistical significance
Table 3.
Association of BMI variability with the incidence of cognitive impairment in the group aged ≤ 50
| Variable | Lowest Tertile | Middle Tertile | Highest Tertile | Per-SD increase | P for trend |
|---|---|---|---|---|---|
| SD | |||||
| Total | 355 | 309 | 272 | ||
| Event | 8 | 13 | 10 | ||
| Model 1 | Reference | 1.91(0.78–4.66) | 1.66(0.64–4.25) | 1.01(0.71–1.44) | 0.364 |
| Model 2 | Reference | 1.73(0.67–4.51) | 1.95(0.69–5.56) | 1.09(0.70–1.69) | 0.229 |
| CV | |||||
| Total | 312 | 312 | 312 | ||
| Event | 8 | 13 | 10 | ||
| Model 1 | Reference | 1.65(0.68–4.04) | 1.26(0.49–3.23) | 0.97(0.66–1.42) | 0.745 |
| Model 2 | Reference | 1.59(0.60–4.19) | 1.41(0.49–4.02) | 1.04(0.66–1.63) | 0.587 |
| ARV | |||||
| Total | 356 | 355 | 225 | ||
| Event | 8 | 15 | 8 | ||
| Model 1 | Reference | 1.04(0.66–1.63) | 1.60(0.59–4.34) | 1.03(0.74–1.43) | 0.410 |
| Model 2 | Reference | 1.92(0.75–4.89) | 1.47(0.49–4.37) | 1.12(0.662–1.89) | 0.557 |
| VIM | |||||
| Total | 312 | 312 | 312 | ||
| Event | 8 | 12 | 11 | ||
| Model 1 | Reference | 1.52(0.61–3.77) | 1.39(0.55–3.50) | 0.95(0.64–1.41) | 0.549 |
| Model 2 | Reference | 1.34(0.50–3.58) | 1.55(0.55–4.35) | 1.02(0.64–1.61) | 0.415 |
Model 1: unadjusted; Model 2: adjusted for age, gender, mean LDL-C, hypertension, diabetes, exercise, smoking, and drinking status. CV: coefficient of variation; SD: standard deviation; VIM: variability independent of the mean; ARV: average real variability
BMI trajectories over 4 years in groups aged ≤ 50 and > 50
In both groups, linear modeling achieved the best model fits (refer to Supplementary Table 5) compared to the higher degrees. Model fit indices supported the adequacy of the linear models, with acceptable values observed in both groups (for participants aged ≤50: RMSEA = 0.006, CFI = 1.00, TLI = 1.00; for those aged > 50: RMSEA = 0.001, CFI = 1.00, TLI = 1.00). Overall, the change in BMI appears opposite between the two groups (refer to Supplementary Table 6). Among those aged ≤50, participants experienced a slight growth in BMI (mean intercept = 24.46; p < 0.001; mean slope = 0.05; p = 0.004). By contrast, participants experienced a slight decline in BMI among those aged > 50 (mean intercept = 25.18; p < 0.001; mean slope =−0.05; p = 0.031). We present the mean and SD of BMI for each group at each time point, details are provided in Supplementary Fig. 1.
Discussion
We observed that substantial BMI variability was inversely associated with the risk of cognitive impairment, a departure from findings in prior studies. This association remained robust after adjusting for key confounders. In the subsequent LGCM, distinct BMI trajectories were identified between the two age groups. While the physiological function of BMI variability independent of BMI changes has often been assumed in previous studies, this assumption may be overly simplistic [21]. The concept of BMI variability was inspired by research on heart rate variability. During the interview, the heart rate rarely continued to increase or decrease, and its shape was close to a sinusoidal function, making the direction of change less relevant [22]. However, BMI may change directionally over time. Therefore, the different effects of BMI variability on cognitive function between the two age groups may be partially explained by differences in BMI trajectories.
So far, greater BMI variability has always been predictive of a greater risk of cognitive impairment, both clinically and pathologically. Bae and colleagues observed a linear trend between BMI variability and cognitive impairment, but this statistically significant association was only noted for mild cognitive impairments [10]. In another study, cognitively unimpaired individuals with more substantial variability in BMI exhibited accelerated cognitive decline [12]. With respect to BMI trajectory, accruing evidence suggests that unintentional weight loss in the elderly is associated with brain atrophy and cognitive impairment, and cognitive integrity is better preserved in those with stable weights [23, 24]Additionally, few studies have simultaneously examined the effects of BMI variability and BMI trajectories on cognitive outcomes. Kang and colleagues reported increased amyloid-beta deposition in individuals with greater BMI variability, regardless of BMI change [9]. Qin and colleagues reported that a declining BMI trend and high variability in late-middleaged individuals were associated with a higher risk of dementia, although BMI ascend could partially offset the adverse effects of high BMI variability on cognitive health [16].
BMI fluctuations arise from dynamic changes in adipose, muscle, and skeletal tissue. Individuals with higher bone mineral density and higher muscle mass have been linked to improved cognitive performance and a reduced risk of dementia over long-term follow-up. Conversely, fat accumulation has been linked to smaller total brain volume, particularly reduced cingulate and hippocampal volumes, leading to poorer performance on memory tasks [25–27]. In later decades of life, unintended weight fluctuations are often due to pathological protein deposition, osteoporosis, and sarcopenia, which increase the risk of developing Alzheimer’s disease [23, 27, 28]. When people get older, lack exercise, have poor appetites, and have bad habits, they experience more lean tissue damage and fat accumulation, leading to greater BMI variability [29, 30].
In our study, we found that in the oleder group, greater BMI variability—despite an overall trend toward slight weight loss—was associated with a reduced risk of cognitive impairment. This unexpected finding raises important questions about the potential underlying mechanisms involved.
Compared with other studies, our participants exhibited better overall health. Notably, individuals aged over 50 years had lower proportions of smokers and drinkers, and a higher proportion of exercise, only 70 participants were over 65 years old, accounting for less than 5% of the total. These population characteristics differ from those of previous studies, as our cohort mainly comprised late-middle-aged participants (50–65 years) rather than elderly adults (> 65 years). One possible explanation is that modest weight fluctuations in late-middle-aged individuals may reflect preserved physiological resilience or adaptive metabolic flexibility, rather than underlying pathology. Additionally, BMI fluctuations in this group may primarily result from reductions in adiposity, possibly accompanied by slight increases in lean mass, suggesting an overall improvement in body composition. This pattern differs from the involuntary weight loss typically observed in older, frailer populations. Nevertheless, this hypothesis requires further investigation using direct measures of body composition. Among individuals under 50 years of age, BMI variability appeared harmful, despite their healthy lifestyles and low likelihood of age-related diseases. Prior studies have demonstrated that persistent obesity and related chronic inflammation in early life may contribute to detrimental effects on the brain [31, 32]. Weight gain and the burden of being overweight in early life may lead to adverse cognitive performance. However, our results did not reach statistical significance, so the effects of BMI variability on cognitive function in early life remain controversial.
Our study has several strengths. First, we controlled potential confounding factors for cognitive function. Second, we introduced LGCM to characterize the time course of BMI. Third, we discussed the role of BMI variability in people younger than 50 years of age, which has been neglected in previous studies. However, there are still some limitations. First, cognitive impairment, commonly observed in geriatric patients, typically begins after the age of 50, fewer people under the age of 50 develop cognitive impairment. Generally, we consider the number of positive events to be at least 10 times greater than the number of covariates in the logistic regression model, since our ORs did not have excessively large confidence intervals, we also performed logistic regression. Second, we used MMSE to evaluate cognitive function, but this instrument has a ceiling effect on identifying cognitive dysfunction. Third, our sample reflects a subset of the general population; the generalizability of our results to the general population should be cautious.
Conclusions
In conclusion, our results indicate that substantial visit-to-visit variability in BMI is associated with a reduced risk of cognitive impairment among participants aged > 50, BMI variability driven by a healthy lifestyle seems to be beneficial for maintaining cognitive integrity, whereas these associations remain unclear among participants aged < 50. Further researches are warranted to explore the effects of BMI variability on brain health, with attention to BMI trajectories and lifestyles of the population. In addition, conducting long-term follow-up studies in younger individuals is essential to elucidate the impact of BMI variability in early life on cognitive performance.
Supplementary Information
Below is the link to the electronic supplementary material.
Acknowledgements
We appreciate all members participating in the China Suboptimal Health Cohort Study (COCAS).
Author contributions
The conceptual foundation of the research: YMZ, Data acquisition and analysis: SHG, YHL, XYC, YMZ, Data interpretation: SHG, YHL, XYC, DLW, Figures and drafting creation: SHG, DLW, Writing the first draft: SHG, Supervision research: YMZ. All contributing authors have examined and endorsed the finalized manuscript and jointly agree to its submission for journal publication.
Funding
Funding for this study was provided by the National Natural Science Foundation of China (81972144, 82372555).
Data availability
The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.
Declarations
Competing interests
The authors declare no competing interests.
Footnotes
Publisher’s note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Change history
11/21/2025
Following the article's publication, Funding statement was added.
Contributor Information
Shuanghui Guo, Email: 18810673986@163.com.
Yumei Zhang, Email: zhangyumei95@aliyun.com.
References
- 1.Lee KW, Hong YJ, Yang EJ, Lee SB, Kim SH, Na S, Kim YD, Park JW. Feasibility and usefulness of cognitive monitoring using a new home-based cognitive test in mild cognitive impairment: a prospective single arm study. BMC Geriatr. 2024;24(1):241. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Wang W, Sun P, Lv T, Li M. The impact of modifiable health metrics on mortality for older adults with low cognitive function. Front Public Health. 2024;12:1304876. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Zhang J, Na X, Li Z, John S, Li G, Yang H, Yang Y, Tan Y, Zhang J, Xi M, et al. Sarcopenic obesity is part of obesity paradox in dementia development: evidence from a population-based cohort study. BMC Med. 2024;22(1):133–133. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Chiesa ST, Norris T, Garfield V, Richards M, Hughes AD. Early-life cumulative exposure to excess bodyweight and midlife cognitive function: longitudinal analysis in three British birth cohorts. Lancet Healthy Longev. 2024;5(3):e204-13. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Ginny N, Zhang Y, William HD, Ap CS. Obesity in Late-Life as a protective factor against dementia and dementia-Related mortality. Am J Alzheimer’s Dis Other Dement. 2023;38:15333175221111658–15333175221111658. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Simchoni M, Derazne E, Pinhas-Hamiel O, Cukierman-Yaffe T, Bendor CD, Bardugo A, Chodick G, Tzur D, Endevelt R, Gerstein HC, et al. Adolescent body mass index and cognitive performance: a nationwide study of 2.48 million Israeli adolescents. Eur J Endocrinol. 2023;188(7):630–40. [DOI] [PubMed] [Google Scholar]
- 7.Zhu L, Lei M, Tan L, Zou M. Sex difference in the association between BMI and cognitive impairment in Chinese older adults. J Affect Disord. 2024;349:39–47. [DOI] [PubMed] [Google Scholar]
- 8.Liang F, Fu J, Moore J, Zhang X, Xu Y, Qiu N, Wang Y, Li R. Body mass index, waist circumference, and cognitive decline among Chinese older adults: a nationwide retrospective cohort study. Front Aging Neurosci. 2022;14: 737532–737532. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Kang SH, Kim JH, Chang Y, Cheon BK, Choe YS, Jang H, Kim HJ, Koh SB, Na DL, Kim K, et al. Independent effect of body mass index variation on amyloid-beta positivity. Front Aging Neurosci. 2022;14:924550. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Mi BE, Min PS. Association between variations in body mass index and cognitive function in older Korean adults. J Obes Metabolic Syndrome 2021:30(3):271–278. [DOI] [PMC free article] [PubMed]
- 11.Zhou R, Liu H, Li F, Yu J, Yuan Z, Zheng J, Zou L, Wu X. Variability in cardiometabolic and inflammatory parameters and cognitive decline. Am J Prev Med. 2021;61(4):e181–9. [DOI] [PubMed] [Google Scholar]
- 12.Schnaider BM, Amir T, Lin H, Sapir G, Ethel B, Mary S, Zhu CW. Stability in BMI over time is associated with a better cognitive trajectory in older adults. Alzheimer’s Dement J Alzheimer’s Assoc. 2022;18(1):2131–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Wang Y, Ge S, Yan Y, Wang A, Zhao Z, Yu X, Qiu J, Alzain MA, Wang H, Fang H, et al. China suboptimal health cohort study: rationale, design and baseline characteristics. J Transl Med. 2016;14(1):291. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Soroush M, Leila C, Niloofar D, Neda C, Maryam T, Fereidoun A, Farzad H. Mean versus variability of lipid measurements over 6 years and incident cardiovascular events: more than a decade follow-up. Front Cardiovasc Med. 2022;9: 1065528–1065528. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Katzman R, Zhang M, Ouang YQ, Wang Z, Liu W, Yu E, Wong S, Salmon D, Grant I. A Chinese version of the mini-mental state examination; impact of illiteracy in a Shanghai dementia survey. J Clin Epidemiol. 1988;41(10):971–8. [DOI] [PubMed] [Google Scholar]
- 16.Qin Z, Liu Z, Li R, Luo Y, Wei Z, He L, Pei Y, Su Y, Hu X, Peng X. Association between BMI trajectories in late-middle age and subsequent dementia risk in older age: a 26-year population-based cohort study. BMC Geriatr. 2023;23(1):773–773. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Singh-Manoux A, Kivimaki M, Glymour MM, Elbaz A, Berr C, Ebmeier KP, Ferrie JE, Dugravot A. Timing of onset of cognitive decline: results from Whitehall II prospective cohort study. BMJ. 2012;344:d7622. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Park S, Freedman ND, Haiman CA, Le Marchand L, Wilkens LR, Setiawan VW. Association of coffee consumption with total and Cause-Specific mortality among nonwhite populations. Ann Intern Med. 2017;167(4):228–35. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Needham BL, Epel ES, Adler NE, Kiefe C. Trajectories of change in obesity and symptoms of depression: the CARDIA study. Am J Public Health. 2010;100(6):1040–6. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Drouard G, Hagenbeek FA, Whipp AM, Pool R, Hottenga JJ, Jansen R, Hubers N, Afonin A, Willemsen G, de Geus EJC, et al. Longitudinal multi-omics study reveals common etiology underlying association between plasma proteome and BMI trajectories in adolescent and young adult twins. BMC Med. 2023;21(1):508–508. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.McMenamin KJ, Harris TB, Baker JF. Weight variability, physical functioning and incident disability in older adults. J Cachexia Sarcopenia Muscle. 2023;14(4):1648–56. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Almeida-Santos MA, Barreto-Filho JA, Oliveira JL, Reis FP, da Cunha Oliveira CC, Sousa AC. Aging, heart rate variability and patterns of autonomic regulation of the heart. Arch Gerontol Geriatr. 2016;63:1–8. [DOI] [PubMed] [Google Scholar]
- 23.Guo J, Wang J, Dove A, Chen H, Yuan C, Bennett DA, Xu W. Body mass index trajectories preceding incident mild cognitive impairment and dementia. JAMA Psychiatr. 2022;79(12):1180–7. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24. Zonneveld MH, Noordam R, Sabayan B, Stott DJ, Mooijaart SP, Blauw GJ, Jukema JW, Sattar N, Trompet S. Weight loss, visit-to-visit body weight variability and cognitive function in older individuals. Age Ageing. 2023;52(1):afac312. [DOI] [PMC free article] [PubMed]
- 25.Booranasuksakul U, Macdonald IA, Stephan BCM, Siervo M. Body composition, sarcopenic obesity, and cognitive function in older adults: findings from the National health and nutrition examination survey (NHANES). 1999–2002 and 2011–2014 J Am Nutr Association 2024;11–4. [DOI] [PubMed]
- 26.Chen L, Zhao S, Wang Y, Niu X, Zhang B, Li X, Peng D. Genetic insights into obesity and brain: combine Mendelian randomization study and gene expression analysis. Brain Sci. 2023;13(6):892. [DOI] [PMC free article] [PubMed]
- 27.Zhang X, Hsu WWQ, Sing CW, Li GHY, Tan KCB, Kung AWC, Wong JSH, Wong IC, Cheung CL. Low bone mineral density with risk of dementia: A prospective cohort study. J Am Med Dir Assoc. 2022;23(10):1719 e1719. [DOI] [PubMed] [Google Scholar]
- 28.Wei S, Nguyen TT, Zhang Y, Ryu D, Gariani K. Sarcopenic obesity: epidemiology, pathophysiology, cardiovascular disease, mortality, and management. Front Endocrinol (Lausanne). 2023;14: 1185221. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Skinner J, Shepstone L, Hickson M, Welch AA. Alcohol consumption and measures of sarcopenic muscle risk: cross-sectional and prospective associations within the UK biobank study. Calcif Tissue Int. 2023;113(2):143–56. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Vanderbilt UM, Center N, Tennessee V, University Medicine Center N, Tennessee V, Center UM, Tennessee N, Education GR, et al. Association of smoking with abdominal adipose deposition and muscle composition in Coronary Artery Risk Development in Young Adults (CARDIA) participants at mid-life: a population-based cohort study. PLoS Med. 2020;17(7): e1003223. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Anstey KJ, Cherbuin N, Budge M, Young J. Body mass index in midlife and late-life as a risk factor for dementia: a meta-analysis of prospective studies. Obes Reviews: Official J Int Association Study Obes. 2011;12(5):e426–437. [DOI] [PubMed] [Google Scholar]
- 32.Leigh SJ, Morris MJ. Diet, inflammation and the gut microbiome: mechanisms for obesity-associated cognitive impairment. Biochim Biophys Acta Mol Basis Dis. 2020;1866(6): 165767. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.