The U-shaped Association between Body Mass Index (BMI) and Cognitive Impairment: Associated Characteristics in Low BMI and High BMI Chinese Adults

Cognitive impairment (CI) is characterized by declines in memory, language, and judgment. The etiology includes both non-modifiable factors and modifiable ones. Among these, body mass index (BMI) has a complex association with CI. Previous studies have reported divergent patterns: one described a J-shaped association [1], while another reported a U-shaped relationship [2]. Notably, optimal BMI levels appear to vary by ethnicity. Among White individuals in the UK Biobank, the optimal BMI was 29.2 kg/m² for men and 28.5 kg/m² for women [3]. In contrast, a complex non-linear relationship was observed in Chinese older adults [4]. Moreover, some reports have described an ‘obesity paradox,’ wherein a higher BMI is associated with a reduced risk of CI [5]. Given the inconsistencies among previous studies, this study aims to examine the relationship between BMI and CI in Chinese adults and to identify associated characteristics within abnormal BMI subgroups.

This study utilized data from 2015 wave of the China Health and Retirement Longitudinal Study (CHARLS), which included 11,582 participants aged 45 years or older. The cognitive assessment is in accorded with the Health Retirement Study and was adapted from the Telephone Interview for Cognitive Status. CI was defined as a score of ≤11 [6]. Study participants were classified into BMI categories according to Chinese guidelines: low BMI (<18.5 kg/m²), normal BMI (18.5 to 23.9 kg/m²), and high BMI (≥24.0 kg/m²) [7]. All continuous variables were categorized as follows: Low-density lipoprotein cholesterol (LDL-C) levels were classified as normal (<3.4 mmol/L) or elevated (≥3.4 mmol/L), high-density lipoprotein cholesterol (HDL-C) levels were categorized as normal (≥1.0 mmol/L) or reduced (<1.0 mmol/L) [8], hemoglobin levels were defined as normal (120–160 g/L for men and 110–150 g/L for women) or reduced (<120 g/L for men and <110 g/L for women), fasting blood glucose levels were classified as normal (<6.1 mmol/L) or elevated (≥6.1 mmol/L) [9].

Restricted cubic spline (RCS) regression analyzed the association between BMI and CI. Univariate logistic regression was used to identify potential associated characteristics, with multiple comparisons adjusted using the false discovery rate method. All statistically significant (P<0.05) categorical variables from the univariate analysis were included in a multivariate logistic regression model to identify independent associated characteristics.

RCS analysis revealed a U-shaped association, with the lowest point observed at a BMI of 27.3 kg/m² (Fig. 1). In low BMI group, hypertension (odds ratio [OR], 1.582; 95% confidence interval, 1.049 to 2.396; P=0.029) and reduced hemoglobin levels (OR, 0.866; 95% confidence interval, 0.778 to 0.962; P= 0.008) were associated with CI. Among individuals in the high BMI group, unmarried (OR, 1.726; 95% confidence interval, 1.408 to 2.115; P<0.001), elevated LDL-C (OR, 4.212; 95% confidence interval, 3.478 to 5.109; P<0.001), elevated fasting blood glucose (OR, 1.053; 95% confidence interval, 1.021 to 1.086; P=0.001), and reduced HDL-C (OR, 0.879; 95% confidence interval, 0.804 to 0.962; P=0.005) were associated with CI. Additionally, female sex, older age, rural residence and depressive symptoms were consistent associated characteristics for CI across both abnormal BMI subgroups.

Fig. 1.

Restricted cubic spline analysis showed the non-linear relationship between body mass index (BMI) and cognitive impairment risk after adjustment for age and sex. A significant negative association was observed at BMI levels below the inflection point (27.3 kg/m²), while a positive association emerged at higher BMI values. Tick marks along the horizontal axis indicate the distribution of BMI in the study population.

This study observed a U-shaped association between BMI and the risk of CI in adults. Specifically, in the low BMI group, hypertension and lower hemoglobin levels were significantly associated with CI. Conversely, in the high BMI group, being unmarried, elevated LDL-C, higher fasting blood glucose, and reduced HDL-C were significantly associated with CI. To our knowledge, this is the first and largest investigation to examine the relationship between comprehensive venous blood biomarkers and CI stratified by BMI subgroup in a Chinese adult population. It is important to note that the Chinese population presents distinct characteristics, not only in its demographic scale but also in its unique BMI distribution and metabolic profiles, which differ considerably from those observed in Western populations.

The significant U-shaped associations observed in our primary analysis are consistent with findings from a previous investigation [10]. Specifically, a recent study reported that the nadir risk of CI occurred at a BMI of approximately 25.5 kg/m² among adults aged 65 years and older. In parallel, our results also indicate a protective effect of overweight against CI in participants over 45 years of age. Notably, the BMI cutoffs identified in our study fell within the overweight category, which further supports the notion that the optimal BMI range for preserving cognitive function in older Chinese adults may lie within the overweight spectrum [3,10]. It should be emphasized; however, that this BMI value reflects the point of lowest probability of CI only in the context of this cross-sectional analysis and may not be optimal for other health outcomes. Thus, these finding should be interpreted within the broader context of an individual’s comorbidities and overall clinical profile.

Our study has several limitations that should be considered. Firstly, the ‘unmarried’ category was heterogeneous, comprising individuals who were separated, divorced, widowed, and never-married. In particular, widowhood may act as potential confounder, as the social and psychological sequelae associated with bereavement could independently influence the risk of CI. However, the small size of widowed subpopulation precluded a stratified analysis of this factor, which should be explored in future research with larger samples. Secondly, the correlational nature of our cross-sectional design precludes causal inference between CI and BMI. Reverse causality remains plausible, whereby early cognitive decline could lead to reduced appetite, impaired self-care, and subsequent weight loss. Thus, these findings warrant cautious interpretation and should be validated in future longitudinal studies.

This study reveals a U-shaped association between BMI and CI among middle-aged and older adults in China. At lower BMI levels, CI was observed associated with nutritional deficiencies and hypertension. Conversely, at higher BMI levels is associated with abnormalities in glucose metabolism and dyslipidemia. Furthermore, female sex, advanced age, rural residence, and depressive symptoms were associated with CI in individuals with either low or high BMI.