Background
Dementias are chronic, progressive neurological diseases characterized by memory loss and cognitive impairment [
1]. Among the dementias, the most common one is Alzheimer's disease (AD), which accounts for approximately 50–70% of all dementia patients [
2]. Critically, its prevalence is rising sharply, owing to the global population aging [
3]. Therefore, reducing the burden of AD has become an important global public health issue [
4].
The main pathological features of AD include amyloid plaques and neuronal filament entanglement [
5,
6]. AD is believed to arise from a combination of genetic and environmental factors and it can be divided into early-onset AD (EOAD) and late-onset AD (LOAD) [
3,
7]. It is important to identify the gene-environment interactions behind the development of AD, which would allow for the development of personalized intervention strategies for early intervention of AD, thereby ultimately reducing the global incidence of AD [
8].
In conjunction with the rising rate of AD, there is also a worrying epidemic of high levels of obesity worldwide [
9]. As an indicator of body nutrition, body mass index (BMI) has been reported to be associated with cerebrovascular adverse events, a variety of cancers, and other diseases [
10‐
13]. However, the association between BMI and AD risk remains controversial [
14]. Several studies have shown that obesity and weight loss in middle age are associated with an increased risk of dementia [
15,
16], while other studies have shown that obesity in old age does not increase the risk of AD [
16,
17]. Moreover, a large UK population study has shown that low BMI across all age groups increases the risk of AD [
18]. Although many previous studies have focused on the association between BMI and AD, these conflicting results suggest that the causal relationship between BMI and AD requires further exploration. Moreover, gene-environment interactions behind the development of AD also need deep investigation. However, little is known about the potential interaction between genetic risk and BMI on AD risk so far.
Understanding the causal relationship between BMI and AD is crucial for AD prevention. However, simple observational studies tend to result in reverse causality and residual confusion [
19]. Mendelian randomization (MR) based on genetic variations is useful to overcome some of these limitations [
20]. Numerous previous studies using MR analysis to assess the causal effect of BMI on AD found that polygenic scores strongly related to a higher BMI are unrelated to higher dementia risk and may even predict a lower dementia risk. This is surprising, however, there has been no further assessment of reverse causality [
21]. Fortunately, bidirectional MR overcomes this limitation [
22]. For the first time, this study used bidirectional MR to assess the causal relationship between BMI and AD. In addition, since we conducted observational studies and MR in the same study population, the conclusions could be more stable and reliable.
Therefore, we sought to use the UK Biobank (UKB) to investigate the relationship between BMI and risk of developing AD. To further assess the relationship between BMI and genetic susceptibility on AD risk, we also explored potential genetic and BMI interactions after calculating the AD genetic risk score (AD-GRS) of each participant. Finally, bidirectional MR was used to further explore the causal relationship between BMI and AD.
Statistical analyses
Comparison of the baseline characteristics between control and AD groups was performed using the Chi-square (or univariate logistic regression) or Wilcoxon rank sum test. The P values were tested and adjusted by Benjamini–Hochberg false discovery rate (FDR) method. Continuous variables are represented as mean ± standard deviation or median ± interquartile range (IQR).
A restricted cubic spline (RCS) was used to further study the potential nonlinear relationship between BMI and AD. Moreover, an age subgroup analysis was also performed. The model was adjusted for age, Townsend deprivation index (TDI), sex, smoking, ethnicity, education level, alcohol use, hypertension, stroke, myocardial infarction, and diabetes. In addition, we divided BMI into three groups (< 23 kg/m2, 23–30 kg/m2, > 30 kg/m2) according to the RCS results.
A Kaplan–Meier survival curve was used to show the risk of AD among the three BMI groups (< 23 kg/m2, 23–30 kg/m2, > 30 kg/m2). The BMI (23–30 kg/m2) group was used as the control group, and the differences between the three groups were evaluated using log-rank tests.
A Cox proportional risk model was used to test the association between BMI and AD. The multivariable model 1 was unadjusted, model 2 was adjusted for age and sex, and model 3 was adjusted for age, TDI, sex, smoking, ethnicity, education level, alcohol use, hypertension, stroke, myocardial infarction, and diabetes.
A Cox proportional risk model was used to estimate the association between BMI and AD-GRS for risk of AD. The multivariable model was adjusted for age, TDI, sex, education level, alcohol use, hypertension, stroke, myocardial infarction, and diabetes.
All statistical analyses were performed with the R package (version 4.1.0). A p value < 0.05 was considered statistically significant.
Discussion
In this large-scale study with an average follow-up time of more than 10 years, we have the following main findings: 1. our results indicated there was a nonlinear relationship between BMI and AD risk in participants aged 60 years or older, but not in participants aged 37–59 years; 2. for participants aged 60 years and older, our results indicated that participants with the BMI (< 23 kg/m2) were associated with a higher AD risk; 3. compared with the BMI (< 23 kg/m2), the higher BMI was associated with lower risk of AD in participants with the same intermediate or high AD-GRS; 4. there was a reverse causality between BMI and AD analyzed using bidirectional MR.
With global incidence of both obesity and dementia increasing year by year, understanding the causal relationship between BMI and AD risk has become a public health priority [
39]. To the best of our knowledge, our study is the first to use bidirectional MR to establish a causal relationship between BMI and AD risk. Our results showed that there is a reverse causality between BMI and AD risk analyzed using bidirectional MR, suggesting that reduced BMI could be one of the early manifestations of AD.
Possible pathogenesis of BMI declines in AD patients has been investigated in previous studies. Reduced hippocampal volume and thinning of the entorhinal and medial temporal cortices are common imaging findings in AD patients [
40]. Imaging data also indicate that brain structural changes, including changes of whole brain and hippocampal atrophy, are associated with alterations in body composition, including reductions in more specific measures of lean mass [
41]. The potential mechanisms underlying the pathophysiological relationship between BMI and AD risk include neuropathological changes occur in regions like hypothalamus that play critical roles in regulation of energy metabolism and food intake [
42]. Behavioral and cognitive changes associated with dementia can also affect weight by interfering with nutrition (forgetting to eat) or by reducing physical activity (a strong predictor of sarcopenia) [
43]. In addition, Apolipoprotein E (
APOE), produced primarily by astrocytes in the central nervous system, is a major cholesterol carrier that transports lipids to neurons to maintain synapses and promote damage repair, which is linked to increased accumulation of cortical amyloid-β (Aβ) [
44]. The E4 allele of
APOE gene (
APOE4) is the strongest genetic risk factor for late AD [
45]. The accumulation of Aβ in
APOE4 + individuals is regulated by leptin signaling in the hypothalamus [
46], and leptin signaling pathway itself could lead to the synthesis and release of anorexia neuropeptides that may contribute to weight loss [
47,
48].
Although our bidirectional MR results showed no positive causal relationship between BMI and AD risk due to the limitations of the MR method, a false-negative result cannot be ruled out. Our MR method mainly studied linear causality. However, our observational study found that the association between higher BMI and AD risk was non-linear. In addition, after the participants were grouped according to AD-GRS, a lower BMI was still associated with a higher risk of AD in the intermediate or high AD-GRS groups. Studies also showed that higher BMI-related genetic variants may slightly reduce the risk of AD [
18], however, their non-linear relationship has been studied. Therefore, we speculated that there might be a non-linear causal relationship between BMI and the risk of AD. However, more research is needed to clarify this.
The possible mechanism underlying that high BMI is associated with lower risk of AD in older individuals remains poorly understood. Blautzik et al. showed that even among
APOE4 carriers, BMI was negatively associated with cortical amyloid load, glucose metabolism in posterior cingulate gyrus, and recent cognitive decline [
47]. Adipose tissue releases molecules such as leptin and adiponectin, which bind to receptors in the hippocampus to regulate neuronal excitability, increase synaptogenesis, and prevent amyloid-induced neuronal cell death [
49,
50]. In addition, microglia are innate immune cells of the central nervous system, which can prevent development of AD by inhibiting accumulation of Aβ [
51]. Adiponectin can inhibit Aβ-induced inflammation and promote anti-inflammatory properties of microglia [
52,
53]. Adiponectin receptor agonists can suppress microglia and astrocyte activation and restore microglia Aβ phagocytosis in mouse models of AD [
54]. Since AD is an age-related disease, this might partially explain why high BMI in old age is associated with a lower risk of AD development [
55]. However, further research will be needed to understand this mechanistic basis.
There is no significant association between BMI and AD risk in observational studies of participants aged 37–59 years old using RCS. Because the preclinical phase of dementia can last for more than 10 years [
56], most of the participants may not have reached the diagnostic criteria for AD, and this population (37–59 years old) requires further follow-up in the future. Genetic and environmental factors have been considered contributors to the progression of AD [
3]. To our knowledge, this is the first study to investigate the interaction between BMI and GRS on development of AD. As expected, for participants aged 60 and older, we observed that participants with high genetic risk had a higher risk of AD. In addition, a lower BMI (< 23 kg/m
2) was associated with a higher risk of AD in the intermediate and high AD-GRS groups. Since genetic factor is an unmodifiable factor for the risk of AD, more attention should be paid to the management of BMI, especially in the populations with intermediate or high AD-GRS. It is considered that increased BMI (greater than 30 kg/m
2) may lead to cardiovascular and metabolic diseases [
57]. In addition, we also found that there was a U-shaped association between BMI and all-cause mortality, and higher BMI (BMI > 30 kg/m
2) and lower BMI (BMI < 23 kg/m
2) were associated with a higher risk of all-cause mortality (Additional file
1: Figure S7). These findings suggest that a higher BMI (BMI > 30 kg/m
2) was not associated with a higher risk of AD, possibly due to the complications that had led to death in participants before AD was diagnosed, further validation is needed in future studies. Therefore, BMI (23–30 kg/m
2) may be a potential intervention for AD without increasing complications and all-cause mortality.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.