The interaction effect of metals exposure and dietary habit on cognitive function in Chinese older adult cohort

Abstract

Background

As the important factors in cognitive function, dietary habits and metal exposures are interactive with each other. However, fewer studies have investigated the interaction effect of them on cognitive dysfunction in older adults.

Methods

2,445 registered citizens aged 60–85 years from 51 community health centers in Luohu District, Shenzhen, were recruited in this study based on the Chinese older adult cohort. All subjects underwent physical examination and Mini-cognitive assessment scale. A semi quantitative food frequency questionnaire was used to obtain their food intake frequency, and 21 metal concentrations in their urine were measured.

Results

Elastic-net regression model, a machine learning technique, identified six variables that were significantly associated with cognitive dysfunction in older adults. These variables included education level, gender, urinary concentration of arsenic (As) and cadmium (Cd), and the frequency of monthly intake of egg and bean products. After adjusting for multiple factors, As and Cd concentrations were positively associated with increased risk of mild cognitive impairment (MCI) in the older people, with OR values of 1.19 (95% CI: 1.05–1.42) and 1.32 (95% CI: 1.01–1.74), respectively. In addition, older adults with high frequency of egg intake (≥30 times/month) and bean products intake (≥8 times/month) had a reduced risk of MCI than those with low protein egg intake (<30 times/month) and low bean products intake (<8 times/month), respectively. Furthermore, additive interaction were observed between the As exposure and egg products intake, as well as bean products. Cd exposure also showed additive interactions with egg and bean products intake.

Conclusions

The consumption of eggs and bean products, as well as the levels of exposure to the heavy metals Cd and As, have been shown to have a substantial influence on cognitive impairment in the elderly population.

1. Introduction

Heavy metals are widespread in the natural environment and contaminated food, water, dust and air. Exposure to metals can exert influence over the development of a range of diseases, including Alzheimer's Disease (AD). Available research suggested that exposure to some metals may lead to cognitive decline in older people, such as lead (Pb), cadmium (Cd), arsenic (As) and manganese (Mn) [[1], [2], [3]]. Although some metal elements are essential trace elements, they may adversely affect the body if they are ingested in excess or transformed into other forms [4]. A cross-sectional study among U.S. adults aged 60 years or older found that L-shaped associations between total copper (Cu) or selenium (Se) intake and low cognitive performance from NHANES data [5]. On the other hand, a study of 317 subjects older than 60 in Malaysia showed that high levels of Pb and Cu in toenails were associated with DNA damage that eventually could be related to cognitive decline [6]. Another study found that higher dietary intake of iron (Fe) and Cu was related to worse cognitive function, but zinc (Zn) and Mn intake levels were not significantly associated with cognitive function [7]. A recent study found a link between reduced Zn levels in hairs and decreased cognitive function [8], despite previous studies showing no significant correlation between Zn and cognitive function [7,9]. Due to the large variety of metals, the precise role of these metals which have been linked to changes in cognitive function, has not yet been clear. Furthermore, the interference and confounding effects on metal exposure need to be considered.

The daily diet provides the body with various nutrients. Inadequate intake of some food can be detrimental to the health of the body and brain. Studies have shown that dietary factors significantly impact human cognitive function [[10], [11], [12]]. Mediterranean diet characterized by a high intake of whole grains, legumes, vegetables, and fruits, as well as moderate alcohol consumption and a high ratio of monounsaturated to saturated fats, was associated with improved cognitive function and slowed down cognitive decline [13]. The study found that adequate carbohydrate supply, appropriately increasing dietary protein intake and reducing fat intake might benefit the cognitive function of elders in rural areas of Qingdao, China [14]. A cross-sectional study included 1176 individuals aged 65–85 years from rural China, compared the correlation between three main dietary patterns and cognitive function. They found that the healthy dietary pattern, which was based on the consumption of rice and flour, red meat, chicken, vegetables, seafood, and fruits, protected against cognitive dysfunction [15]. Currently, most studies are limited to exploring the relevance of individual variables of the environment or single categories of risk factors to older adults with mild cognitive impairment in older adults [1,3,10]. However, it has been found that certain substances or elements in foods may affect metal metabolism in the body [16]. Therefore, it is necessary to assess the effects of environmental metal exposure, individual dietary habits, and their interactions on the risk of developing mild cognitive dysfunction in older adults at an integrated, holistic level.

In the present study, therefore, we employed machine learning, an artificial intelligence technique, which relies on a statistical method known as Elastic-net regression [17], to analyze the associations between various metals exposure and dietary habits and their interactions with the cognitive functions of the older adults. In particular, we aimed to address the following issues: 1) the associations among metals, food intake frequency and cognitive function used by Elastic-net regression; 2) the interaction effect of metals and food intake frequency on cognitive function, as well as establishing a prediction model on the cognitive function.

2. Methods

2.1. Study design and population

This study is a cross-sectional study utilizing baseline survey data obtained from Shenzhen Age-related Disorder (SAD) cohort [18]. All participants were 60–85 years old with Shenzhen household registration who recruited in Luohu District from January 2017 to November 2018. Eligibility criteria of study population included over 60 years of age, performance of a clinical examination, no severe physical or mental diseases, and a permanent community-dwelling resident from one of the 51 community clinics in Shenzhen. Exclusion criteria for the study: no urine metal test results, no cognitive function assessment, and diagnosed with neurodegenerative diseases like dementia. This study presents a baseline data analysis of a cohort study within the framework of a cross-sectional study. All data in this study were obtained between 2017 and 2018 years. Assessment of cognitive function in all study subjects was performed by specialized trained neurologists using Mini-cognitive assessment scale (Mini-Cog). Mini-Cog scale is more acceptable to patients and researchers due to its more time-saving and labor-saving features (shorter time and less language-affected). It is as effective as more extended screening and assessment tools in detecting dementia [19].The Mini-Cog scale has a total score of 5. In this study, those with a Mini-Cog score of 5 were judged to be at low risk for mild cognitive dysfunction, and those with a Mini-Cog score less than 3 were judged to be at high risk for mild cognitive impairment (MCI) [20]. For MCI diagnosis, Mini-Cog has excellent sensitivity [21]. Previous study which involved evaluation of individuals in primary care settings without a documented history of dementia recorded at baseline, reported that the sensitivity of the Mini-Cog was 0.76 and the specificity was 0.73 [22]. A subsequent cohort study, conducted over a 36-month period, showed that the Mini-cog scale exhibited a specificity of 0.85 (0.81−0.89) and a sensitivity of 1.0 (0.84–1.0) in screening for early dementia [23]. According to the cut-off criteria of the Mini-Cog scale, 489 cases of older adults with a high risk of MCI were diagnosed. Older adults with a low risk of cognitive function who participated in the physical examination during the same period were selected as the control group in a 4:1 match according to their age ± 1 year. Finally, we got 2,445 subjects recruited in this study.

Trained interviewers provided each subject with a semi-structured questionnaire through a face-to-face interview, to collect baseline data that including demographic information of the subjects, socioeconomic status, history of diseases, history of falls, smoking status, alcohol use, medication use and physical activity. Marital status was collected by the trained interviewers. Marital status is categorized as either married or other, which included individuals who were unmarried, divorced, remarried, or widowed. All subjects gave their informed consent for inclusion before they participated in the study. The study was conducted in accordance with the Declaration of Helsinki, and the protocol was approved by the Ethics Committee of Shenzhen Center for Disease Control and Prevention (R2007001).

2.2. Food intake frequency

Data on the food behaviors of the respondents were collected by the semi-quantitative food frequency questionnaire (SFQ). A variety of food items were included in the scale, including rice, wheat, coarse grains, fruits, vegetables, eggs, and egg products, milk and milk products, meats (pig, cow, poultry), bean products (all beans, legumes, lentils, chickpeas, split peas, and tofu), river fishes, sea fishes, river shrimps/crab, sea shrimps/crab, river shellfish, sea shellfish, cephalopods (squid, octopus, etc.), cured products and smoked meat. Frequency of all kinds of food intake was calculated in months. Because the frequency of elderly people using the internet shopping is very limited, all the foods basically come from Shenzhen.

2.3. Physical examination

Subjects also underwent a physical examination the same day after taking the community health questionnaire, as well as collecting fasting venous blood and clean mid-stream urine samples on the morning of the physical examination. The physical examination was also conducted at the same time to collect baseline data on standing height and body weight, which were measured in individuals with light indoor clothing and without shoes. Body mass index (BMI) was calculated as weight (kilograms)/standing height (square meters). Given that the study population consists entirely of elderly residents with established health management records within the community, the identification of chronic conditions such as hypertension, diabetes, and hyperlipidemia necessitates consideration not only of physical examination findings but also of their extensive medication history and medical background. For this study, hypertensive was defined as meeting any one of these three criteria: (1) had a past history of hypertension; (2) a history of antihypertensive medication use; (3) had a systolic blood pressure ≥140 mm Hg and a diastolic blood pressure ≥90 mm Hg [24]. According to the Clinical Guidelines for the Prevention and Treatment of Type 2 diabetes in the Elderly in China (2022) [25], diabetes mellitus was defined as a past history of diabetes mellitus or a history of antihyperglycemic medication or fasting blood glucose (FBG) ≥7.0 mmol/L. Based on Chinese adult dyslipidemia prevention and treatment guidelines [26], hyperlipidemic patients were defined as those who reported a diagnosis of hyperlipidemia or a history of taking lipid-lowering drugs or those who met any of the following criteria: (1) total cholesterol (TCHO) ≥5.18 mmol/L, (2) Triglyceride (TG) ≥1.7 mmol/L, (3) Low Density Lipoprotein (LDL) ≥3.37 mmol/L; (4) high density lipoprotein (HDL) ≤1.0 mmol/L.

2.4. Measurement of urinary metal concentrations

21 metals were analyzed in urine samples, including arsenic (As), selenium (Se), zinc (Zn), iron (Fe), lithium (Li), aluminum (Al), titanium (Ti), vanadium (V), chromium (Cr), manganese (Mn), cobalt (Co), nickel (Ni), copper (Cu), rubidium (Rb), strontium (Sr), tin (Sn), barium (Ba), molybdenum (Mo), cadmium (Cd), thallium (Tl) and lead (Pb), by using inductively coupled plasma-mass spectrometry (ICP-MS, NexION 300X, PerkinElmer Inc., Waltham, Massachusetts, USA) [27]. In brief, 0.5 mL urine samples were diluted with 65% (mass fraction) nitric acid to a total volume of 5 mL and mixed thoroughly before analysis.

2.5. Statistical analyses

Descriptive analysis was used to describe demographic characteristics. Comparison of basic characteristics between control and MCI case group were made by using chi-square tests for categorical variables and general Student t-test for continuous variables. The Spearman’s rank correlation test was conducted to investigate correlations among metal concentrations, as well as dietary food intake frequency, due to non normal distributions. All metals concentration were log transformed with 10 as the substrate for the logistic regression and Elastic-net regression analysis. Backward multiple logistical regression model was used to analyzed the association among urine metals, dietary habits and cognitive function. Mann–Whitney U test was used to compare the concentrations of urinary metals and dietary habits between low risk group of MCI and high risk group of MCI. Additionally, the comparison of urinary As and Cd concentration with different dietary intake was examined through Wilcoxon rank sum tests.

Elastic-net regression, also named Resilient network regression, is a model that is good at handling all correlated data. It combines the properties of least absolute shrinkage and selection operator (LASSO) regression model in filtering out essential variables with the advantages of ridge regression in preventing multi-collinearity. In addition, Elastic-net regression can perform well when dealing with correlated parameters [17]. When building the Elastic-net regression model, the total sample (n = 2445) was split into a training set (70% sample, n = 1771) and a test set (30% sample, n = 674). Study subjects of both the training set and test set with a 1:4 ratio of high-risk of MCI group to low-risk of MCI group. The training set was used to build the model, while the test set was used for external validation. The training set was subdivided into 10 equally sized folds (n = 177 each) for tenfold cross validation. One fold was used for internal validation and the others were included as internal training data (n = 1594). After testing the model with the internal training data and calculating the model parameters, the model was validated with the internal validation data. The process above was performed a total of 10 times to obtain the possible pairing of the model’s optimal parameter values. Then, the resulting model was tested with test set data. The most balanced model was selected as the final model to reduce random errors after performing 10 rounds of bootstrapping. In addition, a multivariate logistic regression (MLR) model was established with the same variables to compare with the final model. The predictive performance of the Elastic-net regression model based on MCI risk was evaluated using receiver operating characteristic curve (ROC) analysis. The size of area under the ROC (area under curve, AUC) ranges from 0.5 to 1, which measures the model’s predictive power, with closer to 1 indicating a better effect.

All statistical analyses were performed using the R software (4.1.0) glmnet package (4.1−2). The curves were presented using the pROC package and the ggplot package. The R software's rms package and visreg package (version 4.1.0) were used to construct three times restricted cubic spline and analysis of interaction, respectively.

3. Results

3.1. Characteristics of study population

The demographic characteristics were presented in Table 1. A total of 2445 cases of older adults undergoing physical examination were included in this study, 489 cases in the high-risk group versus 1956 cases in the control group. The mean age for the low-risk group of MCI and high-risk group of MCI were 67.2 ± 5.3 years and 68.0 ± 5.9 years, respectively. There was a statistically significant difference in the distribution of gender, marital status, smoking status, drinking status and height in the high-risk group of MCI compared to the MCI low-risk group (all P < 0.05).

Table 1.

Demographics characteristics of study participants.

	Low Risk of MCI (n = 1956)	High Risk of MCI (n = 489)	p-value
Gender (male/female, n, %)	876/1080 (44.8/55.2)	315/174 (64.4/35.6)	<0.01
Age (years, mean ± SD)	67.3 ± 5.3	68.0 ± 5.9	0.16
Education (n, %)			<0.01
Primary school and below	813 (41.6)	240 (49.1)
Junior high school	642 (32.8)	146 (29.9)
Senior high school	501 (25.6)	103 (21.1)
Marriage (Married/Other, n, %)	1777/179 (90.8/9.2)	418/71(85.5/14.5)	<0.01
Smoking state (n, %)			<0.01
Never smoking	1487 (76.0)	409 (83.6)
Have quit smoking	253 (12.9)	37 (7.6)
Still smoking	216 (11)	43 (8.8)
Passive smoking (yes/no, n, %)	178/1778 (9.1/90.9)	69/420 (14.1/85.9)	<0.01
Alcohol state (n, %)
Never drinking	1586 (81.1)	429 (87.7)	<0.01
Have quit drinking	68 (3.5)	14 (2.9)
Still drinking	302 (15.4)	46 (9.4)
Hypertension (yes/no, n, %)	1030/926 (52.7/47.3)	276/213 (56.4/43.6)	0.14
Hyperlipidemia (yes/no, n, %)	1424/532 (72.8/27.2)	373/116 (76.3/23.7)	0.12
Diabetes (yes/no, n, %)	380/1576 (19.4/80.6)	105/384 (30.7/69.3)	0.31
Height (cm, mean ± SD)	161.0 ± 7.8	157.0 ± 8.0	<0.01
BMI (kg/m2, mean ± SD)	24.3 ± 3.1	24.3 ± 3.1	0.62

SD, standard deviation; BMI, body mass index; MCI, mild cognitive impairment; Continuous variables are presented as mean ± SD; categorical variables are presented as proportion [n (%)]; p-value were derived from the t-test and χ² test.

3.2. Distribution of urinary metals and food intake frequency

The distribution of metal concentrations in urine, along with risk of MCI, were presented in supplement files (Table S1). There were statistically significant differences in the urinary concentrations of Al, As, Sr, Cd, Sn, and Ba in older adults in the high-risk group for MCI compared to those in the low-risk group for MCI. However, no statistically significant differences in the concentrations of other metals were found between the two groups of older adults. Spearman’s rank correlation analyses revealed that most of the metals in urine were positively correlated with each other (p <  0.05).

The frequency of food intake of the subjects in the last year were shown in supplement files (Table S2). There were statistically significant differences in the frequency of monthly intake of wheat, eggs, milk, legumes, fish and, shrimp in the older person between the high-risk of MCI group and low-risk of MCI group (all p < 0.05). Spearman’s correlation analysis of the monthly intake frequency of 18 common human foods showed that a variety of food intakes were correlated with each other (Figure S2).

3.3. The association among urine metals, dietary habits and cognitive function

Variables obtained for demographic characteristics, urine metals concentration values, and dietary habits information were corrected and incorporated into the Elastic-net regression model. Through cross-validation, we derived the optimal parameter pairing of the model as α = 0.7 and lambda.min = 0.007569147. To reduce the random errors, we obtained ten random grouped model results by the bootstrap method, and the model with the most balanced results was taken as the final model. Meanwhile, we used the Akaike information criterion (AIC) as a comparison to build a backward multiple regression model (Table S3). The results and Receiver operating characteristic curve (ROC) of the two models were shown in Fig. 1 and Tables S4–S5, which showed that the Elastic-network regression model had better AUC and stability. The final model had an AUC of 0.754 on the training set and 0.716 on the test set.

Fig. 1.

Comparison of ROC curves of ENR and MLR for screening the variables associated with MCI. The final model of ENR had an AUC of 0.716 on the test set. The final model of MLR had an AUC of 0.672 on the test set. ENR, Elastic-net regression model; MLR, multiple logistical regression model; MCI, mild cognitive impairment; ROC, receiver operating characteristic curve; AUC, the area under the ROC.

The Elastic-net regression model screened out a total of seven significant variables, including education level, gender, height, urinary As and Cd concentration, as well as frequency of egg product and bean products food intake. The forest plot results are shown in Fig. 2. The Elastic-net regression model suggested that the higher education level and female in the older population were protective factors for cognitive function (odd ratio [OR]: 0.83, 95% confidence interval [CI]: 0.74−0.92 and OR: 0.75, 95%CI: 0.63−0.87). Furthermore, urinary concentrations of As and Cd were significantly associated with the high risk of MCI in the study population, and the OR for each additional quartile of these two metals were 1.19 (95%CI: 1.05–1.42) and 1.32 (95%CI: 1.01–1.74), respectively. Interestingly, older people with high frequency of egg intake (≥30 times/month) and bean products intake (≥8 times/month) had a reduced risk of cognitive dysfunction than those with low protein intake (<30 times/month) and low bean product intake (<8 times/month), respectively.

Fig. 2.

Forest plot results for Elastic-net regression model screening variables associated with MCI. Forest plot of MCI associated features identified by screening with ENR. The black squares symbolize odds ratio values, while the line segments denote 95% confidence intervals. MCI, mild cognitive impairment; ENR, elastic-net regression model; OR, odd ratio; 95% CI, 95% confidence interval; Cd, cadmium; As, arsenic. The urinary concentrations of Cd and As values were transformed by 10-logarithm.

We further explored the dose-response of As and Cd concentrations in urine on cognitive function using restricted cubic spline (RCS) bars, as shown in Figure S1. The RCS spline suggested that in the study population, urinary As and Cd concentration were all positively correlated with the risk of cognitive dysfunction (all p for association <0.05), without nonlinear relationship (p for nonlinear > 0.05).

3.4. The interaction effect of urinary metal concentration and diet frequency on cognitive function

Based on the Elastic-net regression model screened out the urinary metals concentration variables and the frequency of dietary intake variables, we further investigated the additive and multiplicative effect between them (Table 2). The additive interactions of urinary As concentrations and the frequency of intake of bean and egg products on the cognitive function in the older adults. Besides, the exposure of Cd was interactive with frequency of intake of bean and egg products, influencing the risk of MCI. However, no multiplicative interactions between the urinary metals (As and Cd) and the frequency of food intake (bean and egg products) were observed on the MCI in older adults (all p > 0.05). To clarify the interaction between metal exposure (As and Cd) and dietary food intake frequency (egg and bean products), we compared the concentrations of urinary As and Cd in populations with different intake of egg and bean products (supplement Table S6). Surprisingly, in the subjects with higher frequency of bean products intake (≥8 times/month), the concentrations of As and Cd in urine were lower than those in the population with lower frequency of intake (<8 times/month). Similarly, among the population with higher intake frequency of egg products (≥30 times/month), the concentration of Cd in urine is also lower than that in the older adults with lower frequency of egg products (<30 times/month).

Table 2.

Interaction effect between urinary metal concentration and dietary frequency on the risk of cognitive impairment.

Characteristics	OR (95% CI)	p-value	RERI	AP	S
As + Egg		0.447	1.02(0.99,1.069)	0.69(0.93,0.52)	2.70 (1.15,3.07)
Low As+High Egg	1
Low As + Low Egg	0.78(0.62,1.01)
High As+High Egg	1.48(1.07,2.05)
High As + Low Egg	1.46 (1.08,1.98)
As + Bean		0.61	0.63 (0.43,0.87)	0.65 (0.33,0.99)	2.37 (2.21,2.53)
Low As+High Bean	1
Low As + Low Bean	0.90(0.68,1.19)
High As+High Bean	1.30 (0.95,1.77)
High As + Low Bean	1.61 (1.21,2.14)
Cd + Egg		0.47
Low Cd+High Egg	1		0.83 (0.61,1.23)	0.79 (0.48,1.35)	2.67 (2.43,2.71)
Low Cd + Low Egg	0.79(0.61,1.03)
High Cd+High Egg	1.17 (0.83,1.65)
High Cd + Low Egg	1.77 (1.32,2.36)
Cd + Bean		0.81	1.14 (1.01,1.22)	1.14 (1.02,1.26)	2.33 (2.21,2.54)
Low Cd+High Bean	1
Low Cd + Low Bean	0.77(0.58,1.02)
High Cd+High Bean	1.47 (0.84,1.57)
High Cd + Low Bean	1.56 (1.19,2.07)

OR, Odd ratio; 95% CI, 95% confidence interval; RERI, Relative excess risk due to interaction; AP, Attributable proportion due to interaction; S, Synergy index; Cd, cadmium; As, arsenic. Models were adjusted for age, education, gender, height, marriage, smoking state, passive smoking and drinking state.

4. Discussion

This study provided evidence that education level, gender, height, urinary As and Cd concentration, and frequency of egg and bean products intake were associated with cognitive function in older adults using the Elastic-net regression model. As and Cd levels in urine were positively associated with cognitive function. Meanwhile, the frequency of monthly intake of egg- and bean-based foods was associated with the decreased risk of MCI. In addition, we identified significant dose-dependent relationships between As and Cd levels and risk of MCI. Particularly, we demonstrated that the intake of egg and bean products interacted additive with the concentrations of As and Cd in urine, influencing the risk of MCI in older adults.

Several studies have indicated that high education level is the protective factor for the risk of mild cognitive dysfunction [[28], [29], [30]], which are consistent with our results. In addition, gender and height also associated with cognitive function selected from Elastic-net regression model in this study. We found that older men had a higher risk of developing cognitive impairment than older women, which was in line with previous research that found men have a higher odds of reporting Subjective Cognitive Decline (SCD) compared to women [31]. The present study also found that height was negatively associated with the risk of developing MCI. Previous studies have confirmed a significant positive correlation between height and cognitive ability across countries [32,33]. Height positively correlates with total cortical surface area in male twins, which underlies the relationship between phenotypic height and general cognitive ability [34].

Based on Elastic-net regression model, we found that the concentration of As and Cd in the urine was associated with the risk of MCI in older adults, which is in line with most of the current studies [35,36]. As a globally recognized heavy metal pollutant, Cd is widely distributed in the biosphere and exposed to humans through various pathways. Epidemiological studies have shown that higher levels of Cd in the urine of older adults are significantly associated with cognitive impairment [36]. Cd could increase the permeability of the blood-brain barrier to accumulate in the brain [37], where it affects attention and cognitive function through activating multiple signaling cascades in neurons, stimulating inflammatory responses and oxidative stress, and leading to neuronal decline and death. Animal studies have shown that chronic Cd administration in mice leads to memory dysfunction, oxidative stress, and accumulation of ROS (Reactive oxygen species) [38]. It has also been suggested that chronic exposure to Cd may be a pathogenic factor inducing AD development. The neurotoxicity of Cd is closely related to hyperphosphorylation of Aβ (Amyloid β-protein) and Tau proteins [39]. In addition, Cd exposure triggers gliosis and activation of other inflammatory receptors and mediators that induce neuroinflammation, leading to neurodegeneration [40].

We also identified As exposure was related with increased risk of MCI in the older population. Multiple studies have shown that exposure to As may affect intelligence and cognitive function in older adults [41], leading to increase susceptibility to neurodegenerative diseases [42,43]. Available information suggests that the mechanism of cognitive deficits in children and adults exposed to inorganic As appears to involve impaired neurotransmitter bio-genesis and glutamate metabolism, resulting in altered synaptic transmission [44,45]. On the other hand, inorganic As induced tau phosphorylation and Aβ over-expression in brain tissue, favoring the formation of intracellular neurofibrillary tangles. The accumulation of extracellular Ab, which may lead to impaired memory and learning processes, also contributed to neurodegenerative pathology [42]. Although the current study demonstrated the neurotoxic effects of As, the mechanisms involved are unclear.

Furthermore, the frequency of monthly intake of egg- and bean-based foods was negatively associated with a high risk of MCI in older adults. This suggested that the intake of egg and bean products may be a protective factor against MCI in older adults, which is consistent with the results of most current studies [46,47]. A 4-year prospective cohort study showed that older adults in the moderate egg intake group (24−63 g/week, half to 1.5 eggs) showed significantly slower rates of memory decline compared with those in the low egg intake group (<23 g/week, about half an egg). Those in the high egg intake group (≥63 g/week, about two or more eggs) also showed slower rates of memory decline [47]. The benefits of egg intake on cognitive function may be related to the high levels of choline, lutein, lecithin and tryptophan which have been associated with improved cognitive function [[48], [49], [50]]. In addition, study conducted on animals [51] have demonstrated that the consumption of egg whites can mitigate the adverse impacts of prolonged heavy metal exposure and diminish the neurotoxic effects induced by these metals. There is limited research involving the relationship between legume intake and cognitive function in older adults. A meta-analysis study has reported that lecithin, rich in both beans and eggs, is beneficial to the brain and memory as a precursor of the neurotransmitter acetylcholine [52]. Moreover, bean products reduce the risk of smoking-induced cardiovascular disease, decreasing the risk of affecting cognitive functions [53,54]. Our findings suggested that the intake of egg and bean products interacted additive with specific urinary concentrations of As and Cd, influencing the risk of MCI in older adults. We further compared the As and Cd concentrations between the subjects with different frequency of egg and bean products intake (Supplement Table S6). Urinary As and Cd concentrations were found to be lower in those with a high frequency of bean or egg products ingestion, respectively. The results suggested that egg and bean products supplied to Shenzhen were not the main sources of dietary exposure to As and Cd for the elderly. The study in Iran evaluated the health risk assessment for human exposure to eleven metals and As via consumption of egg from largest poultry Industry. They found that intake of heavy metal exposure through the consumption of commercial egg product is completely safe for both national and international consumers [55]. In addition, some nutrients in eggs and bean products could promote cognitive functions that counteract the negative effects of metal exposure. As mentioned before, choline, lutein, lecithin and tryptophan, the benefits in eggs, have been associated with improved cognitive function [[48], [49], [50]]. Evidence from clinical and experimental studies supporting a role for soy isoflavones in cognitive function [56,57]. More studies and reports on the interaction between environmental metals exposure and individual dietary habits on MCI in older adults need to be done.

The strength of our study is attributed to our comprehensive analysis of the association between metals, dietary habit and cognitive function in an Elastic-net regression model. In addition, we conducted further analysis to assess the potential interactive effect between metals and dietary habit on cognitive function. However, there are also some limitations associated with our study. First, this study featured a cross-sectional design. Causality in regression could therefore not be proven. Besides, we collected only a single urine sample from each participant in order to measure the levels of metals. Finally, due to the distinctive characteristics of Shenzhen, we should be prudent when extrapolating our findings to other populations.

5. Conclusions

Based on the baseline population of an older adults cohort, this study screened variables significantly associated with MCI in older adults by constructing an Elastic-net regression model. This study provided essential evidence that As and Cd in urine were positively associated with increased risk of MCI. Furthermore, intake of bean and egg products was negatively associated with the risk of MCI in older adults. Also the interaction between bean products intake and urinary As and Cd concentrations may be positively associated with the risk of MCI. These findings may offer a new strategy for explore the mechanism of As and Cd on cognitive function. One approach is to investigate active compounds in food intake to discern the effect on the elimination, absorption, and distribution of As and Cd. In addition, we should aggregate the dietary habits of the population in order to assess potential exposure risks associated with As and Cd. It is crucial to both public and medical professionals to be aware of the interactive effect of egg and bean products consumption with As and Cd exposure in cognitive function. It is also imperative that these findings require confirmation in future prospective studies in a more general population.

Funding

This work was supported by Shenzhen Basic Research Key Project (JCYJ20200109143431341),Shenzhen Key Medical Discipline Construction Fund (SZXK069), andSanming Project of Medicine in Shenzhen (SZSM202211010).

Informed consent statement

Informed consent was obtained from all subjects involved in the study.

Declaration of interests

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Appendix A. Supplementary data

The following is Supplementary data to this article: