Ambient particulate matters are associated with attenuation or reversal of positive association between Folate and cognitive function in Chinese older adults

Abstract

Background

Research on combined effects between ambient particulate matters (PMs) and Folate (FA) with respect to cognitive function remains limited. This study aimed to investigate whether long-term PMs exposure could modulate the association between FA and cognitive function in Chinese older adults.

Methods

The study was composed of 3,880 individuals aged 60 and above. The cognition of the subjects was assessed using the Montreal Cognitive Assessment - Basic Version (MoCA-B). Folate concentration was measured through laboratory serum analysis, and data on ambient particulate matter levels were acquired from the China High Air Pollutant (CHAP) monitoring network. A linear regression analysis was conducted to explore how FA and PMs interact to affect cognitive function, with stratified analysis based on thresholds identified using restricted cubic splines (RCS).

Results

Based on the results of the linear trend analysis, PMs exposure exhibited a negative with MoCA-B scores (all P_trend < 0.001); RCS analysis further revealed that the exposure-response relationships for PM_2.5, PM₁₀ and MoCA-B scores exhibited significant non-linear characteristics (P < 0.05), identifying their effect inflection points. FA was positively associated with MoCA-B scores (β = 0.107, 95% CI: -0.073, 0.287), and this association was effect by PM_2.5 and PM₁₀ exposure levels: as their concentration increased, the positive association was attenuated; When the exposure concentration exceeded 64.54 µg/m³and 103.9 µg/m³, the association even replaced by a negative association.

Conclusion

Our findings indicated that PM exposure were associated with an attenuated positive association between FA and cognitive function, and even a negative association.Therefore, further action must be taken to reduce the harm caused by air pollution to enhance the cognitive benefits of FA.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s12889-026-27926-9.

Introduction

China is undergoing a pronounced demographic shift towards an aging society. Projections indicate that the population aged 65 or above is projected to reach 400 million around the year 2050 [1]. In this profound demographic transition, age-related cognitive decline has become increasingly prominent. In China, more than 50 million elderly people are facing varying degrees of cognitive impairment, among whom about 15 million have developed dementia, and 38 million are presenting with mild cognitive impairment (MCI) [2], approximately 10% to 15% of people diagnosed with MCI each year progress to Alzheimer’s disease(AD) [3]. Since neurodegenerative lesions are mostly irreversible, the cognitive impairment they cause is also difficult to reverse [4]. However, MCI, serving as a bridge between normal cognitive aging and dementia, is generally regarded as a crucial window for intervention [5]. To protect the cognitive function of the elderly, it is necessary to identify modifiable factors that impact cognitive function.

A large body of epidemiological and experimental studies has confirmed that adequate FA levels are associated with better overall and domain cognitive performance, with its neuroprotective mechanisms involving support for neurotransmitter synthesis, maintenance of DNA integrity, and reduction of oxidative stress, among other aspects [6, 7]. FA is broadly recognized for its neuroprotective properties, primarily attributed to its essential involvement in one-carbon metabolism and its regulatory influence on homocysteine levels [8–10].

Air pollution is a key risk factor for dementia [11], and exposure to PM₁, PM_2.5, and PM₁₀ can impair cognitive function [12, 13]. It may elevate vulnerability to neurodegenerative disorders such as Parkinson’s disease and AD, and other conditions by triggering neuroinflammation and oxidative stress [14–16]. Furthermore, there is a positive correlation between PM_2.5 exposure and homocysteine, which itself has been confirmed as a strong risk factor for dementia, further suggesting that air pollution may affect cognitive health through this pathway [17].

However, individual cognitive functions are not determined by a single nutrient but occur within a complex set of exposures. Therefore, the effects of nutrients may be modulated by environmental exposures [18, 19]. Given that PMs and FA are both closely related to cognitive function, they may interact through shared pathophysiological pathways. Therefore, the study is designed to investigate the relationship between PMs, FA and their combined effect on cognitive function: (1) explore the independent effects of PM exposure and FA status on cognitive function in the elderly; (2) whether PM exposure modifies the strength and direction of the association between FA and cognitive function; and (3) identify whether a discernible threshold exists in the interaction between PM exposure and FA regarding cognitive function.

Methods

Study population

The data for this research were sourced from a baseline survey of the Elderly Health and Modifiable Factors Study, conducted between July and September 2018 in Fu yang, Anhui Province, China. For further specifics on the subject enrollment and methodology, readers were referred to our prior publications [20]. Briefly, the study enrolled individuals over 60 years of age with over five years of residency in the region. The study initially enrolled 5,186 elderly participants. Following consent, each individual completed a standardized face-to-face interview and a physical assessment conducted at a community healthcare center. After excluding individuals who had mental disorders (n = 197), missing FA data (n = 946), missing MoCA-B scores data (n = 163), a final sample of 3880 elderly individuals was included in the analysis (FigureS1). The study was approved by the Biomedical Ethics Committee of Anhui Medical University (Approval No. 20190288, No. 83230436), and written informed consent was signed by all participants prior to study enrollment.

Exposure assessment

Estimates for particulate matter (PM₁, PM_2.5, PM₁₀) were derived from the China High Air Pollutants (CHAP) database, providing 10 km resolution data. To estimate spatiotemporal time series, the Spatiotemporal Extra-Trees (STET) method was used. This modeling was enhanced by integrating various big data sources. The STET model showed cross-validated R² (CV-R²) values of 0.77, 0.86, and 0.90 for PM₁, PM_2.5, and PM₁₀, respectively, demonstrating high accuracy [21–23]. Long-term PMs exposure for every individual was derived from their residential address. The addresses, obtained from surveys and subsequently geocoded, were matched with the pollution database. The exposure metric was defined as the five-year average PMs concentration leading up to the study’s baseline assessment.

Folate measurement

Serum Folate levels were measured using a fully automated chemiluminescence immunoassay analyzer (CL-8000i, Shenzhen Mindray Bio-Medical Electronics) according to the manufacturer’s specifications. For each batch of testing, a standard calibration curve (R² ≥ 0.990) was established, and two levels of quality control materials were used for internal quality control (intra-day CV ≤ 5%, inter-day CV ≤ 10%).

Cognitive function assessment

Participant cognition was assessed using the Chinese version of the Montreal Cognitive Assessment-Basic (MoCA-B), a validated screening tool specifically developed for populations with lower educational levels [24, 25]. The MoCA-B evaluates nine cognitive domains: executive function (3 points), verbal fluency (2 points), orientation (6 points), calculation (3 points), abstraction (3 points), delayed recall (5 points), visual perception (4 points), naming (2 points), and attention (2 points), summing to a maximum of 30 points, with elevated scores representing enhanced cognitive status. For participants with lower educational attainment, no additional education correction was required as the test was inherently designed to accommodate educational variability. In our sample, the tool exhibited satisfactory internal consistency, reflected by a Cronbach’s α coefficient of 0.79.

Covariates

Based on previous literature [26, 27], these encompass sociodemographic profiles, for instance, age, gender, region, education level, occupation, economic status, marital status. Additionally, health-related behaviors such as smoking, alcohol intake, physical activity, and diet were considered, along with the presence of chronic disease including diabetes (yes or no) and hypertension (yes or no). Dietary was assessed using a 15-item Food Frequency Questionnaire (FFQ). After standardizing the intake frequency of various foods, five dietary patterns were identified through factor analysis, each characterized by specific food groups (for example, Pattern 1 was mainly animal-based foods, while Pattern 5 was rich in vegetables and fruits).

Statistical analysis

The distributions of continuous and categorical variables are represented by the mean and frequency, respectively. The differences in MoCA-B scores across these characteristics were assessed using independent t-tests or one-way analysis of variance (ANOVA). Bivariate correlations between PMs, FA, and MoCA-B scores were evaluated via Pearson correlation analysis. In subsequent regression models, PMs were modeled as continuous variables with normalization via mean subtraction and SD scaling. Transform FA using the natural logarithm (ln) to meet the model assumption of residual normality and then standardize.

Conduct a linear regression model to quantify the individual correlation between PMs exposure, FA, and MoCA-B scores. Model 1 adjusted only for age and gender. In Model 2, additional adjustments were made for region, education level, occupation, economic status, marital status, physical activity, smoking, alcohol intake, dietary patterns. Model 3 was further adjusted for the effects of hypertension and diabetes based on the previous models. The medians of PMs and FA concentrations in tertiles were also entered as continuous variables into the linear regression model for trend testing. In addition, to investigate the nonlinear association between PMs, FA, and MoCA-B scores, this study employed RCS models were used to visualize the dose-response relationships between PMs, FA, and MoCA-B scores, and to identify potential effect inflection points. Restricted cubic spline (RCS) regression could flexibly fit the nonlinear association between variables without presetting a fixed functional form, and it could objectively identify the curve change characteristics among PMs, FA and MoCA-B scores.

Additionally, an interaction term between PMs and FA was added in Model 3 (FA×PM₁, FA×PM_2.5, FA×PM₁₀). Based on a hierarchical model, the association between PMs, FA, and MoCA-B scores in the elderly was analyzed. Three groups of participants were defined according to their PMs exposure concentrations (low, medium, high), and the association between FA and MoCA-B scores within each group was analyzed using the following tertiles as references. Based on the inflection concentrations of PMs identified by the RCS model, participants were stratified into low-exposure (≤ inflection concentration) and high-exposure (> inflection concentration) groups. Using the covariate settings from the fully adjusted model, the associations between FA and MoCA-B scores within different exposure levels were analyzed. Using an interaction plot to further illustrate how the effect of FA on MoCA-B scores varies with changes in PMs concentration.

Subgroup analysis was conducted stratified by gender (male and female) and region (urban and rural) to evaluate potential heterogeneity in the individual and combined associations between PMs, FA, and MoCA-B scores.

Sensitivity analyses were performed to verify result robustness. Employing alternative modeling approaches and adjustments to variables. (1) The analysis further accounted for body mass index (BMI) and secondhand smoke exposure (categorized as no or yes) into the models as additional confounding factors; (2) Constructed dual-pollutant models, respectively adjusting for the potential confounding effects of nitrogen dioxide (NO₂), sulfur dioxide (SO₂), ozone(O₃); (3) Evaluated the joint-effect by creating a variable combination layer of PMs and FA.

All statistical analyses were performed with IBM SPSS (version 25; Armonk, NY, USA) and R software (version 4.4.3), adopting a two-sided p-value of < 0.05 as the criterion for statistical significance.

Results

Characteristics of participants

Among the 3,880 participants (50.67% of whom were female), the average age was 70.96 ± 5.44 years, 83.81% residing in rural areas, 54.56% being illiterate (Table 1). Included and excluded subjects were compared for baseline characteristics (Table S1).

Table 1.

Characteristics of the study population (n = 3880)

Characteristics		N (%)	MoCA-B (mean ± sd)	t/F	P value
Age groups (years)	60–70	2024(52.63)	21.37 ± 5.77	36.49	< 0.001
	70–80	1611(41.52)	20.60 ± 6.10
	80+	227(5.85)	17.92 ± 6.78
gender	Male	1914(49.33)	22.38 ± 5.53	18.59	< 0.001
	Female	1966(50.67)	19.36 ± 6.11
Region	Rural	3252(83.81)	20.43 ± 6.00	4.13	0.042
	Urban	628(16.19)	23.02 ± 5.68
Education level	Illiteracy	2117(54.56)	18.86 ± 5.87	268.92	< 0.001
	Primary	615(15.85)	21.11 ± 5.58
	Junior	532(13.71)	23.14 ± 5.11
	Senior or above	616(15.88)	25.46 ± 4.16
Occupation	Technical labor-related occupations	148(3.81)	24.47 ± 5.18	20.42	< 0.001
	Manual labor-related occupations	1786(46.03)	20.68 ± 5.70
	No regular job	1547(39.87)	20.88 ± 6.27
	Loss job	399(10.29)	20.15 ± 6.36
Economic status	Poverty	792(21.08)	19.72 ± 6.16	25.69	< 0.001
	General	2892(76.96)	21.12 ± 5.92
	Wealthy	74(1.96)	23.57 ± 5.80
Married status	With a spouse	3598(92.73)	21.00 ± 5.95	7.96	0.005
	Without a spouse	282(7.27)	18.96 ± 6.68
Physical activity	low	1343(34.71)	20.05 ± 6.36	17.66	< 0.001
	middle	1185(30.63)	21.23 ± 5.81
	high	1341(34.66)	21.28 ± 5.81
Smoking status a	Never	2633(68.21)	20.16 ± 6.12	58.20	< 0.001
	Former	461(11.85)	22.82 ± 5.50
	Current	766(19.94)	22.03 ± 5.59
Secondary smoke a	non-passive smoking	2536(65.70)	20.61 ± 6.15	12.82	< 0.001
	passive smoking	558(14.46)	20.30 ± 5.89
	smoking	776(19.84)	22.03 ± 5.59
Alcohol intakea	Never	2692(75.77)	20.31 ± 6.12	33.11	< 0.001
	< Once a week	113(3.18)	22.86 ± 5.50
	Once to six times a week	267(7.51)	22.33 ± 5.51
	≥Once a day	481(13.54)	22.74 ± 5.40
BMI a	Underweight	155(4.01)	20.27 ± 6.38	1.16	0.326
	Normal weight	1636(42.34)	20.77 ± 5.92
	Overweight	1439(37.24)	21.04 ± 5.96
	Obesity	634(16.41)	20.71 ± 6.27
Hypertension	NO	1264(32.58)	21.21 ± 5.79	5.27	0.022
	YES	2616(67.42)	20.68 ± 6.13
Diabetes	NO	3180(81.96)	20.90 ± 5.97	3.24	0.072
	YES	700(18.04)	20.64 ± 6.29
			Median (Interquartile Range)
Dietary patterns	pattern 1		-0.28(-0.49,0.20)
	pattern 2		-0.24(-0.58,0.30)
	pattern 3		-0.19(-0.35,0.03)
	pattern 4		-0.25(-0.60,0.28)
	pattern 5		-0.06(-0.70,0.62)

Bolded values were considered statistically significant

Abbreviations: SD Standard deviation, BMI Body mass index

a Subgroup totals exhibit discrepancies with the overall cohort size because of incomplete covariate information

According to monitoring data from 2014 to 2018, the annual average concentrations of PM₁, PM_2.5, and PM₁₀ were 37.26 ± 1.48 µg/m³, 63.30 ± 1.73 µg/m³, and 102.91 ± 3.29 µg/m³, respectively. The mean serum FA concentration among study participants was 8.46 ± 3.75 ng/mL (Table S2, Figure S2). According to Table S3, PM₁, PM_2.5, and PM₁₀ all showed a negative association with MoCA-B scores, whereas FA showed a positive association, but this positive association did not reached statistical significance (P > 0.05).

Independent associations between PMs, FA, and MoCA-B scores

Results of the linear regression examining the association between PMs, FA, and MoCA-B scores were presented in Table 2. After full adjustment in Model 3, significant inverse correlations were observed between all PMs and MoCA-B scores, with effect estimates (β) of -0.355 for PM₁ (95%CI: -0.525, -0.186), -0.387 for PM_2.5 (95%CI: -0.558, -0.215), and − 0.353 for PM₁₀ (95%CI: -0.525, -0.180). Conversely, FA demonstrated a positive, though not statistically significant, association with MoCA-B scores (β = 0.107, 95%CI: -0.073, 0.287). To investigate the potential nonlinear relationships between PMs, FA and MoCA-B scores, we conducted further analyses. Initially, PMs and FA were divided into tertiles, and the linear trends of MoCA-B scores across the groups were examined (Table 2). No significant trend was observed between FA and MoCA-B scores ( P_trend = 0.505); the overall exposure levels of PM₁, PM₂.₅, and PM₁₀ increased were all significantly negatively associated with MoCA-B scores (all P_trend ≤ 0.001); further testing using RCS models revealed that the exposure-response relationships for PM₂.₅ and PM₁₀ were not strictly linear ( P_nonlinear < 0.05, see Supplementary Figure S3), and inflection points for the effects of both were identified accordingly.

Table 3.

Associations between exposure to FA and MoCA-B scores stratified by PMs

PMs(µg/m3)	FA β (95% CI) ab	P value	P interaction c
5-y average PM1 (µg/m3)
Continuous variable	0.127(-0.052,0.307)	0.165	0.151
Categorical variable by tertile
T1 [≤ 36.56]	0.347(0.082,0.612)	0.010
T2 [36.56–38.40]	0.137(-0.211,0.485)	0.440	0.848
T3 [> 38.40]	-0.071(-0.405,0.262)	0.675	0.069
5-y average PM2.5 (µg/m3)
Continuous variable	0.130(-0.149,0.310)	0.155	0.037
Categorical variable by tertile
T1 [≤ 62.18]	0.370(0.097,0.643)	0.008
T2 [62.18–64.41]	0.114(-0.206,0.434)	0.485	0.337
T3 [> 64.41]	-0.095(-0.444,0.254)	0.593	0.020
5-y average PM10 (µg/m3)
Continuous variable	0.128(-0.052,0.308)	0.163	0.020
Categorical variable by tertile
T1 [≤ 100.71]	0.404(0.128,0.68)	0.004
T2 [100.71–104.77]	0.158(-0.154,0.469)	0.320	0.419
T3 [> 104.77]	-0.125(-0.481,0.231)	0.491	0.024

Bolded values were considered statistically significant

Abbreviations: MoCA-B Montreal Cognitive Assessment - Basic Version, IQR Interquartile range, β Regression coefficient, CI Confidence interval, PMs, PM₁ Particles diameter ≤ 1, PM_2.5 Particles diameter ≤ 2.5, PM₁₀ Particles diameter ≤ 10, FA Folate, T1-T3 First (lowest), second (middle), and third (highest) tertiles

a The value of IQR for PM₁ was 2.83 µg/m³;The value of IQR for PM_2.5 was 3.01 µg/m³; The value of IQR for PM₁₀ was 5.34 µg/m³; The value of IQR for Folate was 5.46 ng/ml

b Covariates in models included age, gender, region, education level, occupation, economic status, marital status, physical activity, smoking, alcohol intake, dietary patterns, hypertension, diabetes

c Interaction terms between PM₁ (dummy variable), PM_2.5 (dummy variable), PM₁₀ (dummy variable) and Folate (continuous variable) were added to the model

PMs as a negative effect modifier of the association between FA and MoCA-B scores

Based on the linear regression model adjusted using Model 3, there is a significant interaction between PM_2.5, PM₁₀ and FA in affecting cognitive function (P < 0.05) (Table 3). Tertile stratification analysis showed that FA was significantly positively correlated with cognition in the groups with PM_2.5 ≤ 62.18 µg/m³ and PM₁₀ ≤ 100.71 µg/m³ (PM_2.5: β = 0.370,P = 0.008; PM₁₀: β = 0.404,P = 0.004), whereas this correlation was not observed in the third tertile groups of PM_2.5 > 64.41 µg/m³ and PM₁₀ > 104.77 µg/m³(PM_2.5: β=-0.095; PM₁₀: β=-0.125), and significant interactions between the groups were found (PM_2.5: P_interaction = 0.020, PM₁₀: P_interaction = 0.024). To further analyze the shift in this effect, we conducted a stratified analysis based on the effect inflection points identified using the RCS model (Fig. 1; Table 4). Exposure to PM_2.5 and PM₁₀ has a threshold effect on cognitive function, with significant differences in effect strength before and after the inflection points, which were consistent with the trend observed in the stratified analysis. In the groups with PM_2.5 ≤ 64.54 µg/m³ and PM₁₀ ≤ 103.9 µg/m³, FA was also significantly positively correlated with cognitive levels (PM_2.5: P = 0.026, PM₁₀: P = 0.002). At the PM_2.5 and PM₁₀ inflection point concentrations (> 64.54 µg/m³, > 103.9 µg/m³), FA is negatively correlated with MoCA-B scores, but the difference is not statistically significant, and a significant interaction between FA and the groups was observed (PM_2.5: P_interaction = 0.030, PM₁₀: P_interaction = 0.003).

Table 2.

Associations between PMs exposure, FA and MoCA-B scores

Variables	Model1 a		Model 2 b		Model 3 c
	β (95% CI)	P value	β (95% CI)	P value	β (95% CI)	P value
5-y average PM1 (µg/m3)
Continuous variable	-0.394(-0.576, -0.213)	< 0.001	-0.353(-0.523, -0.184)	< 0.001	-0.355(-0.525, -0.186)	< 0.001
Categorical variable by tertile
T1 [≤ 36.56]	Reference (0)		Reference (0)		Reference (0)
T2 [36.56–38.40]	-0.057(-0.502,0.389)	0.831	-0.483(-0.905, -0.060)	0.025	-0.483(-0.905, -0.060)	0.025
T3 [> 38.40]	-0.722(-1.166, -0.277)	0.001	-0.769(-1.183, -0.355)	< 0.001	-0.776(-1.190, -0.362)	< 0.001
P for trend		0.001		< 0.001		< 0.001
5-y average PM2.5 (µg/m3)
Continuous variable	-0.558(-0.739, -0.377)	< 0.001	− 0.385(-0.557, -0.214)	< 0.001	-0.387(-0.558, -0.215)	< 0.001
Categorical variable by tertile
T1 [≤ 62.18]	Reference (0)		Reference (0)		Reference (0)
T2 [62.18–64.41]	-1.061(-1.506, -0.616)	< 0.001	-1.106(-1.519, -0.692)	< 0.001	-1.106(-1.521, -0.692)	< 0.001
T3 [> 64.41]	-0.947(-1.391, -0.504)	< 0.001	-0.702(-1.117, -0.287)	0.001	-0.706(-1.122, -0.290)	0.001
P for trend		< 0.001		0.001		0.002
5-y average PM10 (µg/m3)
Continuous variable	-0.533(-0.714, -0.351)	< 0.001	-0.352(-0.524, -0.180)	< 0.001	-0.353(-0.525, -0.180)	< 0.001
Categorical variable by tertile
T1 [≤ 100.71]	Reference (0)		Reference (0)		Reference (0)
T2 [100.71-104.77]	-1.242(-1.685, -0.798)	< 0.001	-1.249(-1.661, -0.837)	< 0.001	-1.248(-1.661, -0.835)	< 0.001
T3 [> 104.77]	-1.132(-1.576, -0.687)	< 0.001	-0.893(-1.313, -0.473)	< 0.001	-0.895(-1.316, -0.475)	< 0.001
P for trend		< 0.001		< 0.001		< 0.001
FA (ng/ml)
Continuous variable	0.105(-0.086,0.297)	0.279	0.110(-0.069,0.290)	0.229	0.107(-0.073,0.287)	0.243
Categorical variable by tertile
T1 [≤ 6.5]	Reference (0)		Reference (0)		Reference (0)
T2 [6.5–10]	0.040(-0.415,0.495)	0.863	0.023(-0.418,0.463)	0.920	0.022(-0.418,0.463)	0.920
T3 [> 10]	0.154(-0.308,0.615)	0.514	0.156(-0.292,0.604)	0.495	0.151(-0.298,0.599)	0.510
P for trend		0.526		0.473		0.505

Bolded values were considered statistically significant

Abbreviations: MoCA-B Montreal Cognitive Assessment - Basic Version, IQR Interquartile range, β Regression coefficient, CI Confidence interval, PMs PM₁: particles diameter ≤ 1; PM_2.5 Particles diameter ≤ 2.5, PM₁₀ Particles diameter ≤ 10, FA Folate, T1-T3 First (lowest), second (middle), and third (highest) tertiles. The value of IQR for PM₁ was 2.83 µg/m³;The value of IQR for PM_2.5 was 3.01 µg/m³; The value of IQR for PM₁₀ was 5.34 µg/m³; The value of IQR for Folate was 5.46 ng/ml

a Model 1: Adjusted for age, gender

b Model 2: Adjusted for age, gender, region, education level, occupation, economic status, marital status, physical activity, smoking, alcohol intake, dietary patterns

c Model 3: Additionally controlled for hypertension and diabetes

Fig. 1.

Estimated effect[β(95%CI)] of Folate on MoCA-B scores after stratification by the inflection points of PM₂.₅ (64.54 µg/m³) and PM₁₀ (103.9 µg/m³) identified by RCS. Covariates in models included age, gender, region, education level, occupation, economic status, marital status, physical activity, smoking, alcohol intake, dietary patterns, hypertension, diabetes

Table 4.

Associations between FA and MoCA-B scores stratified by the inflection points (thresholds) of restricted cubic spline (RCS) of PMs

PMs(µg/m3)	FA β (95% CI) ab	P value	P interaction c
5-y average PM2.5 (µg/m3)
Continuous variable			0.037
Categorical variable by inflection pointt
≤ 64.54	0.234(0.028,0.440)	0.026
> 64.54	-0.095(-0.442,0.252)	0.591	0.030
5-y average PM10 (µg/m3)
Continuous variable			0.020
Categorical variable by inflection pointt
≤ 103.9	0.357(0.136,0.579)	0.002
> 103.9	-0.195(-0.487,0.096)	0.189	0.003

Bolded values were considered statistically significant

Abbreviations: MoCA-B Montreal Cognitive Assessment - Basic Version; interquartile range; β Regression coefficient, CI Confidence interval, PMs, PM_2.5Particles diameter ≤ 2.5, PM₁₀ Particles diameter ≤ 10, FA Folate

a The value of IQR for PM_2.5 was 3.01 µg/m³; The value of IQR for PM₁₀ was 5.34 µg/m³; The value of IQR for Folate was 5.46 ng/ml

b Covariates in models included age, gender, region, education level, occupation, economic status, marital status, physical activity, smoking, alcohol intake, dietary patterns, hypertension, diabetes

c Interaction terms between PM_2.5 (dummy variable), PM₁₀ (dummy variable) and Folate (continuous variable) were added to the model

Figure S4 further illustrates the interaction between PM_2.5, PM₁₀, and FA in relation to MoCA-B scores, that the beneficial effect of FA on cognitive performance progressively declines with rising concentrations of PM_2.5 and PM₁₀. When the concentrations reach 64.34 µg/m³ and 104.59 µg/m³, respectively, the statistical significance of this effect disappears, and this reversal threshold closely aligns with the effect inflection points identified by the RCS model (64.54 µg/m³ and 103.9 µg/m³).

Stratified analysis

The results of subgroup analyses based on gender and urban-rural status are summarized in Table S4. The results of the subgroup analysis indicate that the findings across all populations are consistent with the main analysis conclusions.

Sensitivity analysis

In all sensitivity analyses, the main findings remained robust. (1) Adjusting for BMI and secondhand smoke exposure factors did not change the main findings (Table S5-S7);(2) After respectively controlling for NO₂, SO₂, and O₃, the direction and significance of the interactive association between PMs and cognitive function and FA did not change substantially (Table S5-S7); (3) Joint effect analysis showed that the group with low PM_2.5/PM₁₀ and high FA exposure had the lowest cognitive risk (Table S8).

Discussion

The study explored the independent effects of long-term PMs exposure and FA on cognitive function in Chinese older adults, and the modifying effect of PMs exposure on the cognitive impact of FA. The results indicated that higher exposure to PM₁, PM_2.5, and PM₁₀ was linked to lower MoCA-B scores, although the positive FA-MoCA-B scores correlation was not statistically significant, but it still showed a positive trend. Further analysis found that as the five-year average concentrations of PM_2.5 and PM₁₀ increased, the positive association between FA and MoCA-B scores gradually weakened, disappeared, and even reversed at higher exposure levels. Stratified analysis based on the RCS model breakpoints further clarified this moderating effect’s critical threshold, when PM₂.₅ and PM₁₀ concentrations exceeded 64.54 µg/m³ and 103.9 µg/m³, the positive association was attenuated and even replaced by a negative association. And the interaction between PM₁ and FA did not reach statistical significance.

Existing research has fully demonstrated the benefits of FA for cognitive function [28, 29]. Regarding the biological mechanisms related to FA and cognitive, previous studies have shown that neuroprotective properties of FA were potentially mediated through its role in the modulation of homocysteine (Hcy), with the reduction of Hcy levels contributing to enhanced cognitive outcomes, enhancing vascular health, reducing inflammatory responses, correcting cerebral FA deficiency, and boosting antioxidant capacity [30–32]. FA could contribute to the preservation of blood-brain barrier integrity by lowering serum levels of pro-inflammatory cytokines, thereby further improving cognitive function [33–35]. In addition, FA could help maintain normal DNA methylation status, stabilize genome structure, and regulate the expression of genes for neuroprotection and cognitive function [36, 37]. Consistent with previous research findings, we also observed a positive correlation of FA levels with cognitive function in this study. However, this positive correlation was not significant. One possible explanation was that participants in this study were exposed to high levels of PMs for a long time, which could have masked the significant positive association that FA might have with cognitive function. That was, differences in effects between groups produced a dilution effect, offsetting the overall association.

Studies have shown that long-term exposure to PMs was associated with cognitive decline [38, 39], and exposure to high concentrations of PMs significantly increased the risk of developing cognitive impairment and neurodegenerative diseases [40]. It may also induce a systemic inflammatory response, promoting an increase in inflammation markers associated with the cardiovascular and cerebrovascular systems [41, 42]. PMs induced endoplasmic reticulum stress and neuronal apoptosis through DNA methylation (including hypomethylation caused by disruptions in one-carbon metabolism) [43, 44]. Furthermore, evidence indicated that inhaled PM_2.5 particles could access the brain via the olfactory nerve pathway, effectively circumventing the blood-brain barrier, having a direct impact on the brain [45–47]. PM₁ directly entered the brain, inducing neuroinflammation and Aβ/tau pathology, while PM₁₀ caused cerebrovascular damage and ischemic neuropathy through systemic inflammation [48].

Consistent with previous studies, the increase in air pollution levels was associated with an attenuated positive association of the health benefits brought by FA [17, 18].The different effects of PMs and FA on key pathways such as oxidative stress, inflammatory response, and methylation metabolism suggest that exposure to specific concentrations of PMs could partially offset the neuroprotective potential of FA at the molecular level [49]. We found that there was a clear threshold for the combined effect of PMs and FA: under low PMs exposure, the negative effects of PMs on cognitive function were mild and FA was still positively associated with cognitive function; FA could still exert positive association by regulating antioxidant, anti-inflammatory, and one-carbon metabolism pathways [50, 51]. However, PM exposure above this threshold was associated with an attenuated positive association, with trend toward a negative association. High concentrations of PMs could directly disrupt the one-carbon metabolic pathways in neurons, specifically inhibiting the FA cycle and methionine cycle, and it could also lead to abnormal accumulation of homocysteine and DNA methylation disorder [17, 52]. Thus, the physiological regulatory ability of FA could no longer counteract the multiple neurotoxic damages caused by high PMs. The determined thresholds of PM_2.5 and PM₁₀ offered a relatively manageable intervention cutoff, at which the neuroprotective effect of FA on cognitive function transitioned from effective compensation to decompensated imbalance. For areas with ambient pollutant concentrations exceeding the above thresholds like those in this study, it was recommended that susceptible elderly populations appropriately avoid high-dose FA intake, reduce outdoor activities during peak pollution periods, and strengthen routine personal protection.

The interactions between PM_2.5 and PM₁₀ with FA were significant, while the interaction with PM₁ was not significant. This difference could be explained by PM₁ primarily entered the brain parenchyma directly through the olfactory route or the blood-brain barrier, inducing localized neuroinflammation and protein deposition, with little interference in the circulation and metabolism of peripheral FA [48]; whereas PM_2.5/PM₁₀ deposited in the respiratory tract, induced systemic inflammation and oxidative stress, affected the activity of peripheral one-carbon metabolism enzymes through the circulatory system, and disrupted the blood-brain barrier [53]. As a water-soluble vitamin, FA mainly exerted its anti-inflammatory and methylation-regulating effects through systemic circulation, making it more susceptible to metabolic disruption induced by PM_2.5 and PM₁₀ [54, 55].In addition, components carried by PM_2.5/PM₁₀, such as polycyclic aromatic hydrocarbons and heavy metals, could inhibit key enzymes in the methionine cycle, interfering with the generation of FA-dependent methyl donors [56, 57].

The principal strength of this research lay in its extensive sample, which enabled the detection and characterization of complex interactions. Secondly, this study simultaneously included PM₁, PM₂.₅, and PM₁₀ particles of multiple sizes, which allowing us to reveal their interactive effects with FA under different toxicological mechanisms, thus offering a holistic evaluation of air pollution’s health risks and pinpointing critical hazardous pollutants. Finally, this study overcame the limitations of previous research that relied on dietary questionnaires to indirectly estimate nutrient intake, and in a large elderly population in China, directly measured serum FA levels as an objective biomarker of in vivo bioavailability. This method effectively avoided recall bias and estimation errors, providing a reliable methodological basis for accurately assessing the interaction between PMs and FA. However, our study also had some limitations. First, it employed a cross-sectional design, which limited the inference of causal relationships. Although using the average PM exposure data from the past five years helped to some extent in revealing potential long-term association trends, the causal direction still needed to be verified by future longitudinal cohort studies. Secondly, this study focused primarily on outdoor PMs exposure. In the sensitivity analysis, adjustments were made for secondhand smoke exposure, but there might still have been some bias in the overall estimation of PMs exposure levels. Third, this study only focused on long-term average exposure and did not include short-term time window indicators, so it could not analyze the potential effects of short-term acute exposure. Future research could include both short-term and long-term exposure indicators simultaneously to further compare the similarities and differences in their effects on FA and cognitive function. Fourth, although we adjusted for multiple covariates, it was still difficult to completely rule out residual confounding. Polymorphisms in one-carbon metabolism-related genes independently affected susceptibility to neurodegenerative lesions, and potentially interacted with PMs and FA levels, while there were still unmeasured confounding factors such as health-related lifestyle behaviors that have not been accounted for. Lastly, due to changes in global air pollution levels, climate conditions, socio-economic status, and dietary habits, and given the regional and population-specific characteristics, the PM_2.5 and PM₁₀ thresholds identified in this study might not be universally applicable. Therefore, these findings need to be externally validated for different populations, environmental contexts, and periods in the future.

Conclusions

We found that among the elderly, there was an interaction between FA and PM_2.5/PM₁₀ on cognitive function (MoCA-B scores). As concentrations of PM_2.5 and PM₁₀ increased, the positive association was attenuated and even replaced by a negative association. The thresholds at which this effect disappears were 64.54 µg/m³ and 103.9 µg/m³, respectively. These findings encourage people to take corresponding protective measures to further reduce the harm caused by air pollution, to fully realize the cognitive protective effects of FA. This is of great significance for maintaining cognitive health in the elderly, public health prevention, and urban environmental planning.

Abbreviations

PM

Particulate matter

FA

Folate

MoCA-B

Montreal Cognitive Assessment - Basic Version

MCI

Mild cognitive impairment

AD

Alzheimer’s disease

Hcy

Homocysteine

BMI

Body mass index

RCS

Restricted Cubic Splines

CHAP

China High Air Pollutants

STET

Space-Time Extra-Trees

NO₂

Nitrogen dioxide

SO₂

Sulfur dioxide

O₃

Ozone

IQR

Interquartile range

β

Regression coefficient

SD

Standard deviation

CI

Confidence interval

Authors' contributions

YC, BBY, and YWH wrote the original draft, conducted the investigation, curated the data, designed the methodology, performed the formal analysis, and validated the results. CM, ZHW, LLY, LG, ZWT, WYL, XLL conducted the data analysis and interpretation. FBT, BH conducted surveys and data analysis. LSY, GMC conceived the study design and are responsible for the overall content. All authors read and approved the final manuscript.

Funding

This study was supported by the Research Funds of the Scientific Research Projects for Higher Education of Anhui Province (grant number 2023AH010036), the Center for Big Data and Population Health of IHM (grant number JKS2022015), the Key Scientific Research Fund of the Anhui Provincial Education Department (grant number 2023AH050610), and the Anhui Natural Science Foundation (grant number 1808085QH252).

Data availability

The datasets generated and/or analysed during the current study are not publicly available due [REASON WHY DATA ARE NOT PUBLIC] but are available from the corresponding author on reasonable request.The dataset supporting the conclusions of this article is included within the article (and its additional file).

Declarations

Ethics approval and consent to participate

All study procedures in accordance with the Declaration of Helsinki have been approved by the Research Ethics Committee at Anhui Medical University (No. 20190288, No. 83230436). All individual participants provided written informed consent at enrollment.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.