Cardiovascular disease attenuates the protective effect of folate on global cognitive function in an elderly population: a cross-sectional study

Abstract

Research suggests that folic acid contributes to improving cognitive function. However, there is a lack of systematic research on the association of dietary intake of folate and serum, and red blood cell (RBC) folate levels with global cognitive impairment (CoI) in the elderly population. Importantly, excessive supplementation with folate among American adults at high risk of cardiovascular disease (CVD) may have harmful effects. CVD often leads to worse cognitive function; therefore, it is necessary to explore the characteristics of the association of folate with CoI in both CVD and non-CVD populations. Participants aged ≥ 60 years from the national health and nutrition examination survey (NHANES) 2011–2014 were included. Dietary intake of folate and serum and RBC folate levels were determined through questionnaires or laboratory measurements. Global cognitive function was assessed via the results of three cognitive assessments. Multivariable logistic regression and restricted cubic splines (RCS) were employed to assess the odds ratios (ORs), 95% confidence intervals (CIs), and potential non-linearities of folate with cognition. Additionally, the interaction term of CVD with RBC folate was included in the model, and effect modification was detected through likelihood ratio tests. Finally, several sensitivity analyses were conducted to validate our findings. This study included 2104 participants with complete data and a median age of 68 years, with females comprising 51% of the participants. Of the participants, 444 individuals were defined as having CoI. In the NHANES, Pearson correlation analysis revealed moderate to weak correlations between dietary, serum, and RBC folate levels and CoI (all < 0.6). In addition, when different sources of folate were included separately in the models, fully adjusted logistic regression with continuous variables included in the model revealed that only RBC folate was significantly associated with CoI [odds ratio (OR) 0.86, 95% confidence interval (95% CI) 0.75–0.97, P = 0.02]. According to tertile groups, compared with participants in the lowest tertile, individuals in the highest tertile of total dietary folate levels (OR 0.67, 95% CI 0.48–0.94, P = 0.02), folic acid levels (OR 0.57, 95% CI 0.38–0.86, P = 0.01), and RBC folate levels (OR 0.62, 95% CI 0.44–0.85, P = 0.004) had significantly lower odds of having CoI. The RCS showed a linear negative correlation between RBC folate levels and CoI. Furthermore, interaction analysis suggested that CVD attenuated the protective effect of RBC folate. Several sensitivity analyses also suggested a modifying effect of CVD on the association between RBC folate and CoI. A significant negative correlation exists between RBC folate levels and CoI in the elderly population of the United States, and this association is stronger than that of other folate measures. However, the protective effect of RBC folate on CoI is negated in patients with CVD, and further investigation is needed to explore the underlying mechanisms involved.

Introduction

Cognitive impairment (CoI) is a psychological health disorder that primarily affects cognitive abilities, including learning, memory, perception, and problem-solving, among others. CoI is generally defined as deficiencies in six cognitive abilities: executive function, learning and memory, perceptual motor function, language, complex attention, and social cognition¹. CoI can range from mild to severe, and a significant proportion of mild CoI patients eventually progress to dementia^2–4. In the United States, it is estimated that nearly 14 million people will be diagnosed with dementia in the next 40 years⁵.

Folate is an essential water-soluble B vitamin that participates in cellular metabolism and the biosynthesis of nucleotides and amino acids^6,7. Additionally, folate plays a crucial role in one-carbon metabolism, which is necessary for processes such as immune responses, mitochondrial metabolism, and antioxidant activity^8,9. Folate intake by individuals occurs in two forms: intake of natural folate, which is widely present in plant and animal foods, and intake of folic acid, a synthetic form of folate extensively used in fortified foods and supplements. The bioavailability of folate in the human body is a complex and variable issue that is affected by multiple factors. First, folate in natural foods usually needs to undergo a series of metabolic processes before it can be absorbed and utilized by the human body. After entering the small intestine, these natural folates need to be broken down into small molecular forms and then absorbed through the intestine. However, this process is often incomplete, resulting in a relatively low bioavailability of natural folate, approximately 45–50%^10,11. In contrast, the bioavailability of synthetic folate (i.e., folic acid) is relatively high and is generally considered to be above 85%¹². This is because synthetic folate can be directly absorbed and utilized by the human body without going through complex metabolic steps. Therefore, in many countries or regions, mandatory food fortification programs have been implemented to increase the folate intake of the population, mainly using synthetic folate as a fortificant.

Currently, the relationship between folate and CoI is subject to debate, with most studies suggesting a preventive effect of dietary folate against CoI. A randomized controlled trial from Tianjin, China, revealed that folic acid supplementation significantly improved the cognitive performance of patients with mild cognitive impairment and reduced the levels of peripheral inflammatory cytokines such as IL-6 and TNF-α¹³. Although the main mechanism underlying memory decline remains uncertain, the inflammatory state has been described as a possible pathological mechanism^14–17. Folate deficiency can lead to impaired vitamin B12 metabolism, resulting in an inflammatory state¹⁸. However, some research suggests that high folate intake may increase the risk of CoI in certain individuals^19–24. Therefore, Bailey et al.²⁵ suggested careful monitoring of total folate intake, both natural and synthetic, from foods and dietary supplements. Serum folate is considered an indicator reflecting recent folate nutritional status²⁶. Several forms of folate in serum collectively constitute the metabolic network of vitamin B9 in the body, which is crucial for DNA synthesis, repair, and one-carbon metabolism processes^27,28. Research by Ding et al.²⁹ suggested a U-shaped association between red blood cell (RBC) folate and serum folate levels and subtest scores of CoI. RBC folate levels reflect chronic or long-term (within 4 months) folate nutritional status and are more suitable for evaluating the effectiveness of folate interventions³⁰. Currently, there is limited evidence linking RBC folate to cognition. Lei et al.³¹ suggested that RBC folate concentrations may be a clinical predictor of CoI in patients with abdominal obesity. A study by Min et al.³² indicated that RBC folate appears to be inversely linearly associated with all-cause mortality risk in Alzheimer’s disease patients. However, the relationship between multidimensional folate and global cognitive function has yet to be elucidated.

The relationship between folate and cardiovascular disease (CVD) does not seem as ideal as commonly believed. Xu et al.³³ comprehensively investigated the relationship between dietary, serum, and RBC folate levels and the risk of CVD-specific mortality and all-cause mortality in 14,234 participants with a history of CVD in the NHANES. They found that moderate folate intake had a beneficial effect on long-term survival, whereas high serum folate and RBC folate levels increased the risk of CVD mortality and all-cause mortality in such populations. Liu et al.³⁴ found that both low and high serum folate levels increased the risk of CVD mortality in patients with type 2 diabetes. In addition, hypertension increased the risk of frailty and weakness, especially among elderly individuals^35,36. Cognitive and physical impairments were common in hypertensive patients with prediabetes, especially in frail elderly individuals^37,38. Insulin resistance (IR) may lead to cognitive impairment in frail elderly individuals with prediabetes and hypertension³⁹. Studies have shown that IR not only affects peripheral metabolism but also enters the central nervous system through the blood-brain barrier, influencing glucose metabolism and neuronal function in the brain⁴⁰. For elderly people with hypertension, IR not only exacerbates declines in cognitive function but also may be related to the progression of degenerative diseases such as Alzheimer’s disease. Hypertension can cause damage to the cerebral microcirculation, further strengthening the impact of IR on cognitive function⁴¹. Considering the bidirectional role of folate in CVD patients, CVD may modify the relationship between folate and cognition.

In this study, we systematically analysed the relationship of dietary natural folate, synthetic folic acid, serum folate, and RBC folate levels with global cognitive function, attempting to elucidate the differences in this association between patients with and without CVD.

Methods

Study population

This cross-sectional study investigated participants in the NHANES from 2011 to 2014. The survey obtained ethical approval from the National Center for Health Statistics Ethics Review Board, and all participants provided written informed consent. Details are available at https://www.cdc.gov/nchs/nhanes/index.htm. In the 2011–2014 NHANES cycles, participants aged > 60 years underwent cognitive tests⁴². A total of 29,902 participants over two cycles were included, and 24,369 participants < 60 years, 1,527 with missing exposure variables, 1641 with missing outcome variables, and 261 with missing covariables were sequentially excluded, resulting in a final eligible sample of 2104 participants (Fig. 1).

Fig. 1.

Flow chart for inclusion of participants.

Folate and folic acid

Dietary intake and supplement data were assessed through two 24-hour dietary recall interviews. The first dietary recall interview was conducted in person at the Mobile Examination Center (MEC), and the second interview was conducted by telephone 3 to 10 days later. The dietary folate or folic acid intake from foods was calculated using the Food and Nutrient Database for Dietary Studies of the United States Department of Agriculture⁴³. The total folate intake is the sum of natural folate and folic acid. In this study, folate intake was expressed in dietary folate equivalents (DFE), with units expressed in micrograms (µg). The formula is as follows: DFE (µg) = natural folate (µg) + 1.7 × (food folic acid + folic acid supplements) (µg)⁴⁴. Blood samples were collected from participants at the Mobile Examination Center (MEC), and whole blood and serum samples were properly frozen (− 20 °C) until they were transported to the National Center for Environmental Health for analysis. NHANES uses several methods to monitor the quality of the analyses performed by contract laboratories. In MEC, these methods include performing blind split samples collected during “dry run” sessions. In addition, contract laboratories randomly perform repeated testing on 2% of all samples. Folate content was measured by isotope dilution liquid chromatography-tandem mass spectrometry (LC-MS/MS), and the total serum folate concentration (ng/ml) was one of the parameters measured. For analytes below the limit of detection, a presumed fill value was placed in the result field, which is the square root of the limit of detection divided by 2^26,45. RBC folate (ng/ml) was calculated from the whole blood folate concentration through microbiological assay, adjusting for RBC volume and correcting for total serum folate concentrations, which is the total concentration of various folate forms^46,47. Due to the skewed distribution of the measurements, the original variables were transformed into Z-scores for subsequent analysis.

Cognitive impairment (CoI)

During 2011–2014, a series of assessments in NHANES (variable name prefix CFQ) were re-introduced, including (1) word learning and recall modules from the Consortium to Establish a Registry for Alzheimer’s disease (CERAD); (2) the Animal Fluency test (AFT); and (3) the Digit Symbol Substitution test (DSST). Although cognitive assessments cannot replace a diagnosis based on a clinical examination, they are useful for examining the association of cognitive functioning with many medical conditions and risk factors measured during the NHANES examination (CFQ_G (cdc.gov)). Notably, this testing project targets individuals > 60 years of age. CERAD includes the immediate recall test (IRT) and one delayed recall test (DRT), which reflect immediate and delayed new language learning, respectively⁴⁸. The IRT consists of three consecutive tests, with the sum of the three tests constituting the IRT score. The score ranges for IRT and DRT are 0–30 and 0–10, respectively. The AFT test reflects the fluency of absolute language, which is a component of executive function⁴⁹. As a performance module of the Wechsler Adult Intelligence Scale, the DSST reflects processing speed, sustained attention, and working memory, with scores ranging from 0 to 133⁵⁰. The sum of the scores of these four cognitive function assessments indicates stronger cognitive function as the score increases. Based on previous studies, individuals in the lowest quartile group were defined as having CoI^51,52.

Covariables

The determination of confounding factors that need to be adjusted is based on previously published literature^21,29. (1) Demographic characteristics including age (continuous), sex (male/female), race (Mexican American, non-Hispanic black, non-Hispanic white, and others), educational level (below college, college and higher), and marital status (divorced/separated/widowed, married/cohabiting, never married) were self-reported at the time of the questionnaire; additionally, the ratio of household income to poverty level (< 1.3, 1.3–3.5, > 3.5) was assessed. (2) Lifestyle and physical indicators included alcohol consumption (never, former, current), smoking (never, former, current), total energy intake (classified into high and low groups based on median values), and body mass index (BMI), which was defined as weight in kilograms divided by the square of height in metres (kg/m²). (3) Comorbidities included cancer, liver diseases, CVD (including coronary artery disease, congestive heart failure, heart attack, stroke, and angina), chronic kidney disease (CKD), diabetes, arthritis, and thyroid diseases. In addition, we present the directed acyclic graph in (Supplementary Fig. 1).

Statistical analysis

For the analysis of participant characteristics, continuous variables are represented as the means (standard deviation, SDs) or medians (interquartile ranges, IQRs), whereas categorical variables are represented as sample size (N%). The Student’s t test or Mann-Whitney U test was used to analyse continuous variables; for categorical variables, chi-square (χ²) test was used to compare baseline characteristics between patients with CoI and healthy controls. Pearson’s test was used to detect the correlation of folate indicators. Multiple logistic models were constructed to assess the relationship between folate and CoI risk: (1) unadjusted model (Model 0); (2) Model 1 adjusted for demographic factors (sex, age, race, education, socioeconomic status, and marital status); (3) Model 2 further adjusted for smoking, alcohol consumption, energy intake, and BMI; and (4) Model 3 further adjusted for CVD, CKD, cancer, liver disease, diabetes, arthritis, and thyroiditis. Folate values were included in the logistic regression models as continuous variables (Z-score) or three-category variables based on tertiles to estimate odds ratios (ORs) and confidence intervals (CIs) with CoI, and median values of categorical variables were included in models to test trends (P for trend). Collinearity was tested, and the variance inflation factor (VIF) for all the models in this study was less than 2. Additionally, based on the principle of the minimum Akaike information criterion (AIC), a restricted cubic splines (RCS) regression model with three knots (10th, 50th, and 90th percentiles) was constructed to determine the dose-response relationship between folate and CoI; in the RCS model, the median value of folate served as the reference point, and nonlinear relationships were assessed with the Wald test (P for nonlinear). Subgroup analysis was performed to identify potential effect modifiers. The specific method included the multiplicative terms between RBC folate and stratified variables in the model, and the difference between this model and the original model was compared with the likelihood ratio test (P for interaction).

Several sensitivity analyses were conducted to assess the robustness of the results. (1) The proportion of missing covariables in this study was less than 10% (Fig. 1), and multiple imputation was performed for missing covariables in the first sensitivity analysis. (2) According to previous studies^53,54, a new algorithm was used to redefine global cognitive function. That is, the four sub-items were first converted to Z-scores, and after all Z-scores were added, participants in the lowest quartile group was defined as having CoI. (3) Despite research showing that vitamin B12 and folate do not interact to affect several aspects of cognitive function, we additionally adjusted for serum vitamin B12²⁹. In the sensitivity analysis, in addition to re-evaluating the associated ORs and dose-response relationships, the interaction between CVD and RBC folate was re-evaluated.

Statistical analyses were performed using R_4.3.3 software (https://www.r-project.org/). The “car” package was used to test collinearity, the “rms” package was used to construct restricted cubic spline regression models, and the “mice” package was used for multiple imputation; the “lmtest” package was used to detect interactions. In this study, statistical significance was defined as a two-tailed P value < 0.05.

Ethical approval

The survey received ethical approval from the National Center for Health Statistics Institutional Review Board, and all participants provided written informed consent. For detailed information, please refer to https://www.cdc.gov/nchs/nhanes/index.htm. In the current study, all methods were performed in accordance with the relevant guidelines and regulations.

Results

Population characteristics

After screening, a total of 2,104 participants with complete data were included (Fig. 1). As shown in Table 1, the median age of all participants was 68 years, and 1,083 (51%) of the individuals were female; in this cohort, 444 participants were defined as having cognitive impairment. Compared with the healthy control group, the CoI group had a greater proportions of males, older individuals, individuals with lower educational levels, fewer married individuals, more impoverished individuals, individuals with lower dietary energy intake, and fewer patients with cancer and thyroid disease; however, they had a higher prevalence of arthritis, diabetes, CVD, and CKD. Additionally, several of their folate indicators were significantly lower.

Table 1.

Characteristics of eligible participants.

Characteristics	Total (n = 2104)	Healthy (n = 1660)	Cognitive impairment (n = 444)	P
Sex, n (%)				0.01
Female	1083 (51)	879 (53)	204 (46)
Male	1021 (49)	781 (47)	240 (54)
Age, median (IQR)	68 (63, 75)	67 (63, 74)	72 (66, 80)	< 0.001
Race, n (%)				< 0.001
Mexican American	175 (8)	120 (7)	55 (12)
Non-hispanic black	469 (22)	318 (19)	151 (34)
Non-hispanic white	1099 (52)	953 (57)	146 (33)
Others	361 (17)	269 (16)	92 (21)
Education attainment, n (%)				< 0.001
Less than college	967 (46)	622 (37)	345 (78)
College or higher	1137 (54)	1038 (63)	99 (22)
Marital status, n (%)				< 0.001
Never married	115 (5)	91 (5)	24 (5)
Divorced/separated/widowed	735 (35)	537 (32)	198 (45)
Married/living with a partner	1254 (60)	1032 (62)	222 (50)
Poverty income ratio, n (%)				< 0.001
<1.3	567 (27)	351 (21)	216 (49)
1.3–3.5	826 (39)	659 (40)	167 (38)
>3.5	711 (34)	650 (39)	61 (14)
Alcohol status, n (%)				< 0.001
Never	301 (14)	215 (13)	86 (19)
Former	588 (28)	407 (25)	181 (41)
Now	1215 (58)	1038 (63)	177 (40)
Smoke, n (%)				0.744
Never	1047 (50)	829 (50)	218 (49)
Former	818 (39)	647 (39)	171 (39)
Now	239 (11)	184 (11)	55 (12)
Body mass index, n (%)				0.701
<25 Kg/m2	556 (26)	435 (26)	121 (27)
≥25 Kg/m2	1548 (74)	1225 (74)	323 (73)
Energy intake, n (%)				< 0.001
Lower than median	1053 (50)	772 (47)	281 (63)
Higher than median	1051 (50)	888 (53)	163 (37)
Cancer, n (%)	422 (20)	354 (21)	68 (15)	0.006
Thyroid problem, n (%)	368 (17)	309 (19)	59 (13)	0.011
Liver problem, n (%)	117 (6)	92 (6)	25 (6)	1
Arthritis, n (%)	1041 (49)	796 (48)	245 (55)	0.008
Diabetes, n (%)	689 (33)	490 (30)	199 (45)	< 0.001
CVD, n (%)	461 (22)	314 (19)	147 (33)	< 0.001
CKD, n (%)	727 (35)	503 (30)	224 (50)	< 0.001
Total dietary folate equivalents (µg), median (IQR)	626.5 (375.88, 1072)	665(389.38,1105.62)	494.75(328.38, 869.12)	< 0.001
Nature dietary folate equivalents (µg), median (IQR)	500.25 (185.5, 901.38)	648 (207.5, 914.88)	257.5 (135.88, 809.62)	< 0.001
Folic acid equivalents (µg), median (IQR)	226.75 (128.62, 370.5)	238 (134, 386.5)	181.5 (119.38, 301.88)	< 0.001
Serum folate concentration (ng/mL), median (IQR)	20.9 (13.8, 31.63)	21.5 (14.17, 32.02)	19.6 (12.67, 29.72)	0.007
RBC folate (ng/mL), median (IQR)	552 (400.75, 755)	570 (410.75, 764)	508 (377.75, 737)	0.002

RBC red blood cell, IQR interquartile range, CVD cardiovascular disease, CKD chronic kidney disease.

The estimated ORs per SD increase in folate values with CoI

Figure 2 shows a moderate to low correlation between dietary, serum, and RBC folate levels and CoI, with Pearson coefficients ranging from 0.43 to 0.58. Table 2 presents the associations between five folate indicators and CoI. In the models incorporating continuous variables, after full adjustment (Model 3), only RBC folate significantly decreased the odds of having CoI (OR 0.86, 95% CI 0.75–0.97, P = 0.02). When three-category variables were incorporated into the model, after adjustment (Model 3), the results revealed that compared to the tertile 1, participants in tertile 3 of total dietary folate levels (OR 0.67, 95% CI 0.48–0.94, P = 0.02), folic acid levels (OR 0.57, 95% CI 0.38–0.86, P = 0.01), and RBC folate levels (OR 0.62, 95% CI 0.44–0.85, P = 0.004) had significantly lower odds of having CoI, and all these factors exhibited significant trend effects (P for trend < 0.05, respectively).

Fig. 2.

Pearson correlation coefficients of total dietary folate, serum folate, and RBC folate. RBC red blood cell.

Table 2.

Multivariable logistic regression analysis of the odds of CoI with folate.

		Model 0		Model 1		Model 2		Model 3
		OR (95% CI)	P	OR (95% CI)	P	OR (95% CI)	P	OR (95% CI)	P
Total dietary folate	Continuous (per SD+)	0.68(0.60,0.77)	< 0.0001	0.82(0.70,0.94)	0.01	0.86(0.74,1.00)	0.06	0.87(0.75,1.02)	0.09
	Tertile 1	Ref		Ref		Ref		Ref
	Tertile 2	0.74(0.58,0.94)	0.01	0.83(0.62,1.10)	0.19	0.92(0.69,1.24)	0.59	0.90(0.66,1.21)	0.48
	Tertile 3	0.42(0.32,0.54)	< 0.0001	0.60(0.44,0.82)	0.001	0.67(0.48,0.93)	0.02	0.67(0.48,0.94)	0.02
	P for trend		< 0.0001		0.002		0.02		0.02
Nature dietary folate	Continuous (per SD+)	0.76(0.65,0.87)	< 0.001	0.78(0.65,0.93)	0.01	0.88(0.73,1.05)	0.16	0.87(0.71,1.04)	0.13
	Tertile 1	Ref		Ref		Ref		Ref
	Tertile 2	0.60(0.45,0.80)	< 0.001	0.78(0.55,1.09)	0.14	0.86(0.61,1.21)	0.39	0.87(0.61,1.24)	0.43
	Tertile 3	0.36(0.26,0.49)	< 0.0001	0.62(0.42,0.91)	0.02	0.71(0.48,1.06)	0.10	0.76(0.50,1.13)	0.18
	P for trend		< 0.0001		0.01		0.1		0.17
Folic acid	Continuous (per SD+)	0.64(0.54,0.75)	< 0.0001	0.81(0.67,0.97)	0.03	0.85(0.70,1.01)	0.08	0.86(0.71,1.04)	0.13
	Tertile 1	Ref		Ref		Ref		Ref
	Tertile 2	0.67(0.49,0.90)	0.01	0.56(0.40,0.80)	0.001	0.65(0.45,0.93)	0.02	0.63(0.43,0.91)	0.01
	Tertile 3	0.47(0.34,0.64)	< 0.0001	0.48(0.33,0.69)	< 0.001	0.59(0.39,0.88)	0.01	0.57(0.38,0.86)	0.01
	P for trend		< 0.0001		< 0.0001		0.01		0.01
Serum folate	Continuous (per SD+)	0.93(0.84,1.04)	0.23	0.94(0.83,1.06)	0.34	0.93(0.82,1.05)	0.26	0.91(0.80,1.04)	0.16
	Tertile 1	Ref		Ref		Ref		Ref
	Tertile 2	0.83(0.65,1.07)	0.15	0.94(0.70,1.26)	0.69	0.92(0.68,1.24)	0.58	0.95(0.70,1.29)	0.75
	Tertile 3	0.71(0.55,0.92)	0.01	0.79(0.58,1.09)	0.15	0.75(0.55,1.04)	0.08	0.75(0.54,1.04)	0.09
	P for trend		0.01		0.16		0.08		0.09
Red blood cell folate	Continuous (per SD+)	0.90(0.81,1.01)	0.07	0.92(0.81,1.04)	0.17	0.91(0.80,1.03)	0.14	0.86(0.75,0.97)	0.02
	Tertile 1	Ref		Ref		Ref		Ref
	Tertile 2	0.74(0.57,0.95)	0.02	0.79(0.59,1.06)	0.12	0.81(0.60,1.09)	0.16	0.81(0.59,1.09)	0.16
	Tertile 3	0.65(0.50,0.84)	< 0.001	0.68(0.49,0.92)	0.01	0.69(0.50,0.94)	0.02	0.62(0.44,0.85)	0.004
	P for trend		< 0.001		0.01		0.02		0.004

Model 0: No covariable was adjusted.

Model 1: Adjusted for age, sex, race, education attainment, marital status, and poverty-income ratio.

Model 2: Further adjusted for smoking, drinking status, BMI, and energy intake based on crude Model 1.

Model 3: Further adjusted for arthritis, thyroid problems, cancer, diabetes, liver diseases, CVD, and CKD based on Model 2.

SD standard deviation, CoI cognitive impairment, RBC red blood cell, CVD cardiovascular disease, CKD chronic kidney disease, OR odds ratio, CI confidence interval, ref reference.

Dose-response relationship between folate and CoI

As shown in Fig. 3, total dietary folate, dietary folic acid, serum folate, and RBC folate levels showed linear associations with CoI (P for nonlinearity > 0.05, respectively), except for dietary natural folate levels, which showed a U-shaped association with CoI (P for nonlinearity = 0.019).

Fig. 3.

Dose-response relationship between folate and CoI. The green line in the center represents the effect value, and the blue shaded area indicates the 95% CI. All models were adjusted for age, sex, race, education attainment, marital status, poverty-income ratio, smoking, drinking status, BMI, energy intake, arthritis, thyroid problems, cancer, diabetes, liver diseases, CVD and CKD. When fitting the RCS regression model, the respective median values were used as reference points and the three inflection points were selected at (10, 50 and 90th percentile) based on AIC minimum principle. The Wald Test was used to determine nonlinear P-values. CoI cognitive impairment, RBC red blood cell, CVD cardiovascular disease, CKD chronic kidney disease, RCS restricted cubic spline, AIC akaike information criterion, RBC red blood cell, OR odds ratio, CI confidence interval.

Subgroup analysis and effect modification tests of the relationship between RBC folate levels and CoI

Figure 4 presents the association between RBC folate levels and CoI in different populations. The stratified analysis revealed a significant interaction between CVD and RBC folate levels (P for interaction = 0.035); specifically, there was no significant association between RBC folate levels and CoI in individuals with CVD (OR 0.976, 95% CI 0.799–1.193, P = 0.816), whereas RBC folate levels reduced the odds of CoI in the non-CVD group (OR 0.784, 95% CI 0.659–0.933, P = 0.006). Figure 5 illustrates the dose-response relationship between RBC folate levels and CoI in populations with or without CVD. Specifically, there was a strong negative linear relationship between RBC folate levels and CoI in the non-CVD population (Fig. 5A), whereas this association was not significant in the CVD population (Fig. 5B).

Fig. 4.

Subgroup analysis and effect modification test of RBC folate with the odds of CoI models were adjusted for all covariables other than stratification variables. The significance of the interaction effect was determined by likelihood ratio test. RBC red blood cell, PIR poverty-income ratio, BMI body mass index, CVD cardiovascular disease, CKD chronic kidney disease, OR odds ratio, CI confidence interval, P int P for interaction.

Fig. 5.

Dose-response relationship of RBC folate with CoI among participants without (A) or with (B) CVD. The green line in the center represents the effect value, and the blue shaded area indicates the 95% CI. All models were adjusted for age, sex, race, education attainment, marital status, poverty-income ratio, smoking, drinking status, BMI, energy intake, arthritis, thyroid problems, cancer, diabetes, liver diseases, and CKD. When fitting the RCS regression model, the respective median values were used as reference points and the three inflection points were selected at (10, 50 and 90th percentile) based on AIC minimum principle. The Wald Test was used to determine nonlinear P-values. CoI cognitive impairment, RBC red blood cell, CVD cardiovascular disease, CKD chronic kidney disease, RCS restricted cubic spline, AIC akaike information criterion, RBC red blood cell, OR odds ratio, CI confidence interval.

Sensitivity analysis

Models incorporating multiple imputation for missing covariables, redefining the CoI, or additional adjustment for serum vitamin B12, dietary vitamin B6, and dietary choline further confirmed the robust inverse association between folate levels and CoI (Supplementary Table 1). Moreover, all four sensitivities supported a significant interaction between CVD and RBC folate levels (Supplementary Fig. 2). Additionally, consistent with the main findings, RCS showed a reverse dose-response relationship between RBC folate levels and CoI in non-CVD participants in all four sensitivity analysis models, whereas there was no association between RBC folate levels and CoI in participants with CVD (Supplementary Fig. 3).

Discussion

In this cross-sectional study, we systematically evaluated the relationship between dietary folate, serum folate, and RBC folate levels with CoI. Our results suggest a significant negative linear correlation between RBC folate and CoI, whereas no statistically significant relationship between serum folate levels and CoI was found. Based on the robust association between RBC folate levels and CoI and the advantage of RBC folate levels in assessing long-term folate status, subgroup analyses and interaction tests were conducted³⁰. The results indicated that the association between RBC folate levels and CoI remains consistent across various strata; however, there was a significant interaction between CVD and RBC folate levels, which attenuated the protective effect of RBC folate on CoI. Finally, further support for our findings was provided through multiple imputation of missing covariables, redefinition of CoI, or additional adjustment of the serum vitamin B12 concentration in the model.

In this study, cognitive impairment (CoI) was evaluated with the following three items. (1) Word learning and recall modules from the Consortium to Establish a Registry for Alzheimer’s Disease (CERAD); (2) The Animal Fluency test (AFT); and (3) The Digit Symbol Substitution test (DSST). Zhang et al.²¹ analysed the NHANES database and reported a U-shaped association between total dietary folate levels and CERAD, an inverse dose-response association with AFT, and an overall negative association with DSST. A meta-analysis also revealed that increased dietary intake of folate (rather than B12 or B6) was associated with a reduced risk of dementia in populations without dementia⁵⁵. Cognitive function is a comprehensive concept, and our study shows that both trend tests and the RCS support the tendency of total folate intake to linearly decrease the odds of overall cognition. Additionally, in the present study, a low odds of CoI was observed in tertiles 2–3 when folic acid was used. A study focusing on Americans aged ≥ 80 years reported that after folate fortification, low RBC folate status was rare⁵⁶. These findings support a stronger association of folic acid and RBC folate levels with CoI in our study. Serum folate levels may be a more critical factor during clinical diagnosis, as they provide a rapid reflection of recent folate intake. Ding et al.²⁹ analysed the NHANES database and reported a reverse U-shaped association between serum folate levels and CERAD, AFT, and DSST results. However, unfortunately, in the current study, multiple models did not reveal an association between serum folate levels and CoI. The reason for this difference is the different definitions of the outcome. CoI reflects the comprehensive results of CERAD, AFT, and DSST, and it may be more reasonable to consider these different dimensions of assessment together. Interestingly, why RBC folate and serum folate levels show different associations with CoI may be explained by the higher serum folate levels in the NHANES. According to the World Health Organization definition, serum folate deficiency is considered when folate levels < 4.4 ng/mL (< 10 nmol/L)⁵⁷. In this study, among 2104 participants, only 14 met the diagnostic criteria for serum folate deficiency; this may be because our study population generally did not have low serum folate levels. This finding is reasonable, as Rotstein et al.⁵⁸ previously reported that individuals with serum folate deficiency (< 4.4 ng/ml) presented a greater risk of dementia and all-cause mortality than individuals without serum folate deficiency.

Previous studies have shown that folate concentrations in patients with type 2 diabetes is significantly lower than those in healthy individuals⁵⁹. In frail elderly patients with hypertension, the impact of hyperglycaemia on cognitive impairment is an important research area. Multiple studies have demonstrated that there is a significant correlation between hyperglycaemia and cognitive decline, and this impact is even more pronounced in the elderly population. Hyperglycaemia affects cognitive function through multiple mechanisms. First, hyperglycaemia can damage cerebral microvascular endothelial cells, leading to increased cell permeability and reactive oxygen species production, thereby exacerbating brain damage⁶⁰. In addition, hyperglycaemia can trigger the formation of advanced glycation end products, which can cause deformities and stiffness of protein fibres and subsequently affect neuron function⁶¹. These pathological changes may promote the development of Alzheimer’s disease and other forms of dementia. The sodium-glucose cotransporter 2 (SGLT2) inhibitor empagliflozin has a beneficial effect on cognitive and physical impairments in frail elderly patients with diabetes and hypertension, highlighting how SGLT2 inhibitors can alleviate mitochondrial oxidative stress in human endothelial cells⁶². Moreover, several researchers have confirmed the efficacy and safety of SGLT2 inhibitors in frail elderly individuals^63–66, which highlights the substantial absolute benefits of treatment in these vulnerable patients, who are often unjustly denied corresponding treatments^67,68.

Multiple studies have shown that CVD may lead to the occurrence of CoI. Coronary heart disease causes accelerated cognitive decline^69,70. Atrial fibrillation is the most common type of arrhythmia and can significantly increase CoI risk^71,72. Even if patients with atrial fibrillation do not suffer from stroke, they exhibit significant declines in cognitive function⁷³. Furthermore, research has shown that the prevalence of cognitive impairment in heart failure patients is 25–75%⁷⁴. In our study, CVD attenuated the protective effect of RBC folate on CoI. Studies have shown that high folate levels may exacerbate CVD progression. Xu et al.³³ included 14,234 participants with a history of CVD in NHANES and reported that high RBC folate levels increased the CVD and all-cause mortality rates in this population. Liu et al.³⁴ investigated 7700 adult participants in the United States and found that high serum folate levels also increased the risk of CVD-related death in patients with type 2 diabetes. In addition, as early as 2009, a randomized controlled trial in Norway revealed that folate plus vitamin B12 therapy was associated with better cancer outcomes and all-cause mortality in ischaemic heart disease patients even when food was not fortified with folic acid⁷⁵. Two prospective cohort studies from Denmark and Iowa, USA, suggested that folate supplementation increased the risk of death^76,77. A multicentre clinical cohort study revealed that vitamin B12 and folate levels were not associated with the prognosis of heart failure patients⁷⁸. This evidence suggests that high RBC folate levels may mediate the progression of CVD and more severe symptoms; moreover, the effect of folate is greatly reduced in CVD patients. Importantly, in the NHANES database representing the U.S. population, more than half of American adults at high risk of CVD have already reached the recommended daily intake of dietary folate equivalents (DFE); however, approximately one quarter of these individuals still take additional folate supplements, which may result in folate oversaturation³³. Elevated RBC folate levels often occur in populations who use supplements for a long time, and in combination with the findings of this study, it is possible that as folate levels increase in populations with a history of CVD, worse cardiovascular conditions may occur; poor cardiovascular conditions lead to worse cognitive levels, thereby masking the protective effect of folate on CoI.

Hyperhomocysteinaemia (HHcy) is a pathological condition caused by abnormal folate metabolism and is usually related to methylenetetrahydrofolate reductase (MTHFR) gene polymorphism, especially the C677T mutation. HHcy promotes vascular endothelial injury through oxidative stress, accelerates the formation of atherosclerotic plaques, exacerbates arteriosclerosis, and acts synergistically with hypertension to increase the likelihood of stroke⁷⁹. Intima-media thickness is an important indicator in carotid ultrasound examination and is used to assess the degree of atherosclerosis. HHcy can participate in the occurrence and development of carotid atherosclerosis through a series of pathological mechanisms, such as damaging vascular endothelial cells, increasing vascular smooth muscle, and promoting fibrinogen production, leading to abnormal blood coagulation, platelet aggregation, and thrombosis formation and promoting chronic inflammatory reactions⁸⁰. HHcy has also been found to be a potential independent risk factor for cerebral small vessel disease. It damages the structure and function of the vascular endothelium by increasing proinflammatory cytokine levels and reducing anti-inflammatory cytokine levels, increases vascular permeability, destroys the blood-brain barrier, and ultimately leads to ischaemic and hypoxic damage to white matter and grey matter^81,82. To some extent, this mechanism explains the modifying effect of CVD on the association between RBC folate and CoI.

Advantages and limitations

This study comprehensively explored the association between dietary, serum, and RBC folate levels and CoI in the adult population of the United States. Through interaction analysis, we discovered the potential modifying effect of CVD on RBC folate, providing insights into the potential links between clinical CVD, folate supplementation, and CoI. We comprehensively considered multiple confounding factors and drew relatively robust conclusions from three sensitivity analyses.

However, there are some limitations to consider. Firstly, due to sample size limitations, we were unable to investigate the interaction between CVD subcategories and RBC folate levels, thus providing limited guidance for clinical folate supplementation and testing. To address this issue, we opted to explore the interaction effects among stroke, congestive heart failure, and coronary heart disease when the sample sizes were relatively sufficient. As shown in Supplementary Table 2, the interaction effect was significant in the stroke subgroup. Notably, although the interaction effects were not significant, the association between folate levels and CoI was weaker in populations with congestive heart failure or coronary heart disease than in healthy controls. Second, in the initial study, we attempted to consider the sampling weights of the NHANES; however, weighted logistic regression models showed severe multicollinearity (VIF > 100) among variables. Furthermore, even after feature variable selection through least absolute shrinkage and selection operator regression, multicollinearity persisted despite reweighting; thus, the conclusions of this study represent only the 2104 participants and cannot be generalized to the entire U.S. population. Third, single measurements of folate levels may introduce bias. Fourth, potential confounding factors were inevitably overlooked. Fifth, we must acknowledge the problem of selection bias due to CVD severity. Patients with more severe CVD may have already passed away, which biased our included study subjects towards those with milder CVD. This selection bias may have affected our accurate assessment of the effect of CVD on suppressing the protective effect of RBC folate on cognitive function. In the future, it is necessary to expand the scope and diversity of the research sample and attempt to enrol research subjects from more sources, including CVD patients with different disease severity. We can cooperate with more medical institutions or use a multicentre research approach to increase the possibility of including patients with more severe conditions. Finally, cross-sectional studies can only observe the associations between variables at a specific time point and cannot determine causality. Theoretically, cognitive impairment may lead to low folate levels. For example, patients with cognitive impairment may have an unbalanced diet, resulting in insufficient folate intake. Patients with cognitive impairment may also have metabolic disorders that affect the absorption, utilization, and metabolism of folate. In addition, patients with cognitive impairment may reduce outdoor activities and social activities due to the influence of the disease, thus affecting folate synthesis. To determine the direction of the causal path, more research is needed. A longitudinal study design can be used to follow study subjects and observe trends of change in cognitive impairment and folate levels to determine whether there is a causal relationship.

Conclusion

In summary, our study indicates a robust inverse association between RBC folate levels and overall cognitive function, which is stronger than that of several other folate indicators. Importantly, CVD attenuates the protective effect of RBC folate on CoI, highlighting the need to further explore the underlying mechanisms in future research.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Abbreviations

AFT

Animal fluency test

AIC

Akaike information criterion

BMI

Body mass index

CERAD

Consortium to establish a registry for Alzheimer’s disease

CI

Confidence interval

CIs

Confidence intervals

CKD

Chronic kidney disease

CoI

Cognitive impairment

CVD

Cardiovascular diseases

DFE

Dietary folate equivalents

DRT

Delayed recall test

DSST

Digit symbol substitution test

HHcy

Hyperhomocysteinemia

IQR

Interquartile range

IR

Insulin resistance

IRT

Immediate recall test

MEC

Mobile examination center

MTHFR

Methylenetetrahydrofolate reductase

NHANES

National health and nutrition examination survey

OR

Odds ratio

ORs

Odds ratios

RBC

Red blood cell

RCS

Restricted cubic spline

SD

Standard deviation

SGLT2

Sodium-glucose cotransporter 2

VIF

Variance inflation factor

χ²

Chi-square

Author contributions

Jianqiang Zhang: design the paper, methodology, and writing. Di Yue: statistics, graphing, data cleaning, and manuscript polishing. Huifeng Zhang: funding and designed the paper. All authors have read and agreed to the published the manuscript.

Data availability

Data used for this study are available on the NHANES website: https://wwwn.cdc.gov/nchs/nhanes/.

Declarations

Competing interests

The authors declare no competing interests.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-025-87129-x.