Non-linear association between daily dietary folate intake and cognitive performance in elder America: A population-based cross-sectional study

The association between daily dietary folate intake and cognitive performance in older populations is unclear. This study aimed to investigate whether a dose-dependent association exists between dietary folate levels and cognition performance. The 2011–2014 datasets from the National Health and Nutrition Examination Survey were collected. Data from adults aged ≥ 60 years who completed all three cognitive tests with daily dietary data were analyzed. Weighted smooth curve tting and multiple linear regression models were applied to study the association between folate intake and cognitive performance. Age, sex, race, education, smoking status, alcohol intake, body mass index, hypertension, stroke, diabetes, and depression were adjusted using a weighted multiple regression model. associated with delayed recall (per mcg/day increment, β 0.003, 95% CI -0.002–0.009, P = 0.23). Below the cut-off of 250 mcg/day, dietary folate intake was positively associated with immediate recall (per 10 mcg/day increment, β 0.01, 95% CI 0.00–0.02, P = 0.05) and animal uency test score (per 10 mcg/day increment, β 0.02, 95% CI 0.00–0.03, P = 0.05). However, above this cut-off, the association was not signicant. Dietary folate intake was not signicantly associated with an increase in the digit symbol substitution test score until dietary the folate intake reached 250 mcg/d (per 10mcg/d increment, β 0.29, 95% CI 0.05-0.53, P = 0.04).


Introduction
We are entering an aging era worldwide. It is estimated that, by 2030, people aged >65 years would accumulate to 71 million, with 35 million older people in 2000 in America [1]. Age-associated cognitive decline is common in the older population and affects the quality of life [2]. Nutritional conditions are associated with degenerative cognitive impairment [3,4]. As a feasible monitoring and modifying lifestyle, understanding the association between dietary nutrients and cognition performance could help improve the quality of life of the older population.
As a cofactor in one-carbon metabolism, folate acid may be involved in cognition-related mechanisms. A lack of folate intake causes detrimental gene expression alterations in the hippocampus of mice [5] and memory impairment. In rats with Alzheimer's disease (AD), folate de ciency has been related to memory impairment, with impaired gut microbiota and hippocampal insulin signaling [6]. Dietary folate may play a role in amyloid-β protein (Aβ) metabolism; Aβ is a biomarker of cognitive impairment. Tian et al. [7] found that folate supplementation decreased Aβ deposition by downregulating the expression of βsecretase (β-site amyloid precursor protein-cleaving enzyme 1) and γ-secretase complex catalytic component (presenilin 1) in mice. In vitro, folic supplementation was found to decrease Aβ deposition via stimulation of DNA methyltransferase (DNMT) activity [8]. Folate supplementation also showed dosedependent stimulation of DNMT activity in AD transgenic mice [9]. High serum homocysteine (HHCY) level is commonly secondary to folate de ciency. This secondary metabolic alteration has also been reported to increase tau protein phosphorylation and enhance tau oligomerization and aggregation, leading to AD progression in mice [10].
However, the association between dietary folate intake and cognitive performance remains unclear [11,12]. Evidence on the dose-dependent association between dietary folate and cognitive performance is relatively limited. Thus, in this study, we collected population-based data of older Americans to investigate whether a dose-dependent association exists between dietary folate levels and cognitive performance.

Data collection
The National Health and Nutrition Examination Survey (NHANES) has been a continuous program since 1999, conducted by the National Center for Health Statistics (NCHS). It has a complex, multistage, probability cross-sectional design. 5000 nationally representative American residents are screened each year. Demographic, dietary, physical, health-related, and laboratory information are released in a two-year cycle. The NHANES datasets are available at https://www.cdc.gov/nchs/nhanes/Default.aspx. All participants provided written informed consent. The Research Ethics Review Board of the National Center for Health Statistics approved the NHANES protocol (Protocol #2011-17).
We collected data from the 2011-2012 and 2013-2014 NHANES survey cycles. Trained medical professionals interviewed participants aged ≥ 60 years for cognitive assessment. We collected data from the participants who fully completed three cognition tests. Participants with available dietary information were also collected, as shown in Figure 1.

Daily dietary total folate intake
The Agriculture Department of the U.S. and the Department of Health and Human Services collaborated to obtain dietary information among all participants of the NHANES. Dietary data were collected using the automated multiple pass method, a ve-step interview. This computerized method helps provide precise dietary intake information in a large survey. The 24-h (before the interview) dietary intake recall was collected. Two datasets of daily dietary information were collected. The rst dataset was collected through in-person interviews at the Mobile Examination Center. The second was collected by telephone 3-10 days after the rst visit. The average total folate intake of the two sets was applied in this study.

Cognition performance
Participants aged ≥ 60 years who understood English, Spanish, Chinese, Korean, or Vietnamese were eligible for cognition tests. Participants requiring proxy formants were not eligible. The 2011-2012 and 2013-2014 NHANES cycle applied Consortium to Establish a Registry for Alzheimer's Disease (CERAD), animal uency test (AFT), and digit symbol substitution test (DSST) to assess cognitive performance.
The reliability of these tests has been validated in Americans [13][14][15]. The CERAD evaluates both immediate and delayed word learning capacity, a domain of memory [14]. Three consecutive tests required the recall of ten words per minute. One word was counted for each score. The CERAD-immediate recall score is a summary of these tests. A delayed recall of words was tested after the AFT and DSST or 8-10 min after beginning. Naming as many animals as the participants can in 1 min was performed in the AFT [16]. Each animal named was given a score of one. The AFT assesses categorical verbal uency, an aspect of executive function. The DSST is a part of the Wechsler Adult Intelligence Scale [17]. Participants were required to match the ruled number and symbol from 133 boxes. Each right match was given a score of one. Participants were excluded if they could not complete each task independently. This test evaluated processing speed, sustained attention, and working memory.

Covariates
Age, sex, race, and education were collected from demographic datasets. Participants aged ≥ 80 years were recorded as 80 years old. Education was classi ed as less than 9th grade, 9th-11th grade (including 12th grade with no diploma), high school graduate or equivalent, college degree, college graduate or above. Smoking status was classi ed as never ("never smoked or smoked <100 cigarettes in life"), previous ("smoked ≥100 cigarettes in life and currently no longer smoking"), and current smoker ("smoked ≥100 cigarettes in life and currently smoking"). Demographic information was collected at home. Trained interviewers used the computer-assisted personal interview system in the NHANES. Participants could use English, Spanish, or an interpreter in this interview. Body mass index (BMI) was calculated as the sum of weight(kg)/square of height(m). Underweight referred to BMI <18.9 kg/m 2 ; normal weight, to BMI ≥18.9 and <25 kg/m 2 ; overweight, to BMI≥25 and <30 kg/m 2 ; and obese, to BMI≥30 kg/m 2 . Daily alcohol intake was the average of the rst and second total daily nutrient intake.
Modeled on the U.S. National Health Interview Survey, disease history was obtained from self-reported interviews. If participants reported a diagnosis of hypertension, stroke, or diabetes, the history of disease was recorded as yes. If a participant's health questionnaire-9 (PHQ-9) score >5, depression was recorded as yes [18].

Statistical analysis
Continuous variables are presented as mean ± standard error (SE), while categorical variables are presented as percentages. Among quartile subgroups of daily folate intake, Kruskal-Wallis rank-sum tests and chi-square tests were used for continuous variables and categorical variables, respectively. Regression analyses were performed to study the association between daily dietary folate intake and cognitive function. Daily dietary folate intake was counted as per 10 mcg/day and split into quartiles in regression models. Smooth tting analyses were then applied to study the non-linear relationship between dietary intake and cognitive function. We further used 250 µg/day as the threshold to render the dietary folate intake dichotomous to study non-linear associations in stepwise regression models. Age, sex, race, education, smoking status, alcohol intake, BMI, and the presence of hypertension, stroke, diabetes, and depression were adjusted in regression models, smooth tting analyses, and stepwise regression models.
Weighted clinical characteristics are presented as mean ± standard error (SE) for continuous variables and as percentages for categorical variables according to different survey cycles. They were used to assess differences in clinical characteristics.
All analyses were weighted according to the NHANES analytical guidelines. Statistical signi cance was set at a P-value < 0.05. R software (http://www.R-project.org, The R Foundation) and Free Statistics Version 1.3 were used for analysis.

Results
Weighted clinical characteristics of study participants A total of 199,931 participants were surveyed in the 2011-2012 and 2013-2014 NHANES cycles. A total of 16,299 participants were not eligible for the cognition function tests. A total of 509 participants did not complete all three cognitive tests. A total of 1,766 participants did not complete the 1st day dietary interview, while 1,868 did not ful ll that of the 2nd day. In total, 1,255 participants were nally included in this study, as shown in Figure 1.
As presented in Table 1, the weighted average age of the study participants was 69.3 years. Males accounted for 48.8% of the total study population. Sex, race, education, smoking status, BMI, hypertension, and high serum folate levels were signi cantly different among persons classi ed into the different quartiles of daily folate intake. The lowest dietary folate intake was <240.0 mcg/day, the 2nd was ≥240.0 and <341.0 mcg/day, the 3rd was ≥341.0 and <472.5 mcg/day, and the highest was ≥472.5 mcg/day.

Weighted linear association between daily dietary folate intake and cognitive performance in old Americans
In unadjusted weighted linear regression models, when daily dietary folate intake was treated as a continuous variable, every 10 mcg increment was associated with a 0.023 increase in the CERAD immediate recall score (95% CI 0.010-0.036, P <0.01), 0.043 increase in AFT score (95% CI 0.025-0.060, P < 0.01), and 0.123 increase in DSST score (95% CI 0.056-0.189, P < 0.01). However, daily dietary folate intake was not signi cantly associated with the CERAD delayed recall score (β 0.007, 95% CI -0.001-0.014, P = 0.08). In multiple weighted linear regression models, each 10 mcg increment was associated with a 0.015 increase in the CERAD immediate recall score (95% CI 0.007-0.023, P = 0.02) and a 0.017 increase in AFT score (95% CI 0.006-0.028, P = 0.03 intake in the adjusted model existed in CERAD immediate recall (P for trend 0.01), AFT (P for trend <0.01), and DSST (P for trend 0.03) but not in CERAD delayed recall (P for trend 0.57) ( Table 2). Weighted smooth curve tting of the association between dietary folate intake and cognitive performance in elderly Americans We used a weighted smooth curve-tting method to study the association between daily dietary folate intake and cognitive performance. Age, sex, race, education, smoking status, alcohol intake, BMI, hypertension, stroke, diabetes, and depression were adjusted for. As shown in Figures 2A, 2C, and 2D, the association between dietary folate intake and CERAD immediate recall, AFT, and DSST was not linear.
However, as shown in Figure 2B, the association between daily dietary folate intake and CERAD delayed recall tended to be linear.
Two-piece logistic regression of the association between dietary folate intake and cognitive performance in elderly Americans We further conducted a two-piece analysis to measure the non-linear association between daily dietary folate intake and cognitive function. As presented in Table 3, in daily dietary folate intake <250 mcg/day, 10 mcg increase of folate was independently associated with 0.29 (95%CI, 0.05-0.53, P = 0.04) increment of DSST score. However, when daily dietary folate intake was >250 g/day, DSST was not signi cantly associated with folate intake (β, 0.00; 95% CI, -0.05-0.05; P = 0.98). Nevertheless, every 10 mcg increase in dietary folate intake was associated with a 0.01 increment in CERAD immediate recall (95%CI, 0.00-0.02; P = 0.05) and 0.02 of AFT (95%CI, 0.00-0.03; P = 0.05).

Discussion
In a representative older American population, daily folate intake was nonlinearly associated with cognitive performance. CERAD immediate recall and AFT scores were signi cantly associated with daily folate intake, but not until folate intake was >250 µg/day. However, DSST was signi cantly associated with folate intake until folate intake was <250 g/day. The CERAD-delayed recall was not signi cantly associated with dietary folate intake.
Dietary pattern enrichment in folate was reported to have a positive association with cognitive function.
In a Chinese cohort, participants aged ≥ 60 years with the highest quartile of vegetarian diet had a 28% decrease in cognition in mild cognitive (MCI) risk, compared with those in the lowest quartile [19].
Nutraceutical intervention rich in folate was found to improve cognitive performance in patients with AD and MCI [20,21]. A community-based cohort in America also reported that a low intake combination of multiple vitamins, including folate, was associated with larger total brain volume and better cognition functions [22]. Hence, the question of whether dietary folate alone was independently associated with cognitive performance was raised. However, our study was conducted using population-based datasets, which may provide some information on public cognition health management. We further performed a dose-dependent analysis and found a non-linear association between dietary folate and cognition performance.
An adverse association between dietary folate and cognitive performance has also been reported. A Japanese cross-sectional study found that dietary folate intake was not associated with cognitive function measured by Raven's Colored Progressive Matrices in school children [26]. A prospective cohort enrolling a population of older adults with a sample size of 155 also found that dietary folate intake was not related to cognitive decline measured by the MMSE (OR 2.55, 95% CI 0.78-8.41) [27]. This difference may be due to different age groups and sample sizes.
Our study had several limitations. First, due to temporality bias, this cross-sectional survey could not conclude a causal association between dietary folate intake and cognitive performance. Second, dietary folate intake was assessed based on a 24-h recall, which suffers from intrinsic methodology limitations [17]. However, we averaged the 2-day dietary recall in this study. Third, serum homocysteine may modify the association between dietary folate and cognitive function [28,29]. However, in the 2011-2014 NHANES cycle, serum homocysteine levels were not examined. Thus, subgroup analyses for different serum homocysteine levels could not be conducted. Further well-designed RCTs and prospective studies are needed to validate our ndings.

Conclusion
In an American population aged ≥ 60 years, dietary folate was not signi cantly associated with CERAD delayed recall. Dietary folate intake was not linearly associated with immediate CERAD recall, AFT, and DSST scores. Below the cut-off of 250 mcg/day, dietary folate intake was positively associated with immediate recall and AFT. However, this association was not signi cant. Furthermore, dietary folate intake was not signi cantly associated with the DSST score increase until dietary folate intake reached 250 mcg/day.  Weighted smooth curve tting of association between dietary folate intake and (A) CERAD-immediate recall score, cognitive performance. Weighted distribution of daily dietary folate intake was also presented. (B) CERAD-delayed recall score, (C) AFT score, and (D) DSST score. Age, sex, race, education, smoking status, alcohol intake, BMI, hypertension, stroke, diabetes, and depression were adjusted for weighted smooth curve tting. Red line represented tted association between folate intake and cognition