The National Health and Nutrition Examination Survey (NHANES) is a cross-sectional survey designed to assess the health and nutritional status of a nationally representative sample of the US civilian population.¹ The survey includes questionnaire interviews, laboratory data and physiological examinations. Considering the availability of RBC folate and PhenoAge data, this study utilized data from two NHANES cycles (2007–2010), including 12,153 participants aged 20 years and older. Participants with missing data for RBC folate (N = 1,127) and those lacking biomarkers for the PhenoAgeAccel algorithm (N = 164) were excluded. The final sample size was 8,944 individuals, following the exclusion of participants lacking other covariates (N = 1,918) (Figure 1).

RBC folate is a classical biomarker of folate status. The European Food Safety Authority (EFSA) considers RBC folate concentration to be the most reliable indicator of folate status (18). Whole blood and serum specimens were collected and then stored at ≤ −20°C until transported to the National Center for Environmental Health for analysis (specimens should be frozen at −70°C for long-term storage). RBC folate concentrations were measured using the microbiologic assay (MA) (19).

PhenoAge and PhenoAgeAccel are specifically quantifiable and observable aging indicators developed based on NHANES data to estimate individuals at high risk for multiple diseases and all-cause and disease-specific mortality. The PhenoAge algorithm was developed on chronological age and nine biomarkers from the NHANES III dataset: albumin, creatinine, glucose, C-reactive protein, white blood cell count, lymphocyte percent, red blood cell distribution width, mean cell volume and alkaline phosphatase (7). PhenoAgeAccel was calculated as a residual from a linear regression of PhenoAge against chronological age, a negative PhenoAgeAccel value represents a person who is physiologically younger than their chronological age, while a positive PhenoAgeAccel value indicates a person who is physiologically older (20). The detailed calculations of PhenoAge and PhenoAgeAccel are presented in Supplementary Methods 1.

NHANES collected the following covariates through standardized questionnaires. Sociodemographic factors including age, sex, race (Mexican American, other Hispanic, non-Hispanic White, non-Hispanic Black, other race), education level (< high school, high school, > high school), marital status (married/living with partner, widowed/divorced/separated, never married), ratio of family income to poverty (PIR). The PIR reflects a household’s income relative to federally defined poverty thresholds, calculated by dividing total income by guidelines adjusted for family size, state, and year. Smoking status was categorized as current smokers, formerly smoked, and never smoked. Body Mass Index (BMI) was calculated as measured weight (kg) divided by height squared (m²) and categorized by 25 and 30 kg/m². Alcohol consumption was categorized based on whether participants consumed at least 12 alcoholic drinks per year. Physical activity was classified as inactive (participants who did not engage in either vigorous or moderate physical activity), moderate (activity done for at least 10 min that caused only light sweating or a slight to moderate increase in breathing or heart rate), or vigorous (activity done for at least 10 min in the past 30 days that caused heavy sweating or large increases in breathing or heart rate). The health condition data were composed of hypertension, diabetes, cardiovascular disease and cancer. Hypertension was defined as a prior diagnosis of hypertension, current use of antihypertensive medications, or having a systolic blood pressure level of ≥ 130 mmHg and/or a diastolic blood pressure level of ≥ 80 mmHg (21). Diabetes was defined as a prior diagnosis of diabetes, current use of insulin or diabetes pills, fasting plasma glucose level ≥ 126 mg/dL, or a hemoglobin A1c level ≥ 6.5% (22). Cardiovascular disease and cancer were assessed by self-reported questionnaires.

Based on the NHANES analytic guidelines, appropriate sampling weights were applied to interpret the complexity of survey design in NHANES database during our analysis. The baseline characteristics of the participants included were described using weighted means for the continuous variables or proportions for the categorical variables. One-way ANOVA or Kruskal–Wallis and the χ2 test were conducted to compare the differences between groups. A multivariate linear regression model was used to assess the linear relationship between RBC folate levels and PhenoAgeAccel. Three models were constructed. Model 1 did not adjust for any covariates, while Model 2 was adjusted for age, gender and race. Model 3 further adjusted for education, marital status, poverty income ratio, BMI, alcohol, smoking, physical activity, hypertension, diabetes, cardiovascular disease, and cancer. In addition, generalized additive models and smoothed curve fits were used to examine non-linear relationship between RBC folate and PhenoAgeAccel. Recursive algorithms and two-stage logistic models were utilized to detect any potential inflection points in the relationship. Moreover, subgroup analyses and interaction tests were conducted to explore whether the associations differed across subgroups defined by age, gender, BMI, hypertension, diabetes, cancer, and cardiovascular disease (23). All the analysis were performed with R (version 4.3.1)² and EmpowerStats (version 4.2).³ P < 0.05 was considered statistically significant.

The baseline characteristics of the studied variables across tertiles of RBC folate are presented in Table 1. Our analytical sample included 8,944 participants aged ≥ 20 years. The mean age was 46.85 years, and 50.93% of participants were female. Participants with high RBC folate levels were more likely to be female, non-Hispanic White, have a higher level of education, be married or living with a partner, consume higher amounts of alcohol, never smoked and have lower PhenoAgeAccel value. Additionally, compared to the low RBC folate level group, participants with the high RBC folate levels suffer from hypertension, diabetes, cardiovascular disease and cancer.

The associations between RBC folate and PhenoAgeAccel are presented in Table 2. The results indicated that RBC folate had no significant correlation with PhenoAgeAccel in the non-adjusted model (β: 0.0008, 95% CI: −0.0008, 0.0025), the partially adjusted model (β: 0.000895, 95% CI: −0.0009, 0.0026), and the fully adjusted model (β: 0.0005, 95% CI: −0.0012, 0.0022). After RBC folate was classified into tertiles, in the fully adjusted models, participants in the moderate RBC folate level group were found to be 0.7299 years younger (β: −0.7299, 95% CI: −1.2882, −0.2317) than the low RBC folate level group. However, there was no significant change in PhenoAgeAccel in the high RBC folate level group compared to the low RBC folate level group (β: −0.5898, 95% CI: −1.3481, −0.1686). There was no statistically significant trend between RBC folate levels and PhenoAgeAccel (P for trend > 0.05).

The smooth curve fits model demonstrated a non-linear relationship between RBC folate and PhenoAgeAccel (Figure 2). RBC folate levels exhibited a U-shaped dose-response relationship with PhenoAgeAccel. Threshold effect analysis showed that the inflection point of RBC folate was observed at 732.9 ng/mL (P for likelihood ratio test < 0.001). The PhenoAgeAccel decreased by 0.0027 years per 1 ng/mL increase in RBC folate when RBC folate ≤ 732.9 ng/mL (β: −0.0027, 95% CI: −0.0051, −0.0002), and increased by 0.0058 years per 1 ng/mL increase in RBC folate when RBC folate > 732.9 ng/mL (β: 0.0058, 95% CI: 0.0026, 0.0090) (Table 3).

There was almost no significant difference suggested by the interaction test (P for interaction > 0.05) in the association of RBC folate and PhenoAgeAccel among different subgroups (Figure 3). However, an exception was observed in the subgroup of patients with cardiovascular diseases, where a significant positive correlation was identified between RBC folate and PhenoAgeAccel (β: 0.0027, 95% CI: 0.0003, 0.0050) (P for interaction = 0.0111) (Supplementary Figure 1).