NHANES is an ongoing cross-sectional multistage, stratified program of the general US civilians. It is conducted by the National Center for Health Statistics to assess the health and nutritional status. To obtain a representative sample, non-Hispanic Asians, Hispanics, non-Hispanic Black, older adults, and low-income white individuals are oversampled. The details of the NHANES have been described in the previous study (27). NHANES program is approved by the Centers for Disease Control and Prevention Research Ethics Review Board (28). Documented signed consents are obtained from all adult individuals.

Because serum NfL was measured only in the 2013–2014 cycle of NHANES, this dataset was used for subsequent analyses. A total of 10,175 individuals participated in NHANES during 2013–2014. Among them, 5,769 individuals were 20 years of age and older. Participants without data on NfL (n = 3,698) and those with incomplete or unreliable dietary interview data (n = 380) were excluded. Additionally, pregnant individuals (n = 14), cancer patients (n = 136), and those whose energy intake exceeded ±3 SDs of the mean value of the log-transformed total energy intake (n = 8) were excluded, as these conditions could severely affect dietary intake. Finally, 1,533 participants with a median (IQR) age of 46 (33–59) years were included in the following analyses (Figure 1).

In our study, the regression model in this study included 14 variables. According to empirical rules, it was recommended that each predictor variable in a regression model should be supported by at least 20 observations to ensure reliable and stable estimates. Therefore, the minimum sample size required was 280. Our study included a total of 1,533 participants from the NHANES dataset, which significantly exceeded the required sample size of 280 participants. This large sample size provided ample power to detect meaningful relationships between the variables included in our model.

Blood samples were obtained via venipuncture and collected into blood collection tubes specifically designed to separate serum from other blood products. After collection, the serum was aliquoted into cryogenic vials and stored in a −70°C freezer according to test requirements. Participants aged 20–75 years in a half-sample from NHANES 2013–2014, who had stored surplus or pristine serum samples and consent to future research, were eligible.

In this study, a highly sensitive NfL immunoassay developed by Siemens Healthineers was utilized. This assay leveraged acridinium ester (AE) chemiluminescence and paramagnetic particles and was executed on the automated Atellica immunoassay system. The detailed procedure was as follows: initially, the sample was incubated with AE-labeled antibodies that specifically bound to the NfL antigen. After this binding phase, paramagnetic particles coated with capture antibodies were added to the sample, forming complexes with the AE-labeled antibody–antigen conjugates. Subsequently, the mixture underwent a separation process to remove any unbound AE-labeled antibodies. Following this, acid and base reagents were added to initiate the chemiluminescence reaction, and the resultant light emission was measured. The intensity of the emitted light correlated with the concentration of NfL in the sample.

The lower limit of quantification (LLOQ) of the NfL immunoassay was 3.9 pg./mL, which was determined by replicate testing of low-concentration samples (n = 44). The LLOQ was defined as the concentration at which the coefficient of variation (CV) was not more than 20%. The upper limit of quantification (ULOQ) was 500 pg./mL. For data below the LLOQ, an imputed value was equal to the LLOQ divided by the √2. For data above the ULOQ, an imputed value was equal to the ULOQ multiplied by the √2. Twenty-seven participants had values below the LLOQ. No participants had values above ULOQ.

All participants were eligible for two 24-h dietary recall interviews. The first dietary recall interview was collected in-person in the Mobile Examination Center (MEC) and the second interview was collected by telephone 3–10 days later.

The dietary intake data were used to estimate the types and amounts of foods and beverages consumed during the 24 h before the interview and to estimate intakes of energy, nutrients, and other food components according to the US Department of Agriculture Food and Nutrient Database (29). This database included comprehensive information that was used to code individual foods/beverages and portion sizes reported by participants and included nutrient values for calculating nutrient intakes. Approximately 10 percent of the coder’s work was randomly selected to be independently coded by another coder. Information regarding the consumption of dietary supplements was gathered through a 30-day dietary supplement survey. The mean quantity of vitamin K supplements consumed was ascertained by aggregating the total vitamin K supplements taken and then dividing this sum by 30. The total vitamin K consumption was estimated by combining the daily vitamin K intake from both dietary sources and supplementation.

Participants’ self-reported information included age, sex, race, education, family income, smoking status, alcohol drinking status, and medical history. Alcohol drinkers were defined as those who consume alcohol at least 12 times per year. Smokers were defined as those who had smoked at least 100 cigarettes in their lifetime.

Standing height and weight were measured in the MEC by trained health technicians. Body mass index (BMI) was calculated as weight in kilograms divided by height in meters squared.

Diabetes mellitus was defined as participants who met one of the following items: fasting plasma glucose ≥126 mg/dL; hemoglobin A1c ≥ 6.5%; self-reported physician diagnosis of diabetes; or current taking anti-diabetic drugs (30). Hemoglobin A1c was measured using high-performance liquid chromatography, where hemoglobin components were separated based on ionic interactions, and their absorbance was measured at 415 nm to determine its percentage. Fasting glucose levels were determined using an enzymatic hexokinase method, in which glucose was phosphorylated to glucose-6-phosphate, and the resulting NADPH was quantified photometrically to measure glucose concentration. Detailed methods for measurements of fasting glucose and hemoglobin A1c were reported previously (31). The estimated glomerular filtration rate (eGFR) was calculated using the Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI) equation (32).

Cardiovascular disease was defined as participants who had a self-reported physician diagnosis of either congestive heart failure, coronary heart disease, angina pectoris, heart attack, or stroke.

All analyses were performed using Stata version 15.1 (Stata Corporation, College Station, TX, United States). Considering the multi-stage design of NHANES, appropriate weight and sampling unit variables were designated according to the NHANES analytic guidelines. Qualitative variables were expressed by frequency distributions. Continuous variables were expressed by medians and inter-quartile range (IQR). General characteristics and serum NfL according to dietary vitamin K intake were evaluated using the chi-square test for categorical variables and linear regression for continuous variables. Linear regression models were fitted to assess the relationship between dietary vitamin K intake and serum NfL levels. Covariates were chosen based on literature knowledge. Age, sex, and race were adjusted for the first model. The second model was further adjusted for educational levels, BMI, smoking status, alcohol drinking status, family income, diabetes, and cardiovascular disease (CVD). Further adjusted were made for total daily energy intake according to the nutrient density model (33). The restricted cubic spline (RCS) model was applied to explore the potential dose–response relationship between dietary vitamin K intake and serum NfL levels. A two-tailed value of p < 0.05 is considered statistically significant.

The histogram shows that dietary vitamin K distribution is non-normal (Figure 2). The median (IQR) dietary intake of vitamin K is 81.6 (47.4–142.8) μg/d, and the median (IQR) serum NfL is 12.0 (8.0–18.5) pg./ml. Compared with participants in the lowest quintile of vitamin K intake, those in the highest quartile are more likely to have a higher degree of education, family income, BMI, dietary energy intake, and serum NfL levels; while those in the highest quintile are less likely to be Non-Hispanic White individuals, smokers, and CVD patients (Table 1).

The results show an inverse association between dietary vitamin K intake and serum NfL (Table 2). In the age, sex, and race adjusted model (model 1), compared with participants in quartile 1, dietary vitamin K intake in quintile 4 is negatively associated with serum NfL levels (β = −4.58, 95%CI: −6.34, −2.80, p < 0.001). Associations between dietary vitamin K intake and serum NfL levels still exist after adjusting for potential confounders (model 2), including dietary energy intake (model 3).

The RCS analysis indicate evidence against linearity (P for non-linearity = 0.008). When dietary vitamin K intake is below 200 μg/d, the serum NfL levels decrease rapidly with the increase of dietary vitamin K intake; while this association reach a plateau when the dietary vitamin K intake is higher than 200 μg/d (Figure 3).

The association between dietary vitamin K intake and serum NfL levels is modified by CVD (P for interaction = 0.042). No significant interaction is found between dietary vitamin K intake and other covariates, including age, sex, race, educational levels, BMI, smoking status, alcohol drinking status, family income, and diabetes (Table 3). Stratified analyses show a significant negative relationship between dietary vitamin K intake and serum NfL levels across each subgroup of educational levels, BMI categories, smoking status, alcohol drinking status, family income, and diabetes. These associations are enhanced in some subgroups (middle-aged and older adults, obesity, who have diabetes or CVD) (Table 3).

Given the influence of the digestive system on dietary intake, participants who have liver, stomach, or intestinal diseases are excluded (n = 162). Compared with participants in quartile 1, dietary vitamin K intake in quintile 4 is negatively associated with serum NfL levels (β = −4.58, 95%CI: −7.19, −1.96, p = 0.002). In addition, after adjusting for fat, fiber intake, and eGFR, higher dietary vitamin K intake is also correlated with lower serum NfL levels (β = −4.59, 95%CI: −7.88, −1.30, p = 0.009). A total of 339 participants report that they are taking vitamin K supplements. The intake of vitamin K from both diet and supplements is integrated for a thorough analysis. The median (IQR) total intake of vitamin K is 82.0 (47.9–142.8) μg/d. The finding reveal that the correlation between total vitamin K intake and serum NfL levels persists, albeit with a minor variation (Q4 versus Q1, β = −5.03, 95% CI: −7.76, −2.31, p = 0.001).