NHANES is a national population-based survey program assessing the health and nutritional status of the civilian non-institutionalized general U.S. population. The NHANES data are released every 2 years. Data from the 2017–2018 NHANES cycle were used in this analysis, because this cycle specifically provided data for liver steatosis and fibrosis.

All participants aged 20 years and over were eligible. Participants were excluded if (1) their dietary intakes of vitamins, carotenoids, or serum vitamin C levels are missing; (2) they are infected with hepatitis B or C (defined by the presence of hepatitis C antibodies or the presence of hepatitis B surface antigen); (3) they have significant alcohol consumption (>30 g/day in men and >20 g/day in women) (13); and (4) their data of transient elastography are missing. The protocol was approved by the National Center for Health Statistics Institutional Review Board, and all subjects provided written informed consent.

Dietary intakes of vitamin A, B vitamins (vitamin B1, vitamin B2, niacin, vitamin B6, folate, vitamin B12, and choline), vitamin C and carotenoids (α-carotene, β-carotene, β-cryptoxanthin, lycopene, lutein + zeaxanthin) were included as exposures. All NHANES participants are eligible for two 24-h dietary recall interviews. The first dietary recall interview is collected in-person in the Mobile Examination Center and the second interview is collected by telephone 3–10 days later. In this analysis, dietary intakes of vitamin and carotenoids were assessed using 1-day values for individuals with single recalls and 2-day means for others.

Serum vitamin C concentrations are also available in the dataset, and thus were also included in this analysis. Vitamin C in serum is measured using isocratic ultra-high performance liquid chromatography. Serum specimens were processed, stored, and shipped to the Division of Laboratory Sciences, National Center for Environmental Health, Centers for Disease Control and Prevention, Atlanta, GA for analysis. Vials are stored under appropriate frozen (−30°C) conditions until they are shipped to National Center for Environmental Health for testing. The NHANES quality assurance and quality control protocols meet the 1988 Clinical Laboratory Improvement Act mandates.

The main goals of the transient elastography component are to provide objective measures for two important liver disease manifestations: liver fibrosis and liver steatosis. Liver fibrosis was measured by FibroScan® which uses ultrasound and the vibration controlled transient elastography to derive liver stiffness. The device also simultaneously measures the ultrasound attenuation related to the presence of liver steatosis and records the controlled attenuation parameter as the indicator for the fatness in the liver. The elastography exam was performed by NHANES health technicians, who were trained and certified by NHANES staff and the equipment manufacturer. A detailed description of quality assurance and quality control measures considered for this component can be found in the Procedures Manual (15). We defined liver steatosis as controlled attenuation parameter scores of 263 dB/m or more and liver stiffness values of 8 kPa or higher were considered to have significant fibrosis (11, 16, 17).

According to the previously related publication (13), we included the following covariates that are risk factors of NAFLD: age (in 10-year increments), sex, race/ethnicity (Mexican–American, Other Hispanic, Non-hispanic White, Non-hispanic Black, Other Races), education (≤ high school, >high school), recreational physical activity (at least 10 min continuously per week), smoking (current smoker, former smoker, never smoker), alcohol drinking (continuous), hypertension, diabetes, body mass index (under/normal weight: <25 kg/m², overweight: 25– <30 kg/m², obesity: ≥30 kg/m²), and dietary intakes of cholesterol. Diabetes was defined using a previous diagnosis of diabetes or taking diabetic pills to lower blood sugar or, if diabetes was not previously diagnosed, by a hemoglobin A1c level ≥ 6.5%, or a fasting plasma glucose level ≥ 126 mg/dL (18). Subjects were considered hypertensive if they were taking antihypertensive medication, if their systolic blood pressure exceeded 130 mmHg, or if their mean diastolic blood pressure exceeded 80 mmHg (mean values of three measurements) (19).

Logistic regression was adopted to explore the associations between vitamins and carotenoids and liver steatosis and fibrosis. Subjects were classified into tertiles according to their dietary intakes of vitamins, carotenoids or serum vitamin C concentrations, and the odds ratios (ORs) and 95% confidence intervals (CIs) of liver steatosis and fibrosis for subjects in tertile 2 and tertile 3 were calculated as compared to those in tertile 1. There are three different logistic regression models. Model 1 was adjusted for age, sex and race/ethnicity. Model 2 was adjusted for covariates in model 1, and also education, physical activity, smoking and alcohol drinking. Model 3 was adjusted for covariates in model 2, and also body mass index, hypertension, diabetes, and dietary intakes of cholesterol. The potential non-linear dose-response relationships between vitamins, carotenoids, and liver steatosis and fibrosis were examined by modeling dietary intakes of vitamins, carotenoids, or serum vitamin C concentrations using restricted cubic splines with 3 knots at percentiles 25, 50, and 75 of the distribution (20). A P-value for non-linearity was calculated by testing the null hypothesis that the coefficient of the second spline is equal to 0 (20). We also conducted multivariable linear regression model with the outcomes and exposures expressed as continuous variables. In sensitivity analysis, we explored the associations between vitamin intakes from both diets and supplements and liver steatosis and liver fibrosis. All analyses were conducted using STATA version 12.0, and P ≤ 0.05 was considered statistically significant. Appropriate strata, cluster, and weights were considered in all analyses as suggested by NHANES.

According to the exclusion criteria, individuals who are infected with hepatitis B or C (N = 79), who they have significant alcohol consumption (N = 457) and whose data of transient elastography are missing (N = 630), a total of 4,352 subjects were potentially eligible for this analysis. Among the 4,352 subjects, 4,116 provided data of serum vitamin C concentrations and 3,940 provided dietary intakes of vitamins and carotenoids. The weighted prevalence of was 56.47% for liver steatosis and 15.05% for liver fibrosis, respectively. Patients with liver steatosis and fibrosis showed higher prevalence of diabetes, obesity and hypertension (all P < 0.01). Patients with liver fibrosis and liver steatosis had higher dietary intakes of vitamin B12, choline and cholesterol, while had lower dietary intakes of vitamin C and lutein + zeaxanthin and lower levels of serum vitamin C concentrations (all P ≤ 0.05; Table 1).

Overall, the findings between vitamins and carotenoids and liver steatosis were similar between model 1 and model 2. In model 2, comparing the highest to lowest tertile, higher dietary intakes of vitamin C [OR (95% CI): 0.59 (0.42–0.81)] and β-carotene [0.70 (0.57–0.87)] and serum vitamin C concentrations [0.50 (0.38–0.66)] were inversely associated with liver steatosis, respectively. However, after further adjustment for obesity, diabetes, hypertension and cholesterol intakes (model 3), only dietary intakes of vitamin C [0.68 (0.50–0.93)] and β-carotene [0.71 (0.54–0.93)] were significantly associated with liver steatosis. In addition, a marginally inverse association between lutein + zeaxanthin and liver steatosis were found [0.74 (0.53–1.04)]. No associations were found between other vitamins and carotenoids and liver steatosis in model 3 (Table 2). The findings from multivariable linear regression model are consistent with those from the logistic regression, and the above-mentioned associations between vitamin C, β-carotene, and lutein + zeaxanthin and liver steatosis were also observed (Supplementary Table 2). Similar findings were also found in sensitivity analysis including intakes from Supplementary Table 3.

Overall, the findings between vitamins and carotenoids and liver fibrosis were similar between model 1 and model 2. In model 2, comparing the highest to lowest tertile, higher dietary intakes of vitamin B12 [1.49 (1.01–2.18)] and choline [1.56 (1.11–2.21)] were positively associated with liver fibrosis. In addition, higher dietary intakes of α-carotene [tertile 2 vs. tertile 1: 1.85 (1.16–2.95)] were also positively associated with liver fibrosis. Higher levels of serum vitamin C were inversely associated with liver fibrosis [0.45 (0.29–0.70)]. After further adjustment for obesity, diabetes, hypertension and cholesterol intakes (model 3), higher dietary intakes of choline [tertile 2 vs. tertile 1: 1.43 (1.04–1.98)] were positively associated with liver fibrosis. Higher levels of serum vitamin C were inversely associated with liver fibrosis in model 3 [tertile 2 vs. tertile 1: 0.45 (0.32–0.62)].

No associations were found between other vitamins and carotenoids and liver fibrosis in model 3 (Table 3). The findings from multivariable linear regression model are consistent with those from the logistic regression, and the above-mentioned associations between serum vitamin C, choline, α-carotene and vitamin B12 and liver fibrosis were also observed (Supplementary Table 2). Similar findings were also found in sensitivity analysis including intakes from Supplementary Table 4.

The median values in the first category (tertile 1) of exposures were used as the references values in dose-response analysis. Dose-response analysis with restricted cubic splines showed that the departure from a linear relationship was not significant between vitamin C (P_{fornon−linearity} = 0.30) and lutein + zeaxanthin (P_{fornon−linearity} = 0.75) and liver steatosis, while the departure from a linear relationship was significant between β-carotene (P_{fornon−linearity} = 0.02) and liver steatosis. The departure from a linear relationship was not significant between serum vitamin C (P_{fornon−linearity} = 0.34), choline (P_{fornon−linearity} = 0.36), α-carotene (P_{fornon−linearity} = 0.36) and vitamin B12 (0.45) and liver fibrosis.