Study population
We used public data across three three-year cycles (2013-2018) of NHANES, an ongoing nationwide cross-sectional survey of the non-institutionalized civilians living in the US using complex, multi-stage probability sampling designs. All participants have signed informed consent forms during the period of recruitment. This study participants were restricted to adults (aged ≥ 18 years) who had complete data of blood EO concentrations, inflammatory markers, glycohemoglobin, blood pressure, and information on the definition of NAFLD. Additionally we excluded individuals who were pregnant and had viral hepatitis (hepatitis B surface antigen or hepatitis C RNA). Finally, 2,394 adults were left for the analysis (Figure S1).
Measurement of EO
Considering that HbEO has a longer half-life in contrast to EO in the body, it was used to assess EO exposure in this study. Following NHANES Laboratory/Medical Technologists Procedures Manual (https:// wwwn.cdc.gov/Nchs/Nhanes/2017-2018/ETHOX_J.htm), HbEO levels were determined. The red blood cell samples by washing and packing were stored at -30°C up to ship National Center for Environmental Health for examination. Subsequently, HbEO levels were detected based on a modified Edman reaction using high-performance liquid chromatography coupled with tandem mass spectrometry (HPLC-MS/MS). More details are described in the NHANES Laboratory/Medical Technician Procedures Manual. Measurements below the limit of detection (12.9 pmol/g Hb) were imputed as the LOD divided by √2. Additionally, the assays accord with the quality control and quality assurance performance criteria of the NCEH Division of Laboratory Science for accuracy.
Assessment of NAFLD
We defined NAFLD using US fatty liver index (US FLI) and FLI, highly effective diagnostic indices, which was used to monitor and screen NAFLD cases via a non-invasive method and has been validated in the numerous epidemiological studies [20-22]. Compared with traditional ultrasonography and CT tests, USFLI and FLI were calculated by formulas that included information on age, race, waist circumference, body mass index (BMI), fasting insulin, fasting glucose, gamma glutamyl transferase (GGT), and triglyceride (TG). The cut-off of 30 for USFLI and 60 for FLI were considered to define NAFLD in this study.
Covariates
Demographic information was obtained by questionnaires examinations, anthropometric assessment, and laboratory testing. We considered some variables as potential confounders: age, gender, race, education level, annual household income, BMI, smoking, drinking, hypertension, diabetes, and total cholesterol (TC). Of which, race/ethnicity was classified as Mexican American, Non-Hispanic white, Non-Hispanic black, and others. Education level was classified as lower than 9th grade, 9-11th grade, high school/ GED or equivalent, some college or Associate in Arts degree, and college graduate or above. Annual household income was classified as <$20,000, $20,000–$45,000, $45,000–$75,000, and >$75,000. Current smoking and drinking were categorized into yes and no. Diabetes is assessed by self-reported diagnosis, use of insulin or oral hypoglycemic medication, FBG ≥ 126 mg/dL or HbA1c level ≥ 6.5%. Hypertension was assessed as self-reported hypertension systolic blood pressure (SBP) ≥ 140mmHg or diastolic blood pressure (DBP) ≥ 90mmHg or use of anti-hypertensive medication. Inflammatory markers, including ALP, monocyte count, and lymphocyte count in blood samples were obtained from the laboratory data.
Statistical analysis
The baseline characteristics including means (SD) and frequency (%) were compared by the quartile of HbEO with using ANOVA and χ2 tests. HbEO levels were classified into four groups based on quartiles. Since levels of blood HbEO were in skewed distributions, it was naturally ln-transformed. General linear regression analysis was adopted to assess the effect of HbEO on liver function indices and USFLI, and FLI. The relationship between HbEO and NAFLD was investigated using logistic regression models. Coefficient or odd ratio (OR) and 95% confidence intervals (CI) were shown in three model. Model 1 was adjusted for age and gender. Model 2 was adjusted for covariates in model 1 plus race, education, income, smoking, and drinking. Model 3 was adjusted for covariates in model 2 plus hypertension, diabetes, and TC. In order to investigate the susceptibility of demographic-related differences, we adopted subgroup analyses to evaluate whether the relationship between HbEO and NAFLD were affected by sex. Restricted cubic spline (RCS) was used to model the trend of USFLI and FLI and prevalence of NAFLD across ln-transformed HbEO range. Mediation models were subsequently conducted to explore the role of inflammatory markers as potential mediators of the relationship between ln-transformed HbEO and the prevalence of NAFLD. Considering that CVD and diabetes might influence the association between HbEO and NAFLD, their family history was further adjusted in the sensitivity analyse. All data were analyzed using R 4.3.2, and a two-tailed P <0.05 was regarded as statistical significance.