*2.2. Measurement*

The participant blood samples were collected in EDTA-containing tubes. After mixing, the blood samples were aliquoted into cryo-tubes and stored at −20 ◦C. The blood chemistry markers were measured by Seoul Clinical Laboratories (SCL, Yongin, South Korea), with a reference laboratory service [18]. Briefly, the serum concentrations of total cholesterol, high-density lipoprotein (HDL) cholesterol, and triglycerides (TG) were measured by an enzymatic method using auto analyzer ADVIA 1800 (Siemens Medical Solutions, USA). Serum low-density lipoprotein (LDL) cholesterol concentrations were calculated from Friedewald's equation [19]. Calculated LDL values less than zero were designated as 0 (*n* = 37). The hepatic enzymes (ALT, AST, and GGT) were measured on an auto-analyzer ADVIA 1800 (Siemens Medical Solutions, Malvern, PA, USA).

Total Hg was measured by flow injection cold-vapor atomic absorption spectrometry (DMA 80, Milestone, Bergamo, Italy) using whole blood samples. The limit of detection (LOD) for blood Hg was 0.10 μg/L. A value below the LOD (*n* = 1) was included as LOD divided by the square root of 2. External quality control was performed twice per year by the Korean Association of Quality Assurance

for Clinical Laboratory (KSLM) and the German External Quality Assessment Scheme for analysis of heavy metals in biological materials (G-EQUAS) [17].

#### *2.3. Criteria for Hyperlipidemia and Definition of the Elevated Liver Enzymes*

The criteria for hyperlipidemia were taken from the National Cholesterol Education Program—Adult Treatment Panel III (NCEP-ATP III) [20]. Based on these guidelines, hyperlipidemia was defined as lipid profiles showing high LDL (above 130 mg/dL), high total cholesterol (above 200 mg/dL), or high triglycerides (above 150 mg/dL). Elevated liver enzymes were defined by the reference ranges provided by SCL; ALT concentrations above 49 U/L, AST concentrations above 34 U/L, or GGT concentrations above 73 U/L (for men, above 38 U/L for women) [21].

### *2.4. Statistical Analyses*

Subjects with missing records of blood Hg, lipid profiles, and hepatic enzymes were excluded (*n* = 24). The final dataset contained 6454 personal records, and the distribution of blood Hg was calculated using sampling weights and survey strata information. The blood Hg distribution was right-skewed, so a log-transformation was performed to satisfy the assumptions of normality. Bivariate analyses were initially performed to evaluate the demographic variables, including sex (male, female), age group (19–29, 30–39, 40–49, 50–59, 60–69, and >70), BMI (underweight: <18.5, normal: 18.5–23, overweight: 23–25, and obese: >25), smoking status (non-smoker, past-smoker, or current smoker), alcohol consumption frequency (never, <1 time/month, 1–3 times/month, 1–2 times/week, >3 times/week, or daily), household monthly income (<USD 1500, USD 1500–USD 3000, USD 3000–USD 5000, USD 5000–USD 10000, and ≥USD 10000), and fish consumption (rarely, 1–3 times/month, 1–3 times/week, or 4–6 times/week). We divided the blood Hg levels into three groups based on the interquartile range, low (blood Hg < 25th), middle (25th ≤ blood Hg < 75th), high (blood Hg ≥ 75th), and compared the blood lipid levels and hepatic enzymes among groups. Each marker was regressed on blood Hg with sex, age, BMI, smoking status, alcohol frequency, and income using sampling weights and survey strata information. Analysis of variance (ANOVA) by sex and analysis of covariance (ANCOVA) with age were used to assess the associations between blood Hg and criteria status of each clinical chemistry marker.

Logistic regression analyses were performed to examine the e ffect of Hg on hyperlipidemia and elevated liver enzymes. Self-reported personal medications were considered to adjust for the effect of medicine and individual health status. The corresponding health question was open-ended, so we extracted information for hyperlipidemia-associated diseases by including the following terms: 'hyperlipidemia', 'dyslipidemia', 'high blood pressure', 'hypertension', and 'diabetes'. The terms 'fatty liver', 'hepatitis', 'liver cirrhosis', 'liver disease', and 'elevated liver enzymes' were included to represent personal medications for liver diseases. The final models were selected via model fit scores, such as the Akaike information criteria (AIC) and Bayesian information criteria (BIC). The main e ffects of sex, age, BMI, smoking status, alcohol frequency, and fish consumption, and the two-way interaction of sex and alcohol frequency were included in the final model with the Hg levels. Personal medication information was included in the logistic regression model as a covariate. Finally, the correlations between liver enzymes and lipid profiles were analyzed across the blood Hg groups as part of the sensitivity analysis. The significance level (alpha) was set to 0.05, and all of the statistical analyses were performed in SAS version 9.4 (SAS Institute Inc., Cary, NC, USA, 2013).
