Study participants
We used datasets from the 2014–2016 KNHANES in this study. The KNHANES is a population-based, nationwide, cross-sectional survey conducted by the Korea Center for Disease Control and Prevention since 1998. The study participants were randomly selected using stratified, multistage-clustered probability sampling of the general Korean population. The survey database contains data on demographics and income status (including socioeconomic factors), nutrition, health screening records, and health interview questionnaires; well-trained research assistants perform the health interviews and dieticians administer the nutrition-related questionnaires.
In the 2014–2016 KNHANES, urinary cotinine levels were measured in 16,054 participants. After selecting participants aged 60 years or older, 5,171 participants remained. We then excluded individuals who indicated that they had been diagnosed with various cancers or chronic liver disease to minimize the effect of underlying diseases on lipid profiles. In total, 4,349 participants were included in the final analysis (Figure 1). All data we used were downloaded from the KNHANES homepage (https://knhanes.kdca.go.kr) and were anonymized and de-personalized. The survey participants have provided written informed consent prior to the completion of the survey. The 2014 KNHANES was reviewed and approved by the Institutional Review Board (IRB) of the Korea Center for Disease Control and Prevention (IRB approval no. 2013-12EXP-03-5C) and IRB for the subsequent KNHANES was waived on the grounds of anonymity of the data.
Measures
Demographic variables such as age, level of education (highest level of education), and monthly household income (categorized into quartiles) of the participants were identified. The study subjects who were high school graduates or higher were categorized as the ‘high education level’ group in our study. The study subjects who were lowest 25% for the income according to the National Statistics data were categorized as ‘low income’ group in our study. Furthermore, data on lifestyle variables such as alcohol intake (percentage of individuals who drank more than once a month during the previous year), smoking status (percentage of current smokers), physical activity (percentage of those who exercised moderately for more than 2 hours and 30 minutes a day, who exercised vigorously for more than 1 hour and 15 minutes a day, or who performed an equivalent amount of mixed exercise), and daily energy intake, with daily carbohydrate, protein, and fat intake, were obtained. Daily energy and fat intake (the percentages of fat per total food consumed in 24 hours) were assessed using the 24-hour recall method in the nutritional survey portion of the KNHANES. The medical history, family history of dyslipidemia, and current medications of the participants were evaluated using a questionnaire.
Spot urine samples were obtained by instructing the survey participants to collect a midstream first morning void. Urine cotinine levels were analyzed using a gas chromatography-mass spectrometer detector (GC-MSD) Clarus 600/600T (Perkin Elmer, USA) in the 2014–2015 KNHANES. High-performance liquid chromatography with mass spectrometry (HPLC-MS/MS) using the Agilent 110 series with API 4000 (AB Sciex, USA) was used to quantify urine cotinine levels in the 2016 KNHANES. Based on previous studies, the study subjects with urine cotinine level of <50 ng/mL were classified as non-smokers, and the study subjects with urine cotinine level ≥50 ng/mL were classified as smokers [10,11,17].
Participants were measured for height and body weight only wearing light indoor clothing. Waist circumference was measured to the nearest 0.1cm at the narrowest point between the costal margin and the iliac crest at the end of exhalation. Body weight (kg) was divided by the square of height (m2) to calculate body mass index (BMI). Blood pressure was measured three times at a right arm with each study subject in a seated position, following a five-minute long resting period; the average of the three measurements was calculated for the study subjects.
Blood samples were obtained after fasting for ≥8 hours, from the antecubital vein. Samples were immediately refrigerated, transported under cold storage conditions to the Central Testing Institute in Seoul, Korea. The serum concentrations of serum lipids, TC, HDL-C, LDL-C, and TG, were measured using a Hitachi Automatic Analyzer 7600–210 (Hitachi, Tokyo, Japan) via enzymatic methods using commercially available kits (Sekisui, Japan). All samples were analyzed within 24 hours of transportation.
Statistical analyses
Descriptive analyses were conducted on the baseline characteristics of cotinine-verified smoking group (urinary cotinine level ≥50 ng/mL) and non-smoking group (urinary cotinine level <50 ng/mL). The results were expressed as ‘arithmetic mean ± standard error (SE)’ for the continuous variables except for urine cotinine and serum triglycerides, whose results were expressed as ‘geometric mean (95% confidence intervals [CIs])’ due to the skewness of their distributions; the results for the categorical variables were expressed in ‘weighted percentage (SE).’ The two groups’ baseline characteristics were then compared using Student’s t-tests (for continuous variables) and chi-square tests (for categorical variables).
The distributions of unfavorable quartiles (Q) of lipid profiles and ratios, such as the highest quartile (Q4) of TC, LDL-C, TG, TC/HDL-C, LDL-C/HDL-C, and TG/HDL-C and the lowest quartile (Q1) of HDL-C, were compared between the two groups using chi-squared tests. In addition, subgroup analyses by gender were conducted. Of note, the cut-offs that differentiated between Q1 vs. Q2-4 and Q4 vs. Q3-1 were generated in a gender stratified manner.
In order to quantify the association between urine cotinine-verified smoking status and unfavorable serum lipid profiles and ratios, multivariable logistic regression analyses were performed. In the logistic regression, the binary variables transformed from serum lipid profiles and ratios were used as the dependent variables. For serum TC, LDL-C, TC/HDL-C, LDL-C/HDL-C, and TG/HDL-C, the cut-off divided each variable into Q4 and the rest (Q3-1). For HDL-C, the cut-off divided the variable into Q1 and the rest (Q2-4).
The regression analyses used multiple levels of adjustment; model 1 was adjusted for age, sex, BMI, level of income, and level of education. Model 2 was further adjusted for alcohol consumption, physical activity, family history of dyslipidemia, DM, HTN, lipid-lowering drug use, energy intake, and fat intake. The analyses produced the odds ratios (ORs) and 95% CIs of various unfavorable lipid profiles and ratios for urinary cotinine-verified smoking status. Additional analyses were conducted on male and female subgroups using the same method. All statistical analyses were performed using SAS version 9.3 software (SAS Institute, Cary, NC, USA). Statistical significance was set at two-sided p-values <0.05.