Secular Trends in Lipid Profiles in Korean Adults Based on the 2005–2015 KNHANES

Dyslipidemia is a primary, critical risk factor for cardiovascular disease. Therefore, evaluating the trends in lipid profiles is crucial for the development of health policies and programs. We studied trends in lipid profiles in Korean adults over an 11-year period according to the use of lipid-lowering medications through age-specific analysis. A total of 73,890 participants were included in the Korean National Health and Nutrition Examination Survey III (2005)-VI (2013–2015). The proportion of participants on lipid-lowering medications has increased. This trend was apparent in age groups of over 40 years in both men and women. Lipid-lowering medications successfully reduced mean total cholesterol (TC), but there was no favorable trend in TC in participants not taking lipid-lowering medication in both men and women. Unlike men, triglyceride and non-high-density lipoprotein cholesterol (HDL) decreased in women without lipid-lowering medications. In age-specific hypercholesterolemia, the prevalence of hypercholesterolemia significantly increased in the age groups of 30–59 and 30–49 years in men and women without lipid-lowering medications, respectively. Meanwhile, mean HDL-C levels increased over the 11-year period regardless of lipid-lowering drug use in both men and women. These analyses identified an upward trend in TC and HDL-C over the 11-year period.


Introduction
According to the 2017 statistical update of the American Heart Association, the number of deaths from cardiovascular disease (CVD) has been decreasing since 2004 [1]. However, CVD remains the leading cause of total mortality worldwide, accounting for 30% of all global deaths [1,2]. CVD is responsible for an immense socioeconomic burden on patients and society in general. There are several methods to prevent CVD, including smoking cessation, increased physical activity, and adoption of a healthy diet [3,4]. Dyslipidemia, i.e., abnormal lipid levels in the blood, is considered one of the most important modifiable risk factors of CVD [5]. One recent study found that even moderate hyperlipidemia increases one's risk of coronary heart disease (CHD) later in life, depending on the length of exposure [6]. Therefore, active screening and treatment of dyslipidemia are the first steps in prevention of CVD. Many countries have established public health programs to address dyslipidemia [7]. These efforts have successfully led to a decrease in the prevalence of Automatic Analyzer 7600/7600-210 (Hitachi, Tokyo, Japan) in the 2008,2010,2011,2012,2013,2014, and 2015 surveys.

Lipid Profiles, and Related Variables
According to the criteria of the National Cholesterol Education Program Adult Treatment Panel III (NCEP-ATP III) [16], hypercholesterolemia is defined as serum TC ≥ 240 mg/dL; hypertriglyceridemia is defined as TG ≥ 150 mg/dL; and hypo-HDL-cholesterolemia is defined as serum HDL-C < 40 mg/dL in men and < 50 mg/dL in women. Non-HDL cholesterol was calculated by subtracting the HDL-C value from a TC value [17]. Education duration was divided into four groups: <6 years, 6-9 years, 9-12 years, and ≥12 years. Occupational status was categorized as manual work (clerk, service, and sales workers, skilled agricultural, forestry and fishery workers, and persons who operate or assemble equipment or machines), office work (general managers, government administrators, professionals, and simple office worker), or other (unemployed persons, housekeepers, and students).

Statistical Analysis
All data for continuous variables are presented as means ± standard error (SE). Data for categorical variables are presented as percentages ± SE. All sampling and weight variables were stratified. The SAS survey was used for statistical analysis to account for the complex sampling design and to provide nationally representative prevalence estimates. Analysis of variance (ANOVA) was used to compare the mean values of the continuous variables, analysis of covariance (ANCOVA) was used to compare the age-adjusted lipid level across the KNHANES phases, and the x 2 test was used to compare categorical variables. P for trend values were calculated among the KNHANES phases by logistic regression analyses or linear regression after setting the phase as the continuous variable. We conducted Bonferroni correction for multiple testing. Statistical analyses were performed using SAS version 9.4 (SAS Institute Inc., Cary, NC, USA).

Characteristics of Sample
The unweighted sample sizes for this study in the KNHANES III (2005), IV (2007-2009), V (2010-2012), and VI (2013-2015) were 24,752, 16,359, 17,243, and 15,536, respectively, and the mean ages were 43.6, 44.8, 45.7, and 46.6 years, respectively ( Table 1). The number and proportion of women and men participating in the study were 13,251 (53.5%) and 11,501 (46.5%) in phase III, 9454 (57.8%) and 6905 (42.2%) in phase IV, 9994 (57.9%) and 7249 (42.1%) in phase V, and 8976 (57.8%) and 6560 (42.2%) in phase VI, respectively. Table 2 describes the characteristics of participants after stratifying by sex and use of lipid-lowering medications. Age significantly increased across the KNHANES phases. There were decreasing trends in BMI in men and women with lipid-lowering medication use (β-coefficient = −0.354 and p for trend < 0.001 in men, β-coefficient = −0.363 and p for trend < 0.001 in women). The level of BMI increased in men without lipid-lowering medication use (β-coefficient = 0.107 and p for trend < 0.001), while BMI showed no difference in women without lipid-lowering medication use over time.  Table 3 shows the changes in mean lipid profile levels according to sex and age groups across the KNHANES phases. The mean TC levels increased linearly from 183.4 mg/dL to 187.7 mg/dL (β-coefficient = 1.331, p for trend < 0.001) in men, and from 184.3 mg/dL to 187.9 mg/dL (β-coefficient = 0.805, p for trend = 0.002) in women. The mean HDL-C levels increased over the 11-year period in both men and women (β-coefficient = 1.446, p for trend < 0.001 and β-coefficient = 2.308, p for trend < 0.001), respectively.    Figure 1 presents the age-adjusted and age-specific proportion of those taking lipid-lowering medications across the KNHANES phases. For KNHANES III-VI (2005 and 2007-2015), the proportion of participants on lipid-lowering medication increased from 1.6 to 4.7% in men (β-coefficient = 0.414, p for trend < 0.001) ( Figure 1A) and from 1.5 to 7% in women (β-coefficient = 0.531, p for trend < 0.001) ( Figure 1B). This trend was apparent in age groups over 40 years in both men and women.  Figure 1 presents the age-adjusted and age-specific proportion of those taking lipid-lowering medications across the KNHANES phases. For KNHANES III-VI (2005 and 2007-2015), the proportion of participants on lipid-lowering medication increased from 1.6 to 4.7% in men (βcoefficient = 0.414, p for trend < 0.001) ( Figure 1A) and from 1.5 to 7% in women (β-coefficient = 0.531, p for trend < 0.001) ( Figure 1B). This trend was apparent in age groups over 40 years in both men and women.

Lipid Profile According to Treatment of Lipid Lowering Drug
To clarify the effect of lipid-lowering medications, we also investigated lipid profile trends in participants after stratifying by use of lipid-lowering drugs ( Table 4). The mean TC and non-HDL-C levels declined in men with lipid-lowering drug use (TC: 195.7 to 174.6 mg/dL, p for trend < 0.001; non-HDL-C: 240.4 to 213.0 mg/dL, p for trend < 0.001) over time. Mean TC and TG levels increased

Lipid Profile According to Treatment of Lipid Lowering Drug
To clarify the effect of lipid-lowering medications, we also investigated lipid profile trends in participants after stratifying by use of lipid-lowering drugs ( Table 4). The mean TC and non-HDL-C levels declined in men with lipid-lowering drug use (TC: 195.7 to 174.6 mg/dL, p for trend < 0.001; non-HDL-C: 240.4 to 213.0 mg/dL, p for trend < 0.001) over time. Mean TC and TG levels increased in men without lipid-lowering drug use (TC: 183.6 to 188.9 mg/dL, p for trend < 0.001; TG: 155.0 to 162.8 mg/dL, p for trend = 0.002). In women with lipid-lowering drug use, all lipid profiles levels, except HDL-C, decreased over time. TC was significantly elevated in women without lipid-lowering drug use (TC: 184.4 to 190.1 mg/dL, p for trend < 0.001), while other lipid profiles (TG and non-HDL-C) decreased during the same time period. Bonferroni correction for multiple testing was conducted; significant p-value < 0.0125 and p-for trend < 0.0125. Figure 2 presents the age-adjusted prevalence of hypercholesterolemia (A), hypertriglyceridemia (B), and hypo-HDL-cholesterolemia (C) in participants without lipid-lowering medication use. There was an increasing trend in hypercholesterolemia in men (β-coefficient = 0.123 and p for trend < 0.001) over the examination cycles. By contrast, the prevalence of hypercholesterolemia was not significantly different, and the prevalence of hypertriglyceridemia slightly decreased over time (β-coefficient = −0.064 and p for trend < 0.001) in women. There were significant downward trends in the prevalence of hypo-HDL-cholesterolemia in both men and women (β-coefficient = −0.281 for men, −0.224 for women, and p for trend < 0.001 in both sexes) over the time course of the study period. In age-specific hypercholesterolemia, the prevalence of hypercholesterolemia significantly increased in the age groups of 30-59 and 30-49 years in men and women, respectively, without lipidlowering medication use (Figure 3a). The prevalence of hypertriglyceridemia was not significantly different in all age groups in men and women (Figure 3b). The prevalence of hypo-HDL cholesterolemia significantly decreased in all age groups of men and women (Figure 3c). In age-specific hypercholesterolemia, the prevalence of hypercholesterolemia significantly increased in the age groups of 30-59 and 30-49 years in men and women, respectively, without lipid-lowering medication use (Figure 3a). The prevalence of hypertriglyceridemia was not significantly different in all age groups in men and women (Figure 3b). The prevalence of hypo-HDL cholesterolemia significantly decreased in all age groups of men and women (Figure 3c). In age-specific hypercholesterolemia, the prevalence of hypercholesterolemia significantly increased in the age groups of 30-59 and 30-49 years in men and women, respectively, without lipidlowering medication use (Figure 3a). The prevalence of hypertriglyceridemia was not significantly different in all age groups in men and women (Figure 3b). The prevalence of hypo-HDL cholesterolemia significantly decreased in all age groups of men and women (Figure 3c).

Discussion
Our data revealed that TC and HDL-C level increased and the proportion of taking lipidlowering medications increased over the 11-year study period.
Many clinical and epidemiologic studies have demonstrated that dyslipidemia is an important risk factor of CVD [6, 18,19]. Therefore, many health authorities, including those in Korea, have conducted national health screening programs for early detection and management of dyslipidemia

Discussion
Our data revealed that TC and HDL-C level increased and the proportion of taking lipid-lowering medications increased over the 11-year study period.
Many clinical and epidemiologic studies have demonstrated that dyslipidemia is an important risk factor of CVD [6, 18,19]. Therefore, many health authorities, including those in Korea, have conducted national health screening programs for early detection and management of dyslipidemia [20,21]. The National Cholesterol Expert Panel (NCEP) suggests screening for dyslipidemia in adults over 20 years of age [22], while the United States Preventive Services Task Force (USPSTF) provides positive evidence for screening dyslipidemia in adults [23]. Based on standard guidelines, expert groups also recommend active management, such as therapeutic lifestyle modification and use of appropriate medications (such as statins) in patients with dyslipidemia [21,22,24].
In this study, we determined that the proportion of adults on lipid-lowering drugs gradually increased over the course of KNHANES phases III-VI, and that the proportion of participants on lipid-lowering medication was higher in women than men. Depending on the use of lipid-lowering drug, the lipid profile in participants taking lipid-lowering medications improved, with decreasing TC, non-HDL-C, and TG and increasing HDL-C levels in both men and women. Meanwhile, TC increased in participants who were not taking lipid-lowering drugs. HDL-C level consistently increased regardless of sex, age, and use of lipid-lowering drug. Age-specific analysis revealed an upward trend in TC that was significant in the 30-59 years age group in men and the 30-49 years age group in women without lipid-lowering medication use. The level of lipid-lowering drug usage did not increase significantly in the relatively younger age groups.
Our findings are in agreement with several previous studies. Several studies identified the increasing trend of TC in Japan, Turkey, and China [10][11][12]. Meanwhile, a recent Japanese study reported that serum HDL-C continues to increase using the National Health and Nutrition Survey [25]. A study conducted in a Canadian province also revealed a significant downward trend in the prevalence of hypo-HDL-cholesterolemia over a six-year period [26]. There is no exact reason to explain such trends. HDL-C levels are affected not only by genetic factors, but also by socioeconomic status (i.e., education level, economic level, and occupation) [27].
High socioeconomic status has been regarded as a key determinant of medical intervention accessibility and exposure to modifiable risk factors [28]. The Global Burden of Disease (GBD) 2015 study reported that the prevalence of CVD has declined in regions with a high socioeconomic index (including per capita income and education level) [28]. Korea is one of the most developed countries in Asia, and both household incomes and education levels have improved between 2005 and 2015. Favorable changes in socioeconomic status may lead to increased HDL-C levels.
TG in women without anti-dyslipidemic use has declined. The differences in TG trends between men and women may partly be explained by different rates of obesity between the sexes. According to our previous study on obesity trends, male obesity has increased with time, while female obesity has leveled off, or even decreased [29]. Different trends in obesity prevalence between men and women may also subsequently lead to variable lipid profiles in participants who are not taking lipid-lowering drugs.
TC level has been increasing in adults (30-59 years) not taking lipid-lowering medication. These results suggest that more active and practical interventions are needed to prevent and manage dyslipidemia in these age groups. Screening obese individuals and those with unhealthy lifestyles should be the first step to prevent dyslipidemia and reduce future CVD events [30][31][32]. Many individuals are unaware that they have dyslipidemia. Although the awareness rate of dyslipidemia varies according to prior studies, one recent Korean study reported that only approximately 17% of adults with dyslipidemia over 20 years old were aware of their dyslipidemia status [33]. Effective medical therapy, such as administration of statin drugs, helps to prevent CVD [34,35]. However, recent studies suggested that chronic inflammation is a key risk factor of CVD [36]. Therefore, reducing inflammation manifestation by lifestyle modifications, such as increasing physical activity, adopting healthy diet, and maintaining a healthy body weight, must also be emphasized [4,36].
This study had several limitations. First, we did not include the low-density lipoprotein cholesterol (LDL-C) levels because they were not directly measured in KNHANES. Although LDL-C levels estimated by the Friedwald equation were correlated with direct measurement, underestimation of the LDL-C level is an important problem to evaluate the high-risk group in the presence of TG ≥ 150 mg/dL [37]. Therefore, we did not include LDL levels to minimize bias. Alternatively, we considered non-HDL cholesterol. Several studies supported that non-HDL-C could be a more reliable predictor for CVD [38,39]. Second, we did not distinguish between the types of lipid-lowering medications used by patients. Different drug classes (such as statins, fibric acid, and niacin) have variable effects on specific lipid levels, and in particular on HDL-C [40]. Therefore, the HDL-C level may have been influenced by the type of drug used. Furthermore, we could not consider HDL functionality. Emerging evidence established that HDL functionality is more important than quantity, and HDL-C may no longer be a reliable marker [41,42]. Our authors believe that there is a need to develop inexpensive methods to evaluate functionality of HDL using a large population-based study. Finally, there could have been bias deriving from unequal sex ratio.
Despite these limitations, our study has several advantages. First, we applied sampling weights to all analyses to represent the average Korean citizen. Second, this study investigated an 11-year trend of lipid profiles using nationally representative data. Finally, we examined the lipid profile trend in patients who were and were not treated with lipid-lowering medications.

Conclusions
The analyses identified an upward trend in TC and HDL-C over the 11-year period. There was also an increase in the use of lipid-lowering drugs over time. The positive trends in TC level were only seen in adults who used lipid-lowering drugs. We must continually monitor how long these trends continue, and what type of changes occur. Furthermore, more active strategies for lifestyle changes, as well as more intense screening and treatment of dyslipidemia, remain very important in the reduction of CVD risk in Korean adults.
Author Contributions: Y.-J.K., J.-W.L., and H.-T.K. contributed to the conception and design of the study, advised on all statistical aspects and interpreted the data, reviewed the manuscript and approved the final version to be published, had full access to all the data in the study, and take responsibility for the integrity of the data and the accuracy of the data analysis.