Abstract
This study investigated the association between health-related quality of life (HRQoL) and obesity stratified by sex in 34 935 Korean adults. We used data from the Korea National Health and Nutrition Examination Survey, a cross-sectional, nationwide, population-based survey, from 2007 to 2012. Individuals with higher than average health value scores using the EQ-5D-3L to measure HRQoL were classified as being in good health. Multiple logistic regressions were used to determine the association between obesity and good health. Of the 34 935 adults, 28.0% (9767) were classified as obese, 3.8% (1326) as severely obese, and 23.6% (8249) as overweight. Through multiple logistic regressions after adjustments, women who were severely obese had 31% significantly lower HRQoL than women with normal weight (95% confidence interval = 1.12-1.53). However, the same trend was not found in men. Additional interventional studies would be needed to better understand the causality of the association between obesity and HRQoL in women.
Introduction
Obesity is a global issue in terms of public health for both developed and developing countries. Regardless of the various efforts to decrease the prevalence of obesity, an increase in prevalence is still observed in many countries.1,2 Effective management of obesity as an aspect of public health is important because of the role obesity is known to play in contributing to the development of chronic diseases such as hypertension, diabetes, and hyperlipidemia, and in accelerating the onset and increasing the severity of those diseases.3,4
Obesity has been linked closely not only to physical health problems but also mental health and social problems. 5 Thus, obesity treatment should be evaluated in terms of its impact on physical, mental, and social health. Health-related quality of life (HRQoL) is an important outcome as reflecting the multifaceted dimension in obesity.6,7
Previous studies reported that HRQoL in women was lower than that of men; this trend was shown not only in healthy people but also in patients receiving medical treatment.8,9 The HRQoL of patients with severe diseases was found to be lower in women compared to men.10-13
In investigating the association between HRQoL and obesity, the studies showed that obesity was associated with lower HRQoL.14,15 Even though the direct cause is not clear, obesity may affect HRQoL and may cause chronic diseases to worsen. In addition, women with obesity may have lower HRQoL than men with obesity. 16 The lower HRQoL of women compared with men may be attributed to the higher psychological stress associated with body image in women. 17 While the satisfaction of underweight (body mass index [BMI] <20) women and men with their bodyweight and shape were similar, the normal weight and overweight women were more dissatisfied than men. In particular, women with obesity in Asia were found to be under a heavy burden from severe public disapproval and excessive social pressure or discrimination related to their weight and appearance.18,19 Thus, this study investigated the impact of obesity on HRQoL using the EuroQol-5 dimension (EQ-5D), according to the obesity level in each sex.
Methods
Sample and Data Collection
Data from the Korea National Health and Nutrition Examination Survey (KNHANES) IV (2007-2009) and V (2010-2012), a cross-sectional, nationwide, population-based survey, were used for this study. 20 KNHANES was established by the Korea Centers for Disease Control (KCDC) to investigate the overall health status of people in Korea. Raw data are available to the public online (https://knhanes.cdc.go.kr), and informed consent was obtained from all persons who participated for the blood sampling before this survey. The research ethics committee of the KCDC approved the study protocol.
To obtain a sample nationally representative, multistage, stratified probability sampling of noninstitutionalized households was used. For sampling, the population was first stratified into administrative district, housing type (ie, apartment or house), and so on. Second, several clusters were selected in each stratum. Probability sampling based on the number of households in selected clusters was performed. The survey was conducted every year aimed at obtaining annual trends, beginning in 2007. And the combined sample in every 3 years represented the national population; the data from 2007 to 2009 were collected under KNHANES IV; KNHANES V collected data from 2010 to 2012. It was possible to integrate the KNHANES IV and V data to create a national representative sample using weighted value provided in the database. We could not distinguish the same participants in KNHANES IV and KNHANES V because de-identified IDs were provided in each survey. However, this sampling design was considered to obtain the national representative data and provided the survey data to unite the comprehensive information in several years. It would allow us to obtain the national representative data in overcoming the small sample size.
KNHANES data include demographic data, health examinations, and nutrition survey results. EQ-5D data were only available for survey participants 19 years of age or older. We included the subjects who had both BMI and EQ-5D data for this study.
EQ-5D
We measured the HRQoL using the EQ-5D. The EQ-5D is a set of questions designed to measure health outcomes in the general population, and it was originally intended for self-completion. It consists of 5 dimensions: mobility, self-care, usual activities, pain/discomfort, and anxiety/depression. We used the Korean language version of EQ-5D 3 level (EQ-5D-3L), which was validated as the official language version by the original developing group. 21 EQ-5D-3L has 3-level answers (ie, no problem, some problem, and severe problem) for each of the 5 dimensions, and it can be converted into a health value score to represent health state using a time trade-off (TTO) valuation technique. 22 The health value score of EQ-5D ranged from 0 (death) to 1 (perfect health). A country-specific value set to convert health value score was available. The health value score in this study was obtained by Korea’s national value set, and it was also provided in the same database. 23 Good health state was defined as having a higher health value score than average (average value 0.949 in men and 0.914 in woman) in each sex in our KNHANES data who are 19 years of age or older.
BMI and Comorbidities
The weight, height, and blood pressure readings were performed by trained nurses. Fasting blood samples were collected to investigate fasting levels of plasma glucose (FPG), triglycerides, and high-density lipoprotein (HDL) cholesterol. Quality control for laboratory testing methods was performed. BMI was calculated using the information of weight (kg) and height (m) adjusted for the age. Level of obesity was categorized by BMI based on Asian criteria for obesity3,4: underweight (<18.5 kg/m2), normal (18.5 to <23 kg/m2), overweight (23 to <25 kg/m2), obese (25 to <30 kg/m2), and severe obese (≥30 kg/m2). Asian criteria differ slightly from those of Western countries. 24 Obesity-related comorbidities including hypertension, diabetes, dyslipidemia, coronary heart disease, angina pectoris and myocardial infarction, stroke, and osteoarthritis were investigated by Korean guidelines. 4 Hypertension was defined as physician diagnosis or hypertension criteria for blood pressure (ie, systolic pressure of >140 mm Hg or diastolic pressure of >90 mm Hg) by the National Cholesterol Education Program (NCEP). 25 Diabetes was defined as physician diagnosis or glucose level ≥126 mg/dL by the American Diabetes Association. 26 Dyslipidemia was defined as physician diagnosis or total cholesterol level of ≥240 mg/dL, low-density lipoprotein (LDL) cholesterol level of ≥160 mg/dL, HDL cholesterol level of <40 mg/dL, and triglyceride level of ≥200 mg/dL by the NCEP. 25 Coronary heart disease, including angina pectoris and myocardial infarction, stroke, and osteoarthritis, was defined as physician diagnosis.
Sociodemographic Variables
Socioeconomic status was adjusted for education and house income level from collected data in the survey. Level of education was categorized into 4 levels (ie, less than or equal to elementary school graduation, middle-high school graduation, high school graduation, and bachelor’s degree or higher). Household income was categorized into 4 levels by dividing into 4 annual quartiles (low, mid-lower, mid-upper, and high income).
Statistical Analysis
Descriptive statistics (ie, frequency and percentage) for general characteristics by obesity level were presented. Health value scores were summarized and graphed by sex and obesity level using means and 95% confidence intervals (CIs). To investigate the relationship between obesity level and good health state, multiple logistic regression was conducted after adjustment for socioeconomic variables. In multiple logistic regressions, the dependent variable was divided based on the mean value of normal weights of both men and women. Subgroup analysis according to age group and EQ-5D dimension was also performed.
To convert national representative values, survey sample design and unequal weights were considered in statistical analysis using SAS version 9.3 (SAS Inc, Cary, NC) and STATA 12 (Stata Corp, College Station, TX).
Results
We included 34 935 adult survey participants with complete BMI and EQ-5D data. Of the 34 935 adults, 28.0% (9767) were classified as obese, 3.8% (1326) as severely obese, and 23.6% (8249) as overweight (Table 1). Among severely obese people, the percentage of younger people who were aged 19 to 40 years was highest (38.8%) when compared with the other age groups (41-60 years and ≥61 years). People who were overweight (60.1%), obese (72.9%), and severely obese (76.8%) had more chronic diseases than people who were underweight (25.5%) and normal weight (41.9%). The mean EQ-5D score of all subjects was 0.949 and 0.914 in men and women, respectively.
General Characteristics of the Study Population Classified by Weight.a
Abbreviation: BMI, body mass index.
In general, the greater the degree of obesity, the lower HRQoL scores (Figure 1). The mean HRQoL was highest in individuals who were normal weight (0.937), followed by individuals who were underweight (0.934), overweight (0.931), obese (0.920), and severely obese (0.906). When grouped by gender, women showed a decrease of HRQoL as BMI increased. However, men did not show the same pattern; HRQoL of men became more increased slightly as BMI increased.
EQ-5D according to the BMI and sex.
The HRQoL for women by age group are presented in Figure 2. Obese and severely obese women aged 60 years or older had lower utility weight than women at normal BMI.
EQ-5D according to the BMI and age range in women.
Multiple logistic regression after adjustments for age, education, income, and chronic diseases showed that women who were obese had marginally lower HRQoL and women who were severely obese had 31% lower HRQoL than women of normal weight with statistical significance (95% CI = 1.12-1.53; Table 2). Men did not show similar trends. Lower education level, lower income level, and the presence of chronic disease were statistically significant factors for worse HRQoL.
Multiple Logistic Regression Analyses for Poor Health-Related Quality of Life (EQ-5D Male <0.946, Female <0.932) and Obesity.
Abbreviation: BMI, body mass index.
Women who were overweight, obese, and severely obese had higher odds ratio (OR) for having problems in each dimension (ie, mobility, self-care, usual activities, pain, and anxiety/depression; Table 3). The greater the increase in obesity level, the higher the odds ratio to have problems.
Logistic Regression Analyses of Preventing Any Problems in Each Dimension of the EQ-5D According to the Level of Obesity in Women.
Abbreviation: BMI, body mass index.
P < .001. *P < .05.
Discussion
Through this analysis, we found that women showed a decrease in HRQoL as obesity increased, but this trend was not shown in men. In multiple logistic regression after adjustments for age, education, income, and chronic diseases, women who were obese had marginally lower HRQoL and women who were severely obese had 31% greater lower HRQoL than women of normal weight, with statistical significance. The negative association between obesity and HRQoL was well known.14,15 Our study adds to what is known by demonstrating that HRQoL decreases as the extent of BMI increased. Subgroup analysis in men and women showed different results by sex. In a recent study by Choo et al in Korean population, obesity was associated with lower score of HRQoL in women, like in our study. 16 Our study has merits to use the recent data on large population with sufficient sample size and representation of the Korean population. The sample size of our study was more than 2.5 times the sample size in the study by Choo et al. We expanded the sample size using the KNHANES IV (2007-2009) and KNHANES V (2010-2012), while Choo et al included the KNHANES IV. We categorized the group according to the obesity level (underweight, normal weight, overweight, obesity, severe obesity), while the previous study included 3 groups (nonoverweight, overweight, obese). Thus, the trend could more clearly be shown by obesity stage and gender in our study. In the method of statistical analysis, we analyzed the HRQoL using the multiple logistic regression based on the mean value of men and women. On the other hand, the Choo et al study used multiple linear regressions for the HRQoL. Linear regression may be not appropriate for data having nonnormal distribution such as HRQoL (as data have celling effects because most people have perfect health of 1). Our study considered the characteristics of HRQoL data, and we analyzed the data using logistic regression.
Although the reasons for sex difference are not clear, we assumed some causes. First, obesity is defined as excess fat. 27 However, due to the difficulty of measuring fat, BMI is generally used as surrogate marker for obesity. BMI does not differentiate weight between muscle and fat. Usually, men have a higher percentage of muscle than women. Though a man and a woman could have identical BMIs, the man would most likely have a smaller percentage of body fat than the woman. 28 It seems more likely that BMI may be a poor indicator of levels of obesity in men. Second, the influence of obesity for mental health may be less in men. The social and peer pressures women face regarding their bodies and self-image is higher in Asian societies, as discussed earlier in the article. Third, difficulties in social activities including job opportunity based on obesity level may be more severe in women than in men. 29 Employment discrimination for obese women may affect those with less HRQoL. Fourth, chronic illness with pain such as osteoarthritis may have impacts on HRQoL. 30 Osteoarthritis is related with obesity and more prevalent in women. 31 It also affects those with HRQoL in obese women. Further research should be performed for identifying the reason.
For in-depth investigation between HRQoL and obesity by age group in women, mean value score of EQ-5D-3L for the divided group based on the age and obesity level was presented. The HRQoL of women ages 19 to 40 years was very similar regardless of the severity of obesity. While the HRQoL of people who were obese or severely obese was significantly decreased when compared with women of normal weight in the 41 years and older age group (Figure 2). As age increased, BMI of women has greater influence on HRQoL.
Obese women had more problems in all dimensions of the EQ-5D: mobility, self-care, usual activities, pain/discomfort, and anxiety/depression. The likelihood of having any problems increased as obesity levels increased in all dimensions. In patients who are severely obese, the likelihood of mobility problems was highest (OR = 3.20, P < .001), followed by usual activity (OR = 2.43, P < .001), self-care (OR = 2.35, P < .001), pain/discomfort (OR = 1.86, P < .001), and anxiety/depression (OR = 1.48, P < .001). The results for some dimensions were similar to a previous study by Sach et al. 15 In the previous study, mobility (OR = 2.76, P < .001) and pain (OR = 1.94, P < .001) had a statistically significant relationship with obesity.
Through the present study, we identified that mean HRQoL of men was higher than that of women (0.946 in men and 0.932 in women). In addition, the utility weight of all obesity levels except underweight was lower in women. This result was similar to the results of previous studies.9-12 Female patients with brain tumors, cardiac diseases, chronic diseases, or bipolar disorder had poorer HRQoL than men. The previous studies suggested that the reason for the lower HRQoL of women was lower physical function than men.11,13 Women are more likely to report lower physical function and experience greater impairments in physical life quality than men, and depression may be also proposed as the cause of lower HRQoL.10,12 In the study by da Rocha et al, HRQoL was lower and depressive symptoms were more severe in women with chronic diseases than in men; however, there was no significant difference found between healthy men and women. 12
The lower HRQoL in men and women was correlated with lower education, lower income, older age, and chronic disease after controlling for other factors. This result was in accordance with previous studies.14,15,32 Serrano-Aguilar et al reported that the lower HRQoL was correlated with lower education, older age, and chronic disease. 14 Sach et al and Lin et al also mentioned the correlation between HRQoL and chronic diseases.15,32 It seemed that excess weight as a chronic condition was one of the contributing factor for low HRQoL, because obesity was associated with various cardiovascular risk factors and diseases such as hypertension, diabetes, and heart disease.14,26 Among chronic diseases, osteoarthritis significantly affected the HRQoL. People with osteoarthritis had 3 times lower HRQoL than those without that disease. Previous studies have showed that the dimension of pain significantly affected the HRQoL of patients with osteoarthritis. 30 In particular, the effect of pain on lowering HRQoL was noticeable in individuals who were severely obese. Interestingly, disease-specific differences on HRQoL by sex were observed in our study. Whereas diabetes, chronic heart disease, stroke, and osteoarthritis affected HRQoL in both men and women, men with stroke and osteoarthritis tend to have higher odds ratios to have lower HRQoL than women. As far as we know, the study of disease-specific differences on HRQoL by sex is scarce. In Fang et al study, women with osteoarthritis had lower HRQoL than men, unlike our study. 33 In our study, osteoarthritis was prevalent in women (16.8% in women vs 5.1% in men), but women with osteoarthritis had lower odds ratios to have lower HRQoL than men. The sex difference between disease prevalence and impact of disease on HRQoL may show the complex aspects. Further research would be required.
The present study has certain limitations. First, this study was designed as a cross-sectional survey so it is impossible to confirm causality. All presented results showed the association not causality. Thus, our study does not fully explain the reason and mechanism for the association. The cause-and-effect relationship of obesity and HRQoL should be confirmed using longitudinal data. Second, even though we included main confounding factors such as demographic information, sociodemographic variables, and comorbidities in previous studies, unmeasured confounding factors (ie, unmeasured illness) were not adjusted. Third, we used the general instrument such as EQ-5D to identify HRQoL. EQ-5D did not include the questions on some specific obesity-related problems (ie, sexual life). In interpreting our results in particular regarding men, this limitation would be considered. Fourth, generally there was no clear definition of good health state. In our study, good health state was defined as having an average value or a higher value score than the average value score using EQ-5D-3L in our data. Readers should consider the limitation of subjective definition for good health state when they interpreted our study results. However, the subjective definition may not affect critically to see the association of obesity and HRQoL.
Conclusions
The present study highlights the sex differences in the association between obesity and HRQoL using the EQ-5D instrument in Korean population. Women were associated with decreasing HRQoL by increasing levels of obesity. On the contrary, the HRQoL of men was not significantly affected by obesity level. In addition, lower education level, lower income level, and the presence of chronic diseases were statistically significant factors for lower HRQoL in accordance with previous studies. Additional interventional studies would be needed to better understand the causality of the association between obesity and HRQoL in women.
Footnotes
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by the National Research Foundation of Korea(NRF) grant funded by the Korea government(MSIP) (No. 2015R1C1A2A01052768)