Prolonged or Transition to Metabolically Unhealthy Status, Regardless of Obesity Status, Is Associated with Higher Risk of Cardiovascular Disease Incidence and Mortality in Koreans

The risk of chronic disease and mortality may differ by metabolic health and obesity status and its transition. We investigated the risk of cardiovascular disease (CVD) and cancer incidence and mortality according to metabolic health and obesity status and their transition using the nationally representative Korea National Health and Nutrition Examination Survey (KNHANES) and the Ansan-Ansung (ASAS) cohort of the Korean Genome and Epidemiology Study. Participants that agreed to mortality linkage (n = 28,468 in KNHANES and n = 7530 adults in ASAS) were analyzed (mean follow-up: 8.2 and 17.4 years, respectively). Adults with no metabolic risk factors and BMI <25 or ≥25 kg/m2 were categorized as metabolically healthy non-obese (MHN) or metabolically healthy obese (MHO), respectively. Metabolically unhealthy non-obese (MUN) and metabolically unhealthy obese (MUO) adults had ≥1 metabolic risk factor and a BMI < or ≥25 kg/m2, respectively. In KNHANES participants, MUN, and MUO had higher risks for cardiovascular mortality, but not cancer mortality, compared with MHN adults. MHO had 47% and 35% lower risks of cancer mortality and all-cause mortality, respectively, compared to MHN. Similar results were observed in the ASAS participants. Compared to those persistently MHN, the risk of CVD was greater when continuously MUN or MUO. Transitioning from a metabolically healthy state to MUO also increased the risk of CVD. Few associations were found for cancer incidence. Using a nationally representative cohort and an 18-year follow-up cohort, we observed that the risk of CVD incidence and mortality and all-cause mortality, but not cancer incidence or mortality, increases with a continuous or a transition to an unhealthy metabolic status in Koreans.


Introduction
Cardiovascular disease (CVD) and cancer are the leading causes of death globally, accounting for 18.6 million and 10.0 million deaths, respectively, in recent years [1,2]. Obesity is positively associated with hypertension, diabetes, dyslipidemia, and increased risk of CVD, some cancers, and mortality [3][4][5]. However, a portion of obese adults has metabolically healthy phenotypes (normal blood pressure, glucose, and lipid levels). Whether these metabolically healthy obese (MHO) individuals are at higher risk of cardiovascular events, cancer, and mortality is not clear [6][7][8][9][10][11][12][13][14]. In addition, obesity and metabolic health status are status was missing (n = 35) and those with a history of CVD or cancer (n = 291) at baseline. None of the participants died within the first year of follow-up. Accordingly, 7530 participants were included in the analysis for a mortality follow-up. The flow charts of study participants for each study are shown in Figure 1.

Assessment of Metabolic Health Status and Obesity
Metabolic health status was defined by the presence of any of the three metabolic health risks: diabetes mellitus, hypertension, or dyslipidemia [28]. Diabetes mellitus was defined as having at least one of the following: fasting glucose ≥ 126 mg/dL, self-reported use of anti-diabetic medication, or a self-reported physician diagnosis of diabetes mellitus. The presence of hypertension was defined as a systolic blood pressure of ≥140 mmHg, diastolic blood pressure of ≥90 mmHg, or a self-report of taking antihypertensive mediations or a medical diagnosis of hypertension. Dyslipidemia was defined as a total cholesterol of ≥240 mg/dL, self-reported antihyperlipidemic medication use, or a self-reported medical diagnosis of dyslipidemia. The participants were classified as metabolically healthy (MH) or metabolically unhealthy (MU) based on the number of metabolic health risks (0 vs. 1-3, respectively). BMI was calculated as body weight (kg) divided by the square of the height (m 2 ). The participants were classified as non-obese (N) or obese (O) when the BMI was < or ≥25 kg/m 2 , respectively, according to the Asia-Pacific guidelines by the World Health Organization [40] and the guidelines by the Korean Society for the Study of Obesity [41]. Participants were then classified as metabolically healthy non-obese (MHN), metabolically unhealthy non-obese (MUN), metabolically healthy obese (MHO), or metabolically unhealthy obese (MUO). For KoGES participants, the combination of metabolic health and obesity status was defined at each visit. The transition of metabolic

Assessment of Metabolic Health Status and Obesity
Metabolic health status was defined by the presence of any of the three metabolic health risks: diabetes mellitus, hypertension, or dyslipidemia [28]. Diabetes mellitus was defined as having at least one of the following: fasting glucose ≥ 126 mg/dL, self-reported use of anti-diabetic medication, or a self-reported physician diagnosis of diabetes mellitus. The presence of hypertension was defined as a systolic blood pressure of ≥140 mmHg, diastolic blood pressure of ≥90 mmHg, or a self-report of taking antihypertensive mediations or a medical diagnosis of hypertension. Dyslipidemia was defined as a total cholesterol of ≥240 mg/dL, self-reported antihyperlipidemic medication use, or a self-reported medical diagnosis of dyslipidemia. The participants were classified as metabolically healthy (MH) or metabolically unhealthy (MU) based on the number of metabolic health risks (0 vs. 1-3, respectively). BMI was calculated as body weight (kg) divided by the square of the height (m 2 ). The participants were classified as non-obese (N) or obese (O) when the BMI was < or ≥25 kg/m 2 , respectively, according to the Asia-Pacific guidelines by the World Health Organization [40] and the guidelines by the Korean Society for the Study of Obesity [41]. Participants were then classified as metabolically healthy non-obese (MHN), metabolically unhealthy non-obese (MUN), metabolically healthy obese (MHO), or metabolically unhealthy obese (MUO). For KoGES participants, the combination of metabolic health and obesity status was defined at each visit. The transition of metabolic health-obesity phenotype was determined by comparing phenotypes at baseline and the fourth follow-up (2009-2010; n = 6665).

CVD and Cancer Incidence and Mortality Ascertainment
The biennial follow-up design of the KoGES cohort allows for the detection of a new onset of disease. The incidence of CVD was defined by a new self-report of at least one of the following in KoGES participants: physician diagnosis of CVD (i.e., myocardial infarction, congestive heart failure, coronary artery disease, cerebrovascular disease, or peripheral artery disease), use of stroke medication, or treatment for CVD. The new onset of cancer was also defined by a new self-report of at least one of the following in KoGES participants: physician diagnosis of cancer (i.e., lung cancer, gastric cancer, hepatocellular carcinoma, colorectal cancer, pancreatic cancer, uterine cancer, breast cancer, thyroid cancer, prostatic carcinoma, gallbladder cancer). Mortality data were ascertained by death records provided by Statistics Korea. The causes of death were classified according to the codes from the International Classification of Diseases, tenth version (ICD-10). We identified cardiovascular mortality (I00-I99; 778 deaths in KNHANES and 134 deaths in KoGES), cancer mortality (C00-D48; 1101 deaths in KNHANES and 227 deaths in KoGES), and all-cause mortality (3426 deaths in KNHANES and 624 deaths in KoGES). For sensitivity analyses, unnatural deaths due to external causes (V01-Y98; 330 deaths in KNHANES and 61 deaths in KoGES) were additionally excluded for all-cause mortality analyses.

Covariate Assessment
Demographic and behavioral data were obtained from questionnaires. Education level was categorized into four groups: elementary school graduate and below, middle school graduate, high school graduate, and university graduate and above. Household income was categorized into low (≤1,500,000 Korean won in KNHANES and <2,000,000 Korean won in KoGES) and high (>1,500,000 Korean won in KNHANES and ≥2,000,000 Korean won in KoGES). Smoking and drinking status were classified as current smokers/drinkers and non-smokers/drinkers. Current smokers/drinkers were defined as those who answered as smoking cigarettes or drinking alcoholic beverages at the time of the survey. The metabolic equivalents (METs), as a representative of physical activity, were calculated by summing the METs of each type of activity type (2.4 for light, 5.0 for moderate, and 7.5 for intense activities) multiplied by frequency/week.

Statistical Analysis
The characteristics of the four metabolic health and obesity status groups are presented as the mean (standard error of the mean) for continuous variables or as the number (percentage) for categorical variables and were compared by one-way analysis of variance or a chi-square test, respectively. For the assessment of mortality risk, participants were followed from survey entry to the date of death or the end of follow-up (31 December 2019), whichever occurred first. To investigate the risk of CVD and cancer incidence, KoGES participants were followed until the development of CVD or cancer or their last examination. The retention rates for KoGES were 85.8% (n = 8603), 74.9% (n = 7515), 66.7% (n = 6688), 66.5% (n = 6665), 62.2% (n = 6238), 58.9% (n = 5906), 63.0% (n = 6318), and 61.4% (n = 6157) at each follow-up survey, sequentially. All participants completed at least one follow-up. The Cox proportional hazard regression model was used to calculate hazard ratios (HRs) and 95% confidence intervals (CIs) for cancer incidence, cardiovascular events, cardiovascular and cancer mortality, and all-cause mortality according to metabolic health and obesity status. Additionally, the prevalence of each phenotypic group at each visit was examined in KoGES participants who completed all eight follow-up examinations (n = 3995). The association between the transition of phenotype from baseline to the fourth visit and cancer or cardiovascular event was investigated in participants with data for both visits (n = 6665). Cox models were adjusted for potential confounders: sex, age, residential area (urban or rural), education status, household income, current smoking status, current drinking status, and physical activity assessed as METs. For sensitivity analysis, we excluded underweight participants (BMI < 18.5 kg/m 2 ), as the higher mortality risk of underweight adults compared to normal-weight adults may have confounded the association with mortality [42][43][44]. The complex survey design was taken into account for all analyses of KNHANES. All analyses were performed with Stata SE 13.0 (Stata Corp, College Station, TX, USA), and a two-tailed p-value < 0.05 was considered statistically significant.

Baseline Characteristics of KNHANES Study Participants
Among the 28,468 participants in the nationally representative KNHANES study, 34.9% (n = 9944) of the participants were obese, and 54.0% (n = 15,377) were metabolically unhealthy ( Results were similar when excluding underweight participants (data not shown). When obesity status was stratified as underweight, normal weight, and obese, metabolically unhealthy adults had a higher risk for cardiovascular mortality regardless of BMI (data not shown). In addition, the number of metabolic risks was positively associated with a risk for cardiovascular and all-cause mortality, regardless of obesity status (Table S1). These results indicate that metabolic health may be a greater determinant of cardiovascular and all-cause mortality than obesity, but not of cancer. Obesity may rather be protective with respect to cancer mortality when metabolically healthy (HR: 0.53 [95% CI: 0.34-0.82]) for MHO compared to MHN). Results were similar when excluding deaths due to external causes (data not shown) and were similar between men and women when stratified by sex (Table S2).

Baseline Characteristics of KoGES Study Participants
Currently, KNHANES does not have longitudinal follow-up data of participants, thus we are unable to measure the transition of metabolic health and obesity status or assess the incidence of disease. To investigate the effects of transition of metabolic health and obesity status and disease incidence using the community-based ASAS cohort of KoGES, we first replicated the results of KNHANES in the ASAS cohort. Participants (n = 7530) were followed for a mean of 17.4 years. Among the participants, 2937 (39.0%), 1356 (18.0%), 1579 (21.0%), and 1658 (22.0%) were classified as MHN, MUN, MHO, and MUO, respectively, at baseline (Table 3). Similar to the KNHANES participants, the metabolically healthy participants were younger than the metabolically unhealthy participants. The mean BMI was approximately 22.5 kg/m 2 in the non-obese groups and 27.4 kg/m 2 in the obese groups, which is comparable to KNHANES participants. The prevalence of diabetes mellitus, hypertension, and dyslipidemia was 19.5%, 73.8%, and 26.2% among subjects in MUN, and 18.8%, 79.0%, and 26.1% among MUO subjects, respectively. The mean number of metabolic health risks was higher in MUO compared to MUN adults (1.24 ± 0.01 vs. 1.20 ± 0.01, p < 0.001).

Transition of Metabolic Health and Obesity Phenotypes and the Risk of CVD and Cancer
Participants with both baseline and the fourth KoGES follow-up data were included in the analyses of transition of metabolic and obesity status and disease incidence (n = 6665). These participants had similar characteristics to the above total KoGES mortality study population (data not shown). The prevalence of the four metabolic health-obesity phenotypes at each visit during the 16-year follow-up is presented in Figure 2. The prevalence of obesity was consistent at approximately 43% during the years. Notably, the proportion of MHN and MHO adults decreased by 57.3% and 49.1%, respectively, resulting in an increase in metabolically unhealthy adults (66.3% of the population by the eighth follow-up). The eight-year transition (from baseline to the 4th follow-up) of metabolic health and obesity phenotypes shows that the proportion of maintaining each metabolic health-obesity phenotype was relatively high in metabolically unhealthy participants (70.7% and 71.4% for  Figure 3). However, only 45.9% of the MHO participants maintained this phenotype while 33.2% progressed to MUO. study population (data not shown). The prevalence of the four metabolic health-obesity phenotypes at each visit during the 16-year follow-up is presented in Figure 2. The prevalence of obesity was consistent at approximately 43% during the years. Notably, the proportion of MHN and MHO adults decreased by 57.3% and 49.1%, respectively, resulting in an increase in metabolically unhealthy adults (66.3% of the population by the eighth follow-up). The eight-year transition (from baseline to the 4th follow-up) of metabolic health and obesity phenotypes shows that the proportion of maintaining each metabolic health-obesity phenotype was relatively high in metabolically unhealthy participants (70.7% and 71.4% for MUN and MUO, respectively; Figure 3). However, only 45.9% of the MHO participants maintained this phenotype while 33.2% progressed to MUO.   The risk of CVD incidence was assessed according to the transition of metabolic obesity phenotypes (Figure 4). Compared to individuals consistently MHN, those persisted to be metabolically unhealthy were at more than a two-fold greater risk for incidence (  The risk of CVD incidence was assessed according to the transition of metabolic and obesity phenotypes (Figure 4). Compared to individuals consistently MHN, those that persisted to be metabolically unhealthy were at more than a two-fold greater risk for CVD incidence  Compared to adults with consistent MHN, stable MHO had a lower risk of cancer incidence, though the HR did not reach significance (HR: 0.76 [95% CI: 0.48-1.18]; Table  5). Few associations were observed between the transition of metabolic health and obesity status and cancer incidence. Compared to those with a stable MHO phenotype, adults that transitioned from MHO to MHN had a higher risk for cancer (HR: 1.95 [95% CI: 1.02-3.76]). These results indicate that MHO may be more protective of cancer incidence compared to MHN. Compared to adults with consistent MHN, stable MHO had a lower risk of cancer incidence, though the HR did not reach significance (HR: 0.76 [95% CI: 0.48-1.18]; Table 5).
Few associations were observed between the transition of metabolic health and obesity status and cancer incidence. Compared to those with a stable MHO phenotype, adults that transitioned from MHO to MHN had a higher risk for cancer (HR: 1.95 [95% CI: 1.02-3.76]). These results indicate that MHO may be more protective of cancer incidence compared to MHN. Table 5. Hazard ratios (HRs) and 95% confidence intervals (CIs) for cancer incidence according to the combination of metabolic health risks and obesity status at baseline and fourth follow-up.

Discussion
The risks of CVD, cancer and all-cause mortality, according to metabolic health and obesity status, were examined in Koreans using two large databases-a nationally representative database and the ASAS cohort of KoGES. Additionally, we examined the risks of the incidence and mortality of CVD and cancer and the risk of all-cause mortality according to metabolic health and obesity status and its transition in Koreans using a biennial 18-year follow-up cohort. In both cohorts, individuals with metabolically unhealthy phenotypes exhibited a higher risk for cardiovascular and all-cause mortality than MHN individuals. Among metabolically healthy adults, transition to MUO increased the risk of CVD incidence. In metabolically unhealthy adults, changing to a healthy metabolic profile or non-obese weight decreased the risk of CVD incidence. On the other hand, MHO adults did not have an increased risk for cancer incidence, cardiovascular or cancer mortality, or all-cause mortality compared to MHN.
The relationships of baseline metabolic health and obesity status with cardiovascular, cancer, and all-cause mortality were similar in both KNHANES and KoGES cohorts, despite some discrepancies in baseline characteristics and length of follow-up. Compared to the nationally representative KNHANES participants, the KoGES ASAS cohort participants were physically active, more likely to live in rural areas, be less educated, have lower household income, and refrain from alcoholic beverages. Still, overall, the directions of the HR were very similar between the two cohorts. Our analyses of nationally representative data (KNHANES) with 8 years of follow-up were similar to that of KoGES (17 years of follow-up). The results were also in line with some previous studies despite the participants' younger mean age [30] or shorter follow-up within 5 years [24,29] or 10 years [31,32] in other studies. The use of KNHANES ensures to capture a nationally representative sample of noninstitutionalized adults. Institutionalized adults may have underlying health conditions, which may introduce bias to the present results. On the other hand, although previous large studies in Koreans using the National Health Insurance System database excluded patients previously diagnosed with CVD or cancer, they did not exclude hospitalized patients or adults with other severe medical issues [24,29]. In addition, these previous studies utilized medical expense claim data of which approximately 30% do not match the primary diagnostics codes, resulting in possible errors in identifying CVD and cancer incidence [27,28]. Still, similar results indicate that metabolically unhealthy Korean adults have a higher risk of mortality.
The present study shows that both long-term and short-term metabolic health and obesity are associated with cardiovascular morbidity and mortality, in addition to all-cause mortality. Compared to those constantly MHN, adults that maintained MUN or MUO phenotypes from baseline throughout the follow-up had higher risks of CVD incidence. However, the risk of CVD was weaker in adults that transitioned from MHN to a metabolically unhealthy state (MUN or MUO; Figure 4) indicating that a longer duration of a metabolically unhealthy state may worsen the risk of CVD. Still, relatively short-term metabolic health status cannot be ignored. Among MHO adults, those that transitioned to MUO had an increased risk of CVD compared to adults that were continuously MHO, in line with previous studies in Chinese with a shorter follow-up [18]. In addition, recovery from a metabolically unhealthy phenotype to a metabolically healthy phenotype decreased the risk of CVD incidence. This may explain why studies with relatively short follow-ups have similar results [18,24,29]. Our results show that a prolonged metabolically deleterious status may increase CVD incidence and mortality risk, but this risk may be attenuated if the duration is shortened. Therefore, continuous efforts to improve and maintain metabolic health are required to prevent cardiovascular and all-cause mortality.
We show that compared to adults MHN at baseline, those that were MHO have a higher risk of CVD incidence, but a lower risk of cardiovascular mortality, cancer incidence, and mortality, and all-cause mortality. Several previous studies, including those with Asians, have similarly reported that MHO individuals have an increased risk for incident CVD than MHN controls [10,11,24,[32][33][34] but lower risk of cardiovascular mortality [24,[29][30][31]. The inconsistency between the higher risk of CVD incidence and the lower risk of CVD mortality in MHO adults compared to MHN adults is not fully understood. One possible explanation is that as a high proportion of adults that were MHO at baseline transition to MUO, the risk of CVD also increases. However, due to the relatively short duration of being metabolically unhealthy, the risk of death is similar to MHN and not as high as those that were MUO from baseline. Another possibility is that cardiovascular events may be worsened by detrimental metabolic status resulting in fatal cardiovascular events in MUO adults, but not in MHO individuals. Regarding cancer, we found that the MHO phenotype may be protective of cancer incidence and mortality. These results are consistent with some previous reports in the US [34,45,46]. In Koreans, previous studies resulted in a higher risk of breast cancer or colorectal cancer incidence in MHO compared to MHN [35,36]. We did not observe a higher risk of cancer incidence, possibly owing to that we were unable to assess incidence by cancer type. In addition, breast and prostate cancers have a relatively high survival rate [47]. Therefore, the incidence of cancer may not have affected cancer mortality, reflecting the lower risk of total cancer mortality of MHO adults in our study and previous studies with a shorter follow-up [28,37]. With respect to all-cause mortality, the risk was similar to or higher in MHO compared to metabolically healthy and normal-weight Whites [31,48,49]. However, a meta-analysis in Asians resulted in a lower risk of all-cause mortality in MHO adults compared to MHN adults [28,50], similar to the results of the current study. The mechanism for the protective effect of obesity on cancer and mortality in metabolically healthy adults is not clear [12]. Our results, in addition to that of others [50], indicate that racial differences may exist, especially for all-cause mortality. For instance, in the US, Asian-Americans have a higher rate of diabetes mellitus compared to non-Hispanic Whites despite their lower BMI [51]. As diabetes mellitus is affected by genetic factors in addition to obesity and other lifestyle factors [52], the relatively low proportion of Asians that remain metabolically healthy despite obesity in their older years may also be genetically protected from cardiovascular, cancer, and all-cause mortality. More research on genetic or epigenetic effects is needed. Nevertheless, although obese, attaining metabolic health may prevent cardiovascular deaths, but not events, and cancer morbidity and mortality in Koreans.
Our study has some limitations. Although the follow-up period was as long as 18 years for KoGES participants, we did not have a sufficient number of deaths to calculate the effect of metabolic health and obesity status transition on mortality. In addition, due to the limited number of events, we were not able to assess the risk according to specific cardiovascular events or specific cancer types. As participants were community-dwelling adults, adults in nursing homes or other care facilities were not surveyed. Therefore, the results may not be generalizable to institutionalized adults. However, adults residing in care facilities are likely to have multiple health complications which may compromise the results. Obesity was assessed by BMI, not the proportion of body fat, a more precise measure of obesity. In addition, KNHANES and KoGES do not provide data on fat distribution, such as visceral fat. Despite these limitations, we believe that our study has its strengths. This is the first study to prospectively utilize the nationally representative KNHANES database, linked with mortality data, to assess the association between metabolic health and obesity status and mortality. In addition, the use of the follow-up data of the KoGES database enabled us to assess the transition of metabolic health and obesity status on CVD and cancer incidence. To reduce misclassification of obesity and metabolic health status, we adopted a strict MHO definition using anthropometric and biochemical measurements, as well as the use of medications, and self-reported doctor-diagnosis of each metabolic health risk. These may be more accurate than national health insurance claim data as used in some previous studies. The exclusion of those with CVD or cancer at baseline and the long follow-up suggests the possible causal relationship of metabolic health and obesity with CVD incidence and mortality. The use of the two databases, the long mean follow-up of KoGES (>17 years), and the consistent results strongly support the conclusions. Excluding deaths from external causes resulted in similar results, strengthening our confidence in the association between the risk of death and participants' metabolic and obesity status.

Conclusions
In conclusion, we demonstrated through a nationally representative database and an 18-year biennial prospective cohort that poor metabolic health is associated with a higher risk of cardiovascular and all-cause mortality, while MHO is associated with a lower risk of cancer incidence and mortality and all-cause mortality. A persistent or a transition to unhealthy metabolic profiles increases the risk of cardiovascular incidence, emphasizing the importance of management of metabolic health.
Supplementary Materials: The following are available online at https://www.mdpi.com/article/10 .3390/nu14081644/s1, Table S1: Hazard ratios (HRs) and 95% confidence intervals (CIs) for cardiovascular, cancer, and all-cause mortality according to the number of metabolic health risks and obesity status in KNHANES participants, Table S2: Hazard ratios (HRs) and 95% confidence intervals (CIs) for the all-cause mortality and cause-specific mortality according to the combination of metabolic health risks and obesity status by sex in KNHANES.