Association of Metabolic Health and Central Obesity with the Risk of Thyroid Cancer: Data from the Korean Genome and Epidemiology Study

Korea has the highest incidence of thyroid cancer in the world, according to the Global Cancer Observatory in 2018 (1). Since 1999, the incidence has increased by 22.3% per year; since 2009, thyroid cancer has been one of the most prevalent cancer in Korea (2). Estimates indicated that thyroid cancer would remain one of the cancer types with the highest incidence in Korea until 2020, especially among women (3).

The increase in the incidence could be attributed to screening and overdiagnosis during the same period in Korea (4–6). However, the increase in thyroid cancer incidence cannot be fully explained by improved screening and diagnosis, which leave out the vital role of behavioral and biological factors of which body mass index (BMI) is the most well-known factor (7–9). Likewise, from 1998 to 2018, the age-standardized prevalence of obesity in Korea, especially severe obesity, increased (10), suggesting a possible relationship of obesity with cancer development.

Many prospective studies have reported the long-term associations between obesity and thyroid cancer (11–16). The evidence from meta-analyses (17, 18) and large-scaled pooled analyses (19–21) have indicated that obesity is associated with thyroid cancer, although minor inconsistent results were reported in study populations in Korea and Japan (13, 14). Obesity has been proposed to closely relate to thyroid cancer through many factors: thyroid hormones, insulin resistance, adipokines, inflammation, sex hormones, and serum triglyceride concentrations (22–25). In addition, nonobese individuals with metabolic abnormalities, such as central obesity, glucose intolerance, increased blood pressure (BP), or dyslipidemia, have a higher cancer risk than those without obesity and metabolic abnormalities (26, 27). Individuals who are “metabolically healthy obese” are defined as those with an increased BMI (≥25 kg/m²), but with no or few metabolic abnormalities, such as central obesity (waist circumference ≥90 and ≥85 cm for men and women, respectively), glucose intolerance (fasting glucose concentration ≥100 mg/dL), increased blood pressure (BP; systolic and diastolic BP ≥130 and ≥85 mmHg, respectively), and dyslipidemia (triglyceride level ≥150 mg/dL, high-density lipoprotein [HDL]-cholesterol levels <40 and <50 mg/dL for men and women, respectively; refs. 28, 29). Conversely, metabolically unhealthy, normal-weight individuals, who have a normal BMI (<25 kg/m²), but various metabolic abnormalities, have a high risk of various diseases and a high mortality rate (30–34). Metabolically healthy obese and metabolically unhealthy normal weight groups have not been sufficiently investigated with regard to the risk of thyroid cancer. Thus, the risk of thyroid cancer among metabolically healthy obese individuals has gained interest. It is unknown whether the risk of thyroid cancer differs between metabolically healthy and obese and metabolically unhealthy and obese individuals. There is limited information on the potential differential association between metabolic abnormality and thyroid cancer among people with normal BMI or overweight/obesity.

In this context, this study investigated the associations between metabolic abnormality and obesity, and the risk of thyroid cancer using data from a population-based prospective cohort. The association of obesity, measured by BMI and metabolic components (including waist circumference), with thyroid cancer risk was assessed.

This study analyzed data collected from the population-based cohort of the Korean Genome and Epidemiology Study (KoGES), including the KoGES_Ansan and Ansung Study, the KoGES_Cardiovascular Disease Association Study (CAVAS), and the KoGES_Health Examinee Study (HEXA), which was conducted by the Korean government through the National Research Institute of Health, Centers for Disease Control and Prevention, and the Ministry of Health and Welfare. The KoGES_HEXA is one of three cohorts that were included in the Korean Genome and Epidemiology Study. To collect participants' genetic, environmental, and lifestyle information that could affect cancer expression, potential participants from the Korean population, both men and women (age ≥40 years), were recruited from the National Health Examinee Registry. Detailed information on the study's design and data-collection procedures were reported previously (35). Participants ages 40 to 79 years were recruited from the National Health Examinee Registry, which is part of the national insurance program that provides fully paid biannual health check-ups for subscribers. Participants visited medical institutions nationwide, which were mainly general hospitals in the metropolitan areas and major cities of Korea. Eligible individuals were invited to voluntarily participate in the study. For baseline and follow-up data collection (including questionnaires, physical examinations, and clinical investigations), participants were invited to visit the survey sites and were invited to complete the follow-up surveys by postal mail and telephone periodically.

A total of 173,343 participants were recruited between 2004 and 2013 for the baseline survey and were asked to attend follow-up visits between 2007 and 2016 (mean follow-up duration, 7.4 years). Data generated from the baseline to the initial follow-up were analyzed in this study. We excluded participants who had thyroid cancer (n = 959), other cancers (n = 3,035), or indeterminate cancer status (n = 1,596) based on the questionnaire completed at the baseline. We excluded individuals without data on anthropometric measurements related to BMI (n = 912); serum-based measures on triglycerides, HDL-cholesterol, and fasting glucose; hypertension; or waist circumference (n = 6,191). A total of 160,650 participants (55,252 men and 105,398 women) were included in the final analysis (Fig. 1). For sensitivity analysis, to reduce the potential bias from reverse-causal effects of tumors prior to diagnosis, we excluded individuals diagnosed with thyroid cancer <12 months after enrollment (n = 88, incident thyroid cancers).

Data were collected by trained personnel at survey sites during the baseline and follow-up phases. Demographic information, menopausal status, lifestyle (smoking, alcohol drinking, physical activity, and undergoing menopausal hormone therapy), and a history of cancers or other diseases (i.e., disease status of the participants and his/her family members) were obtained using structured interviewer-administered questionnaires.

All the study participants provided samples for biomarker measurements on the same day as the questionnaire survey according to standard procedures. Anthropometric data and clinical characteristics were obtained by trained personnel to evaluate participants' current health statuses. The body height and weight of the participants were measured barefoot while wearing light indoor clothing; BMI was calculated as: weight (kg)/height (m)². Waist circumference was measured in the horizontal plane midway between the lowest rib and the iliac crest. BP in the right arm was measured using a standard mercury sphygmomanometer (Baumanometer; W.A. Baum Co. Inc.) after the participant had rested in a quiet room for 5 minutes.

Blood samples were collected from all the study participants after at least 10 hours of fasting, and all samples were transported to the National Biobank of Korea for evaluations in the same core clinical laboratory that is accredited and participates annually in inspections and surveys conducted by the Korean Association of Quality Assurance for Clinical Laboratories. Blood concentrations of fasting glucose, TC, HDL-cholesterol, and TG and serum concentrations of HDL-cholesterol and TG were measured using enzymatic methods (ADVIA 1650 and ADVIA 1800; Siemens Healthineers, Deerfield and ADVIA 1650 Chemistry System). Sociodemographic information, family and personal medical history, nutritional intake, and consistency of physical activity were obtained during the examination (35).

An individual was considered “obese” if he/she had a BMI ≥25 kg/m² (36). In reference to the modified principles of the National Cholesterol Education Program Adult Treatment Panel III, metabolic abnormality was defined as the presence of three or more of the following unhealthy factors (37): triglyceride level ≥150 mg/dL or treatment for hyperlipidemia; systolic and diastolic BP ≥130 and ≥85 mmHg, respectively, or receiving antihypertensive treatment; HDL-cholesterol levels <40 and <50 mg/dL for men and women, respectively; fasting glucose concentration ≥100 mg/dL or antidiabetic treatment; waist circumference ≥90 and ≥85 cm for men and women, respectively, defined according to the optimal cutoff values for abdominal obesity among Korean adults (38), which was selected due to its stringent criteria. There are four phenotypes in the combination of obesity and metabolic status: metabolically healthy and normal weight (BMI <25 kg/m²; reference group); metabolically healthy and obese (BMI ≥25 kg/m²); metabolically unhealthy and normal weight; and metabolically unhealthy and obese. Thyroid cancer risk was further investigated by combining general obesity (BMI ≥25 kg/m²) and central obesity (waist circumference ≥90 and ≥85 cm for men and women, respectively) or HDL-cholesterol (<40 and <50 mg/dL for men and women, respectively).

Information on incident thyroid cancer was collected by the interviewer based on self-reported status based on diagnosis by a medical doctor. Participants were queried through follow-up questionnaires at least annually about new medical events.

Baseline characteristics are described as means and standard deviations (SD; for continuous variables) or frequencies and percentages (for categorical variables). We used multivariate Cox proportional hazards (PH) regression to assess HRs and their 95% CIs to investigate the association of metabolic health and obesity with the risk of thyroid cancer. The PH assumption was checked by evaluating the correlation between Schoenfeld residuals and time. There was no major violation of the PH assumption (P > 0.05). The time variable was measured from the date of the baseline visit to the date of diagnosis of thyroid cancer, date of death or development of other cancers, or date of the last follow-up visit (all rounded according to month), whichever occurred first. The model was initially adjusted for age at recruitment into the study (as continuous data, by year) and sex and then further adjusted for smoking (never, past, and current smoker), education (elementary school graduation; lower school, middle school, or high school graduation; and college graduation or higher), alcohol consumption (<4 drink and ≥4 drink per week), physical activity (<3 times/week or <30 minutes each time; ≥3 times/week and ≥30 minutes each time), and undergoing menopausal hormone therapy (never, past, and current). Sex-stratified subgroup analysis was conducted using similar covariates when the sexes were combined (excluding the sex variable). The data of women were further adjusted for menopausal status (premenopausal and postmenopausal). In addition, the models were further refined by excluding individuals who were diagnosed with thyroid cancer <12 months from the baseline visit to avoid reverse causality. BMI, triglycerides, blood pressure, HDL-cholesterol, blood glucose, and waist circumference were shown to be correlated in multicollinearity test. Therefore, those variables were not included in the same model.

The likelihood ratio test was performed to investigate whether metabolic healthy status and high BMI had multiplicative joint effects on the risk of thyroid cancer in the entire population and sex strata. We compared the log-likelihoods of models with and without the interaction term of metabolic health*BMI. For all the models, the tests showed insignificant results (P > 0.05). Therefore, we did not include that interaction term for further analysis. All statistical analyses were performed using R version 4.0.0 (The R Foundation for Statistical Computing, Vienna, Austria) with a predetermined two-sided significance level of 0.05.

All participants provided written informed consent before the study and the study protocol was approved by the Institutional Review Board of the National Cancer Center (IRB no. NCC2018–0164).

Raw data were generated at the National Research Institute of Health, Centers for Disease Control and Prevention and the Ministry of Health and Welfare, Republic of Korea. Derived data supporting the findings of this study are available from the corresponding author M.K.K. on request.

During 299,386 person-years of follow-up, we identified 471 cases of first primary thyroid cancer, 44 of which were diagnosed among men and 427 among women. From 1 year after enrollment, 383 thyroid cancer cases (men, 39 cases; women, 344 cases) were diagnosed. Among all the phenotype groups, men had higher blood pressure, triglycerides, fasting glucose, waist circumference, and lower HDL-cholesterol than women. There were no remarkable distinctions in the extent of physical activity among the groups. In women, the proportion of postmenopausal individuals increased in obese and/or metabolically unhealthy groups. The statuses of menopausal hormone therapy were similar among groups of women. Overall, metabolic unhealthy people had a higher mean age, lower educational level, and higher rate of alcohol consumption of four times or more per week (Table 1).

The associations between metabolic phenotypes and incident thyroid cancer risks are shown in Table 2. After multivariate adjustment, the metabolically unhealthy and obese phenotype was associated with a higher risk (HR = 1.97; 95% CI, 1.46–2.63) of thyroid cancer, compared with the metabolically healthy and normal weight phenotype among women. After excluding the observations with a follow-up time of less than 1 year, individuals with the metabolically unhealthy and normal weight or metabolically unhealthy and obese phenotype had a similar higher risk of thyroid cancer (HR = 1.57; 95% CI, 1.02–2.40 and HR = 1.71; 95% CI, 1.21–2.41, respectively), compared with the metabolically healthy and normal weight phenotype among women. No significant associations were observed between metabolic phenotypes and thyroid cancer among men. Although estimations were not highly reliable due to the small number of cases, obese and/or metabolically unhealthy men consistently had elevated HRs of thyroid cancer, especially the men with metabolically unhealthy and obese (HR = 1.17; 95% CI, 0.43–3.02).

Figure 2 presents the Cox proportional hazard regression estimates of survival curves for the time to incident thyroid cancer, stratified by phenotypes of metabolic health status and obesity (excluding thyroid cancer cases diagnosed less than 1 year from the baseline visit). Among women, along with participants with metabolically unhealthy and obese, those with metabolically unhealthy and normal weight appeared to have higher thyroid cancer risks, compared with women with metabolically healthy and normal weight. In men, the hazard of thyroid cancer development was not distinct among the phenotypic groups.

After multivariate adjustment for covariates and excluding those who had been enrolled for less than 1 year, a high BMI (HR = 1.27; 95% CI, 1.00–1.62), low HDL-cholesterol (HR = 1.49; 95% CI, 1.19–1.87), and high waist circumference (HR = 1.34; 95% CI, 1.03–1.74) yielded a significantly increased risk of thyroid cancer among women (Table 3). No significant association was found among men.

As shown in Table 4, the combined impacts of BMI and waist circumference were assessed in overall subjects and across sex strata. In the overall population, regardless of participants followed-up for less than 1 year, participants with normal BMI and high waist circumference, and vice versa had higher risks of thyroid cancer than participants with normal BMI and normal waist circumference (HR = 1.58; 95% CI, 1.02–2.43 and HR = 1.42; 95% CI, 1.05–1.91, respectively). In women, before the exclusion of thyroid cancer cases diagnosed with 1 year from the baseline visit, compared with women with normal BMI and low waist circumference, women with normal BMI and high waist circumference had a significantly higher HR (1.60; 95% CI, 1.05–2.42). After the exclusion, women with normal BMI and high waist circumference had still significantly elevated risk of thyroid cancer (HR = 1.62; 95% CI, 1.03–2.56). In contrast to the women, men with high BMI and normal waist circumference had higher HRs for thyroid cancer (HR = 2.34; 95% CI, 1.03–5.28), compared with the reference group. Similar associations were observed before and after the exclusion.

In Table 5, the combination associations between HDL-cholesterol, BMI and thyroid cancer were evaluated. Women with low HDL-cholesterol retained a significantly higher HR after the exclusion, regardless of BMI status (HR = 1.49; 95% CI, 1.13–1.96, and HR = 1.75; 95% CI, 1.26–2.42), compared with those with normal BMI and high waist circumference. Men with both high BMI and low HDL-cholesterol had a significantly elevated risk of thyroid cancer (HR = 2.60; 95% CI, 1.05–6.47), compared with those with normal BMI and high waist circumference.

In this prospective study, we found that women who were metabolically unhealthy (defined as having a normal weight with three or more metabolic abnormalities, including central obesity, type 2 diabetes, dyslipidemia, and an elevated BP) had a significantly increased risk of thyroid cancer, regardless of obesity status. Moreover, women with low HDL-cholesterol and normal weight and high waist circumference had an increased risk of thyroid cancer. Different results were observed in men: participants with high-BMI and low waist circumference or low HDL-cholesterol had elevated thyroid cancer risk, although the estimates for some phenotypes were unreliable due to small numbers of cases.

The role of obesity (determined by abnormalities of BMI, waist circumference, and metabolic status) on the incidence of thyroid cancer has been reported in previous large-scale prospective studies across different populations. Meta-analyses and pooled analyses yielded results that agree regarding the influences of generalized and central obesity on thyroid cancer incidence (17–21). According to a European cohort study, a high blood glucose level was associated with a decreased risk of thyroid cancer among women, whereas BMI and metabolic factors did not affect the outcome among men (39). However, in an Austrian cohort, a high glucose level was related to an increased risk of thyroid cancer when both sexes were combined, whereas no significant results were observed in the women's groups (40). Another cohort study conducted in Austria showed that an elevated serum triglyceride concentration may lead to an increase in thyroid cancer development in both sexes but not in men or women separately (23). Moreover, in this study, besides the metabolically unhealthy and obese, the metabolically unhealthy and normal weight, but not the metabolically healthy and obese, was shown to increase the risk of thyroid cancer among women, suggesting metabolic unhealthiness as a more important risk factor than overall obesity. This is consistent with the results of another cohort study conducted in Korea (41) with different criteria for phenotype classification, wherein metabolically unhealthy persons were defined as those having at least one of the following metabolic abnormalities: (i) fasting glucose level ≥100 mg/dL or current use of glucose-lowering agents; (ii) BP ≥130/85 mmHg or current use of BP-lowering agents; (iii) elevated triglyceride level (≥150 mg/dL) or current use of lipid-lowering agents; (iv) low HDL-C (<40 mg/dL in men or <50 mg/dL in women), and (v) insulin resistance, defined as a HOMA-IR score ≥2.5. This Korean cohort study indicated that women with metabolically unhealthy obesity but not metabolically healthy obesity were found to have an increased risk of thyroid cancer. However, a Korean cohort study, which utilized the same criteria with this study (37) showed that both metabolically unhealthy and normal weight and metabolically healthy and obese had a significant association with thyroid cancer development among women (42). The metabolic features of many diseases, such as diabetes, cardiovascular, neurological, obesity, and aging, overlap with cancer metabolism (27). The metabolic overlap between cancer and the abovementioned diseases might elucidate how metabolic health is intertwined with cancer development. Individuals with metabolically unhealthy and normal weight, compared with those with metabolically healthy and normal weight or even those with metabolically healthy and obese, have a higher risk of several diseases such as cardiovascular disease (43) and type 2 diabetes (44). Persons with metabolically unhealthy and normal weight exhibited a greater all-cause mortality rate than those with metabolically healthy and obese in a prospective study of Korean elderly participants (45).

Obesity coexists with metabolic abnormalities. However, it is unclear whether the increased risk of thyroid cancer is associated with obesity per se or the presence of co-existing metabolic abnormalities (39, 46, 47) because most previous studies evaluated the association between BMI and the risk of thyroid cancer without considering the metabolic status associated with obesity. In our study, women with a BMI within the normal range and metabolic abnormality (metabolically unhealthy and normal weight phenotype) had an increased risk of thyroid cancer (HR = 1.58; 95% CI, 1.03–2.24). Furthermore, the risk of thyroid cancer was consistently elevated among women with a normal BMI and waist circumference ≥85 cm (central obesity phenotype; HR = 1.61; 95% CI, 1.02–2.54). Central obesity might be the driver of the association between metabolic phenotype and the risk of thyroid cancer. Similar results were observed among persons with breast cancer (48). Central body fat has been consistently shown to be closely related to the occurrence of type 2 diabetes mellitus, independently of BMI (49, 50). In this study, there was no significant difference in the association between central obesity and the risk of thyroid cancer risk among women with a high BMI. It was suggested that BMI may result in the misclassification of body fat distribution due to the different contributions of bone mass, muscle mass, and fluid to the body weight (51). Therefore, the findings of the present study suggest that excessive central obesity may be a better predictor of metabolically related diseases beyond BMI. Regarding men, there was a very distinct result: high BMI was associated with thyroid cancer in low waist circumference men. This was similar with the statement of a previous study in Korea (41): obesity was associated with an increased risk of incident thyroid cancer in both metabolically healthy and metabolically unhealthy men. Little is known about the independent impact of HDL-cholesterol on thyroid cancer. This study may be the first to suggest that HDL-cholesterol increased the risk of thyroid cancer in women (regardless of BMI) and men with high BMI.

Possible mechanisms have been suggested for the association between obesity, metabolic health, and the risk of thyroid cancer. Circulating levels of monocyte chemoattractant protein-1 (MCP-1) and IL8 are positively related to both BMI and metabolic factors such as waist circumference, C-reactive protein level, and the homeostasis model assessment score and is negatively associated with the HDL-cholesterol level (52), suggesting that their role may underlie inflammation-related complications, including cancer. Insulin resistance, adipocytokines, and inflammation, which result from the expansion of adipose tissue, interact to create a vicious cycle that contributes to the pathogenesis of thyroid carcinoma (53). Furthermore, reproductive factors, such as parity, menstruation, and breastfeeding, influenced the risk of thyroid cancer among women in a pooled analysis based on 25 studies (54). This may explain why women have a higher incidence of thyroid cancer than men.

This study had several limitations. First, the incidence of thyroid cancer identified in this study may differ from the incidence of thyroid cancer in the general population because the participants regularly participated in health screening examinations. However, a small discrepancy was expected given the universal health coverage in Korea (55), to ensure the representativeness of the study population for people aged 40 and older. Finally, due to the low number of cases among men, the results related to sex differences should be interpreted with care. In this study, the ratio of men to women thyroid cancer incidence is similar to its respective ratio in the Korean Central Cancer Registry (2). Therefore, these results can be extrapolated to the general Korean population.

Nevertheless, the study had several strengths, in terms of the standardized data collection, examiner-measured anthropometry, and comprehensive information on the risk factors of thyroid cancers. Furthermore, the study findings were obtained from population-based follow-up data across the country. Moreover, conclusions were drawn after excluding observations among persons who had undergone follow-up for less than 1 year, thereby avoiding the influence of reverse causation in data interpretation.

In conclusion, we found evidence that metabolically unhealthy phenotype in women was associated with an increased risk of thyroid cancer, regardless of BMI. Furthermore, women with central adiposity may have an increased risk of thyroid cancer despite having a normal BMI. Women with low HDL-cholesterol are likely to have elevated thyroid cancer risk regardless of BMI. Thus, our findings have implications for thyroid cancer intervention programs in normal weight women: central obesity and HDL-cholesterol may be a better indicator for target groups rather than BMI. Further research is needed to confirm the impact of BMI on thyroid cancer in men, considering the waist circumference and HDL-cholesterol.

No disclosures were reported.

D.N. Nguyen: Software, formal analysis, visualization, methodology, writing–original draft, writing–review and editing. J.H. Kim: Software, methodology. M.K. Kim: Conceptualization, resources, data curation, supervision, funding acquisition, investigation, writing–original draft, project administration, writing–review and editing.

The data assessed in this study were obtained from the Korean Genome and Epidemiology Study (KoGES; 4851-302), National Research Institute of Health, Centers for Disease Control and Prevention, Ministry for Health and Welfare, Republic of Korea. The authors are thankful to the National Cancer Center (NCC), Republic of Korea for funding the project (grant no. NCC-1910180). The sponsor had no role in the study design, data collection, analysis, and data interpretation.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.