Abstract
The potential differences by sex in the association between body mass index (BMI) and thyroid cancer risk remain unconfirmed.
Data from the National Health Insurance Service-National Health Screening Cohort (NHIS-HEALS; 2002–2015; population size = 510,619) and the Korean Multi-center Cancer Cohort (KMCC) data (1993–2015; population size = 19,026) were used. We constructed Cox regression models, adjusted for potential confounders, to evaluate the association between BMI and thyroid cancer incidence in each cohort, and assessed the consistency of the results.
In the NHIS-HEALS, 1,351 and 4,609 incident thyroid cancer cases occurred during the follow-up among men and women, respectively. Compared with BMI of 18.5–22.9 kg/m2, BMI of 23.0–24.9 [no. of cases = 410; HR, 1.25; 95% confidence interval (CI), 1.08–1.44], 25.0–29.9 (no. of cases = 522; HR, 1.32; 95% CI, 1.15–1.51), and ≥30.0 kg/m2 (no. of cases = 48; HR, 1.93; 95% CI, 1.42–2.61) were associated with an increased risk of incident thyroid cancer among males. Among females, BMI of 23.0–24.9 (no. of cases = 1,300; HR, 1.17; 95% CI, 1.09–1.26) and 25.0–29.9 (no. of cases = 1,406; HR, 1.20; 95% CI, 1.11–1.29) were associated with incident thyroid cancer. The analyses using the KMCC demonstrated results consistent with wider CIs.
Sex differences in the associations between BMI and thyroid cancer incidence were observed in Korean cohorts.
BMI <23 kg/m2 may help prevent incident thyroid cancer, especially among men.
Introduction
Higher body mass index (BMI) values, rapidly becoming common worldwide (1), are reportedly a risk factor for several cancers, including postmenopausal breast and colorectal cancers (2). It has also been suggested that a higher BMI may be a risk factor for thyroid cancer, the most common malignancy of the endocrine system, accounting for up to 2% of all cancers (3, 4). However, although the association between BMI and thyroid cancer has been reported to be consistent overall (5), there remains some inconsistency: some studies showed a positive association (6–8), while others failed to detect any association (9–12).
In addition, although sex-specific differences have been suggested in the association between BMI and thyroid cancer risk, this possibility has not been thoroughly investigated and remains unconfirmed. For example, associations between BMI and thyroid cancer incidence have been reported only among men (13), only among women (14, 15), or among both men and women (16, 17).
Therefore, this study aimed to investigate the sex-specific associations of BMI with thyroid cancer incidence based on longitudinal data. We investigated the associations using the National Health Insurance Service-National Health Screening Cohort (NHIS-HEALS) and confirmed the robustness of the results using the Korean Multi-center Cancer Cohort (KMCC). Although the KMCC has a smaller cohort size and slightly different covariate information, we used the analytic approach using two independent cohorts to confirm the robustness of the findings and enhance the reliability of causal inferences (18).
Materials and Methods
Study participants
The study protocol was approved by the Ethics Review Board of the Seoul National University Hospital (H-01–10–084–002 and H-1310–082–528) and complied with the tenets of the Declaration of Helsinki. Informed consent was not required for use of the NHIS-HEALS claims data, which were deidentified before being provided to researchers. All participants in the KMCC had provided written informed consent.
The NHIS-HEALS included deidentified claims data from 514,866 individuals, accounting for a 10% random sample of participants in the free-of-charge national health screening program between 2002 and 2003. The national health screening program was administered by the National Health Insurance Service (NHIS), the sole mandatory insurance system in the Republic of Korea (19, 20). Participants were censored at the date of thyroid cancer diagnosis, death, or December 31, 2015, whichever comes first. Details regarding cohort design and key variables were reported elsewhere (19). Self-administered questionnaires were used to collect data regarding sociodemographic factors, lifestyle factors, and medical history, while anthropometric factors were measured during the health screening.
The KMCC was a population-based cohort study, which recruited 20,636 participants from six urban and rural areas in the Republic of Korea between 1993 and 2004. We censored the participants of the KMCC at the date of thyroid cancer diagnosis, death, or December 31, 2015, whichever comes first. The outcome identification was available until 2015 because record linkage to the national cancer registry was impossible after 2015 due to revisions in related laws in Korea. The KMCC was constructed to investigate the relationships between environmental and lifestyle factors and cancer (21). Detailed methods for the data collection were previously reported (21). Structured questionnaires were administered by trained interviewers to collect information regarding sociodemographic factors, lifestyle factors, and medical history. Physical examinations were performed to collect data on anthropometric factors, including height and weight, measured using standard methods (22).
This study excluded individuals with a history of cancer at the time of enrollment, those who were diagnosed with thyroid cancer within 1 year after enrollment, and those with missing data regarding weight or height. On the basis of these criteria, the analyses of the NHIS-HEALS data excluded 4,247 of 514,866 individuals (0.8%), resulting in 510,619 individuals (99.2%). The analyses of the KMCC data excluded 984 of 20,636 participants (4.8%) resulting in a final sample of 19,652 participants (95.2%).
Detection of thyroid cancer incidence
In the NHIS-HEALS, incident thyroid cancer was defined as a principal diagnosis of thyroid cancer [International Classification of Diseases, 10th revision (ICD-10) code C73] that occurred first until December 31, 2015. In the KMCC, incident thyroid cancer cases were identified until December 31, 2015, via record linkage to the national cancer registry. This method was assessed to be efficient, accurate, and complete (23).
Statistical analysis
According to the World Health Organization stratification for Asians, the BMI values were categorized into <18.5 kg/m2, 18.5–22.9 kg/m2, 23.0–24.9 kg/m2, 25.0–29.9 kg/m2, and ≥30.0 kg/m2. Alternatively, we collapsed some strata to obtain enough sample sizes, categorizing the BMI values into <23.0 kg/m2, 23.0–29.9 kg/m2, and ≥30.0 kg/m2 or <18.5 kg/m2, 18.5–24.9 kg/m2, 25.0–29.9 kg/m2, and ≥30.0 kg/m2 (Western BMI cutoff).
The associations between BMI and thyroid cancer incidence were evaluated using Cox proportional hazards models. The Cox proportional hazards models used follow-up time as the time-scale and were adjusted for potential confounders selected on the basis of previous reports (24–37): age (years), sex, cigarette smoking status [ever smoker (total of ≥400 cigarettes smoked) vs. never smoker (<400 cigarettes smoked)], frequency of alcohol consumption (never, <4 times/week, 4–6 times/week, or ≥7 times/week), regular exercise (yes or no), height (m), and medical history of type 2 diabetes or hypertension. Analyses using the NHIS-HEALS data were also adjusted for low household income [yes (Medicaid beneficiary or 1st to 2nd household income decile) vs. no (≥3rd household income decile)]. Because up to 4.2% of covariate information was missing in the NHIS-HEALS data, we used a missing data category for them. The analyses using the KMCC data were adjusted for low education levels [yes (no education or primary school graduate) vs. no (more than primary school graduate)] instead of low household income to control for socioeconomic status because the information on household income was not available. There was no missing covariate information in the KMCC data. After assessing the association between sex and thyroid cancer incidence, we performed stratified analyses by sex to assess potential sex differences in the associations between BMI and thyroid cancer incidence. We also evaluated the interactions between sex and BMI by testing the product terms between sex and BMI added to the models mentioned above.
We conducted a sensitivity analysis using the latency period of 5 years instead of 1 year as used in the main analyses to confirm the robustness of the results. In addition, to assess the possible period effect, we evaluated the impact of the calendar year enrolled in the KMCC study by estimating the regression coefficients of the corresponding terms in similar Cox models with age as a time-scale. All analyses were performed using SAS software (version 9.4; SAS Institute Inc.).
Data availability
Researchers can access the NHIS-HEALS data after approval by the NHIS through the following website (https://nhiss.nhis.or.kr/bd/ab/bdaba000eng.do). The KMCC data are available on reasonable request to the corresponding author after obtaining approval from an appropriate ethics review board.
Results
The NHIS-HEALS included 510,619 participants, 5,960 of whom were diagnosed with thyroid cancer during the follow-up period. In the NHIS-HEALS, 60.9% of participants were aged between 45 and 64 years, and 54.3% were men. Those with BMI ≥23.0 kg/m2 were more likely to have a medical history of type 2 diabetes and hypertension compared to those with BMI <23.0 kg/m2. In the KMCC, 50.1% of participants were aged between 45 and 64 years, and 40.1% were men. Those with BMI ≥23.0 kg/m2 were more likely to be women, have low education levels, be never smokers and never drinkers, and have a medical history of type 2 diabetes and hypertension compared with those with BMI <23.0 kg/m2 (Table 1). Compared with individuals who did not develop thyroid cancer (n = 504,659), those who were diagnosed with thyroid cancer during the follow-up period (n = 5,960) were more likely to be young, women, earning a lower income, never smokers, and never drinkers (Supplementary Table S1). Similar patterns were observed in the KMCC. The KMCC included 19,652 participants, 118 of whom were diagnosed with thyroid cancer during the follow-up period. Compared with those who did not develop thyroid cancer (n = 19,534), those who did (n = 118) were more likely to be young, women, highly educated, never smokers, and never drinkers (Supplementary Table S1). The person-years were larger among men than among women in the NHIS-HEALS and larger among women than among men in the KMCC due to the difference in the proportion of sex included in each cohort (Table 1; Supplementary Table S1).
. | NHIS-HEALS (person-years: 5,221,293) . | KMCC (person-years: 238,357) . | ||||||
---|---|---|---|---|---|---|---|---|
. | . | BMI (kg/m2) . | . | BMI (kg/m2) . | ||||
. | Total (N = 510,619) . | <23.0 (N = 192,313) . | 23.0–29.9 (N = 303,778) . | ≥30.0 (N = 14,528) . | Total (N = 19,652) . | <23.0 (N = 9,290) . | 23.0–29.9 (N = 9,707) . | ≥30.0 (N = 655) . |
Agea | ||||||||
<45 years | 125,625 (24.6) | 52,410 (27.3) | 70,142 (23.1) | 3,073 (21.1) | 4,946 (25.2) | 2,505 (27.0) | 2,269 (23.4) | 172 (26.3) |
45–64 years | 311,200 (60.9) | 107,738 (56.0) | 193,994 (63.9) | 9,468 (65.2) | 9,843 (50.1) | 3,979 (42.8) | 5,477 (56.4) | 387 (59.1) |
≥60 years | 73,794 (14.5) | 32,165 (16.7) | 39,642 (13.0) | 1,987 (13.7) | 4,863 (24.8) | 2,806 (30.2) | 1,961 (20.2) | 96 (14.7) |
Men | 277,456 (54.3) | 100,417 (52.2) | 171,136 (56.3) | 5,903 (40.6) | 7,883 (40.1) | 4,168 (44.9) | 3,577 (36.9) | 138 (21.1) |
Low household incomeb | 81,556 (16.0) | 32,292 (16.8) | 46,645 (15.4) | 2,619 (18.0) | NA | NA | NA | NA |
Low educationc | NA | NA | NA | NA | 4,121 (21.2) | 2,193 (24.0) | 1,804 (18.7) | 124 (19.1) |
Ever smokers | 161,506 (33.0) | 62,438 (33.9) | 95,682 (32.9) | 3,386 (24.3) | 9,117 (47.1) | 4,995 (54.7) | 3,916 (40.8) | 206 (32.0) |
Ever drinkers | 217,913 (43.5) | 78,652 (41.7) | 133,953 (44.9) | 5,308 (37.3) | 6,919 (35.2) | 3,337 (35.9) | 3,400 (35.0) | 182 (27.8) |
Not regularly exercised | 285,506 (57.7) | 114,729 (61.5) | 162,205 (55.1) | 8,572 (61.0) | 5,910 (30.1) | 2,743 (29.5) | 2,925 (30.1) | 242 (37.0) |
Medical history | ||||||||
Type 2 diabetes | 21,663 (4.2) | 6,589 (3.4) | 14,155 (4.7) | 919 (6.3) | 864 (4.4) | 266 (2.9) | 562 (5.8) | 36 (5.5) |
Hypertension | 42,842 (8.4) | 9,718 (5.1) | 30,363 (10.0) | 2,761 (19.0) | 2,233 (11.4) | 661 (7.1) | 1,433 (14.8) | 139 (21.2) |
. | NHIS-HEALS (person-years: 5,221,293) . | KMCC (person-years: 238,357) . | ||||||
---|---|---|---|---|---|---|---|---|
. | . | BMI (kg/m2) . | . | BMI (kg/m2) . | ||||
. | Total (N = 510,619) . | <23.0 (N = 192,313) . | 23.0–29.9 (N = 303,778) . | ≥30.0 (N = 14,528) . | Total (N = 19,652) . | <23.0 (N = 9,290) . | 23.0–29.9 (N = 9,707) . | ≥30.0 (N = 655) . |
Agea | ||||||||
<45 years | 125,625 (24.6) | 52,410 (27.3) | 70,142 (23.1) | 3,073 (21.1) | 4,946 (25.2) | 2,505 (27.0) | 2,269 (23.4) | 172 (26.3) |
45–64 years | 311,200 (60.9) | 107,738 (56.0) | 193,994 (63.9) | 9,468 (65.2) | 9,843 (50.1) | 3,979 (42.8) | 5,477 (56.4) | 387 (59.1) |
≥60 years | 73,794 (14.5) | 32,165 (16.7) | 39,642 (13.0) | 1,987 (13.7) | 4,863 (24.8) | 2,806 (30.2) | 1,961 (20.2) | 96 (14.7) |
Men | 277,456 (54.3) | 100,417 (52.2) | 171,136 (56.3) | 5,903 (40.6) | 7,883 (40.1) | 4,168 (44.9) | 3,577 (36.9) | 138 (21.1) |
Low household incomeb | 81,556 (16.0) | 32,292 (16.8) | 46,645 (15.4) | 2,619 (18.0) | NA | NA | NA | NA |
Low educationc | NA | NA | NA | NA | 4,121 (21.2) | 2,193 (24.0) | 1,804 (18.7) | 124 (19.1) |
Ever smokers | 161,506 (33.0) | 62,438 (33.9) | 95,682 (32.9) | 3,386 (24.3) | 9,117 (47.1) | 4,995 (54.7) | 3,916 (40.8) | 206 (32.0) |
Ever drinkers | 217,913 (43.5) | 78,652 (41.7) | 133,953 (44.9) | 5,308 (37.3) | 6,919 (35.2) | 3,337 (35.9) | 3,400 (35.0) | 182 (27.8) |
Not regularly exercised | 285,506 (57.7) | 114,729 (61.5) | 162,205 (55.1) | 8,572 (61.0) | 5,910 (30.1) | 2,743 (29.5) | 2,925 (30.1) | 242 (37.0) |
Medical history | ||||||||
Type 2 diabetes | 21,663 (4.2) | 6,589 (3.4) | 14,155 (4.7) | 919 (6.3) | 864 (4.4) | 266 (2.9) | 562 (5.8) | 36 (5.5) |
Hypertension | 42,842 (8.4) | 9,718 (5.1) | 30,363 (10.0) | 2,761 (19.0) | 2,233 (11.4) | 661 (7.1) | 1,433 (14.8) | 139 (21.2) |
Note: Data are reported as numbers (column %).
Abbreviation: NA, not available.
aAge at study entry.
bMedicaid beneficiary or 1st–2nd household income decile.
cNo education or primary school graduate.
In the NHIS-HEALS, compared with BMI of 18.5–22.9 kg/m2, a higher risk of incident thyroid cancer was associated with BMI of 23.0–24.9 kg/m2 (HR, 1.17; 95% confidence interval, 1.10–1.25), BMI of 25.0–29.9 kg/m2 (HR, 1.20; 95% CI, 1.13–1.28), and BMI ≥30.0 kg/m2 (HR, 1.24; 95% CI, 1.08–1.43). Female sex was associated with a higher risk of thyroid cancer (HR, 5.90; 95% CI, 5.34–6.53). When the NHIS-HEALS participants were stratified according to sex, compared with BMI of 18.5–22.9 kg/m2, BMI of 23.0–24.9 kg/m2 (HR, 1.25; 95% CI, 1.08–1.44), BMI of 25.0–29.9 kg/m2 (HR, 1.32; 95% CI, 1.15–1.51), and BMI ≥30.0 kg/m2 (HR, 1.93; 95% CI, 1.42–2.61) were associated with a higher risk of incident thyroid cancer among men. Meanwhile, among women, BMI of 23.0–24.9 kg/m2 (HR, 1.17; 95% CI, 1.09–1.26) and BMI of 25.0–29.9 kg/m2 (HR, 1.20; 95% CI, 1.11–1.29) were associated with a higher risk of incident thyroid cancer (Table 2). The interaction was found between BMI and sex on thyroid cancer incidence (Pinteraction = 0.04). The results were similar in analyses using common Western BMI cutoffs (Supplementary Table S2).
. | Total . | Men . | Women . | ||||||
---|---|---|---|---|---|---|---|---|---|
. | Person-years (Cohort = 510,619) . | No. of cases (N = 5,960) . | HR (95% CI) . | Person-years (Cohort = 277,456) . | No. of cases (N = 1,351) . | HR (95% CI) . | Person-years (Cohort = 233,163) . | No. of cases (N = 4,609) . | HR (95% CI) . |
BMI (kg/m2) | |||||||||
<18.5 | 112,621 | 75 | 0.68 (0.54–0.85) | 60,225 | 12 | 0.59 (0.33–1.05) | 52,396 | 63 | 0.69 (0.54–0.89) |
18.5–22.9 | 1,835,938 | 2,039 | 1.00 | 948,652 | 359 | 1.00 | 887,286 | 1,680 | 1.00 |
23.0–24.9 | 1,445,599 | 1,710 | 1.17 (1.10–1.25) | 813,488 | 410 | 1.25 (1.08–1.44) | 632,111 | 1,300 | 1.17 (1.09–1.26) |
25.0–29.9 | 1,678,421 | 1,928 | 1.20 (1.13–1.28) | 954,352 | 522 | 1.32 (1.15–1.51) | 724,069 | 1,406 | 1.20 (1.11–1.29) |
≥30.0 | 148,714 | 208 | 1.24 (1.08–1.43) | 60,881 | 48 | 1.93 (1.42–2.61) | 87,833 | 160 | 1.14 (0.97–1.35) |
<23.0 | 1,948,559 | 2,114 | 1.00 | 1,008,877 | 371 | 1.00 | 939,682 | 1,762 | 1.00 |
23.0–29.9 | 3,124,020 | 3,638 | 1.21 (1.15–1.28) | 1,767,840 | 932 | 1.32 (1.17–1.49) | 1,356,180 | 2,706 | 1.20 (1.13–1.28) |
≥30.0 | 148,714 | 208 | 1.26 (1.09–1.46) | 60,881 | 48 | 1.96 (1.45–2.66) | 87,833 | 160 | 1.16 (0.99–1.37) |
. | Total . | Men . | Women . | ||||||
---|---|---|---|---|---|---|---|---|---|
. | Person-years (Cohort = 510,619) . | No. of cases (N = 5,960) . | HR (95% CI) . | Person-years (Cohort = 277,456) . | No. of cases (N = 1,351) . | HR (95% CI) . | Person-years (Cohort = 233,163) . | No. of cases (N = 4,609) . | HR (95% CI) . |
BMI (kg/m2) | |||||||||
<18.5 | 112,621 | 75 | 0.68 (0.54–0.85) | 60,225 | 12 | 0.59 (0.33–1.05) | 52,396 | 63 | 0.69 (0.54–0.89) |
18.5–22.9 | 1,835,938 | 2,039 | 1.00 | 948,652 | 359 | 1.00 | 887,286 | 1,680 | 1.00 |
23.0–24.9 | 1,445,599 | 1,710 | 1.17 (1.10–1.25) | 813,488 | 410 | 1.25 (1.08–1.44) | 632,111 | 1,300 | 1.17 (1.09–1.26) |
25.0–29.9 | 1,678,421 | 1,928 | 1.20 (1.13–1.28) | 954,352 | 522 | 1.32 (1.15–1.51) | 724,069 | 1,406 | 1.20 (1.11–1.29) |
≥30.0 | 148,714 | 208 | 1.24 (1.08–1.43) | 60,881 | 48 | 1.93 (1.42–2.61) | 87,833 | 160 | 1.14 (0.97–1.35) |
<23.0 | 1,948,559 | 2,114 | 1.00 | 1,008,877 | 371 | 1.00 | 939,682 | 1,762 | 1.00 |
23.0–29.9 | 3,124,020 | 3,638 | 1.21 (1.15–1.28) | 1,767,840 | 932 | 1.32 (1.17–1.49) | 1,356,180 | 2,706 | 1.20 (1.13–1.28) |
≥30.0 | 148,714 | 208 | 1.26 (1.09–1.46) | 60,881 | 48 | 1.96 (1.45–2.66) | 87,833 | 160 | 1.16 (0.99–1.37) |
aCox proportional hazards regression models were adjusted for age, sex, household income status, physical activity, smoking pack-years, frequency of alcohol consumption, height, and medical history of type 2 diabetes or hypertension.
In the KMCC, compared with BMI <23.0 kg/m2, BMI of 23.0–29.9 kg/m2 was associated with a higher risk of incident thyroid cancer (HR, 1.49; 95% CI, 1.00–2.22). Female sex was also associated with a higher risk of thyroid cancer (HR, 4.86; 95% CI, 2.26–10.46). When the KMCC participants were stratified by sex, BMI of 23.0–29.9 kg/m2 was associated with a higher risk of incident thyroid cancer among women (HR, 3.30; 95% CI, 1.00–10.87). Among men, the point estimates of the associations between higher BMI and incident thyroid cancer were larger than those among women; while the associations were found to be less precise (HR, 2.71; 95% CI, 0.71–10.38 for BMI of 23.0–24.9 kg/m2; HR, 3.29; 95% CI, 0.90–12.06 for BMI of 25.0–29.9 kg/m2; compared with BMI of 18.5–22.9 kg/m2), possibly due to lower number of cases among men than women (Table 3). Although the point estimates for the associations between BMI and thyroid cancer differed between men and women, the interaction between BMI and sex was not observed in the formal test (Pinteraction = 0.74). The results were consistent in analyses using common Western BMI cutoffs (Supplementary Table S3).
. | Total . | Men . | Women . | ||||||
---|---|---|---|---|---|---|---|---|---|
. | Person-years . | No. of cases . | HR (95% CI) . | Person-years . | No. of cases . | HR (95% CI) . | Person-years . | No. of cases . | HR (95% CI) . |
BMI (kg/m2) | |||||||||
<18.5 | 18,929 | 4 | 0.79 (0.24–2.57) | 7,472 | 0 | NA | 11,457 | 4 | 0.92 (0.28–3.03) |
18.5–22.9 | 100,774 | 36 | 1.00 | 42,045 | 4 | 1.00 | 58,728 | 32 | 1.00 |
23.0–24.9 | 57,656 | 35 | 1.58 (0.99–2.53) | 21,951 | 5 | 2.71 (0.71–10.38) | 35,705 | 30 | 1.44 (0.87–2.38) |
25.0–29.9 | 67,337 | 37 | 1.36 (0.85–2.17) | 21,798 | 7 | 3.29 (0.90–12.06) | 45,539 | 30 | 1.13 (0.68–1.89) |
≥30.0 | 8,532 | 6 | 1.60 (0.67–3.82) | 1,648 | 0 | NA | 6,884 | 6 | 1.57 (0.65–3.77) |
<23.0 | 119,704 | 40 | 1.00 | 49,518 | 4 | 1.00 | 70,186 | 36 | 1.00 |
23.0–29.9 | 124,994 | 72 | 1.49 (1.00–2.22) | 43,750 | 12 | 3.30 (1.00–10.87) | 81,245 | 60 | 1.28 (0.84–1.96) |
≥30.0 | 8,532 | 6 | 1.63 (0.69–3.89) | 1,648 | 0 | NA | 6,884 | 6 | 1.59 (0.66–3.80) |
. | Total . | Men . | Women . | ||||||
---|---|---|---|---|---|---|---|---|---|
. | Person-years . | No. of cases . | HR (95% CI) . | Person-years . | No. of cases . | HR (95% CI) . | Person-years . | No. of cases . | HR (95% CI) . |
BMI (kg/m2) | |||||||||
<18.5 | 18,929 | 4 | 0.79 (0.24–2.57) | 7,472 | 0 | NA | 11,457 | 4 | 0.92 (0.28–3.03) |
18.5–22.9 | 100,774 | 36 | 1.00 | 42,045 | 4 | 1.00 | 58,728 | 32 | 1.00 |
23.0–24.9 | 57,656 | 35 | 1.58 (0.99–2.53) | 21,951 | 5 | 2.71 (0.71–10.38) | 35,705 | 30 | 1.44 (0.87–2.38) |
25.0–29.9 | 67,337 | 37 | 1.36 (0.85–2.17) | 21,798 | 7 | 3.29 (0.90–12.06) | 45,539 | 30 | 1.13 (0.68–1.89) |
≥30.0 | 8,532 | 6 | 1.60 (0.67–3.82) | 1,648 | 0 | NA | 6,884 | 6 | 1.57 (0.65–3.77) |
<23.0 | 119,704 | 40 | 1.00 | 49,518 | 4 | 1.00 | 70,186 | 36 | 1.00 |
23.0–29.9 | 124,994 | 72 | 1.49 (1.00–2.22) | 43,750 | 12 | 3.30 (1.00–10.87) | 81,245 | 60 | 1.28 (0.84–1.96) |
≥30.0 | 8,532 | 6 | 1.63 (0.69–3.89) | 1,648 | 0 | NA | 6,884 | 6 | 1.59 (0.66–3.80) |
Abbreviation: NA, not available.
aCox proportional hazards regression models were adjusted for age, sex, low education, physical activity, cigarette smoking status, frequency of alcohol consumption, height, and medical history of type 2 diabetes or hypertension.
In the sensitivity analysis using the latency period of 5 years instead of 1 year, the results were consistent with those from the main analyses, while point estimates of the associations increased slightly (Supplementary Table S4). Any prominent pattern of the impact of the calendar year was not found, with point estimates of the regression coefficients ranging between 0.61 in 1996 and 1.64 in 2000 compared with 1993 (Supplementary Table S5).
Discussion
This study revealed that higher BMI was associated with an increased risk of incident thyroid cancer among both men and women. Although the point estimates of the associations were generally larger among men than among women, the CIs overlapped between men and women in most cases.
Some previous studies reported stronger associations between BMI and incident thyroid cancer among men than among women (16, 26, 27). However, other studies revealed no association (9–12), while some revealed positive associations exclusively among men (13), exclusively among women (14, 15), or for both sexes with similar magnitude (8, 28). Meanwhile, in this study, larger point estimates of the associations between BMI and thyroid cancer incidence among men than among women were found, while conclusive conclusions could not be drawn due to disproportionate sample sizes among male and female cases and overlapped CIs in sex-stratified analyses. These heterogeneous findings may be related to underlying differences between the study cohorts for race, age, medical history, diet (including iodine intake), and other lifestyle factors, like cigarette smoking and alcohol consumption status. Thus, while our results appear to be robust based on the consistent associations in two independent cohorts, further studies conducted among diverse populations and with sufficient male cases are needed to explain the apparent contradictions reported in the literature and the mechanisms that dictate the sex-specific difference in the association between BMI and incident thyroid cancer.
The relationship between adiposity and thyroid cancer incidence may be explained via several mechanisms, including hyperinsulinemia, changes in circulating levels of leptin and adiponectin, chronic inflammation, and oxidative stress (38–40). The stronger association between BMI and thyroid cancer incidence in the male participants observed in this study may be explained by the excess production of estrogen in the adipose tissue, resulting in estrogen/androgen imbalance, especially among men (26, 41). Because adipose tissue can contribute to estrogen/androgen imbalance in postmenopausal women (42), we assumed that the heterogeneity of the association between BMI and thyroid cancer by sex, observed in this study, was compromised to some degree by including postmenopausal women in the study populations.
This study has several weaknesses. First, we only performed the analyses based on BMI and did not consider other body fat indices. This is because the data sources did not include information regarding waist and hip circumferences, waist-to-hip ratio, skin-fold thickness, and body fat percentage. Although BMI is a widely used index, it does not reflect the relative proportions of fat and muscle tissue. Second, the clinicopathological features of thyroid cancer could not be assessed due to the lack of relevant data. It is possible that overweight or obese individuals could have a delayed diagnosis of thyroid cancer because accumulated fat around the neck complicates the diagnosis, leading to poor outcomes (43). Third, because the proportion of participants with BMI ≥30.0 kg/m2 was low in the NHIS-HEALS and KMCC, the estimates of the association between BMI and thyroid cancer incidence in this stratum were unstable, especially for sex-stratified analyses (Tables 2 and 3), which warrants cautious interpretation. Fourth, data limitations precluded an analysis of whether diet, including iodine intake, might influence the risk of developing thyroid cancer (44, 45). Fifth, analyses for thyroid-stimulating hormone, leptin, adiponectin, insulin resistance, insulin-like growth factor-1, thyroid hormones, and thyroid antibodies could not be conducted due to data limitations. Sixth, although analysis by the histological types and cancer stage would help reduce the risk of surveillance bias (which is a concern due to the high surveillance rate for thyroid cancer in Korea), the information on the histologic types and cancer stage was unavailable in the NHIS-HEALS and the KMCC. Therefore, future studies with more detailed outcome data, including the histologic types and cancer stage, are warranted.
However, this study also has notable strengths. First, the longitudinal study design may help establish the causality of the relationships, including confirmation of temporal relevance. Second, we evaluated the consistency of the results using two independent cohorts. Third, the sample sizes were relatively large and permitted sufficiently powered analyses stratified by sex and analyses for BMI ≥30.0 kg/m2.
In conclusion, BMI was positively associated with incident thyroid cancer among both men and women. Although the point estimates of the associations were found to be larger among men than among women, the CIs overlapped with each other in most cases. These findings suggest that maintaining BMI <23 kg/m2 may help prevent incident thyroid cancer, especially among men.
Authors' Disclosures
No disclosures were reported.
Authors' Contributions
K.-N. Kim: Formal analysis, methodology, writing–original draft. K. Kim: Formal analysis, writing–original draft, writing–review and editing. S. Lee: Formal analysis, writing–original draft, writing–review and editing. S.K. Park: Conceptualization, formal analysis, supervision, funding acquisition, methodology, writing–original draft, project administration.
Acknowledgments
S.K. Park was supported by a grant from the Korea Health Technology R&D Project through the Korea Health Industry Development Institute (KHIDI), funded by the Ministry of Health & Welfare, Republic of Korea (HI16C1127).
The publication costs of this article were defrayed in part by the payment of publication fees. Therefore, and solely to indicate this fact, this article is hereby marked “advertisement” in accordance with 18 USC section 1734.
Note: Supplementary data for this article are available at Cancer Epidemiology, Biomarkers & Prevention Online (http://cebp.aacrjournals.org/).