The mean age of subjects in the entire groups of cohorts (n=848,932) was 58 years at baseline, with 51% of the participants being male, and 20% being obese (≥30 kg/m2) (). There were different entry and exit dates for each of the cohorts within the period from 1979 to 2009, during which 1,156 participants (388 male and 768 female) were diagnosed with a first primary thyroid cancer. There were 132 thyroid cancer cases with missing information on histology. Of the 1,024 thyroid cancer cases with complete histological information, 810 (79%) were papillary, 164 (16%) were follicular, 34 (3%) were medullary, and 16 (2%) were anaplastic.
General characteristics of the cohorts
Among women, but not men, BMI was positively associated with baseline age (). Compared to their normal-weight (18.5–24.9 kg/m2) counterparts, obese (≥30 kg/m2) participants were less likely to be current smokers and have a post-high school education. The proportions of women who were white, married, and reported drinking alcohol were lower in the obese compared to normal-weight category.
Age-adjusted means and percentages of select baseline characteristics according to BMI category
When BMI was modeled continuously (per 5 kg/m2, ), significant positive associations were observed in men from two out of four cohorts (NIH-AARP and PLCO) and, in women, four out of five cohorts (NIH-AARP, USRT, PLCO, and BCDDP). The pooled HR for thyroid cancer was 1.21 (95% CI, 0.97–1.49) in men and 1.16 in women (95% CI, 1.08–1.24); results were not significantly different by sex (P-interaction=0.34). When we excluded the first two years of follow-up, the pooled HR became slightly stronger and statistically significant in men (HR, 1.30 [95% CI, 1.15–1.47]) but remained the same for women (HR, 1.17 [95% CI, 1.08–1.27]). Despite a slightly stronger pooled HR in men compared to women, there was a non-significant inverse association among men in the AHS cohort (HR, 0.69 [95% CI, 0.41–1.15]). For approximately 31% of the AHS cohort, missing or extreme BMI values were assigned using data from a five-year follow-up questionnaire, and to a lesser extent drivers' license information on height and weight. However, exclusion of these individuals did not materially alter the HR for men in the AHS cohort, and in women, the positive association became slightly stronger.
Multivariable-adjusted HRsa and 95% CIs for BMI (per 5 kg/m2) and thyroid cancer risk by sex: pooled analysis
In the aggregate dataset (which combined men and women in the five cohorts), the relationship between BMI and thyroid cancer was approximately log-linear (per 5 kg/m2
: HR, 1.17 [95% CI, 1.10–1.24]) (); the inclusion of high-order terms did not significantly improve the data fit (P
>0.05). We also examined the association using broader BMI categories, which correspond to the WHO criteria for normal-weight (18.5–24.9 kg/m2
), overweight (25.0–29.9 kg/m2
), and obesity (≥30 kg/m2
). Compared to normal-weight, the HRs for overweight and obesity were 1.20 (95% CI, 1.04–1.38) and 1.53 (95% CI, 1.31–1.79).
Multivariable-adjusted HRsa and 95% CIs for BMI categories and thyroid cancer risk: aggregate analysis
We observed no significant heterogeneity in the association between BMI (per 5 kg/m2) and thyroid cancer risk by baseline age, birth cohort, education level, smoking status, alcohol intake, or physical activity level (). Though not significantly different, the results were stronger among never smokers (HR, 1.21 [95% CI, 1.11–1.31]) compared to former or never smokers. The results were stronger for papillary (n=810 cases) compared to follicular thyroid cancer (n=164 cases), but this difference was also not statistically significant, likely due to the smaller number of follicular tumors (P-heterogeneity=0.27). We also separately examined the associations between BMI (per 5 kg/m2) and medullary (n=34 cases, HR, 0.87 [95% CI, 0.59–1.28]) and anaplastic (n=16 cases, HR, 1.45 [95% CI, 0.95–2.22]) cancers, though the relative risk estimates were unstable due to small numbers. The differences in the results for the four histologic types were not significant (P-heterogeneity=0.23). When we restricted the results to Caucasians (n=1,075 cases), the HR was 1.19 (95% CI, 1.12–1.26).
Multivariable-adjusted HRsa and 95% CIs for BMI (per 5 kg/m2) and thyroid cancer risk, stratified by selected risk factors: aggregate analysis
Additional adjustment for physical activity level and medical history of diabetes had little influence on the results (Supplementary Table
). Adjustment for other factors available only from the USRT cohort, including personal and medical exposure to radiation and medical history of benign thyroid conditions also did not change the results (Supplementary Table
The strength of the association for BMI in young adulthood (18–20 years old) (per 5-kg/m2 increase: HR, 1.18 [95% CI, 1.03–1.35]) was very similar to that of baseline BMI (per 5-kg/m2 increase: HR, 1.17 [95% CI, 1.11–1.24]). However, mutual adjustment for baseline BMI slightly attenuated the association for young adulthood BMI (HR, 1.08 [95% CI, 0.93–1.25]), while that of baseline BMI remained similar (HR, 1.14 [95% CI, 1.04–1.25]).
We additionally examined the association between height (per 5 cm) and thyroid cancer risk in men and women, separately, and found a significant positive association in women (HR, 1.06 [95% CI, 1.00–1.12]) but no association in men (HR, 1.01 [95% CI, 0.94–1.08]). We observed significant heterogeneity between studies in men (P-heterogeneity=0.01), but the results were less heterogeneous after the exclusion of the AHS cohort (HR, 1.03 [95% CI, 0.96–1.11], P-heterogeneity=0.34).