Body size and risk of differentiated thyroid carcinomas: Findings from the EPIC study



Results from case-control and prospective studies suggest a moderate positive association between obesity and height and differentiated thyroid carcinoma (TC). Little is known on the relationship between other measures of adiposity and differentiated TC risk. Here, we present the results of a study on body size and risk of differentiated TC based on a large European prospective study (EPIC). During follow-up, 508 incident cases of differentiated TC were identified in women, and 58 in men. 78% of cases were papillary TC. Cox proportional hazard models were used to estimate hazard ratios (HRs). In women, differentiated TC risk was significantly associated with body mass index (BMI, kg/m2) (HR highest vs lowest quintile = 1.41, 95% CI: 1.03–1.94); height (HR = 1.61; 95% CI: 1.18–2.20); HR highest vs lowest tertile waist (HR = 1.34, 95% CI: 1.00–1.79) and waist-to-hip ratio (HR = 1.42, 95% CI: 1.05–1.91). The association with BMI was somewhat stronger in women below age 50. Corresponding associations for papillary TC were similar to those for all differentiated TC. In men the only body size factors significantly associated with differentiated TC were height (non linear), and leg length (HR highest vs. lowest tertile = 3.03, 95% CI: 1.30–7.07). Our study lends further support to the presence of a moderate positive association between differentiated TC risk and overweight and obesity in women. The risk increase among taller individuals of both sexes suggests that some genetic characteristics or early environmental exposures may also be implicated in the etiology of differentiated TC.


Thyroid carcinoma (TC) represents about 2% of all malignancies worldwide, and it is two-to-three times more common in women than in men.1 TC incidence rates, but not mortality rates,2, 3 have been rapidly growing over the last decades in many countries.4, 5 Obesity has also been increasing in the westernized world, reaching pandemic proportion. Although the prognosis for differentiated carcinomas is very good (over 90% 5-year survival in Europe6), TC treatment (thyroidectomy; radioactive iodine; and lifelong thyroid hormone replacement therapy) are associated with substantial life long side effects.7, 8 The only well-established risk factor for TC is ionizing radiation exposure, especially during childhood, but it is unclear to which extent increases in medical radiation can explain the current upward trends in TC incidence.9 Results from case-control studies,10–13 as well as from prospective studies,14–18 suggested a moderate positive association between increased body size, namely body mass index (BMI) and weight, and TC risk, notably in women. An association with height was also consistently reported10–13, 19–22 whereas the relationship between measures of adiposity other than BMI (e.g., waist, hip and waist-to-hip ratio) and TC risk was seldom explored.22, 23

Here, we present findings on the influence of different body size measures and the risk of differentiated TC in women and men who participated in a large European prospective study, the European Prospective Investigation into Cancer and nutrition (EPIC) cohort. Body measures investigated include exhaustive body size measurements, as weight, height, BMI, waist, hip, waist-to-hip ratio, as well as leg length, a variable that to our knowledge has never been investigated before in a prospective way.

Material and Methods

The European Prospective Investigation into Cancer and nutrition

The EPIC cohort consists of about 370,000 women and 150,000 men, mainly aged 35–69 years, recruited from 1992 to 1998 in 23 research centers in 10 European countries (Denmark, France, Germany, Greece, Italy, Norway, Spain, Sweden, the Netherlands, and the United Kingdom). Baseline questionnaire data and anthropometric measurements were collected from study participants in the period 1991–2000. The questionnaires included detailed questions about current habitual diet, menstrual and reproductive history, current and past use of oral contraceptives in women, a history of previous illness and surgical operations, lifetime history of tobacco smoking and consumption of alcoholic beverages, and physical activity. In addition, blood samples were also collected for most participants. Extensive details about EPIC recruitment procedures, questionnaires, anthropometric measurements, and biologic sample collection are given elsewhere.24 All participants had given their written informed consent, and the Internal Review Boards of IARC and all boards from recruitment centers approved the EPIC project and the current study.

Study population

This study is based on data from a population of 343,765 female and 146,824 male subjects. Subjects were excluded if they had prevalent cancer other than nonmelanomatous skin cancer (n = 23,785), or no follow-up information (n = 4,366). Subjects were also excluded if they did not complete the lifestyle questionnaire (n = 1,272) or if data on baseline anthropometric variables were unavailable (n = 1,257).

Identification and selection of thyroid cancer cases

In all EPIC centers, except those in Greece and Germany, data on vital status are collected by record linkage with regional and/or national mortality registries. In Greece and Germany, data on vital status are continuously collected through active follow-up. In all centers except those in Greece, Germany, and France, incident cases of cancer are identified through record linkage with regional cancer registries. In France, Germany, and Greece, follow-up for cancer incidence is based on a combination of methods, including the use of health insurance records, contacts with cancer and pathology registries, and active follow-up through study participants and their next-of-kin. Closure dates for the present study were defined as the latest date of complete follow-up for both cancer incidence and vital status. This spans between 31 December 2007 and 31 December 2009 in most of the centers in Denmark, Italy, Netherlands, Norway, Spain, Sweden, and United Kingdom whereas it corresponds to December 2006 in France and December 2009 in Germany and Greece.

A total of 604 incident TC cases (defined according to the International Classification of Diseases, ICD-10 code C73) were identified within the EPIC cohort at the time of the present report. The few anaplastic (n = 6), or medullary (n = 28) TC and those defined as lymphoma (n = 1), or as “other morphologies” (n = 3) were excluded from our present study leaving 566 first primary differentiated TC.25

Anthropometric measurements

At recruitment, body weight, height, waist, and hip circumferences were measured according to standardized procedures.26 In brief, weight was measured to the nearest 0.1 kg, and height was measured without shoes. Sitting height was assessed as the length from the seat to the top of the head; leg length was computed by subtracting sitting height from standing height. BMI was calculated as weight in kilograms divided by height expressed in meter squared (kg/m2). Waist circumference was measured either at the narrowest torso circumference or at the midpoint between the lower ribs and iliac crest. Hip circumference was measured either at the widest point, or over the buttocks. Waist-to-hip ratio (WHR) was calculated as the ratio between waist and hip circumferences. In Umeå (Sweden), only weight and height were collected. For a part of the Oxford cohort, linear regression models were used to predict sex- and age-specific values in subjects with both measured and self-reported body measures.26, 27 In France, weight and height were requested by questionnaires for all subjects, and subsequently weight, height, and waist and hip circumferences were measured for a subgroup of 29% of the participants. In Norway, height and weight were self-reported. Sitting height was measured in all centers except at Bilthoven (Netherlands subcohort), Sweden, Norway, and the UK study centers, and in 29% of the French cohort, leaving 331 differentiated TC cases with available data. Waist and hip measures were not collected in Norway, and among 71% of the French participants, leaving 386 differentiated TC cases for the analysis of these variables.

Statistical analyses

Statistical analyses were performed separately for men and women. Tertiles (or quintiles for some of the variables, where appropriate) of different anthropometric variables were calculated on the sex-specific population distribution. Anthropometric variables were also assessed as continuous variables, and associations with differentiated TC risk was evaluated in strata of selected variables. Heterogeneity of the hazard ratios (HRs) between strata of selected variables was tested by comparing the overall maximum-likelihood estimate of differentiated TC to stratum-specific maximum likelihood estimates. The test statistic was compared to the chi-squared distribution with degrees of freedom equal to the number of strata minus one. Country-specific incidence rates among EPIC participants were computed, and age standardization was made using the direct method and was based on the world standard population.

Cox proportional hazard models were used to estimate HRs and 95% confidence intervals (95% CI) of differentiated TC for each body measure. Age was used as underlying primary dependent time variable, with entry and exit time defined, respectively, as the subject's age at recruitment and age at differentiated TC diagnosis, death or last complete follow-up. All multivariate models were stratified by study center and by year of age at recruitment to be less sensitive against violations of the proportional hazard assumptions. Correlations between anthropometric variables adjusted for age were calculated as Spearman's partial correlation coefficients. BMI was also categorized as follows: <18.5, 18.5–24.9, 25.0–29.9, ≥ 30 for underweight, normal weight, overweight and obese, respectively, according to WHO categories.28 Tests for trend were based on the likelihood-ratio test between the models with and without a linear term for each body measure's tertile or quintile. Adjustments for alcohol drinking (never, ever, unknown), education (primary school -including no school degree- and secondary school or more), physical activity (“low - inactive and moderately inactive” vs. “high - moderately active and active” categories) and, in women, menopausal status (pre/peri menopause, postmenopause, oophorectomy/hysterectomy), exogenous hormone use (never, ever, unknown), number of pregnancies (none, more than 1, unknown), age at menarche (<13 years, 13–14, >15) did not sensibly alter the associations between the anthropometric variables and differentiated TC risk, therefore these covariates were not included in the final model. As current and past smoking may have a strong opposite influence on anthropometry, this variable was retained in the final model, and was categorized as follows: never, former, current, and unknown. Appropriate sensitivity analyses were performed excluding the subjects with self-reported body size measures and women from France who had contributed 40.6% of female differentiated TC cases, and have already been included in a previous publication on BMI and differentiated TC risk.15


Selected characteristics of the population included in the present study are shown in Table 1. Of the 566 differentiated TC cases that were identified, 508 (89.7%) were female and 58 were male. The number of cases in women was 10-fold higher than the number of cases in men. The reason for this large difference between female and male cases relies on the fact that in some of the EPIC centers only women were recruited.24 Mean age at enrolment in the cohort was 50.9 years in women, and 52.3 years in men. TC cases included 442 papillary, 77 follicular, and 47 not otherwise specified (NOS) TC. TNM stage was known for 62% of female and 48% of male cases, but the availability of TNM information greatly varied between countries, and was completely absent in four countries. Age-standardized incidence rates of differentiated TC in the EPIC subcohorts in different countries varied substantially between countries. The highest rate in women was in France (10.2) and the lowest in Denmark (0.9). In men, the highest was found in Italy (3.6), and the lowest in Spain (0.5) (Table 1). The differences observed in age-standardized incidence rates are consistent with population-based incidence rates in the corresponding countries.29

Table 1. Cohort characteristics among 343,765 women and 146,824 men by country in EPIC
inline image

All anthropometric variables were correlated to each other, in both genders: BMI was strongly, positively correlated to waist and hip circumferences (Spearman's r = 0.83 and 0.83, respectively, in women; Spearman's r = 0.84 and 0.75 respectively, in men), and moderately to WHR (Spearman's r = 0.47 in women, and Spearman's r = 0.58 in men). A weak inverse correlation was found between height and BMI in women (r = −0.18) and men (r = −0.15).

HRs according to different anthropometric variables in women are shown in Table 2, separately for all differentiated TC and for papillary TC cases only. For all differentiated TC, significantly increased risks were found for weight (HR for highest vs. lowest quintile = 1.72, 95% CI: 1.262.34); height (HR = 1.61, 95% CI: 1.182.20); BMI (HR = 1.41, 95% CI: 1.031.94); waist (HR for highest vs. lowest tertile = 1.34, 95% CI: 1.001.79); and WHR (HR = 1.42, 95% CI: 1.051.91). BMI was also categorized into WHO categories confirming the association with differentiated TC risk (χ2 trend; p = 0.015). When adjusting WHR by BMI, as an indication of abdominal fat independent from the overall fat, the associations with WHR was slightly reduced and became of borderline statistical significance (HR highest vs. lowest tertile = 1.33, 95% CI: 0.971.81, p = 0.049) (data not shown). No statistically significant increase in differentiated TC risk was found with increasing leg length. Associations were similar when restricting analyses to papillary TCs, or when analyzing anthropometric variables as continuous variables (Table 2). Increasing leg length was significantly associated with an increase in papillary TC risk (HR for highest vs. lowest tertile = 1.47, 95% CI: 1.032.09).

Table 2. Hazard ratios (HR) and corresponding 95% confidence interval (CI) of differentiated thyroid carcinomas (TC) in women by anthropometric variables at recruitment in EPIC
inline image

When statistical analyses were repeated excluding subjects with self-reported anthropometric measures, positive associations with weight, height, and WHR were not materially changed (HR per a 5 kg increase in weight = 1.06, 95% CI: 1.011.10, for all differentiated TC, HR = 1.06, 95% CI: 1.011.12 for papillary cases only; HR per a 5 cm increase in height = 1.13, 95% CI: 1.041.24, for all differentiated TC, HR = 1.18, 95% CI: 1.071.31 for papillary cases only; HR per a 0.1 increase in WHR = 1.16 (1.011.33) for all differentiated TC, and HR = 1.16, 95% CI: 0.991.35 for papillary cases only). The exclusion of French differentiated TC cases did not eliminate the associations with height and BMI (HR per a 5-cm increase in height = 1.15, 95% CI: 1.051.27, for all differentiated TC, HR = 1.22, 95% CI: 1.091.36 for papillary cases only; HR per a 5 kg/m2 increase in BMI = 1.07, 95% CI: 0.951.21, for all differentiated TC, HR = 1.07, 95% CI: 0.921.23 for papillary cases only). Likewise, the exclusion of the first year of person-time did not substantially alter the associations between the anthropometric variables and differentiated TC risk (data not shown).

HRs for all differentiated TC and papillary TC only according to different anthropometric variables in men are shown in Table 3. No significant associations were observed with weight, BMI, waist, height, or WHR. A statistically significant increase in differentiated TC risk in men was observed with increasing leg length (HR for highest vs. lowest tertile = 3.03, 95% CI: 1.307.07, for all differentiated TC and 3.57, 95% CI: 1.2210.5 for papillary TCs only).

Table 3. Hazard ratios (HR) and corresponding 95% confidence interval (CI) of differentiated thyroid carcinomas (TC) in men by anthropometric variables at recruitment in EPIC
inline image

As there was a moderate negative correlation between height and BMI in both sexes, we repeated the analyses adjusting height also for BMI, and BMI also for height. We did not observe any significant changes in the corresponding HRs (in women, HR per a 5-cm increase in height adjusted by BMI = 1.13, 95% CI: 1.051.22, for all differentiated TC, HR = 1.16, 95% CI: 1.071.27 for papillary cases only; and HR for a 5 kg/m2 increase in BMI adjusted by height = 1.14 (1.031.26) for all differentiated TC, and HR = 1.17, 95% CI: 1.041.30 for papillary cases only. In men, HR for a 5-cm increase in height adjusted by BMI = 1.32, 95% CI: 1.081.60, for all differentiated TC, HR = 1.21, 95% CI: 0.961.53 for papillary cases only; and HR for a 5 kg/m2 increase in BMI adjusted by height = 0.82 (0.561.18) for all differentiated TC, and HR 0.83, 95% CI: 0.531.30 for papillary cases only).

Table 4 shows the variations in differentiated TC risk by height and BMI in women from different study countries and in separated strata of selected variables. HRs associated with 5 cm increase in height or 5 kg/m2 increase in BMI showed some fluctuations across the strata considered, but none of the differences in HRs across strata were significantly heterogeneous. However, the association with height tended to be stronger for smaller tumors (T1) (p for heterogeneity = 0.074), and the association with BMI tended to be stronger in women below age 50 (p for heterogeneity = 0.053).

Table 4. Hazard ratios (HR) and corresponding 95% confidence interval (CI) of differentiated thyroid carcinomas (TC) in women according to height (per 5 cm) and BMI (per 5 kg/m2) stratified by selected variables in EPIC
inline image


In the present large cohort study, we observed a moderate increase in differentiated TC risk with increasing body size. Among women, differentiated TC risk increased with increasing height, weight, BMI, waist circumference and WHR. Among men, BMI and body size did not seem to influence differentiated TC risk but positive associations were found with leg length. However, in this population, the number of cases was too small to draw firm conclusions on associations.

Our findings on BMI in women are consistent with previous reports from case-control10–13 and prospective14-18, 30 studies. A pooled analysis of five United States-based prospective studies including 768 female and 388 male TC cases14 showed an overall 16% increase in TC risk per 5 kg/m2 increase in BMI in women. The French E3N cohort,15 based on 317 women with differentiated TC cases (62% of whom, are included in the present study showed a 21% risk increase per 5 kg/m2 increase in BMI. Compared to our present report, the proportion of obese individuals (BMI ≥ 30) was larger in the pooled analyses from the United States but lower in the French prospective study.

Findings from case-control and prospective studies of TC in men showed less consistent associations with BMI than in women.10, 11, 14, 16, 18, 31 In our present study, we did not observe a statistically significant association between increasing BMI and differentiated TC risk but the number of male TC cases was nearly ten-fold smaller than the number of female cases.

We also found a statistically significant association with differentiated TC risk with increasing WHR in women, although this association was slightly reduced by adjustment for BMI. If confirmed, an association with WHR may suggest a special role of abdominal fat, i.e., the type of fat most strongly associated with insulin resistance and hyperinsulinemia.32, 33 In men, we did not observe any associations between increasing WHR and differentiated TC risk. In the National Institutes of Public Health-American Association for Retired Persons (NIH-AARP) study23 weak associations with WHR were shown but they did not show statistical significance in either women or men. In the study from the Women's Health Initiative,22 little association was observed between increasing WHR and thyroid cancer risk in post-menopausal women.

The mechanisms that may explain the association between obesity and overweight and differentiated TC risk are ill-understood. Clinically manifest thyroid dysfunction severely affects body weight, and there are suggestions that obesity can influence thyroid hormone levels.34, 35 In vitro experiments have shown that thyroid stimulating hormone (TSH) enhances the proliferation of TC cells,36, 37 but evidence in humans is scarce.38, 39 Obesity is associated with insulin resistance and increased production of insulin and insulin-like growth factors which in turn have been reported to be associated to thyroid disorders.40, 41 In post-menopausal women, obesity is strongly related to higher levels of circulating estrogens which have been suggested to stimulate TC cell proliferation in vitro.42

An association between height and TC risk was suggested by case-control studies10-13 and confirmed by two large cohort studies from Norway19 and Korea,21 as well as by the results of the study from the Women's Health Initiative.22 Our present cohort study lent further support to the possibility that taller people have an increased risk of differentiated TC. Thyroid hormones regulate, among other things, the growth of long bones, in synergy with growth hormone.43 A significant association with height was observed in women whereas in men the association was statistically significant only in the second height tertile. Interestingly, leg length was associated with papillary TC risk in both sexes. An especially strong association with height was reported for medullary TC,44 which however are not included in our present study. Height per se may be involved, on account of possible increased organ size in the tallest individuals45, 46 as it has been proposed to explain the relationship between height and colorectal cancer risk.21 However, it is also plausible that genetic or environmental factors (e.g., diet during childhood or adolescence factors) that influence adult height may have a role in the etiology of differentiated TC. The pre-pubertal stature increase is mainly due to an increase in leg length than from an increase in trunk length, and leg length has been recognized as a marker for growth before puberty.47 The association that we found with leg length would strongly point, if confirmed, to genetic characteristics or environmental factors that influence pre-pubertal growth.

We also evaluated whether individual's or tumor's characteristics had an influence on the associations with height and BMI and differentiated TC. No significant heterogeneity emerged but heterogeneity of borderline statistical significance was found by tumor stage and age. The association with BMI, but not with height, with differentiated TC tended to be stronger in women below age 50 years. The association with height, but not with BMI, tended to be stronger in women with T1 cases than in more advanced TNM stages. This might suggest that BMI is more influential in women below age 50, or more strongly associated with greater use of healthcare services by younger women (because of events related to reproduction and peri- and postmenopausal symptoms).5 Similarly, taller women may have an earlier diagnosis. Nevertheless, the American pooled analysis had previously presented findings on BMI and TC risk in strata of age, education level, smoking status and found no significant heterogeneity.14

Strengths of our present study include the large size of the EPIC cohort and the large number of differentiated TC cases among women. The broad range of anthropometric measurements available (weight, height, BMI, waist, hip, WHR), including some never evaluated before in respect to differentiated TC risk (as leg length), and of possible confounding and modifying risk factors are additional assets of our present study. Histological subtype classification was available for the vast majority of differentiated TC cases and information on TNM stage for more than half of the cases. Weaknesses of our study include, however, lack of systematic information on the best-established risk factors for TC, i.e., history of radiation and of benign thyroid diseases.48 The exclusion of individuals with history of cancer other than nonmelanomatous skin cancer provides however reassurance on the absence of individuals who had received radiotherapy. Differentiated TC cases from France represented 40% of the total of female cases in our present report and findings on BMI and cancer risk for these female cases had been already published.15 Most importantly, lack of information on the modalities of TC diagnosis does not allow us to rule out the possibility that heavier women or taller individuals underwent more intense thyroid gland investigation. Although education did not alter the association between weight and height and thyroid cancer risk, residual confounding by socioeconomic status may still have influenced our results given that increased medical surveillance could have contributed to the associations observed. Differences in diagnosis practices may have also induced differences in incidence rates among countries. Autopsy series showed that papillary TC can be found in a majority of histologically examined thyroids49 and the increased clinical surveillance of neck masses and use of ultrasounds and CT scans are responsible for at least a fraction of recent rises in papillary TC incidence rates especially among young women4, 5. A study of 259 TC cases in the United States showed a negative association between BMI and nodal metastasis and tumor invasion.50

In conclusion, the findings of our study lend further support to the existence of a moderate positive association of differentiated TC risk with overweight and obesity in women but, possibly, not in men. A similarly modest risk increase was found in both sexes among the tallest individuals and suggests that some genetic or early environmental determinants of height may be implicated in the etiology of differentiated TC.


We thank Mr Bertrand Hémon for his precious help with the EPIC database.