Phytoestrogens and Thyroid Cancer Risk: A Population-Based Case–Control Study in Connecticut

Over the past few decades, the age-adjusted incidence rate of thyroid cancer has been rapidly increasing worldwide (1). Compared with all other cancer types, the age-adjusted incidence rate of thyroid cancer has increased the fastest, from 8.74/100,000 in women and 3.38/100,000 in men in 1994 to 21.43/100,000 in women and 7.371/100,000 in men in 2015 (https://seer.cancer.gov/archive/csr/1975_2015/). It is now the fifth most common cancer among women in the United States (2). Although improvements in diagnostic technology such as fine-needle aspiration with ultrasound guidance may be associated with increased detection of microcarcinomas (diameter ≤ 1 cm), the incidences of both smaller-size tumors and larger-size tumors are increasing (1, 3, 4). Furthermore, studies have shown that approximately 50% of the variability in thyroid cancer incidence rates in the United States could not be explained by the theory of “overdiagnosis” (5, 6). Established risk factors for thyroid cancer include female gender, radiation exposure to the head and neck, high body mass index (BMI), history of benign thyroid diseases, and family history of thyroid cancer (7–10). It has been suggested that nutritional factors are associated with the development of thyroid cancer (11, 12). However, research on the effects of consuming foods rich in phytoestrogens (PE) has been controversial (7, 8, 13–18).

PEs include isoflavonoids (genistein, daidzein, and glycitein), lignans, and coumestans (including coumestrol). Dietary PEs mostly come from beans, soy products, and foods with added soy protein or soy flour (19). Previous studies have indicated that PE-rich diets may be associated with a decreased risk of breast cancer and prostate cancer (12, 20–24). As endocrine-disrupting compounds (EDC), PEs could also potentially affect thyroid cancer risk by affecting synthesis of thyroid hormones (TH), altering thyroid stimulating hormone (TSH) levels (25), and interacting with estrogen receptors (ER; refs. 24, 26–28).

After the FDA declared that PEs have a protective effect against coronary heart disease in 1999, the sale of soy foods (only those marketed as soy products, not including foods with added soy flour and/or protein) expanded from $1 billion in 1996 to $4.5 billion in 2009 (12). Therefore, the possible influence of soy foods and their major component PEs on thyroid cancer risk warrants further investigation.

The study population has been described elsewhere (29–34). In brief, cases were histologically confirmed thyroid cancer patients (ICD-O-3: 8021, 8050, 8052, 8130, 8260, 8290, 8330–8332, 8335, 8340–8346, 8450, 8452, 8510) in Connecticut diagnosed between 2010 and 2011. Subjects eligible for the study were aged between 21 and 84 years at diagnosis, alive at the time of interview, and had no prior cancer diagnoses, except for nonmelanoma skin cancer. Cases were identified through the Yale Cancer Center's Rapid Case Ascertainment Shared Resource (RCA). A total of 701 eligible thyroid cancer cases were identified during the study period with 462 (65.9%) completing in-person interviews. Population-based controls with Connecticut addresses were recruited using a random digit dialing method based on both cell phone and landline numbers. A total of 498 subjects participated in the study, with a participation rate of 61.5%. Cases and controls were frequency matched by age (≤5 years apart). Distributions of age, gender, and race were similar between the participants and nonparticipants for both cases and controls.

All procedures were performed following a protocol approved by the Human Investigations Committee at Yale and the Connecticut Department of Public Health. Eligible participants were reached by letter and then by phone once approved by the hospitals and by each subject's physician (for cancer cases), or following selection through random sampling (for the control population). After obtaining written consent, eligible participants were interviewed by trained interviewers and a food frequency questionnaire (FFQ) was administered to determine their diverse diets. The frequency and amount of consumption of previously validated PE-rich food were captured (35). These include soy-based foods (including tofu, soy burgers or soy meat-substitutes, soymilk, bean soup, and beans), foods with added soy flour (including doughnuts, sweet rolls, Danish, pop-tarts, white bread, pancakes, waffles and French toast), as well as foods with added soy protein (including canned tuna). Additionally, a standardized, structured questionnaire was used to obtain information on major known or suspected risk factors that might confound the association between PE-rich food intake and risk of thyroid cancer. Specific PE amounts were calculated by using Diet*Calc Analysis Software version 1.5.0, which generates nutrient estimates based on the FFQ. The six compounds examined represent two major PE types found in foods: the isoflavones: genistein, daidzein, biochanin A, formononetin, and glycitein; and the coumestan: coumestrol.

Unconditional logistic regression was used to estimate the odds ratios (OR) and 95% confidence intervals (95% CI). All cutoff points of each PE subtype and of PE-rich food intake were based on the distribution among controls. We also restricted our analyses to papillary and well-differentiated thyroid cancer subtypes (the papillary and follicular subtypes were combined due to the small sample size of follicular carcinoma). In order to exam the effects of PEs on tumor size, analyses were conducted by dividing the cases into two groups according to the tumor size (microcarcinomas ≤ 1 cm, macrocarcinomas > 1 cm) among papillary carcinomas. We also analyzed thyroid cancer among women because the number of men was too small for a meaningful analysis.

After excluding participants who did not respond to the FFQ, 387 cases (83.8%) and 433 controls (86.9%) were included in the final analysis. Age, gender, race, family history of cancer of any kind, education, BMI, previous benign thyroid diseases, history of radiation exposure, previous alcohol consumption, smoking history, daily caloric intake, percentage of calories from fat, total carotenoid intake, vitamin C intake, vitamin E intake, and fiber intake were considered as potential confounding variables. Among female subjects, additional controlled confounders included age at menarche, use of oral contraceptives, age at first full-term pregnancy, menopausal status, and estrogen use. Decisions on which covariates to include in the final model were based on a greater than 10% change in the estimates. Tests for linear trends were calculated by using each individual PE variable as continuous variables in the multivariate unconditional logistic regression models. Statistical analyses for this study were performed using SAS version 9.4 (SAS Institute Inc.). Results were considered statistically significant when two-sided P values were <0.05.

As illustrated in Table 1, cases tended to be less educated and had a higher BMI when compared with controls. Cases were more likely to have prior benign thyroid diseases and less likely to consume alcohol. Cases were also more likely to be female compared with controls. Distributions of race, family history of cancer of any kind, smoking, previous radiation exposure, daily caloric intake, percentage of calories from fat, vitamin C intake, vitamin E intake, and fiber intake were similar between cases and controls. The baseline characteristics of those who had available PE data were similar to the overall study population.

Compared with those whose consumption of coumestrol was less than 40 μg/day (Table 2), individuals who consumed 40 to 80 μg/day (OR = 2.48, 95% CI, 1.39–4.43), 80 to 130 μg/day (OR = 2.41, 95% CI, 1.32–4.40), or >200 μg/day (OR = 2.38, 95% CI, 1.26–4.50) experienced an increased risk of thyroid cancer, although the linear trend was not statistically significant (P for linear trend = 0.25). Other specific PEs, total isoflavones, and total PEs were not statistically significantly associated with thyroid cancer risk. Similar patterns were noticed for papillary and well-differentiated carcinomas. No significant associations between intake of soy-based foods, foods with added soy flour, and foods with added soy protein and thyroid cancer risk were observed (Supplementary Tables S1–S5).

After stratification by tumor size (Table 3), statistically significant associations with coumestrol intake were observed mainly for microcarcinomas (OR = 4.33, 95% CI, 1.86–10.07 for 40–80 μg/day; OR = 4.38, 95% CI, 1.79–10.71 for 80–130 μg/day; and OR = 3.11, 95% CI, 1.22–7.91 for >200 μg/day compared with <40 μg/day; P for linear trend = 0.34). Individuals who consumed 40 to 80 μg/day also showed an increased risk of macrocarcinomas (OR = 2.31; 95% CI, 1.09–4.86).

When the analyses were restricted to females (Table 4), a statistically significant increase in risk was only seen among those who had a coumestrol intake of 40 to 80 μg/day (OR = 2.08, 95% CI, 1.01–4.29) compared with those who had an intake of less than 40 μg/day. Similar patterns were observed for well-differentiated and papillary thyroid cancers (Supplementary Table S6). After stratification by tumor size (Table 5), an increased risk of papillary microcarcinomas was observed among those with a coumestrol intake of 40 to 80 μg/day (OR = 2.94, 95% CI, 1.12–7.73) and a decreased risk of papillary thyroid cancers with tumor size > 1 cm (OR = 0.26, 95% CI, 0.08–0.85) among those with a genistein intake of 1,860 to 3,110 μg/day.

This population-based case–control study suggests that an increased risk of thyroid cancer is associated with coumestrol consumption, which is the most estrogenic PE (22). The observed associations varied by tumor size (≤1 cm vs. >1 cm). Our study results also suggest that moderate genistein intake is associated with a decreased risk of thyroid cancer with larger tumor size among females.

There is only one previous study, by Horn-Ross and colleagues in San Francisco, that has investigated coumestrol intake and thyroid cancer risk (7). They did not find an association between coumestrol intake and thyroid cancer risk. The inconsistency could be due to the different cutoffs of coumestrol consumption. Their reference level was less than 81.3 μg/day, which contained both our reference (<40 μg/day) and low-exposure (40–80 μg/day) groups. When our analysis was performed applying the same cutoffs as the San Francisco study (<81.3, 81.4–123.1, 123.2–167.7, 167.8–270.4, ≥270.5 μg/day), we found no significant association between coumestrol intake and thyroid cancer risk, which is consistent with their findings. The paradoxical results could be due to the reference group in the Horn-Ross' study incorporating that which was the exposure group (40–80 μg/day) in our study, which showed more than two times the risk of thyroid cancer.

Coumestrol is the most estrogenic compound and has the highest receptor binding affinity to both ER-α and ER-β, followed by genistein (36). The proliferation of thyroid cancer cells is positively stimulated by ER-α agonists and inhibited by ER-β agonists (37) in particular ER-β1 (38, 39). Moreover, in a prior experimental study including subjects with small well-differentiated thyroid cancers (average tumor size 9.4 mm), there was a greater expression of ER-α among tumor tissues than normal tissues (40). It has also been suggested that there are fewer ER-β1 receptors in smaller tumors (<2 cm) compared with larger tumors (38). Therefore, it is possible that coumestrol increases microcarcinoma risk by agonizing overexpressed ER-α in smaller tumors, in addition to lessening the protective effect from ER-β, thus leading to microcarcinoma proliferation.

Notably, we did not observe a linear relationship between coumestrol intake and thyroid cancer risk. The possible explanations include: (i) some agents are carcinogenic at low-exposure levels, but their carcinogenic potential does not increase with higher exposure such as radiation to the thyroid (41); (ii) a higher dose exposure could paradoxically be protective, as the mutagenic rate is overwhelmed by apoptosis; (iii) the sample size in each category is insufficient to demonstrate small differences.

In addition, thyroid microcarcinomas have shown the fastest increase in incidence and approximately half of all newly diagnosed thyroid cancers were microcarcinomas (4). An observational study of papillary microcarcinomas from Japan demonstrated that patients with papillary microcarcinomas that showed enlargement of 3 mm or more constituted 6.4% and 15.9% of patients at 5-year and 10-year follow-ups, respectively, and that the enlargement process was unrelated to patient background or clinical features (42). Thus, thyroid microcarcinoma appears to be a distinct entity with pathophysiologic features that are distinct from macrocarcinomas.

The anticarcinogenic effect of genistein on thyroid cancer has been demonstrated by previous studies (24, 43, 44). Evidence has shown that genistein could function as an anticarcinogen due to its higher binding affinity to ER-β than to ER-α (36), thereby leading to a more antiproliferative effect, inducing apoptosis and inhibiting angiogenesis (43, 44). Although it has been suggested that increased TSH levels are associated with an increase in thyroid cancer risk by causing less differentiation and more malignant cell transformation (45–47), multiple randomized controlled trials (RCT) have consistently shown that genistein supplementation did not change TSH levels with various doses (25, 48–51), suggesting that the protective effect of genistein was not through a TSH pathway. The San Francisco study (7), though not statistically significant, showed a 30% thyroid cancer risk reduction in the highest genistein intake group (OR = 0.70; 95% CI, 0.44–1.1). Among our female population, low-to-moderate genistein intake was associated with a 56% decrease in papillary thyroid cancer risk. When the analyses were stratified by tumor size, decreased risk was observed mainly among larger tumors. The amount of ER-α expression in macrocarcinomas is about 1.3 times the amount in microcarcinomas (52), whereas the binding capacity of genistein to ER-β is five to seven times higher than to ER-α (36). Therefore, it is possible that the binding capacity of ER-β overcomes the number of ER-α and thus leads to a greater anticarcinogenic effect on macrocarcinomas (40). Again, this suggests that microcarcinomas and macrocarcinomas possess different pathophysiologies.

Support for an association between total intake of isoflavones and thyroid cancer risk has been inconsistent (7, 15, 18). Two case–control studies have suggested that high isoflavone intake has a protective effect against thyroid cancer (7, 15). One prospective cohort study (18) found no association, which is consistent with our findings. Previous RCTs have concluded that isoflavone soy protein supplements do not affect serum TSH levels (25, 50, 53). Therefore, it is unlikely that total isoflavone intake affects thyroid cancer risk by influencing TSH levels.

Historically, the major consumers of soy foods were Asian countries such as China, Korean, and Japan (54). The estimated daily soy consumption was found to be higher in China, Korean, and Japan (soy and soy foods: 23.5–135.4 g/day, 21.07 g/day, and 50.7–102.1 g/day, respectively; soy protein: 2.5–10.3 g/day, 7.4–8.5 g/day, and 6–11.3 g/day, respectively; ref. 55) than in the USA with one previous study showing an average soy protein intake of 9.25 g/day, with higher consumption found among vegetarians (55, 56). The concentrations of genistein and coumestrol measured in various foods differed by area and population (35, 57, 58). In general, the genistein mainly comes from soy-based foods (such as soybeans, soymilk, tofu, soy protein, and black bean sauce) and foods with added soy flour or protein (such as doughnuts, “power”-type bars, etc.). The main sources of coumestrol include sprouts (such as alfalfa sprouts, clover sprouts, and mung bean sprouts), soy-based products (such as soybeans, soymilk and tofu), and foods with added soy flour or protein (such as canned tuna, doughnuts, pancakes and waffles). Overall, the level of genistein in the same food category is much higher than the level of coumestrol (35, 57).

Previous observational studies have shown inconsistent results regarding the effects of soy-based food consumption on thyroid cancer risk, with some studies suggesting a protective effect (7, 14, 15) and other studies reporting no associations (8, 13, 18). Our study found no statistically significant relationship between soy-based food consumption and thyroid cancer risk. Studies that reported a protective effect were mainly conducted in Asian populations (7, 14, 15), where soy food consumption was higher than in Western populations. This may suggest that soy-based food consumption, especially at a relatively higher amount, could be beneficial in the prevention of thyroid cancer, and genistein may play an important role in the anticarcinogenic process.

Our study, consistent with a study conducted in San Francisco, has shown no association between foods with added soy flour or protein and thyroid cancer risk (7). It has been suggested that there is a positive relationship between refined cereals and thyroid cancer (16, 17). Though foods with added soy flour, such as doughnuts, pancakes, and waffles, have a relatively high genistein content, the potential anticarcinogenic effects of genistein may be cancelled out by their high sugar components, which can cause glycemic overload, elevated insulin and insulin-like growth factor level, and can subsequently promote tumor cell growth (16, 59). It has been suggested that fresh fish consumption is a protective factor against thyroid cancer, but consumption of processed fish, such as canned tuna, has been linked to an increased risk of thyroid cancer (17, 60). The proposed mechanism is that the additives in processed foods may affect iodine absorption and potentially cause iodine deficiency (60). Iodine deficiency is associated with increased thyroid cancer risk (61). Therefore, foods with added soy flour or protein, due to their mixed components and other possible carcinogenic factors, may promote tumor proliferation.

To the best of our knowledge, this is the first study to investigate the effects of both specific phytoestrogenic compounds and soy-based foods on thyroid cancer risk by tumor size. Our study has several strengths. First, Diet*Calc Analysis Software was adopted to estimate PE compound levels. The accurate measurement of specific nutrients could decrease the possibility of exposure misclassification. Furthermore, carefully reviewed pathologic reports allowed for stratified analysis based on the pathologic information.

However, the present study also has limitations. Dietary intake was self-reported and thus might lead to potential recall bias. Because the relationship between PE intake and thyroid cancer risk has not been well established, potential recall bias is likely to be nondifferential and result in an underestimation of the true association. Due to limited sample size, the study was unable to investigate the association between PE intake and rare histologic subtypes, such as medullary and anaplastic thyroid cancer. The subgroup analyses, which were stratified by histology, gender, and tumor size, might yield unstable results. The FFQ in our study captured the vast majority of previously validated PE-rich food; however, certain food items, mainly Asian-style bean products and sprouts, that contain high levels of genistein [such as Chinese black bean sauce, Natto (fermented Japanese beans), and soybean sprouts] and coumestrol (such as clover sprouts, soybean sprouts, alfalfa sprouts and mung bean sprouts) were not included (https://data.nal.usda.gov/dataset/usda-database-isoflavone-content-selected-foods-release-21-november-2015/resource/1de757af). Nevertheless, according to a Centers for Disease Control and Prevention (CDC) population survey (62), among the Connecticut respondents (with 88.6% being of white race), only 3.9% of them reported consuming alfalfa sprouts, 6.0% bean sprouts and 8.7% other sprouts, which included clover sprouts. The consumption was significantly lower than in Asian countries, where sprouts have been common foods since ancient times (63). Hence, the uncaptured PE-rich food items were unlikely to make a significant contribution to our final results. Lastly, our study did not include lignans, the quantities of which were undetected or of trace amounts in most foods (35), which could potentially result in an underestimation of total PE exposure.

This population-based case–control study supports the hypothesis that coumestrol is associated with an increased risk of thyroid cancer with small tumor size and that genistein is associated with a decreased risk of thyroid cancer with larger tumor size among females. The novel finding that risks varied by tumor size warrants further investigation.

No potential conflicts of interest were disclosed.

Conception and design: Q. Wang, X. Ni, R. Udelsman, Y. Zhang

Development of methodology: R. Udelsman, Y. Zhang

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): Y. Zhang

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): Q. Wang, N. Zhao

Writing, review, and/or revision of the manuscript: Q. Wang, N. Zhao, X. Ni, R. Udelsman, Y. Zhang

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): Q. Wang, H. Huang

Study supervision: X. Ni, Y. Zhang

This study was supported by American Cancer Society (ACS) grant RSGM-10-038-01-CCE (Y. Zhang) and NIH grant R01ES020361 (Y. Zhang). Certain data used in this study were obtained from the Connecticut Tumor Registry located in the Connecticut Department of Public Health. The authors assume full responsibility for the analyses and interpretation of these data. The cooperation of the Connecticut hospitals, including Charlotte Hungerford Hospital, Bridgeport Hospital, Danbury Hospital, Hartford Hospital, Middlesex Hospital, Hospital of Central Connecticut, Yale/New Haven Hospital, St. Francis Hospital and Medical Center, St. Mary's Hospital, Hospital of St. Raphael, St. Vincent's Medical Center, Stamford Hospital, William W. Backus Hospital, Windham Hospital, Eastern Connecticut Health Network, Griffin Hospital, Bristol Hospital, Johnson Memorial Hospital, Greenwich Hospital, Lawrence and Memorial Hospital, New Milford Hospital, Norwalk Hospital, MidState Medical Center, John Dempsey Hospital, and Waterbury Hospital, in allowing patient access, is gratefully acknowledged. Rajni Mehta from the Yale Comprehensive Cancer Center's RCA provided great help in both Institutional Review Board (IRB) approvals and field implementation of the study. Helen Sayward, Anna Florczak, and Renee Capasso from the Yale School of Public Health did exceptional work in study subject recruitment.

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.