Reproductive and Hormonal Risk Factors for Thyroid Cancer in Los Angeles County Females¹

Because the female excess in thyroid cancer incidence peaks in the reproductive years, hormonal changes related to the menstrual cycle, pregnancy, and menopause may be relevant etiological factors. The hypothesis that chronic elevation of serum TSH³ and related hormones are important etiological factors in thyroid carcinogenesis (1) provides a related motivation for investigating reproductive and hormonal exposures. Experimentally, chronic elevations of TSH induce thyroid growth and produce thyroid hyperplasia and tumors (2). Thyroid volume varies up to 50% throughout the menstrual cycle (3) and increases further in pregnancy and delivery (4, 5). TSH is elevated in pregnancy (4, 6, 7), and pregnancy-related elevations in serum human chorionic gonadotropin may directly stimulate the thyroid (4, 8, 9).

Epidemiological studies of thyroid cancer have focused on reproductive and hormonal factors that might explain the female excess in incidence as well as provide further insights into these cancers (10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29). Those exposures related to increased risk for thyroid cancer in females include infertility (11, 14), irregular menstrual cycles (15), occurrence of a miscarriage (12, 13, 14, 20, 21), multiple pregnancies or live births (12, 13, 15, 16, 20, 22, 23, 25), use of lactation suppressants (10, 12, 28), use of OCs (10, 12, 13, 15, 21, 29), and use of other noncontraceptive (usually peri- and postmenopausal) estrogens (10, 20). In one study, hormonal exposures were risk factors for thyroid cancer in younger (age <35 years) but not in older women (12).

We previously reported the results of a case-control study of 108 matched case-control pairs with cases diagnosed with thyroid cancer at age ≤40 years in 1980 or 1981 in Los Angeles County (13). In that study, we also interviewed women who were 41–54 years of age and extended case identification to include patients diagnosed through 1983. This report describes results for the expanded study.

All incident cases of histologically confirmed thyroid cancer diagnosed between January 1, 1980, and December 31, 1983 among women who were 15–54 years of age at diagnosis were identified through the Los Angeles County population-based registry, the University of Southern California Cancer Surveillance Program. The study was restricted to English-speaking white women (including Spanish-surnamed) who were born in the United States, Canada, or Europe. The Cancer Surveillance Program identified 377 patients meeting these criteria, and their attending physicians granted permission to contact 348 (92%) of these women. Cases that moved out of the area between the time of diagnosis and attempted contact were included among the 377 eligible patients. We were unable to locate 24 patients, and 22 refused to be interviewed. Because we used a detailed exposure history regarding reproductive events and OC use, only living cases were eligible for inclusion. Nine otherwise eligible cases were deceased. Completed questionnaires were thus obtained on 302 patients (94% of those contacted; 80% of all eligible cases). The average time between the case diagnosis and the interview was 2.7 years.

A single neighborhood control that was matched on birth year (within 5 years) was sought for each interviewed patient. To be able to ascertain exposures over a comparable referent period (defined as the age at diagnosis of the case minus 1 year), controls also had to be at least as old at the interview as the case was at the diagnosis. Neighborhood controls were identified by using a procedure that defines a sequence of houses on specified blocks in the neighborhood in which the patient lived at the time of her thyroid cancer diagnosis. We sought to interview the first female resident in this sequence who matched the case on race and age. If no one was home, we left an explanatory letter and made a follow-up visit. For each case, up to 80 housing units were visited and 3 return visits were made before failure to obtain a matched control was conceded. We were able to obtain matched controls and complete control questionnaires for 296 of the 302 cases. The first appropriate woman agreed to participate in 263 instances. Three of the controls were later found to be ineligible due to a prior thyroidectomy, and one control was younger than the matched case was at the diagnosis. Thus completed questionnaires on 292 case-control pairs were available for analysis. The average interval between the case and the matched control interview was 0.3 years. We have previously reported results on reproductive factors and prior thyroid disease on the subset of 108 case-control pairs in which the case, who was ≤40 years of age, was diagnosed in 1980 or 1981 (13).

Interviews were conducted by telephone for diagnosis years 1980 and 1981 and in-person for diagnosis years 1982 and 1983. Two interviewers conducted interviews, and both members of a case-control pair were interviewed by the same person and by the same method (telephone versus in-person). Because of the nature of the study, interviewers were not blinded to disease status; however, a standard structured questionnaire was used to reduce possible interviewer bias. Exposure data on the questionnaire were obtained up to the time of the interview. Analyses were limited to exposures accrued up to the reference age (case’s age at diagnosis minus 1 year). Similarly, exposure information for controls was limited to that accrued up to the reference age of her matched case.

Information was collected on menstrual and reproductive history, hormone use, OC use, diet (fresh and saltwater fish, caffeinated beverages), alcohol, smoking, prior thyroid conditions, radiation treatment, dental and diagnostic X-rays, occupational exposure to radiation, and history of thyroid disease, thyroid cancer, and other cancers in first degree relatives. This paper focuses on the history of benign thyroid diseases, reproductive history, and use of exogenous hormones.

Conditional logistic regression was used to estimate matched ORs and their 95% CIs. In addition to reproductive and hormonal exposures, a history of benign hyperplastic thyroid disease (including adolescent thyroid enlargement, goiter, or benign thyroid nodules) was included in all models. Because of the strength of the association of thyroid cancer with prior benign thyroid disease and the potential association between benign thyroid disease and certain reproductive variables, we conducted statistical analyses excluding the subjects with prior benign thyroid disease and performed unmatched analyses adjusting for age (in 5-year age groups) and study period (1980–1981 versus 1982–1983) by using unconditional logistic regression. In general, the two methods of analysis provided similar results, and we report results for the conditional logistic regression analyses. Differences between results for the unmatched analysis excluding subjects with prior benign thyroid disease and results for the matched analyses are described in the text.

We were particularly interested in possible differential effects of reproductive and exogenous hormone exposures by age at diagnosis (12). Analyses were therefore conducted separately for younger (age <35 years at diagnosis) and older (age ≥35 years at diagnosis) women. Results are reported for the entire sample; when differential effects are noted by age, they are reported in a separate table. We also conducted all analyses on the papillary subtype alone (n = 240 case-control pairs; Table 1). Small sample size among other histologies precluded separate analyses. All tests for trends used exposure factors as continuous variables when possible. Joint effects of variables and effect modification were also examined using conditional logistic regression. Specific variables examined as potential modifiers of reproductive associations included: radiation treatment to the head or neck, prior hyperplastic thyroid disease, and age at diagnosis. All Ps reported are two-sided.

Over one-half (52%) of the 292 cases were papillary, and 30% were mixed papillary/follicular histologies (Table 1). The remaining cases were classified as follicular or other. Of the 292 cases, 174 (60%) were <35 years of age and 118 (40%) were ≥35 years of age at diagnosis (Table 1). Cases and controls had comparable levels of education. Although cases and controls were generally comparable on religion practiced as a child, a higher proportion of cases were Jewish (49; 17%) cases versus 30 (10%) controls (OR, 2.3; 95% CI, 1.2–4.2; relative to Protestant religion).

Consistent with the host of prior literature, radiation treatment to the head or neck at least 4 years prior to the diagnosis was an exceptionally strong risk factor for thyroid cancer (28 cases and 2 controls exposed; OR, 14.0; 95% CI, 3.5–121.3). Of the 28 exposed cases, 10 (36%) were treated prior to age 13, 14 (50%) were treated between ages 13 and 18, and 4 (14%) were treated after age 18. One of the two exposed controls was first treated at age 13, and the other was treated as an adult. Conditions for which the subjects were treated were: acne (14 cases; 2 controls), other skin conditions (4 cases), hypertrophy of tonsils or adenoids (4 cases), thymic enlargement (3 cases), ear infection (1 case), asthma (1 case), and tumors of the head (2 cases). The majority of the cases exposed to therapeutic radiation were diagnosed after age 35 (24 of 28 cases), and both exposed controls were in the older age group (for women <35 years of age, 4 cases and 0 controls were exposed; for women ≥35 years of age, 24 cases and 2 controls were exposed).

We evaluated benign thyroid conditions that occurred 1, 5, and 10 years before diagnosis (Table 2). A diagnosis of hypothyroidism was not related to thyroid cancer, and a history of hyperthyroidism was associated with a nonsignificantly elevated cancer risk. Thyroid cancer risk was significantly elevated among subjects reporting benign thyroid conditions characterized by hyperplasia, including physician-diagnosed thyroid enlargement at an adolescent age, goiter, and thyroid nodules or benign tumors. The effect was apparent in both younger and older age groups (for any hyperplastic thyroid disease, the OR was 20.5 and the 95% CI was 5.3–175.0 in women <35 years of age; the OR was 11.3 and the 95% CI was 3.6–57.7 in women 35–54 years of age at diagnosis). Elevated risk associated with these hyperplastic conditions persisted when considering 5- and 10-year latency intervals. However, relative risk estimates tended to decrease with longer latency, and those associated with adolescent thyroid enlargement were not statistically significant.

Age at menarche was not related to thyroid cancer incidence. Women who had never menstruated regularly were at increased risk for thyroid cancer (Table 3), and the elevation in risk was particularly apparent in younger women (Table 4).

Whereas a history of any miscarriage was not related to thyroid cancer in the total sample (Table 3), risk was nonsignificantly elevated in younger women (Table 4). In the unmatched analyses excluding subjects with prior benign thyroid disease, the risk for prior miscarriage among women <35 years of age at diagnosis was significantly increased (the OR was 2.2 and the 95% CI was 1.1–4.5 for any miscarriage; the OR was 2.7 and the 95% CI was 1.1–6.8 for miscarriage in the first pregnancy).

In the total sample, the number of pregnancies was not associated with thyroid cancer risk. Younger women with at least three pregnancies had elevated risk; the test for trend on number of pregnancies was not statistically significant (Table 4). The apparent increased risk for number of pregnancies in younger women was stronger in the unmatched analysis excluding women with prior benign thyroid disease (OR, 2.1; 95% CI, 1.1–4.0 ; for three or more pregnancies, P trend = 0.08). Age at first live birth and number of live births were not associated with disease in this case-control sample. We also found no association of thyroid cancer risk with ever having an induced abortion (OR, 1.1; 95% CI, 0.7–1.9). After adjusting for prior hyperplastic thyroid disease and number of live births, we found that women who had ever breastfed had a decreased risk of thyroid cancer, and thyroid cancer risk decreased by total months of breastfeeding (P trend = 0.04).

An elevated risk for thyroid cancer was apparent in women who had a surgical menopause (Table 5). The elevated risk for surgical menopause appeared to be related to whether or not the ovaries were also removed (P trend = 0.04 for number of ovaries removed). Women who had a hysterectomy without oophorectomy had no increase in risk relative to women who were still menstruating. Thyroid cancer risk was significantly elevated among women who had both ovaries removed. Results were not altered by adjustment for age at hysterectomy or use of replacement estrogens. The median age at surgical menopause was slightly lower for cases (median age, 36 years) than controls (median age, 38 years). Thyroid cancer risk did not significantly vary by age at either surgical or natural menopause.

In the total sample, any OC use was not associated with thyroid cancer risk (Table 6). Risk was also not related to duration of use, age at first use, age at last use, or time since last use of OCs. The use of OCs was analyzed separately among women with and without a history of irregular menstruation by using unconditional logistic regression and adjusting for age and prior benign hyperplastic thyroid disease. Any use of OCs was not significantly associated with thyroid cancer risk in either group (among women with a history of irregular menstruation, the OR was 0.8 for any OC use and the 95% CI was 0.2–2.6; among women without a history of irregular menstruation, the OR was 1.4 and the 95% CI was 0.9–2.2).

Any treatment for nausea during pregnancy was not related to thyroid cancer (OR, 1.2; 95% CI, 0.7–2.0; adjusted for prior benign hyperplastic thyroid disease and number of pregnancies). The use of other specific pregnancy-related hormones was asked only of the women in the 1982–1983 study period (a total of 147 case-control pairs; 85 pairs <35 years of age; 62 pairs ≥35 years of age). Analyses were adjusted for prior hyperplastic thyroid disease and number of births. Women who had ever had hormones to induce labor showed no clear trend in disease risk by the number of hormone-induced labors (Table 6). An elevated thyroid cancer risk was associated with any use of lactation suppressants, and a significant trend by number of pregnancies using lactation suppressants was seen in the total sample and in the older age group. Among older women, thyroid cancer risk more than tripled for any use of labor-inducing hormones and lactation suppressants (Table 4). Postmenopausal use of exogenous estrogens, duration of use, age at first use, and years since first use were not related to thyroid cancer risk (Table 6).

Results of analyses conducted among the 240 case-control pairs of papillary histology closely mirrored those derived from the total sample, and ORs showed only slight variations from the analyses of the total sample. There were two exceptions to this observation. Although thyroid cancer risk was elevated in women having a hysterectomy plus ovaries removed (relative to women who were still menstruating, the OR was 0.8 and the 95% CI was 0.2–2.4 for hysterectomy with no oophorectomy; the OR was 2.1 and the 95% CI was 0.3–14.5 for hysterectomy with partial oophorectomy; the OR was 5.4 and the 95% CI was 0.9–32.9 for hysterectomy with complete oophorectomy), the trend in number of ovaries removed among women with surgical menopause was not statistically significant in the papillary subsample (P trend = 0.09). For any use of lactation suppressants (adjusted for prior hyperplastic thyroid disease and number of births), the OR was 2.7 (95% CI, 1.2–6.4; for number of pregnancies in which lactation suppressants were used, P trend = 0.005).

The strongest risk factors were evaluated in a multivariate model (Table 7). Whereas a history of radiation treatment to the head or neck and hyperplastic thyroid disease each showed a very strong relationship to thyroid cancer in females, irregular menstrual cycles and surgical menopause each independently doubled the risk of thyroid cancer. No effect modification by prior radiation treatment to the head/neck or by prior benign hyperplastic thyroid disease was apparent for reproductive factors or exogenous hormone use.

In agreement with a host of studies, prior thyroid disease, including adolescent thyroid enlargement, goiter, and thyroid nodules or benign tumors, was a strong risk factor for thyroid cancer in Los Angeles County females. Histories of hypothyroidism and hyperthyroidism were not associated with thyroid cancer risk. The past occurrence of thyroid nodules or benign tumors was a particularly strong risk factor. These findings are supportive of the hypothesis that exposures or medical conditions related to increased cell division may increase the likelihood of either a mutational event or division of already mutated cells (30). Our estimates of relative risk are in agreement with earlier studies conducted in the United States, Europe, and Asia, which note relative risks around 5–7 for goiter and of much higher magnitude (≥20) for thyroid nodules (12, 21, 22, 31, 32, 33). In contrast, one study from Hawaii reported a relative risk of only 2.4 for women and did not examine prior thyroid nodules (14). These estimates of relative risk may be upwardly biased due to surveillance bias because women with thyroid conditions may be more actively followed subsequent to the detection of benign disease. The large number of thyroid cancers diagnosed among cases within 1–5 years of detection of a benign thyroid nodule suggests that this may, in part, be the case. However, the persistence in elevated thyroid cancer risk over 5- and 10-year latency intervals and the fact that hypothyroidism is not associated with any elevation in cancer risk suggest that certain benign thyroid conditions are truly related to subsequent development of thyroid cancer. A direct etiological relationship is, however, difficult to prove. Although we cannot completely discount recall bias as a possible explanation for these findings, we carefully designed the structured interview to include only conditions that were medically diagnosed.

Another strong risk factor in this study was prior radiation treatment to the head or neck, although only 28 of 292 cases were so exposed (23 of the 28 were born before 1945). Therapeutic radiation for benign conditions was much more prevalent among older than younger cases (20% of cases ≥35 years of age compared to 2% of cases <35 years of age at diagnosis), thereby reflecting the lesser use of this treatment in recent years. Consistent with previous case-control studies, the effect appears to be related to earlier age at exposure (12, 31). Although we did not obtain medical record validation of radiotherapy, we did perform a substudy in which mothers of a sample of the case (n = 150) and control (n = 138) subjects were interviewed regarding their daughters’ childhood exposures (34). In this subsample, all daughters who reported a history of radiotherapy in childhood were verified as positive by their mothers’ reports. Three case mothers and one control mother reported radiotherapy that was not reported by their daughter. Although we cannot consider the mothers’ reports as a “gold standard,” it is encouraging that we saw no evidence of over-reporting of radiotherapy. We also did not see an excess of under-reporting in control subjects (relative to their mothers’ reports). Such an excess, if apparent, would upwardly bias the estimate of relative risk obtained in the overall case-control sample.

Because of the dramatic female to male excess in the incidence of well-differentiated thyroid cancers observed in the reproductive years, reproductive exposures in women have been vigorously studied in recent epidemiological studies. Our findings are in general agreement with these studies, which find, at best, a small contribution of reproductive factors to the etiology of thyroid cancer. A number of case-control studies have reported increased risk with miscarriage or abortion, particularly in the first pregnancy (11, 13, 14, 20, 21), and this is the most established reproductive risk factor for thyroid cancer in women. More inconsistent results have been found for number of pregnancies or live births (20, 28) and later age at first full term pregnancy (14, 15, 28). One study reported increased risk for thyroid cancer with earlier age at first pregnancy (22). Consistent with a number of previous epidemiological studies, we found no association with age at first live birth (12, 16, 19, 20, 23, 25, 28) or age at menarche (14, 20, 21, 22). The reproductive variables for which we did find an association were: an elevated risk associated with never menstruating regularly, miscarriage (among younger women), an increasing risk with number of pregnancies (significant only in younger women without a history of benign hyperplastic thyroid disease), and a decreased risk with breastfeeding.

The associations between thyroid cancer and irregular menstrual cycles and miscarriage have some clinical basis. Dysfunctional menstrual cycles, anovulation, infertility, and miscarriages may be due to underlying thyroid disorders (35). Miscarriage is more frequent in women with mild thyroid abnormalities (including goiter, benign nodules, and autoimmune thyroid disease; Ref. 36). These thyroid disorders are also associated with abnormally elevated TSH levels at delivery (in a small percentage of such women) and with elevated serum thyroglobulin levels throughout gestation (which may suggest an oversensitivity to thyroid stimulatory factors; Ref. 36).

We also found an elevated risk with surgical menopause, which has been reported by others (14, 20, 28). McTiernan and colleagues (10) reported a relative risk of 1.5 for menopause before age 40. Although these results were not reported by type of menopause, it is likely that a substantial proportion of these were due to surgical rather than natural menopause. In accord with others (15, 28), we found no real trends with age at menopause. The association with surgical menopause in our case-control sample was limited to women who had ovaries removed, with the elevated risk particularly apparent when both ovaries were removed. The issue of oophorectomy has been evaluated in one other epidemiological study (12), which found no association with thyroid cancer risk. Surgical menopause in relation to oophorectomy should be examined in the future to validate our results.

Our relative risk estimate of 1.5 for any OC use in younger women <35 years of age agrees with relative risk estimates from other studies that have also found no effect of duration (10, 12, 15, 21, 29). Studies that did report results by age at diagnosis found similar results to ours; the modest effect of OCs was apparent in the younger but not older women (12, 15, 29). A recent case-control study of papillary thyroid cancer demonstrated a reduction in disease risk with OC use among women <45 years of age and no association in women >45 years of age (27). Other studies have found no effect of OC use but have not evaluated this exposure by age at diagnosis (14, 20).

In our total case-control sample, we found an elevated risk with any use of lactation suppressants, and thyroid cancer risk significantly increased with the number of pregnancies using lactation suppressants. Among older women, an elevated risk associated with labor inducing hormones was not related to the number of pregnancies in which they were used. A few studies have also reported increased risk with lactation suppressants (10, 12, 29); in one of these, the association was apparent only in younger women (12). In contrast, the effects of both labor inducing hormones and lactation suppressants in our study were evident primarily in older women. Prior to the 1980s, the most common pharmacological methods of lactation suppression were long-acting estrogens and estrogen-androgen combinations (37, 38). The effects of a single injection postdelivery of estradiol valerate (20 mg) or estradiol valerate (16 mg) plus testosterone enanthate (360 mg) versus placebo were studied in relation to serum hormone levels (38). Women on the estrogen-androgen regimen had average serum testosterone levels higher than adult males up to 2 weeks postinjection and remained elevated to five times the normal level for females at 6 weeks postinjection (38). Serum estradiol was elevated above placebo in both treated groups up to 2 weeks postinjection. Use of dopamine agonists such as bromocriptine to suppress lactation were introduced in the 1970s and approved in the 1980s after the conduct of our study. It is therefore likely that few of our women would have used these medications for lactation suppression. However, the relative lack of effect of lactation suppressants on thyroid cancer risk observed in women <35 years of age may be in part due to the fact that some of these women were using the newer dopamine agonists. Prior to the 1980s, oxytocin was the primary pharmacological means of induction of labor (39).

Although >99% of thyroid cancers diagnosed in Los Angeles County were histologically verified in the period of case ascertainment of this study (40), we did not conduct a centralized pathology review of tumor tissue. Although we may be confident in the fact of thyroid cancer, variation between pathologists in classification of histological subtypes likely led to some misclassification by histology.

In summary, this case-control study of thyroid cancer in Los Angeles County females suggests that reproductive and exogenous hormonal exposures appear to confer only a small to moderate elevation of thyroid cancer risk. The large female excess in thyroid cancer incidence that is particularly apparent in the reproductive years remains largely unexplained.