Reproductive and Menstrual History and Papillary Thyroid Cancer Risk: The San Francisco Bay Area Thyroid Cancer Study¹

Gender differences in thyroid cancer incidence suggest that the etiology of this disease may be related to hormonal and reproductive function. More women than men are diagnosed with thyroid cancer at a ratio of ∼3:1 (1), and this pattern is consistently observed across geographic location and ethnicity (2, 3, 4). Also as illustrated in Fig. 1 ³ for papillary thyroid cancer, the most prevalent histological type, incidence for women peaks within the female reproductive period, whereas for men, it increases steadily with age.

Nevertheless, epidemiological studies examining menstrual and reproductive factors in relation to risk of thyroid cancer have yielded mixed results (5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22). A recent pooled analysis of 14 case-control studies found late menarche, late age at first birth, miscarriage as outcome of first pregnancy, postmenopausal status among women ≤55 years old or recent menopause among women ≥65 years old, current OC⁴ use, and treatment to suppress lactation to confer increased cancer risk (23, 24, 25). However, the ORs associated with these risks were often of small magnitude. In presenting data from a population-based, case-control study of thyroid cancer conducted in the San Francisco Bay Area, we further examine these relationships and explore potential differences by age and race/ethnicity.

Study methods have been described previously (26). Briefly, women diagnosed with thyroid cancer between June 1, 1995, and May 31, 1998 (June 1, 1992, and May 31, 1998 for Asian women), at ages 20–74 were identified through the Greater Bay Area Cancer Registry, a population-based cancer registry affiliated with the Surveillance, Epidemiology, and End Results program of the National Cancer Institute and with the California Cancer Registry. Of the 817 cases identified, 608 (74%) were interviewed and 106 (13%) declined to participate. The remainder were not interviewed for other reasons, including death, illness, and the inability to converse in English, Spanish, Mandarin, Cantonese, Tagalog, or Vietnamese.

Controls were women identified through RDD and were frequency matched to cases on 5-year age groups and race/ethnicity (i.e., white, African American, Latina, Asian, or Native American). A total of 9756 telephone numbers were dialed, of which 3898 were nonresidential and 856 were not answered despite ten attempts made on varied days and times. Enumeration was completed for 3928 (79%) of the 5002 identified residences. Of the 793 controls selected from the enumerated residences, 558 (70%) were interviewed, 154 (19%) declined to participate, and the remainder were not interviewed for other reasons.

All of the participants were interviewed in person using a standardized structured questionnaire. Information was collected on a variety of factors, including menstrual history (i.e., age at menarche, menopausal status, and age at menopause), reproductive history (i.e., number of pregnancies, age at each pregnancy, and outcome of each pregnancy), and exogenous hormone use (i.e., duration of use, age at first use, and age at last OC and HRT use).

Histological type was classified according to an expert review of tumor specimen slides. Of the 608 interviewed cases, 544 (88%) were classified as papillary histology, which included three subtypes: papillary carcinoma alone, papillary carcinoma with follicular features, and mixed papillary and follicular carcinoma (27). Analyses presented in this paper have been limited to women with papillary thyroid cancer. The concordance between the classification of papillary cancer reported by the cancer registry and that determined by expert review was 99%.

Only events occurring before the age at diagnosis for cases or the age at selection (established through matching) for controls were included for analysis. A FTP was defined as a pregnancy lasting at least 28 weeks regardless of its outcome, and parity was based on the total number of full-term pregnancies. Unconditional logistic regression was used to compute ORs and their 95% CIs (28). Variables, including age, race/ethnicity (i.e., Asian, white, Latina, mixed/other) history of radiation to the head or neck, history of goiter or nodules, first-degree family history of proliferative thyroid disease (i.e., thyroid cancer, goiter, or nodules), education level, parity, OC use, recency of last FTP, and birthplace, were considered to be possible confounders and were factored into the logistic regression models when deemed appropriate.

To examine potential risk differences separately by age at diagnosis/selection and by race/ethnicity, stratified analyses were performed for women of reproductive age (20–44 years old) and of postreproductive age (45–74 years old) and for Asian and white women, respectively. Results by age or race/ethnicity have been reported only when such differences were noteworthy.

The potential impact of surveillance or diagnostic bias was also evaluated by conducting analyses stratified by how the cancer came to medical attention. Women with thyroid cancer were asked whether their cancer came to medical attention because (a) they went to see their physician after experiencing a problem with their neck or throat or (b) their physician discovered their tumor during examination for something else. According to whether they agreed with either the first or second choice, cases were respectively classified into two subgroups: “patient-found tumor” or “physician-found tumor.”

The age and race distributions for interviewed cases and controls were similar (Table 1). The mean age ± SD was 42.1 ± 12.6 and 43.1 ± 13.3 years old, respectively, for cases and controls. Approximately one-half of the participants identified themselves as white and 35% as Asian. Women of Filipina, Chinese, or Vietnamese descent comprised the majority of the Asian group. Most participants received at least a high school diploma; this was true for more controls than for cases. Women with thyroid cancer were more likely to report history of goiter or nodules, radiation to the head or neck region, and first-degree family history of proliferative thyroid disease.

Relative to menarche at age 12, papillary thyroid cancer risk was associated with both earlier and later reported ages of menstrual onset (Table 2). When stratifying by age, this U-shape pattern was only observed among women of postreproductive age (≥45 years old; Table 3). Among Asian women, self-reported menarche at age 15 or older conferred greatest risk of thyroid cancer (OR, 1.9; 95% CI, 0.93–3.7), and risk increased per year of later menarche (OR, 1.10; 95% CI, 0.97–1.24). Adjusting for place of birth, as within or outside of the United States, did not change these risk estimates; about 85% of the Asian women reported being foreign born. A contrasting trend was noted among white women (OR, 0.95; 95% CI, 0.84–1.06 per year) with the highest risk group being those experiencing menarche before age 12 (OR, 1.5; 95% CI, 0.91–2.6). These ethnic-specific patterns were most pronounced among women of postreproductive age, with the greatest risk observed among older Asian women (age ≥45) who had experienced later menarche (age ≥15; OR, 6.5; 95% CI, 2.2–19.2).

When analyses for age at menarche were conducted in reference to menarche before age 13, as cited in the recently published pooled analysis by Negri et al. (24), the U-shape pattern of risk was not apparent (Tables 2 and 3). Overall, women reporting menarche at ages 15 and older were at 23% increased risk for thyroid cancer (95% CI, 0.84–1.8). The magnitude of such risk was about 2-fold among women of postreproductive age (OR, 2.2; 95% CI, 1.2–4.1) and of Asian descent (OR, 1.7; 95% CI, 0.93–3.1).

Approximately 70% of participants were premenopausal and 21% were postmenopausal. The menopausal status of the remaining 9% could not be determined because of the use of HRT before the cessation of menses. Neither menopausal status nor age at menopause was related to risk of papillary thyroid cancer (Table 2).

Relative to nulligravid women, papillary thyroid cancer risk estimates for ever having been pregnant (OR, 0.94; 95% CI, 0.69–1.3), ever having a miscarriage (OR, 1.0; 95% CI, 0.68–1.5), and ever having an abortion (OR, 0.87; 95% CI, 0.71–1.2) were all near unity. Miscarriage as a woman’s first pregnancy outcome (Table 4), a risk factor for thyroid cancer according to several other studies (11, 13, 17, 24), was associated with a nonsignificant increase in risk among reproductive-aged women (OR, 1.4; 95% CI, 0.74–2.6) and Asian women (OR, 1.9; 95% CI, 0.75–4.8); a significant, albeit unstable risk estimate was noted for reproductive-aged Asian women (OR, 14.0; 95% CI, 1.7–114.3).

As shown in Table 4, the relationship between parity and papillary thyroid cancer risk was mixed. When adjusting only for matching variables, established risk factors, and parity, having a recent pregnancy (whether defined by either late age at first FTP, later age at last FTP, time since last pregnancy or FTP, or the number of pregnancies or FTPs in the last 5 years) seemed to confer a 2-fold or greater increase in risk among the younger group of parous women. The effect for age at first FTP was slightly diminished in further adjusting for any one of the other variables reflecting a recent pregnancy (e.g., OR, 2.9; 95% CI, 1.3–6.8, and OR, 2.2; 95% CI, 0.85–5.9, for age at first FTP between 25 and 29 and ≥30, respectively, in relation to <20, adjusting for the number of recent pregnancies), whereas the effects of these other variables, when adjusting for age at first FTP, were reduced more substantially (e.g., OR, 1.6; 95% CI, 0.82–3.3, for 2 or more recent pregnancies compared with none, adjusting for age at first FTP). In analyses stratified by how the cancer came to medical attention (Table 5), the effect for age at first FTP was greater among women whose physician found the tumor. In contrast, the effect for the number of pregnancies in the last 5 years was stronger among women who discovered their own tumors. Across both strata, however, there was considerable overlap in the CIs for each variable.

Although more controls (68%) than cases (62%) reported ever having used OCs, the average duration of OC use for controls and cases was comparable, 63 and 67 months, respectively. Women who used OCs, whether former or current users, were at about 25% lower risk for papillary thyroid cancer (Table 6). When examining this association stratified again by how cancer came to medical attention, the effect was slightly stronger among women who discovered the tumor themselves (Table 5). No consistent patterns of risk were observed for duration of OC use or age at first or last OC use.

Analyses for HRT were restricted to the 438 women of postreproductive age (Table 6). About 46% reported ever having used HRT. There was some suggestion that women who used HRT may be at lower risk for papillary thyroid cancer. However, no distinct trend was observed for duration of HRT use, and age at first HRT use was unrelated to thyroid cancer risk.

We examined numerous aspects of hormonal and reproductive events but noted few associations with papillary thyroid cancer risk. Women who reported either early or late age at menarche, relative to menarcheal onset at age 12, were at about 50% increased risk for papillary thyroid cancer; however, age and ethnic differences in this pattern were noted. Risk was also elevated among young, parous women with a recent pregnancy, although the exact nature of that relationship remains somewhat unclear. Although ever use of OCs appeared to be protective, no trend was observed for duration of use.

With some exceptions, our results reflect those of recently published work, a case-control study conducted in western Washington of 410 papillary cases and 574 controls (20, 21), a matched case-control study conducted in Los Angeles County of 292 white female pairs (82% papillary carcinoma; Ref. 13), and a pooled analysis of data on 2247 cases (80% papillary carcinoma) and 3699 controls amassed from 14 case-control studies of thyroid cancer (24, 25). Like Rossing et al. (21) and Negri et al. (24), we observed that risk was related to later age at menarche; Mack et al. (13) found no association. Yet, we also noted differences in effects among subgroups of women. Asian women with later menarche were at increased risk of papillary thyroid cancer, whereas among white women, early age conferred some elevation in risk. The reasons for this difference are unclear. Furthermore, age at menarche (both early and late) seemed to be more important among older (i.e., age 45) rather than younger women. When examining the effects for age at menarche by age and ethnicity jointly, the effects were strongest for late menarche among older Asian women.

In the pooled analysis (24), miscarriage as the outcome of a woman’s first pregnancy was a risk factor for papillary thyroid cancer (OR, 1.7; 95% CI, 1.1–2.7). A similar finding was reported by Mack et al. (13) but specifically for white women younger than 35 years old with no prior benign thyroid disease (OR, 2.7; 95% CI, 1.1–6.8). We observed this association only for Asian women of reproductive age, but the CI around our estimate was very wide. For most other pregnancy-related events, our findings for the entire study population were also similar to those of Negri et al. (24) with respect to weak associations.

Among parous women, the pooled analysis as well as the more recent studies among young women, including our own, found late age at first birth to confer some increased risk (13, 21, 24). Like Rossing et al. (21), we observed this association to be attenuated slightly when other variables of recency of pregnancy were measured simultaneously. However, in our data, there was evidence that the assessment of age at first FTP may reflect a certain amount of surveillance bias. The strongest finding of Rossing et al. (21) was for the number of births in the last 5 years, but they do not report the effects of adjusting this variable for age at first FTP. Our other measures of recency of pregnancy were also attenuated when adjusting for age at first FTP. Nevertheless, surveillance bias is not a likely explanation for these latter findings because the effects were stronger or equally strong among women who discovered their tumors themselves compared with those whose tumors were discovered by a medical professional. Such findings, overall, suggest that there is some, probably relatively short-term, effect of pregnancy on thyroid cancer risk; however, the exact nature of this effect remains unclear.

Contrary to the pooled analysis (25), which found OC use to confer a small increase in risk (OR, 1.2; 95% CI, 1.0–1.4), particularly among current users (OR, 1.6; 95% CI, 1.1–2.4), and to Mack et al. (13), who observed no association, we found OC users to be at reduced risk of papillary thyroid cancer. Rossing et al. (20) also reported a decrease in risk (OR, 0.6; 95% CI, 0.4–0.9) with OC use among younger women (age, <45), but no association among older women. None of these studies, however, showed an association with duration of use or age at first use. The lack of trend with duration of use suggests that this association may not reflect an etiological effect but a possible confounding by yet unidentified characteristics of the women who did or did not use OCs in these various studies.

Despite fair consistency between the results of the present study and others, several biases may influence the observed associations. First, women who participated in these epidemiological studies may have been characteristically different from women who did not participate. For factors whose effects on risk are relatively consistent across studies, nonparticipation would have had to been related to the same factors in each of the diverse populations which have been studied to produce consistently biased results, but this seems unlikely. In our study, White women (79% of cases, 77% of controls) were more willing to participate than Asian women (66% of cases, 62% of controls). However, our response rates for Asian women are fairly similar to those reported for Asian women in the San Francisco-Oakland area who participated in a case-control study of breast cancer a decade earlier (68% of cases, 75% controls; Ref. 29). Yet, whether or not such variation in response might affect the risk differences observed by ethnicity is unknown. Second, inherent to the case-control design, the accuracy of recall between women with and without thyroid cancer may have differed. However, the role of menstrual and reproductive factors in thyroid cancer etiology is not well established and, therefore, consistent preconceived ideas are not likely held in the general population, decreasing the probable impact of recall bias. Recall of reproductive events should be relatively good among young women because these events would have been memorable and recent.

Perhaps of greatest concern is surveillance/diagnostic bias because thyroid cancer is often an asymptomatic, slow-growing disease, detected by health practitioners during examination for other conditions. As McTiernan et al. (14) have suggested, it is plausible for women who often seek medical care to be more likely diagnosed with thyroid cancer and also to use exogenous estrogens or to have had more pregnancies. In our analysis of 220 papillary thyroid cancer cases among parous women of reproductive age, 47% were self-detected and 53% were discovered by a physician. Although differences in medical utilization were not directly measured, a similar percentage of these cases (i.e., 98%) and controls (i.e., 96%) reported having medical insurance coverage within the last 5 years before the interview. Evidence of surveillance bias was observed only for age at first FTP; for all other variables examined in the stratified analysis by how cancer was diagnosed (including other measures of recent pregnancy), risk estimates were slightly stronger for women who self-detected their cancer than for those whose cancer was discovered by a physician. Two previous studies likewise failed to note substantial differences in risk estimates between similarly stratified groups (14, 20).

Overall, the present study suggests that hormonal changes associated with a recent pregnancy may be worth further exploration in elucidating the role of sex hormones in thyroid cancer etiology. The substantially and consistently higher incidence rates of thyroid cancer in women compared with men, which emerge at puberty and disappear after menopause, also suggest that although the associations between self-reported menstrual and reproductive events and cancer risk are not strong, further examination of the potential role of sex hormones in thyroid carcinogenesis is necessary.