Thyroid cancer is a relatively rare form of cancer that occurs 2–3 times more frequently among females. In most countries, it accounts for approximately 1–5% of all cancers in females and <2% in males.1 This female predominance, which is greatest during reproductive ages, is observed in all geographical areas and ethnic groups. The age-standardized incidence rates (per 100,000) of thyroid cancer, across most populations, vary from approximately 2–10 in females and 1–3 in males.1
As thyroid cancer and the majority of benign thyroid disorders (e.g., Graves disease, non-endemic goitre, Hashimoto thyroiditis) are significantly more common in women, a major effort has focused on examining the influence of reproductive and hormonal factors in the etiology of diseases of the thyroid gland. An international collaborative group has recently published a series of overviews by pooling data from all the case-control studies of thyroid cancer published during the period 1980–97 (n = 12) and 2 unpublished studies.2–4 The epidemiology of thyroid cancer has been comprehensively reviewed by Franceschi et al.,5 Salabe,6 Ron7 and Mack and Preston-Martin.8 Along with the epidemiological observations, there is more direct clinical evidence for the role of hormonal factors in thyroid cancer. Elevated levels of thyroid stimulating hormone (TSH) are associated with thyroid hyperplasia and are thought to increase the risk of neoplastic transformation.9, 10 Increased secretion of TSH occurs during puberty, pregnancy, delivery and puerperium, as well as while using oral contraceptives.11–13 In healthy pregnant women, there is a progressive increase in TSH levels, basal metabolic rate and thyroid volume (pseudo-goitre) throughout pregnancy. It has also been shown that serum human chorionic gonadotrophin, with its marked elevation in early pregnancy, may directly stimulate the thyroid gland through its TSH-like activity.14 Moreover, oestrogen receptors have been demonstrated in thyroid cancers, particularly in well-differentiated carcinomas and oestrogen has been shown to promote thyroid tumors in animal models.15, 16 Based on the findings from descriptive and analytical epidemiology, laboratory and animal studies and individual case reports, there is sufficient evidence to suggest that hormonal and biochemical changes related to reproductive events and patterns could be relevant to the etiology of benign thyroid dysfunction or disease and neoplasia.
Little is known about the epidemiology of thyroid cancer in the Middle East. During the 5-year period 1994–1998, thyroid cancer (ICD-10 code: C73) was the second most common cancer (after breast) among women in Kuwait, accounting for 8.1% of all cancers among Kuwaiti and 8.7% among non-Kuwaiti (expatriate) women. The average annual incidence rate (per 100,000) was 5.3 in Kuwaiti and 4.7 in non-Kuwaiti women; the age-standardized rate was 7.7 in Kuwaiti and 5.0 in non-Kuwaiti women (Kuwait Cancer Registry, unpublished data). Similarly high relative frequency and rates of thyroid cancer among women have also been reported from other countries in the Gulf region. Data from recently established cancer registries in Oman, Qatar, Saudi Arabia and the United Arab Emirates show that thyroid cancer is also the second most common cancer among women in these countries, accounting for 7.7%, 10.9%, 8.8% and 12.4% of all female cancers, respectively.17–19 To our knowledge, there has been no previous attempt to study the etiology of thyroid cancer in the Middle East. We have conducted a population-based case-control study in Kuwait to examine potential relationships between reproductive and hormonal factors and thyroid cancer.
For administrative purposes, Kuwait, with its population of about 2.2 million and area of 17,818 square km, is divided into 6 governorates. Each governorate has a well-defined area and population and is accordingly divided into a number of districts. Medical services in each governorate comprise a network of primary health care clinics (at least 1 in each district) and a general public hospital. In addition, there are a number of centralized specialty hospitals, including the Kuwait Cancer Control Centre (KCCC). Health care at all levels is provided free of charge by the government to all residents; except for some services for which non-Kuwaitis (expatriates) pay a nominal fee.
Our study was conducted at the KCCC, which is the only specialized cancer treatment hospital in the country and offers modern facilities in various modalities of cancer diagnosis, treatment and follow-up. Almost all suspected cancer patients are referred to the centre; those who are initially diagnosed or treated elsewhere are also referred for further treatment or follow-up. A thyroid cancer clinic is held every week for current and newly diagnosed patients for regular follow-up and consultation. A population-based cancer registry (Kuwait Cancer Registry) has been established at the center since 1979. The medical records and reports of the pathology department at the centre serve as the principal source of information to the registry. Notification of cancer to the registry is compulsory through a Ministerial Decree. The registry collects information on malignant neoplasms according to the recommendations of the International Agency for Research on Cancer (IARC) and has regularly contributed data in the IARC monographs.1
Ascertainment of cases and controls
As a first step, we extracted information on all the 774 patients with thyroid cancer (ICD-9 code: 193; ICD-10 code: C73) who were registered at the cancer registry between 1 January 1981 and 31 December 1996. Of these, 238 (30.7%) patients were Kuwaitis (188 females, 50 males) and 536 (69.3%) were non-Kuwaitis (348 females, 188 males). Among these patients, 66 were known to have died; 43 were Kuwaitis (30 females, 13 males) and 23 were non-Kuwaitis (12 females, 11 males). As at the time of the study (1 May 1998–30 June 1999) the cancer registration data were available up to 1996, details of 100 additional patients diagnosed with thyroid cancer between 1 January 1997 and 30 June 1999 were obtained from the medical records and registers of the thyroid cancer clinic at the KCCC. Of these, 50 patients were Kuwaitis (43 females, 7 males) and 50 were non-Kuwaitis (32 females, 18 males). To achieve sufficient statistical power for a meaningful analysis, we decided to include as many of the cases as possible that were diagnosed since 1 January 1981. This rationale was based on the fact that there were relatively small number of new cases each year, particularly among the Kuwaiti nationals and that many non-Kuwaiti patients return to their home country after diagnosis or initial treatment. Furthermore, during the period of Iraqi occupation of the country and subsequent Gulf War (August 1990–February 1991) a large majority of non-Kuwaitis returned to their home countries; and those with chronic diseases, particularly cancer, were most unlikely to have returned to Kuwait after the liberation.
Cases were eligible for inclusion in the study if thyroid cancer had occurred as the first primary cancer and if they were alive, aged ≤70 years and resident in Kuwait at the time of the study. Confirmation of diagnosis and histopathological characteristics for all patients included in the study was obtained by a systematic review of the medical records and, where necessary, consultation with the attending oncologist. A bilingual female researcher, proficient in Arabic and English languages and not aware of the epidemiology of thyroid cancer, was trained to conduct personal interviews with cases and controls. All eligible patients who attended the weekly thyroid cancer clinic or other in- and out-patient departments at the KCCC during the study period (1 May 1998–30 June 1999) were solicited to participate in the study. Other eligible patients, who did not visit the KCCC during the study period were traced through the information available in medical records and where possible, were invited to visit the hospital for an interview. The study finally included a total of 313 patients with thyroid cancer (238 females, 75 males). Of these, 270 (86.3%) patients were recruited from the thyroid cancer clinic or other departments at the KCCC and 43 (13.7%) were successfully traced through the medical records; with participation rates of 96.8% (270/279) and 89.6% (43/48) respectively.
The district of residence was determined for each case at the time of the interview and the local primary health care clinic was visited to select a suitably matched control subject. Primary care clinics are conveniently located in the centre of each district, usually along with other public amenities (such as supermarkets, shopping malls, banks and restaurants); and most clinics operate 7 days a week from 7:00 AM to 21:00 PM. All clinics have separate sections for men, women and children; and residents in the district can seek consultation with a General Practitioner on first-come-first-serve basis as prior appointment is not required. Besides these consultations, a typical primary care clinic offers services such as a nurse clinic (for minor medical/surgical procedures) ante-natal and well-baby clinic, vaccination, school health department, preventive medicine department (essentially for control of infectious diseases), dental clinic, laboratory (for basic blood and urine examination) and a pharmacy. Due to the convenient location and timings and easy access to a wide variety of services, it is generally agreed that all residents in the district have an equal opportunity to visit the clinic. We were therefore confident that we would obtain a reasonably representative sample of controls from primary care clinics during the study period.
One control subject was individually matched to each case, based on year of birth (within 3 years), gender, nationality and district of residence. Subjects were considered eligible to serve as controls if they were visiting the primary care clinic for minor complaints (upper respiratory tract infections, skin rash/infection, acute gastrointestinal complains, headache, back/joint pain, minor trauma/injuries and lower urinary tract infections) or accompanying such persons (e.g., a woman accompanying her child/husband/friend) or visiting for any other purpose. Of the 340 eligible controls subjects, solicited for inclusion in the study, 27 refused to be interviewed, yielding a response rate of 92.1% (313/340). All cases and controls included in the study were initially contacted and interviewed in-person during the study period that extended from 1 May 1998–30 June 1999.
Assessment of exposure and analysis
Information from all cases and control subjects was obtained by using a structured questionnaire, administered by the same interviewer. A verbal informed consent was obtained from all participants before initiating the interview. The questionnaire was divided into 6 parts and collected information on: (i) sociodemographic characteristics (age, gender, marital status, level of education, nationality, area of residence, occupation, smoking status); (ii) gynecological and reproductive history (age at menarche and menopause, number and outcome of pregnancies, use of female hormones for contraception, hormone replacement therapy and other indications, surgical removal of uterus and ovaries); (iii) medical history (history of autoimmune, chronic, neoplastic and benign thyroid disease; and history of exposure to diagnostic and therapeutic radiation); (iv) family history (parental consanguinity, history of benign and malignant thyroid disease and other cancers); (v) diet (frequency of consumption of 13 food items); and (iv) clinical and histopathological information (abstracted from the database of the cancer registry and medical records). In this report, we present a detailed analysis of the reproductive and hormonal factors among the 238 women with thyroid cancer and their control subjects.
For each control a “pseudo-diagnosis” date was determined, the date on which the subject was the same age as her matching case was at diagnosis. The analyses of data on reproductive and hormonal factors were restricted to events before the diagnosis (cases) or pseudo-diagnosis (controls) date. The analysis of data on smoking was based on smoking status at the time of diagnosis/pseudo-diagnosis. All data management and analyses were conducted using the SPSS and STATA statistical programs. We used standard statistical methods for individually matched case-control studies, as described by Breslow and Day,20 for the estimation of odds ratios (OR), adjusting for confounding variables where necessary. Unconditional logistic regression was used for the analyses of reproductive and hormonal factors because never-married women were excluded from the analysis of variables related to pregnancy, miscarriage and use of female hormones and this affected the matching (n = 26 pairs). The results obtained, however, did not differ substantially when the conditional method was applied. For all other variables, we used conditional logistic regression. Results are presented as odds ratios with corresponding 95% confidence intervals (95% CI). Unless otherwise indicated, all statistical tests are based on the likelihood ratio test procedure; and for the trend in odds ratios are based on a χ2 test for trend.
Characteristics of cases and controls
The distribution of 238 women with thyroid cancer and their individually matched controls, according to various sociodemographic variables and smoking status, is shown in Table I; 144 (60.5%) were Kuwaiti nationals and 94 (39.5%) were non-Kuwaitis. Among the non-Kuwaitis, the majority (71.3%) were of Arabic origin and 25.5% were from Southeast Asia. For the cases, the average age (±SD) at the time of interview was 40.9 ± 11.9 years (range 12–69 years); the median age was 41 years. There was no difference in the average age at the time of interview between Kuwaitis and non-Kuwaitis (41.0 ± 12.7 vs. 40.7 ± 10.6). Table II shows the distribution of women with thyroid cancer according to age at diagnosis, year of diagnosis and histology. The majority (76.8%) of women were diagnosed with thyroid cancer during their reproductive age (15–44 years). The average age at diagnosis was 34.7 ± 11 years (range 10–65 years); the median age was 35 years. There was no difference in the average age at diagnosis between Kuwaitis and non-Kuwaitis (34.7 ± 11.6 vs. 34.8 ± 11.0). The majority (71%) were diagnosed during the period 1991–99 and 29% during 1981–90. Papillary carcinoma (including mixed papillary/follicular variant) was the most common histopathological type accounting for about 83% of all cases; the remaining were classified as follicular (9.7%), medullary (0.4%), or other (0.8%).
Table I. SocioDemographic Characteristics of 238 Women with Thyroid Cancer and Their Individually Matched Controls
| Kuwaiti||144 (60.5)||144 (60.5)|
| Non-Kuwaiti||94 (39.5)||94 (39.5)|
| Arabs1||44 (46.8)||44 (46.8)|
| Southeast Asian||24 (25.5)||24 (25.5)|
| Bedouin2||23 (24.5)||23 (24.5)|
| Other||3 (3.2)||3 (3.2)|
| Single||19 (8.0)||21 (8.8)|
| Married||195 (81.9)||180 (75.6)|
| Widowed||18 (7.6)||30 (12.6)|
| Separated/divorced||6 (2.5)||7 (2.9)|
| No formal education||94 (39.7)||73 (30.7)|
| Primary||9 (3.8)||14 (5.9)|
| Intermediate/secondary||82 (34.6)||79 (33.2)|
| Professional diploma||23 (9.7)||31 (13.0)|
| University +||29 (12.2)||41 (17.2)|
| Professional||5 (2.1)||6 (2.5)|
| Semi-professional||53 (22.3)||60 (25.3)|
| Skilled||3 (1.3)||1 (0.4)|
| Semi-skilled||5 (2.1)||9 (3.8)|
| Manual labourer||17 (7.1)||8 (3.4)|
| Housewife||139 (58.4)||144 (60.8)|
| Unemployed||16 (6.7)||9 (3.8)|
|Cigarette smoking status|
| Non-smoker||217 (91.2)||226 (95.0)|
| Former smoker||10 (4.2)||4 (1.7)|
| Current smoker||11 (4.6)||8 (3.4)|
Table II. Distribution of 238 Women with Thyroid Cancer by Age at Diagnosis, Year of Diagnosis, and Histology
|Age at diagnosis (years)|
| 10–14||3 (1.3)|
| 15–24||47 (19.7)|
| 25–34||65 (27.3)|
| 35–44||71 (29.8)|
| 45–54||42 (17.6)|
| 55–65||10 (4.2)|
|Year of diagnosis|
| 1981–85||31 (13.0)|
| 1986–90||38 (16.0)|
| 1991–95||77 (32.4)|
| 1996–991||92 (38.7)|
| Papillary||139 (58.4)|
| Papillary/follicular variant||58 (24.4)|
| Follicular||23 (9.7)|
| Medullary||1 (0.4)|
| Other||2 (0.8)|
| Unknown||15 (6.3)|
Association with demographic factors
Table III shows the association between selected demographic factors and thyroid cancer. There was an increase in risk of thyroid cancer among women who had ever smoked cigarettes (current and former smokers) compared to those who had never smoked (OR = 2.1; 95% CI: 0.9–5.3). With regard to educational level, we found a significant inverse association with thyroid cancer (p-trend <0.05): women with 12+ years of education (professional diploma/university degree) were at a significantly reduced risk of thyroid cancer compared to those who had no formal education (illiterate/read and write) (OR = 0.4; 95% CI: 0.2–0.8). There was no association between marital status and thyroid cancer.
Table III. Risk of Thyroid Cancer Among Women Associated with Selected SocioDemographic Factors1
| Never married||19/21||1.0||—||—|
| Ever married||219/217||1.2||0.5–2.5|
| No formal education||94/73||1.0||—||<0.05|
Association with reproductive factors
Table IV presents a descriptive analysis of selected reproductive and hormonal factors, such as age at first and last pregnancy, miscarriage, parity and use of oral contraceptives, among cases and controls. Table V shows the association between various reproductive events and the risk of thyroid cancer. Overall, events such as age at menarche, pregnancy, menopausal status and age at menopause were not associated with incidence of thyroid cancer. There seemed to be an association, however, between age at first and last pregnancy and risk of thyroid cancer. Young age at first pregnancy seemed to have a protective effect, as the odds ratios were consistently below unity for ages 16–24 years. Women who had their last pregnancy at ages ≥30 years were at a significantly increased risk compared to those who had their last pregnancy at earlier ages (OR = 2.1; 95% CI: 1.2–3.8). There was also a significant trend in risk with increasing age at last pregnancy. With regard to parity, the odds ratios were consistently above unity for women who had more than 2 live births. Considering the relatively high number of children borne by the women in our study, women with ≥5 births were at a modestly increased risk compared to those who had fewer number of children (OR = 1.5; 95% CI: 0.9–2.5). There was also some indication of an increasing trend in risk with increasing number of live births (p-trend = 0.08). In contrast, miscarriage seemed to have a protective effect on the incidence of thyroid cancer as the odds ratios were consistently below unity, compared to the reference category, for all variables related to the experience (i.e., as an outcome of first and last pregnancy, ever had a miscarriage and number of miscarriages). There was a significant decreasing trend in odds ratios with increasing number of miscarriages. The maximum protective effect was observed for women who had a miscarriage as the outcome of first pregnancy (OR = 0.1; 95% CI: 0.03–0.4) and for those who had experienced ≥3 miscarriages (OR = 0.3; 95% CI: 0.1–0.8). These findings are almost entirely based on the experience of spontaneous abortion. It is noteworthy that induced abortion, due to religious and sociocultural reasons, is prohibited by Kuwaiti law.
Table IV. A Descriptive Analysis of Selected Reproductive and Hormonal Factors Among 238 Women with Thyroid Cancer and Their Individually Matched Controls
|Menarche (age in years)||13.2 ± 1.6||10–19||13||13.1 ± 1.6||9–17||13|
|Pregnancies (n)||5.4 ± 3.9||0–15||5||5.0 ± 3.8||0–16||5|
|First pregnancy (age in years)||20.2 ± 4.5||12–33||19||20.3 ± 3.9||12–31||20|
|Last pregnancy (age in years)||32.0 ± 6.9||18–51||32||30.9 ± 7.0||18–56||30|
|Births (n)||4.9 ± 3.5||0–15||5||4.3 ± 3.3||0–14||4|
|Miscarriages (n)||0.5 ± 1.1||0–9||0||0.7 ± 1.2||0–6||0|
|First use of female hormones (age in years)||25.1 ± 6.7||16–46||23||25.4 ± 6.3||15–50||24|
|Duration of female hormone use (months)||45.7 ± 39.4||0.5–144||36||44.0 ± 36.7||1–180||36|
|First use of OC1 (age in years)||25.2 ± 6.9||16–46||23||24.9 ± 6.1||15–50||23|
|Duration of OC1 use (months)||47.7 ± 39.6||0.5–144||36||44.7 ± 37.3||1–180||36|
|Menopause (age in years)||46.7 ± 6.6||35–60||46||44.9 ± 7.9||30–60||45|
Table V. Risk of Thyroid Cancer Among Women Associated with Various Reproductive Factors1
|Age at menarche (years)|
|Age at first pregnancy (years)3|
|Outcome of first pregnancy3|
| Live birth||187/159||1.0||—|
|Age at last pregnancy (years)3|
|Outcome of last pregnancy3|
| Live birth||178/173||1.0||—|
|Number of live births|
|Time since last birth (years)3|
|History of post-partum thyroiditis3|
|Ever had a miscarriage34|
|Number of miscarriages34|
|Age at menopause (years)|
We also determined whether there was an association between last birth and subsequent diagnosis of thyroid cancer (by subtracting age at last birth from age at diagnosis). The average time between the two events was 6.1 ± 6.7 years (range 0–35 years); the median time was 3 years. Women in the second and third year after a birth were at a substantially increased risk compared to those who had their last child more than 10 years before the diagnosis (OR = 2.0; 95% CI: 1.0–4.1). An increased risk of similar magnitude was observed in the second and third year after a birth when we used women in the first year of birth as a referent group (OR = 2.4; 95% CI: 1.3– 4.2) (data not shown). We also asked women about the history of post-partum thyroiditis, a benign, often self-limiting and transient, episode of thyroid dysfunction occurring 2–8 months after delivery. Women who had experienced episodes of post-partum thyroiditis were at a significantly increased risk of developing thyroid cancer (OR = 10.2; 95% CI: 2.3–44.8). Two cases who also had a history of hyperplastic thyroid disease (1 with nodule, 1 with goitre) were excluded from this analysis; none of the women included in the analysis had history of any other benign thyroid disease.
As high parity and older age at last pregnancy were positively associated with thyroid cancer, we conducted a further analysis to examine: 1) which factor (i.e., multiple exposures to pregnancy or being old when exposed to pregnancy) was more important; and 2) the joint effects of these exposures in thyroid cancer (Table VI). The analysis showed that childbearing during the latter half of reproductive life had a substantial effect on the incidence of thyroid cancer; for any given level of parity, there was about a 2-fold increased risk if the age at last pregnancy was ≥30 years.
Table VI. Risk of Thyroid Cancer Among Women Associated with the Joint Effects of Age at Last Pregnancy and Parity1
Association with hormonal and gynecological factors
Among women who had ever used female hormones (for any indication), the large majority had used them exclusively for contraceptive purpose (Table VII). There was no difference in the number of cases and controls who had used oral contraceptives (89 vs. 91, respectively). Overall, any female hormone use was not associated with thyroid cancer risk. Based on small numbers, an OR of 3.2 was observed for women who had ever used lactation suppressants. Only a small number of cases and controls had hysterectomy or oophorectomy; an OR of 3.1 was observed for women who had an oophorectomy.
Table VII. Risk of Thyroid Cancer Among Women Associated with Various Hormonal and Gynaecological Factors1
|Ever used female hormones|
|Age at first use of female hormones (years)|
|Duration of female hormone use (months)|
|Ever used OC3|
|Age at first OC3 use (years)|
|Duration of OC3 use (months)|
|Ever used hormones for infertility|
|Ever used lactation suppressants|
|Ever used HRT4|
|Ever had hysterectomy|
|Ever had oophorectomy|
To our knowledge, this is the first epidemiological study to ascertain the association of reproductive and hormonal factors in the incidence of thyroid cancer in the Middle East. Since the establishment of population-based cancer registry in the late 1970s, thyroid cancer has consistently been the second most commonly recorded neoplasm among Kuwaiti women.1, 21 Similarly high relative frequency and incidence of the disease has also been observed in other countries in the Gulf region.17–19 Concerning the reproductive factors, these countries also have relatively high birth and total fertility rates. For example, in 1998, the total fertility rate (i.e., number of children per women) in Bahrain, Kuwait, Oman, Qatar, Saudi Arabia and the United Arab Emirates was 3.4, 5.2, 6.5, 5.1, 5.0 and 4.9, respectively.22 Other reproductive health patterns include relatively high prevalence of consanguinity, young age at first marriage and childbearing, short birth intervals and older ages at last birth.23 All of the above factors result in a relatively long reproductive life span. The Kuwaiti women in our study showed this reproductive pattern (see Tables IV,V).
A number of case-control studies, mostly from western Europe and USA and 2 from Asia (China and Japan) have examined the influence of reproductive factors in thyroid cancer. The data are limited, however, and the findings have been inconsistent, partly because of the relatively small number of cases in some of the studies. A meta-analysis by a consortium2 showed that parity, spontaneous or induced abortion and history of infertility were not associated with the risk of thyroid cancer. The OR was above unity for women who had their first birth at ages ≥30 years (OR = 1.3; 95% CI: 1.0–1.8). It was concluded that associations of menstrual and reproductive factors with thyroid cancer were generally weak, but seemed stronger among women diagnosed with the disease at younger ages (≤35 years).3 A similar conclusion was reached concerning the role of exogenous female hormones: there was no evidence of a persistent excess risk of thyroid cancer after use of oral contraceptives and hormone replacement therapy.4 It is also noteworthy that all the studies in the combined analysis have been conducted in countries with relatively much lower birth and fertility rates than Kuwait. Moreover, the mean age at diagnosis of thyroid cancer (about 35 years) among women in our study was relatively much lower than that reported in other series (45–54 years).7
Case-control studies are subject to a variety of biases. These may include issues of case and control ascertainment, misclassification, representativeness and participation rates; recall and information bias between cases and controls and survival bias. The free and easy access to heath care, KCCC being the only cancer hospital in the country and the weekly thyroid cancer clinic, provided equal opportunity to all thyroid cancer patients, resident in the country, to visit the KCCC during the 14-month study period. Patients who were living abroad or had emigrated (particularly non-Kuwaitis), those with incomplete or unavailable case-notes or invalid telephone contact had no opportunity to be included in the study. Nevertheless, the age, gender and histological distribution of cases included in our study (see Table II) is similar to the population-based data from the cancer registry.
Sociodemographic differences in the characteristics of women who did or did not agree to participate in the study may influence our results; as may errors in recall of various aspects of reproductive life and events, particularly for cases diagnosed in the 1980s. The possibility of recall bias is always a potential problem in case-control studies. Based on our experience with other studies conducted in Kuwait, certain reproductive variables, particularly parity, ages at births and miscarriage, are unlikely to be greatly affected by recall or information bias. It is noteworthy that the total fertility rate among women in our study was similar to the national rate; and the reproductive patterns and events among our cases and controls (see Table IV) are consistent with the findings of a large population-based survey conducted by the Ministry of Health.22, 23 Furthermore, the distinctive network of primary care clinics in Kuwait, with free and easy access to a variety of services, provided equal opportunity to all women, with varying reproductive or other characteristics, to be selected as control subjects.
Among all cancers, thyroid cancer has one of most favorable prognosis with all-stage relative 5-year survival rate of around 95% for women.7 The study included cases diagnosed over an extended period; 66 patients were known (in the records of the cancer registry) to have died by the time our study was started. Some other patients may have died abroad or from causes other than thyroid cancer. The exclusion of these cases from the study might influence the results if some reproductive or hormonal exposure is associated with relatively worse prognosis of thyroid cancer. Furthermore, there have also been changes in histological classification and diagnostic criteria of thyroid cancer over this period.5, 24 Data from our cancer registry show that over 97% of all thyroid cancers diagnosed in Kuwait were histologically verified. Variation between pathologists in classification of histological subtypes, however, is likely to have led to some misclassification by histology.
In our study, age at pregnancy seemed to be significantly related to the risk of thyroid cancer. Women who started childbearing at a young age had a substantially reduced risk of thyroid cancer. Women who had their last pregnancy at ages ≥30 years were at about 2 times increased risk. There was also a significant trend in risk with increasing age at last pregnancy. Considering that the reproductive life (ages 15–45 years) of women can be divided approximately into 2 equal halves, these findings (along with recency of diagnosis in the second and third year after a birth and the mean age at diagnosis of about 35 years) suggest that childbearing during the first half of reproductive life offers some protective effect; whereas, having a large number of children during the latter half substantially increases the risk of thyroid cancer. These findings are biologically plausible and are consistent with the hypothesis concerning the possible role of thyroid stimulating hormone in thyroid cancer.
Only 3 other case-control studies have examined the effect of age at pregnancy in some detail. An increased risk of similar magnitude was reported in a study from Italy (relative risk (RR) = 2.2; 95% CI: 1.3–3.7 for women aged ≥30 years at last birth);25 whereas, no association was found in studies from Switzerland26 and Washington State in the US.27 In a nested case-control study of a nation-wide Swedish cohort, there was a significant trend in excess risk among women who had a live birth at ages ≥25 years (OR = 1.6; 95% CI: 1.1–2.5 for women aged ≥35 years at last birth).28 A positive association, of similar magnitude, with late last birth (≥35 years) was also demonstrated in a prospective cohort study of 63,090 women in Norway.29
Among the various reproductive factors, parity has received most attention. Most case-control studies have demonstrated that women with thyroid cancer have had relatively more pregnancies or live births than controls.25, 26, 28, 30–36 These studies have shown modest increase in risk with parity (varying approximately between 1.5–2.5 for parity ≤3 vs. >3). Some case-control studies, however, did not show any association with parity.27, 37–39 In a prospective birth cohort study of 1.1 million Norwegian women of reproductive age, there was a progressive increase in risk of thyroid cancer with parity (RR = 1.0, 1.13, 1.30, 1.39, 1.46 for women with 0, 1, 2, 3 and 4+ live births, respectively).40 The risk estimates for parity 2, 3 and 4+ were statistically significant. Another cohort study of 63,090 Norwegian women with 124 cases of thyroid cancer, however, did not find any increase in risk with parity.29 It is plausible that pregnancy at young age induces some protective modification in risk and this threshold is overwhelmed by prolonged and repeated hormonal (TSH, hCG and sex hormones) stimulation due to large number of pregnancies and subsequent metabolic stress and hypertrophy of the thyroid gland. It is also possible that thyroid gland of older women, who have also borne a relatively large number of children, may become more susceptible to some of the carcinogenic stimuli occurring during pregnancy. The relatively large number of children borne by the women in our study population provided a unique opportunity to ascertain the role of age at pregnancy and parity in thyroid cancer. In agreement with most studies, we also found a modest association between high parity and the risk of thyroid cancer.
For history of spontaneous or induced abortion, the data are conflicting; with some reports of increased risk of thyroid cancer with history of miscarriage or abortion, particularly as an outcome of first pregnancy26, 31–33, 37 and others of none.29, 38, 39, 41 Moreover, a decreased risk was found in a study from Italy for women who had ever experienced an abortion (RR = 0.6; 95% CI: 0.4–1.0).25 Our data showed an inverse relationship between miscarriage and risk of thyroid cancer, particularly for women who had a miscarriage as outcome of first pregnancy and for those who have had ≥3 miscarriages; there was also a significant trend in decreasing risk with increasing number of miscarriages. It is possible that these findings could well be due to chance, particularly when large number of associations are examined; differential recall or information bias between cases and controls; or due to interaction with some indigenous factor(s), the findings could be more relevant to our study population. It is plausible that early termination of pregnancy may lower the risk of cancer as the thyroid gland is relieved of the prolonged and recurrent stimulation by TSH and other hormones. As shown in the present and most other studies, this would be consistent with the tendency toward increasing risk of thyroid cancer in relation with increasing number of full-term pregnancies.
Only 3 epidemiological studies have examined the recency of diagnosis of thyroid cancer in relation to time since last birth.27, 28, 40 All studies found higher risk with shorter time intervals. In agreement with these studies, our data showed that women in the second and third year after the last birth were at a substantially increased risk of being diagnosed with thyroid cancer. Rossing et al.27 conducted a sub-analysis of their data to examine whether parity in the last 5 years had any influence on incidence of thyroid cancer; they found that women with parity ≥2 had a significantly increased risk of thyroid cancer (OR = 4.2; 95% CI: 2.0–8.9). These findings have led to the suggestion that hormonal, metabolic, or other biochemical changes during both pregnancy and lactation may exert a transient influence on risk of thyroid cancer.
Women who had a history of post-partum thyroiditis were at about 10 times increased risk of developing thyroid cancer. Numerous epidemiological studies have consistently demonstrated a strong link between history of benign thyroid disease such as adenoma, nodule(s) or goitre and thyroid cancer. In most case-control studies, the odds ratios for these conditions have varied approximately between 2.5–30.0. Post-partum thyroiditis, is an often self-limiting and transient episode of autoimmune thyroiditis that occurs in about 2–10% of women during the 2–8 months after delivery.42 In 10–25% of the women who had experienced 1 episode, the condition recurred after subsequent delivery. In our study, 10.5% of ever-pregnant cases had a history of post-partum thyroiditis. The possible link between this condition and thyroid cancer has not been studied before. It can be argued that this finding could be due to differential recall or information bias between cases and controls, as cases may have a better recollection of their history of thyroid-related conditions. The magnitude of risk, however, is consistent with similarly high risk reported for other benign thyroid conditions and could not be entirely attributed to possible minor differences in recall between cases and controls. The findings of further similar studies, showing consistent results, would finally permit more definite conclusions about the possible link between post-partum thyroiditis and thyroid cancer.
Women on oral contraceptives have higher levels of TSH compared to women with normal menstrual cycle. Among Kuwaiti women, there has been a decline in the use of oral contraceptives over the last two decades: the prevalence of current use among married women decreased from 79% in 1984 to 45% in 1999.43 In a recent household survey, about 62% of the married Kuwaiti women had ever used oral contraceptives; and they were used more often for spacing than limiting the number of children (Dr. Shah, personal communication). There has been no report of the prevalence of use of hormone replacement therapy in Kuwait, which we believe is relatively lower than the Western countries. In concordance with the majority of epidemiological studies,44 we did not find any association between use of exogenous female hormones and thyroid cancer.
A number of epidemiological studies have examined the association of thyroid cancer with cigarette smoking and body size or weight gain. Most studies examining smoking have shown a (usually non-significant) decrease in the risk of thyroid cancer among women.35–38, 45 Based on smoking status at the time of diagnosis/pseudo-diagnosis date, we found that women who had ever smoked were at a non-significant increased risk of thyroid cancer compared to those who had never smoked. In a recent cross-sectional survey, the prevalence of smoking among adult Kuwaiti women, aged 18–60 years, was around 2%.46 Most studies examining body size or weight gain have shown a positive association with thyroid cancer among women. These factors were not examined in the present study. It may be of interest, however, to note that compared to other populations, Kuwaiti women have one of the highest reported prevalence of obesity. In a cross-sectional survey, the prevalence of overweight (BMI >25) among adult Kuwaiti women was about 73% and the prevalence of obesity (BMI >30) was 41%.47
In conclusion, the results of our study, conducted on a Middle Eastern population with relatively high birth and fertility rates and incidence of thyroid cancer, support the hypothesis that reproductive factors and patterns, particularly childbearing at older ages and high parity may influence, or contribute to, the risk of developing thyroid cancer. On the contrary, miscarriage or a woman's underlying biology that leads to miscarriage may confer some protective effect. The results provide perhaps the first indication in the literature of a possible link between history of post-partum thyroiditis and thyroid cancer. The findings of the study may also be relevant to other countries in the Middle East, particularly those with similarly high fertility rates and incidence of thyroid cancer.