Cohort study of thyroid cancer in a San Francisco Bay area population

Authors


Abstract

Using data from a large health plan, we performed a cohort study of thyroid cancer among 204,964 persons (aged 10–89 at baseline in 1964–1973, 54% female) followed for a median of 20 years. There were 196 incident thyroid cancers (73 in men, 123 in women). Risk was independently and positively related to female gender [relative risk (RR) = 1.56, 95% confidence interval (CI) = 1.12–2.19], Asian race (RR = 2.86, 95% CI = 1.76–4.65), completed college or post-graduate education (RR = 1.76, 95% CI = 1.20–2.59), history of goiter (RR = 3.36, 95% CI = 1.82–6.20), radiation of the neck region (RR = 2.33, 95% CI = 1.28–4.23) and family history of thyroid disease (RR = 2.18, 95% CI = 1.17–4.05). An inverse association was found for black race (RR = 0.55, 95% CI = 0.33–0.91). Cigarette smoking, alcohol consumption, personal history of hyperthyroidism, hypothyroidism, overweight or obesity, weight gain since age 20, height, occupational exposures, reproductive factors, oral contraceptives and hormone use did not show statistically significant relations to thyroid cancer. These results provide further evidence for a role of female gender, radiation, goiter, Asian race, high educational attainment and family history of thyroid disease in the etiology of thyroid cancer. © 2001 Wiley-Liss, Inc.

Thyroid carcinoma is rare, representing 0.5% of cancers among U.S. men and 1.7% cancers among U.S. women, but remains the most common malignancy of the endocrine system.1 Its incidence increased from 1987 to 1991 in the United States,2 is greater in women than in men and peaks between the ages of 25 and 65.3, 4 The prognosis of well-differentiated, localized thyroid tumors is extremely good.1

Persons considered to be at high risk include those with a history of radiation treatment during infancy and childhood,5, 6 those with exposure to nuclear testing7 or to nuclear reactor accidents8 and those with benign conditions of the thyroid gland.9–13 Other putative determinants include diet (notably iodine deficiency or excess),14–20 and genetic factors.21, 22 Other less well-studied potential risk factors include smoking, alcohol intake,23 body size and weight gain.18, 24

In this report, we evaluated several potential predictors of thyroid cancer (including demographic, behavioral, medical and reproductive factors as well as occupational exposures) in a large, well-defined, multi-ethnic cohort of men and women enrolled in a health-maintenance organization (HMO) in northern California and followed for up to 33 years.

MATERIAL AND METHODS

Study design and procedures

We used a historical cohort study design with measurement of characteristics at baseline (1964–1973) and prospective follow-up of events from baseline through the end of 1997. The cohort was composed of 95,058 male and 112,838 female subscribers of the Kaiser Permanente Medical Care Program of Northern California, aged 10–89, who underwent multiphasic health checkups between 1964 and 1973 in San Francisco or Oakland. Of the original cohort, 28 women and 6 men were excluded for having a prior diagnosis of thyroid cancer before the multiphasic checkup. An additional 2,395 women and 503 men were excluded for having a self-reported history of thyroid operation (though the extent of surgical removal was unknown). Thus, the final sample of persons free of thyroid cancer and with an intact thyroid gland consisted of 94,549 men and 110,415 women.

Northern California Kaiser Permanente members constitute an ethnic and socio-economic mix that is generally representative of the local population, except that the extremes of wealth and poverty are under-represented.25 If study participants attended more than 1 checkup, only the data from the first were used.

The multiphasic checkup collected information on several demographic factors (age, gender, race/ethnicity and educational attainment), anthropometric measurements, health behaviors (cigarette smoking and alcohol consumption), physiological characteristics (including reproductive factors), medical conditions and occupational exposures using standardized instruments and procedures previously described.26, 27

Body mass index was estimated as weight (in kilograms) divided by height (in meters squared) and weight gain as measured weight at the multiphasic checkup minus self-reported least weight since age 20. Medical conditions of interest included personal history of hyperthyroidism, hypothyroidism, goiter and treatment of tonsils or the neck region with X-rays (though no information was collected on radiation dose or age at the time of treatment). The checkup also elicited information on maternal and/or paternal history of major diseases, including diseases of the thyroid gland, and women were queried about oral contraceptives (“pills to control your periods and/or pills to prevent pregnancy”) and hormone use in the past year. Although women were not specifically asked about estrogen use, an affirmative response to use of any hormones would almost always be indicative of estrogen use. The multiphasic questionnaire contained a section on reproductive factors that included parity, miscarriages, age at menarche, cessation of menstrual periods and hysterectomy (but did not inquire about oophorectomy status). The checkup also assessed 11 types of occupational exposure in the past year and before 1 year previously: (i) chemicals, cleaning fluids or solvents; (ii) insect or plant sprays; (iii) ammonia, chlorine, ozone or nitrous gases; (iv) engine exhaust fumes more than 2 hr/day; (v) plastics or resin fumes; (vi) lead or metal fumes; (vii) asbestos, cement or grain dusts; (viii) silica, sandblasting, grinding or rock dust; (ix) x-ray or radioactivity; (x) UV radiation and (xi) radar or microwave. Responders indicated whether their work environment involved daily or frequent exposure to different occupational hazards, but no information was given on duration of exposure.

Follow-up for incident thyroid cancer was accomplished through our Division of Research Cancer Incidence File, which started in 1969 among residents of the 5 San Francisco Bay Area counties. In 1973, the Cancer Incidence File became part of the local Surveillance, Epidemiology and End Results (SEER) program. In 1988, its coverage was extended to members living in outlying counties and to hospitalized patients in the Sacramento–Stockton area. Starting in 1990, the File encompassed all Kaiser Permanente subscribers in northern California.

Besides year of diagnosis, the File provides information on histology and disease stage according to the International Classification of Diseases for Oncology (ICD-O) morphological coding system. Vital status through the end of 1997 was ascertained using the California Automated Mortality Linkage System.28 Mortality data were available regardless of continuing health plan membership, and out-of-state mortality was estimated to be <2%.28

Person-time for each person ended either when he or she developed thyroid cancer, underwent total thyroidectomy (n = 58) or left the health plan for any reason including death or on 31 December 1997, whichever came first. Total thyroidectomy during follow-up was ascertained using ICD-8 procedure code 22.2 (between 1971 and 1978) and ICD-9 procedure codes 06.4, 06.6 and 06.52 (between 1979 and the end of 1997). The termination of health-plan membership was determined as failure to appear in the mid-year membership rosters for 2 consecutive years (even if the subscriber rejoined the health plan thereafter), with censoring date at the end of the year before the 2-year membership gap. Thirty-five percent of study participants were followed until the closing date (31 December 1997). Attrition due to termination of health-plan coverage and death was of the order of 2% per year, and the median follow-up time was 19.9 years (range <1–33 years).

Statistical analysis

We computed gender-specific thyroid cancer incidence rates per 100,000 person-years according to 10-year age intervals from ages 10 through 89. Cox proportional hazards models were used to estimate age- and gender-adjusted relative risks (RRs) and 95% confidence intervals (CIs) of thyroid cancer associated with each of the study variables.29 To assess independent associations, multivariate Cox proportional hazards models were constructed with simultaneous entry of all study covariables. To prevent loss of subjects, we modeled missing observations of each predictor with indicator (dichotomous) variables. The proportions of missing data for each variable are given in Table I.

Table I. Cohort Characteristics, by Gender. Northern California Kaiser Permanente Multiphasic Cohort, 1964–73
CharacteristicMean ± SD or %
Men (n = 94,549)Women (n = 110,415)
  • 1

    Height was missing in 8,207 men (8.6%) and in 9,819 women (8.7%); mean and SD were estimated among those with non-missing values.

  • 2

    Measured body weight at the checkup minus self-reported least weight since age 20.

  • 3

    Mean number of children and frequency distribution estimated assuming that missing responses were equal to no children; the mean (SD) number of children among women who had 1 or more children (n = 63,796) was 2.5 ± 1.5.

  • 4

    Mean number of miscarriages and frequency distribution estimated assuming that missing responses were equal to no miscarriages; the mean (SD) number of miscarriages among women who had one or more miscarriages (n = 23,701) was 1.6 ± 1.1.

Age (years)40.3 ± 14.038.9 ± 14.4
Race
 White77%75%
 Black14%16%
 Asian4%4%
 Other or missing5%5%
Level of education
 No college46%53%
 Partial college28%28%
 Completed college or post-graduate19%12%
 Missing8%7%
Cigarette smoking
 Never31%44%
 Former21%12%
 Current39%35%
 Missing9%9%
Alcohol consumption (drinks/day)
 016%25%
 1–256%53%
 3–511%4%
 6 or more4%1%
 Missing13%17%
Self-reported personal history
 Hyperthyroidism1%3%
 Hypothyroidism3%12%
 Goiter1%4%
 Treatment of tonsils or neck region with X-rays2%3%
 Uterus operation5%
Family history of thyroid disease1%3%
Body mass index (kg/m2)25.4 ± 3.523.9 ± 4.4
 <204%13%
 20–2540%50%
 >25–3039%19%
 >308%8%
 Missing9%9%
Height (m)11.75 ± 0.071.62 ± 0.07
Weight gain (kg)29.2 ± 7.98.8 ± 8.3
 ≤533%37%
 6–1029%22%
 >1027%23%
 Missing11%18%
Number of children31.4 ± 1.7
 043%
 1 or 234%
 ≥323%
Number of miscarriages40.3 ± 0.8
 079%
 1 or 218%
 ≥33%
Age at menarche (years)12.9 ± 1.6
 <1216%
 12–1463%
 ≥1512%
 Missing9%
Self-reported cessation of menstrual periods
 No64%
 Yes35%
 Missing1%
Oral contraceptives use
 No56%
 Yes27%
 Missing17%
Hormone use
 No83%
 Yes15%
 Missing2%
Occupational exposures
 Chemicals, cleaning fluids or solvents6%3%
 Insect or plant sprays2%2%
 Ammonia, chlorine, ozone or nitrous gases3%2%
 Engine exhaust fumes (>3 hr/day)7%1%
 Plastic or resin fumes2%1%
 Lead or metal fumes4%1%
 Asbestos, cement or grain dusts4%1%
 Silica, sandblasting, grinding or rock-drilling dust6%2%
 X-ray or radioactivity3%2%
 UV radiation1%1%
 Radar or microwave1%1%

To study potential differential associations by histological type, we also analyzed papillary and follicular carcinomas separately. All statistical analyses were run in SAS for UNIX, version 6.11 (SAS Institute, Cary, NC).

RESULTS

The mean age of the cohort was 40 years for men and 39 years for women (SD = 14 years) (Table I). Combined, about 76% of the cohort was white (including Hispanics), 15% black and 4% Asian; in 5%, race/ethnicity was classified as “other” or missing. Men had slightly higher educational attainment than women. Former or current cigarette smoking was more frequent in men (60%) than in women (47%). Likewise, consumption of 3 or more alcoholic drinks per day was more common in men (15%) than in women (5%). However, the prevalence of all medical conditions (notably hypothyroidism) and of family history of thyroid disease was higher in women than in men. About 2% of men and 3% of women reported receiving X-ray treatment of the neck region in the past. This proportion was greater among those born before 1945 (2.4% of men, 3.2% of women) than among those born in 1945 or later (1.8% of men, 1.9% of women).

The distributions of body mass index, weight gain, reproductive factors and occupational exposures are also shown in Table I. Roughly one-fourth of men and women gained 10 kg or more between the least weight since age 20 and the baseline weight. Fifty-seven percent of women had 1 child or more, and the mean number of children (among all women in the study) was 1.4 (SD = 1.7). Twenty-one percent of women reported 1 miscarriage or more. Mean age at menarche was 12.9 (SD = 1.6), and 35% of women reported cessation of menstrual periods. About 27% and 15% of women reported use of oral contraceptives and hormones in the past year, respectively. The most frequent occupational exposures were engine exhaust fumes in men and chemicals, cleaning fluids or solvents among women.

A total of 196 newly diagnosed thyroid cancers (73 among men, 123 among women) were identified. Overall, the predominant histological type was papillary (77%), followed by follicular (13%) (Table II). Medullary, anaplastic and other types were rare. The distribution of histological types varied by race: the proportion of papillary thyroid carcinomas was greatest among Asians (91%), whereas the proportion of follicular carcinomas was greatest among blacks (23%).

Table II. Distribution of Thyroid Cancer by Histological Type and Race: Kaiser Permanente Multiphasic Cohort, 1964–1973
Histological typeWhiteBlackAsianOther or missingAll
  1. The distribution of thyroid cancer by histological type and race was similar in men and women; thus, pooled results are presented.

Papillary114 (77%)12 (71%)21 (91%)4 (50%)151 (77%)
Follicular19 (13%)4 (23%)1 (4%)1 (12%)25 (13%)
Medullary1 (1%)1 (6%)1 (4%)1 (12%)4 (2%)
Anaplastic0 (0%)0 (0%)0 (0%)1 (12%)1 (1%)
Other14 (9%)0 (0%)0 (0%)1 (12%)15 (7%)
Total148 (75%)17 (9%)23 (12%)8 (4%)196 (100%)

Thyroid cancer incidence by age at diagnosis showed an inverted U-shaped pattern in women, with a peak in the 40–49 age group (Fig. 1). Conversely, incidence rates tended to increase with age in men. The female-to-male rate ratio was 2.5, 1.9 and 2.4 for ages 30-39, 40-49 and 50-59, respectively (Fig. 1).

Figure 1.

Age-specific thyroid-cancer incidence rates per 100,000 person-years through 1997 by age at diagnosis and gender. Kaiser Permanente Multiphasic Cohort, 1964–1973. Rates were estimated using an algorithm that allocated person-years to corresponding age groups as each cohort member aged during follow-up.

In the analysis with age as a covariate, the following baseline variables showed statistically significant increased risk of thyroid cancer: female gender, Asian race, complete college or post-graduate education, personal history of goiter, treatment of tonsils or the neck region with X-rays and family history of thyroid disease. An inverse association was found for black race (Table III). All of these relationships were largely maintained in the multivariate model. No significant associations were noted for age, smoking, alcohol consumption, body mass index, weight gain, height and occupational exposures.

Table III. RR and 95% CI for Thyroid-Cancer Incidence through 1997 Associated with Selected Baseline Variables: Kaiser Permanente Multiphasic Cohort, 1964–1973 (n = 204,964, 196 events)
Baseline variables (unit or referent category)Age- and sex-adjustedMultivariate-adjusted
Age (10–29 years)
 30–590.96 (0.70–1.32)0.93 (0.66–1.31)
 60–890.78 (0.42–1.46)0.87 (0.45–1.69)
Female gender (male gender)1.39 (1.04–1.86)1.56 (1.12–2.19)
Race (White)
 Black0.52 (0.31–0.86)0.55 (0.33–0.91)
 Asian2.35 (1.51–3.65)2.86 (1.76–4.65)
 Other or missing0.93 (0.45–1.89)1.09 (0.52–2.25)
Level of education (No college)
 Partial college1.37 (0.98–1.93)1.24 (0.88–1.76)
 Completed college or post-graduate2.02 (1.39–2.92)1.76 (1.20–2.59)
 Missing1.33 (0.73–2.44)2.68 (1.30–5.51)
Cigarette smoking (Never)
 Former1.16 (0.77–1.74)1.13 (0.75–1.70)
 Current0.97 (0.70–1.34)1.01 (0.71–1.42)
 Missing0.60 (0.33–1.08)0.56 (0.26–1.21)
Alcohol consumption (1–2 drinks/day)
 00.85 (0.59–1.22)0.90 (0.61–1.34)
 3–50.92 (0.51–1.68)0.98 (0.54–1.79)
 ≥60.83 (0.26–2.63)0.95 (0.30–3.02)
 Missing0.71 (0.47–1.10)0.98 (0.59–1.61)
Self-reported personal history (Negative history)
 Hyperthyroidism1.48 (0.60–3.60)0.87 (0.34–2.23)
 Hypothyroidism0.94 (0.56–1.58)0.68 (0.40–1.17)
 Goiter3.29 (1.86–5.82)3.36 (1.82–6.20)
 Treatment of tonsils or neck region with X-rays2.56 (1.43–4.59)2.33 (1.28–4.23)
Family history of thyroid disease2.55 (1.38–4.70)2.18 (1.17–4.05)
Body mass index (≤25 kg/m2)
 >25 kg/m20.88 (0.64–1.20)1.08 (0.74–1.56)
 Missing0.95 (0.55–1.62)1.53 (0.67–3.46)
Weight gain (≤5 kg)
 6–10 kg1.17 (0.82–1.66)1.20 (0.83–1.73)
 >10 kg0.85 (0.57–1.24)0.88 (0.57–1.38)
 Missing0.69 (0.44–1.09)0.56 (0.27–1.15)
Height (<1.72 m in men, <1.59 m in women)
 1.72–1.78 m in men, 1.59–1.65 m in women1.16 (0.80–1.68)1.31 (0.89–1.93)
 >1.78 m in men, >1.65 m in women1.22 (0.85–1.77)1.40 (0.94–2.09)
 Missing1.15 (0.64–2.05)1.16 (0.77–1.74)
Occupational exposures (Negative history of exposure)
 Chemicals, cleaning fluids or solvents1.16 (0.73–1.84)1.15 (0.72–1.83)
 Insect or plant sprays1.45 (0.76–2.77)1.48 (0.78–2.82)
 Ammonia, chlorine, ozone or nitrous gases1.28 (0.70–2.36)1.30 (0.71–2.41)
 Engine exhaust fumes (>3 hr/day)0.72 (0.36–1.44)0.79 (0.40–1.60)
 Plastic or resin fumes0.72 (0.28–1.89)0.68 (0.26–1.78)
 Lead or metal fumes0.88 (0.40–1.97)0.95 (0.43–2.12)
 Asbestos, cement or grain dusts1.57 (0.76–3.24)1.68 (0.81–3.48)
 Silica, sandblasting, grinding or rock-drilling dust1.04 (0.46–2.34)1.12 (0.49–2.52)
 X-ray or radioactivity1.47 (0.73–2.98)1.33 (0.66–2.69)
 UV radiation0.40 (0.05–3.08)0.34 (0.04–2.61)
 Radar or microwave1.90 (0.26–13.8)1.99 (0.27–14.5)

In the age-adjusted analysis among women, there were no statistically significant relationships of thyroid cancer risk with parity (RR of 1 or 2 live births vs. nulliparous women = 1.09, 95% CI = 0.71–1.69; RR of 3 live births or more vs. nulliparous women = 1.47, 95% CI = 0.94–2.30), miscarriages (RR of 1 or 2 vs. none = 0.58, 95% CI = 0.24–1.36; RR of 3 or more vs. none = 0.38, 95% CI = 0.07–1.64), age at menarche (RR of <12 vs. 13–14 years = 1.13, 95% CI = 0.71–1.80; RR of ≥15 vs. 13–14 years = 0.80, 95% CI = 0.42–1.51), cessation of menstrual periods among women with no history of surgery on the uterus (RR = 0.60, 95% CI = 0.35–1.02), use of oral contraceptives (RR = 1.07, 95% CI = 0.69–1.67) or hormone use (RR = 0.82, 95% CI = 0.47–1.43).

When the analysis was performed according to histological type, the same risk factors were found for papillary thyroid cancers as for all thyroid cancers combined, with the exception of no significant independent association with family history of thyroid disease (multivariate RR = 1.43, 95% CI = 0.63–3.27). In both the age- and sex-adjusted analyses and the multivariate-adjusted analysis, the risk of follicular thyroid carcinomas was significantly elevated only in the presence of history of goiter (multivariate RR = 7.03, 95% CI = 2.45–20.3) and family history of thyroid disease (multivariate RR = 8.63, 95% CI = 3.08–24.13).

DISCUSSION

This report of a large, representative cohort of San Francisco Bay Area residents enrolled in a large HMO confirms the known gender difference in thyroid cancer incidence (greater in women than in men) and the increased incidence of thyroid cancer in women during reproductive years. We were also able to confirm previously identified risk factors for thyroid cancer: Asian race, high educational attainment, radiation exposure, history of goiter and family history of thyroid disease.5, 6, 8–11 That persons of Asian ethnicity have elevated rates of thyroid cancer is a well-described phenomenon.14, 30, 31 A markedly elevated incidence of thyroid cancer has been reported among Filipino residents in Los Angeles County.32 Our results also suggest lower risk among blacks.

Our data show a greater proportion of radiation exposure in those born before 1945. In the 1930s and 1940s, it was common medical practice to administer radiation for benign conditions of the head and neck (such as enlarged thymus, acne or tonsilar hypertrophy).33, 34 Malignancies of the thyroid gland can develop as early as 5 years following radiation exposure but can appear 30 or more years later.35

Our findings are consistent with data from Switzerland showing that thyroid cancer tended to occur among more highly educated persons36 and with the finding in Los Angeles County that the incidence of thyroid cancer was elevated among persons with high socio-economic standing.34 In the analysis presented here, the association with high educational attainment was not explained by radiation exposure. Moreover, the setting of the present study (HMO with broad representation of persons from diverse socio-economic levels and equal access to care) makes it less likely that the relation of thyroid cancer with high level of education is due to detection bias as a result of inequalities in diagnostic practices.

We could not confirm several prior studies reporting a positive association between alcohol consumption and thyroid cancer23, 37 and an inverse association with cigarette smoking.23

The role of hormonal and reproductive factors in the etiology of thyroid cancer remains unclear. In agreement with our results, a study among 63,090 Norwegian women revealed no strong associations with reproductive factors.38 The lack of association with reproductive factors is also consistent with a report among Los Angeles County women39 and with pooled analyses of case-control studies of thyroid cancer.40, 41 However, some modest associations have also been reported, including bilateral oophorectomy,39 early first childbirth (before age 20 or <5 years after menarche), artificial menopause42 and miscarriage as outcome of first pregnancy.19, 43, 44

Our results also indicate that family history of thyroid disease may be a predictor of the incidence of non-papillary thyroid cancers. Several reports have demonstrated familial aggregation of medullary thyroid carcinoma, either in isolation or as part of a multiple endocrine neoplastic syndrome.45, 46

Anthropometric measures have not been included in the majority of thyroid-cancer studies. Dal Maso et al.24 reported that both height and weight were moderately related to thyroid cancer risk. In our study, no associations were found for body weight, height or weight gain since age 20.

Our data showed no significant associations with 11 different occupational exposures. Relatively few studies have investigated the role of occupational or specific chemical hazards in relation to thyroid cancer. In a death certificate study in 24 U.S. states, women with probable silica exposure (both occupational and from an industry associated with silica) had elevated thyroid cancer mortality.47 In a Swedish case-control study, increased risks were seen for women who had worked as a dentist/dental assistant, teacher, shoemaker or warehouse worker and for occupational contacts with undefined chemicals, X-rays or video display terminals.48 Work with diagnostic X-rays and exposure to electromagnetic fields constituted risk factors for thyroid cancer of the papillary type in another Swedish case-control study.49

Some limitations should be noted in the current investigation. First, our automated database did not allow identification of different ethnic subgroups within the Asian ethnic group. Thus, we could not confirm the finding of higher thyroid cancer rates among Filipinos.14, 35 Second, no information was available on benign thyroid nodules or adenomas (a known risk factor for thyroid cancer), diet or genetic markers. Third, changes in exposures after baseline were not considered. Fourth, since the Kaiser Cancer Incidence File expanded its geographic coverage over time, it is likely that some early thyroid-cancer cases (i.e., pre-1988) that occurred in the outlying counties were missed. However, this is unlikely to have biased the results since supplemental separate analysis of early (i.e., before 1989, 140 cases) and late (in 1989 or later, 56 cases) thyroid-cancer cases showed similar results, with the exception of a weaker and non-statistically significant association with history of X-ray treatment (RR = 1.35, 95% CI = 0.32–5.64) in the analysis of later cases. Finally, no information was available on the specific types of menopausal replacement hormone used or on use before 1 year prior to baseline.

In conclusion, these data from a large HMO cohort confirm previously known risk factors for thyroid cancer: female gender, high education, family history of thyroid disease, radiation, benign thyroid disease and Asian race. Our findings also suggest lower thyroid-cancer risk among blacks and no link to occupational exposures. The null findings for hormonal and reproductive factors are consistent with the results of case-control studies. However, further investigation into the elevated rates of thyroid cancer among women is necessary.

Ancillary