Health Reports
Sociodemographic characteristics associated with thyroid cancer risk in Canada

View the most recent version.

Archived Content

Information identified as archived is provided for reference, research or recordkeeping purposes. It is not subject to the Government of Canada Web Standards and has not been altered or updated since it was archived. Please "contact us" to request a format other than those available.

Archived

This page has been archived on the Web.

by Tracey Bushnik and William K. Evans

Thyroid cancer incidence in Canada has increased rapidly over the past 25 years.^{Note 1} The age-standardized incidence rate for women has increased five-fold to a projected 29.1 per 100,000 in 2017. At the same time, there has been a four-fold projected increase for men to 8.8 per 100,000.^{Note 1} Although the increased risk in women has been widely reported,^{Note 2Note 3Note 4Note 5} studies in the United States and England have also found variations in risk according to sociodemographic characteristics such as ethnicity, education and income.^{Note 6Note 7Note 8Note 9Note 10} In Canada, the examination of these associations has been limited to provincial-level data only or area-based measures,^{Note 11Note 12Note 13} as little information on these characteristics has been available at the national level.

The Canadian Census Health and Environment Cohorts (CanCHECs) now make it possible to examine Canada-wide cancer incidence according to a variety of individual-level sociodemographic characteristics. Using the 1991 and 2001 CanCHECs, the present study examines thyroid cancer incidence over nine years of follow up; presents estimates of the sex-specific relative risk of thyroid cancer according to age, immigrant status, ethnicity, educational attainment and family income; and examines whether these relative risks changed over time. The sex-specific relative risk of thyroid cancer by histology (papillary versus non-papillary) according to the same characteristics is also described.

Methods

Data source

1991 and 2001 CanCHECs

The 1991 and 2001 CanCHECs are datasets formed through the linkage of census long-form questionnaire responses (completed in 1991 and 2001 by about 20% of Canadian households) to the Canadian Mortality Database (CMDB), the Canadian Cancer Registry (CCR), the Canadian Cancer Database (CCDB), and tax files. Standard deterministic and probabilistic linkage techniques were used, and details about these datasets have been published elsewhere.^{Note 14Note 15}

The 1991 CanCHEC had 2,644,400 cohort members (count rounded to nearest 100) who consented to having their data linked. Cohort members were aged 25 or older on Census Day (June 4, 1991) and were not residents of institutions. Follow-up for death and for cancer was conducted until December 31, 2011, and December 31, 2010, respectively. The 2001 CanCHEC had 3,537,500 cohort members. They were aged 19 or older on Census Day (May 15, 2001) and were not residents of institutions. Follow-up for death and cancer was conducted until December 31, 2011, and December 31, 2013, respectively. Both cohorts are considered to be representative of the Canadian population, but did contain high percentages of people who were married or in a common-law union, had higher levels of education and income, and were labour force participants. This is mainly because the linkage methodology relied on respondents being tax filers. All counts based on the CanCHECs were rounded to the nearest 100.

Thyroid cancer

The CanCHECs included CCR and CCDB data that were linked to cohort members. The CCR is a person-oriented, population-based registry established in 1992. It includes data collected by each provincial/territorial cancer registry and reported to Statistics Canada.^{Note 16} The CCDB is a historical tumor-oriented database containing cancer cases diagnosed from 1969 to 1991.^{Note 17} For this study, thyroid cancer (TC) was identified in the CCR with variables for site (ICD-O-2/3 code C739), behaviour (ICD-O-3 code 3) and histology (ICD-O-3 codes 8050, 8260, 8340, 8341, 8343, 8344 or 8350 for papillary; 8290, 8330, 8331, 8332 or 8335 for follicular; 8345, 8346 or 8510 for medullary; 8012, 8020, 8021, 8030, 8031 or 8032 for anaplastic; and all remaining codes for “other,” excluding 9050 to 9055, 9140, and 9590 to 9992). In the CCDB, TC was identified with the ICD9 code 193. Since Quebec cancer data are not available after 2010, cancer follow-up for each cohort was conducted for nine years. A 10-year washout period was used to identify and exclude individuals who were diagnosed with TC prior to cohort inception.

Inclusion criteria for the present analysis included individuals aged 25 to 89 on Census Day without a TC diagnosis during the washout period. From the 1991 CanCHEC, 5,500 cohort members were excluded because the baseline age was under 25 or 90 or older. A further 1,000 were excluded because of a TC diagnosis during the washout period, leaving 2,637,900 people in the 1991 cohort. The cancer follow-up period extended from January 1, 1992, to December 31, 2001. The same exclusion criteria were applied to the 2001 CanCHEC (325,900 excluded due to age; 2,400 due to washout), resulting in 3,209,200 people in the 2001 cohort. The cancer follow-up period for this cohort extended from January 1, 2002, to December 31, 2010.

Covariates

All covariates were based on census data at baseline. Age group on Census Day was categorized as 25 to 34, 35 to 44, 45 to 54, 55 to 64, 65 to 74, and 75 to 89. Marital status was categorized as married or common-law; separated, widowed or divorced; or single. Immigrant status was defined as non-immigrant (including non-permanent resident), immigrant with 10 years or less in Canada, or immigrant with more than 10 years in Canada. Immigrant status was then dichotomized as immigrant or non-immigrant for the hazard model analyses. Ethnicity was determined from the question on visible minority status and categorized as Caucasian, Black, East Asian (Chinese, Korean, Japanese), Southeast Asian (e.g., Filipino, Cambodian, Indonesian, Vietnamese), South Asian (e.g., East Indian, Pakistani, Sri Lankan), or other (includes all respondents not included in the other categories). Highest level of education was categorized as less than secondary graduation, secondary graduation, postsecondary diploma or certificate, or university degree. Economic family income adequacy quintiles were estimated from the ratio of a family’s 12-month total pre-tax post-transfer income to the Statistics Canada low-income cut-off (pre-tax post-transfer for the year prior to the census collection year) for the applicable economic family and community size group. These ratios were ranked, and quintiles were constructed within each census metropolitan area, census agglomeration, or rural and small town area (outside census metropolitan areas or census agglomerations) to account for regional differences in housing costs. Highest level of education and income quintiles were dichotomized (secondary graduation or less versus postsecondary diploma or higher; first to fourth quintile versus fifth (highest) quintile, respectively) for the histology-specific analysis.

Statistical analysis

Descriptive statistics were used to describe the characteristics of the cohorts and their age at TC diagnosis. Age-standardized incidence rates of TC per 100,000 (standardized to the 1991 population aged 25 to 89^{Note 18}) were estimated by sex, across covariates and by histology type. Standard Cox proportional hazard models with death as the censoring event were used to examine associations between covariates and incidence of TC.^{Note 19} Unadjusted and adjusted hazard ratios (HR) and their 95% confidence intervals (95% CIs) for TC were estimated for each cohort by sex. The cohorts were pooled to examine the association between the covariates and the risk of papillary thyroid cancer (PTC) versus non-papillary thyroid cancer (NPTC) by sex. A cohort indicator variable was included in all pooled models. Interactions were tested between the cohort indicator variable and each covariate, and confirmed that the direction of the association of each covariate with the outcome did not vary significantly by cohort. Cochran’s Q was used to test the homogeneity of the estimated HRs across ethnicity categories.^{Note 20} Because of the heterogeneity of those included in the “other” ethnicity category, TC incidence rates and hazard ratios were not reported.

Results

Compared with the 1991 cohort, the 2001 cohort members were older and more likely to be immigrants to Canada, to be of an ethnicity other than Caucasian, and to have a university degree (Table 1). In both cohorts, women were more likely than men to be aged 75 to 89; to be separated, widowed or divorced; or to be in the lowest income quintile. Women were less likely than men to be Caucasian or have a university degree.

Thyroid cancer incidence

In the 1991 and 2001 cohorts, there were about 1,700 and 4,800 reported cases, respectively, of TC during nine years of follow-up. The average age at diagnosis for men was the same in both cohorts (56 years), whereas the average age for women was higher in the 2001 cohort—52 years, versus 49 years in 1991. For both sexes and in both cohorts, age-standardized TC incidence rates observed among immigrants, East Asians and Southeast Asians (except for men in the 1991 cohort) were noticeably higher than average (Table 2). There was a gradient of higher TC incidence rates for women with higher levels of education in the 1991 cohort, and for both sexes in the 2001 cohort. In general, TC incidence rates increased between 1991 and 2001 across all characteristics examined: rates were almost double for men in the 2001 cohort compared to the 1991 cohort (8.4 versus 4.4 per 100,000), and more than double for women (24.9 versus 10.8 per 100,000).

In both cohorts, women were at greater risk of TC than men (1991 cohort unadjusted HR = 2.59; 95% CI: 2.33 to 2.88; 2001 cohort unadjusted HR = 2.90; 95% CI: 2.71 to 3.10). Table 3 presents adjusted HRs by sex. Compared with Caucasians, East and Southeast Asian women were at greater risk of TC in both cohorts (p ≤ 0.05 for heterogeneity), while East Asian men were at marginally greater risk of TC in the 1991 cohort only. Immigrant status was associated with increased risk for women in both cohorts, and for men in the 2001 cohort. Women with a university degree in both cohorts, and men with a postsecondary diploma or higher in the 2001 cohort were at increased risk of TC compared with those with less than a secondary school education.

Histology type

Papillary thyroid cancer accounted for 71% and 78% of thyroid cancer cases for men and women, respectively, in the 1991 cohort. Its share increased to 80% and 85% for men and women, respectively, in the 2001 cohort.

The age-standardized incidence rate for PTC was more than double for men in the 2001 cohort compared with the 1991 cohort (6.7 per 100,000; 95% CI: 6.2 to 7.1 versus 3.0 per 100,000; 95% CI: 2.6 to 3.3), and close to triple for women (21.4 per 100,000; 95% CI: 20.6 to 22.2 versus 8.3 per 100,000; 95% CI: 7.8 to 8.9) (Figure 1). For men, there was a marginal increase in the age-standardized incidence rate of the “other” histology type between the two cohorts. For women, there was a slight increase in the rate of follicular, medullary and “other” thyroid cancers.

Compared with men aged 25 to 34, men aged 45 to 74 were at increased risk of PTC, and men at older ages were at even greater risk for NPTC (Table 4). Although women of older ages were also at greater risk of NPTC, the risk of PTC among women declined with age.

For both men and women, there was significant variation in risk of PTC across ethnicity categories (p ≤ 0.02 for heterogeneity) that was not present for NPTC. East Asian men (HR = 1.39; 95% CI: 1.08 to 1.80), East Asian women (HR = 1.38; 95% CI: 1.20 to 1.58) and Southeast Asian women (HR = 1.59; 95% CI: 1.33 to 1.90) were at increased risk of PTC compared with Caucasian men and women. Men and women who were immigrants, had a postsecondary diploma or higher, and were in the highest income quintile were at increased risk of PTC. Immigrant women (HR = 1.51; 95% CI: 1.27 to 1.80) and Southeast Asian women (HR = 1.65; 95% CI: 1.07 to 2.56) were also at increased risk of NPTC.

Discussion

The present study found that thyroid cancer risk varied across several sociodemographic characteristics in two population-based cohorts, 10 years apart. Regardless of the cohort, women were at greater risk of TC than men. Moreover, women who were immigrants, who were of East or Southeast Asian descent, or who had a university degree were all at increased risk of TC. In the 2001 cohort, men were at greater risk of TC if they were an immigrant or had a postsecondary diploma or higher. When TC was considered by histology, immigrant status, ethnicity, education level and income were each associated with the risk of papillary thyroid cancer for both sexes while—for women only—immigrant status and ethnicity were associated with the risk of non-papillary thyroid cancer.

Women had higher overall incidence rates of TC than men, regardless of marital or immigrant status, ethnicity, education or income, and this gap between men and women widened over time. Although this gender disparity has been widely reported,^{Note 2Note 3Note 4Note 5} it is not completely understood. Potential factors include increased detection through much higher rates of diagnostic imaging tests among women in general compared with men,^{Note 21} and hormonal or reproductive factors.^{Note 22Note 23Note 24} However, hormonal or reproductive factors cannot readily explain the shift from no difference in to an inverted “U” in risk among women in the 2001 cohort, where women aged 35 to 54 were at higher risk of TC compared with women under the age of 35, or 65 or older. It has been suggested instead that the introduction of ultrasonography in gynecologic and obstetric settings has resulted in more frequent examinations of the thyroid gland in women of reproductive age.^{Note 25} This, coupled with increasing rates of diagnostic imaging tests among women as a whole,^{Note 21} could help explain the increased risk of TC, and particularly of PTC, in women during the reproductive years.^{Note 26Note 27Note 28}

Compared with non-immigrants, immigrant women in both cohorts, and immigrant men in the 2001 cohort, were at increased risk of TC, regardless of age, ethnicity, education or income. Shah et al. also found increased risk of TC among Asian immigrants to Ontario, results that were independent of primary care visits and intensity of diagnostic radiology tests.^{Note 13} Moreover, they reported that the immigrants in their study generally had fewer contacts with the health care system, which suggests that increased detection did not explain the increase in risk for this population group. Although the present study did not have any information regarding health service use, it did control for education and income, which some consider to be markers of access to and use of health care services.^{Note 9} It follows that the increased risk among immigrants in the present study, like the findings of Shah et al., cannot be attributed solely to differences in detection. Rather, differences in diet, behaviour, comorbidities, prior environmental exposures including radiation, or pre-existing thyroid disease may be contributing factors.^{Note 29Note 30Note 31Note 32Note 33Note 34}

East and Southeast Asian women in both cohorts were at increased risk of TC compared with their Caucasian counterparts, regardless of age, immigrant status, education or income. These results are consistent with studies that have examined Asian ethnic groups in England and the United States.^{Note 8Note 30} An examination of the results by histology type showed that the increased risk among East Asian women in the present study was attributable to PTC, whereas the increased risk among Southeast Asian women was attributable to both PTC and NPTC. Interestingly, there was also an increased risk among Southeast Asian men for NPTC. That these findings persisted, despite the study having controlled for immigrant status, suggests that factors unrelated to country of origin may play a role. Differential exposure to factors such as familial thyroid disease, or iodine deficiency or excess, might contribute to the increased risk for these population groups.^{Note 30Note 35}

Like others who have found an association between higher socioeconomic status (SES) and TC incidence,^{Note 6Note 9Note 10Note 11} the present study found that, regardless of other characteristics, men and women who were in the highest income quintile or who had a postsecondary diploma or higher were at greater risk of being diagnosed with PTC, but not NPTC. A possible explanation is the increased use of medical diagnostic imaging among people with higher SES. Hall et al. found that areas in Ontario with a greater proportion of highly educated people had higher rates of diagnostic ultrasounds as well as higher rates of thyroid cancer.^{Note 12}

Lastly, the observation of a significant increase in TC incidence, and particularly the PTC histology between the two cohorts across all socioeconomic characteristics, is consistent with worldwide trends.^{Note 2Note 3Note 4Note 36Note 37} This speaks not only to the changes in diagnostic practices that have resulted in increased detection of these tumours,^{Note 2Note 25Note 31Note 38Note 39} but also to the greater relative risk among certain groups (e.g., immigrants) in the present study who represent an increasing proportion of the Canadian population. That this variation in relative risk persisted across time suggests that more than increased detection is responsible for the observed differences between groups. The previously mentioned risk factors of diet, environmental exposures, comorbidities, and heredity,^{Note 29Note 30Note 31Note 32Note 33Note 34Note 35} and other factors such as proximity to health care services and physician characteristics, all likely play a role.

Strengths and limitations

The present study has a number of strengths. It is a population-based analysis of two large cohorts, each with nine years of follow-up. The datasets allowed for the examination of several characteristics beyond age and sex, including individual-level measures of socioeconomic status. Unlike other studies of race or ethnicity, the present study was able to examine both ethnicity and immigrant status in an effort to disentangle their individual associations with TC risk. By pooling the two cohorts, it was possible to undertake a more detailed examination of risk factors related to PTC versus NPTC.

However, this study also has several limitations. The number of thyroid cancer cases was relatively small, particularly in the 1991 cohort, which prevented a more detailed analysis by histology type. Analysis of less-aggregated ethnicity categories was not possible because of changes over time in the census’ aggregate classification of visible minority status. Furthermore, TC incidence and hazard ratios were not estimated for the “other” ethnicity category in this analysis because it included a wide range of ethnicities, likely with differing levels of risk of thyroid cancer. Pacific Islanders, for example, who have been grouped with Southeast Asians in previous research,^{Note 30} were included in the “other” category, as were Latin Americans, a group that includes people from many countries who have been shown to have a range of risk profiles in previous research.^{Note 3} The cohorts did not contain information on tumour size or staging for the period under analysis, and information was not available on diet, behaviour, comorbidity or family history of TC. The 10-year washout period may have been insufficient to prevent the capture of some recurrent rather than new thyroid cancer cases in the follow-up period.

Conclusion

Thyroid cancer incidence has increased significantly in Canada. While increased detection, particularly of papillary thyroid cancer, has certainly played a role, it does not fully account for the higher relative risk of thyroid cancer among the immigrant population and certain ethnic groups. Knowing that certain groups are at greater risk can help target awareness and treatment programs, but more research is needed to better understand the determinants of the increased risk in these populations.