Site-Specific Cancer Incidence by Race and Immigration Status in Canada 2006–2015: A Population-Based Data Linkage Study

The collection of race-based and Indigenous identity data is essential to measure and address health inequalities. Most evidence regarding racial inequalities in cancer health outcomes comes from the United States, where important disparities by race in cancer outcomes have been documented (1, 2). However, findings from the United States may not be transposable to other health systems with different populations and contextual factors influencing health. The examination of race-based outcomes in other countries can help to identify commonalities and potential differences in drivers of cancer inequalities.

Canada is a country with a public healthcare system and universal health coverage. It is ethnically and racially diverse due to Canada's long history of immigration, with 23% of the population being immigrants in the 2021 census (3). Unlike in the United States, where yearly statistics on cancer incidence by race and ethnicity are published and available in cancer registries (2), the Canadian Cancer Registry (CCR) does not collect demographic data beyond age and sex, making it difficult to monitor inequalities in cancer incidence and mortality. There is a legacy of colonialism and a historical reluctance to gather race-based data in Canada, which have led to a current paucity of data for monitoring racial health inequalities (4, 5). There have been recent calls to provide more racially disaggregated health data (1, 4). Previous studies which have examined racial and ethnic differences in cancer outcomes in Canada have tended to be based in a single province or examined only the most incident cancer types (6–12). Recent national-level estimates of site-specific cancer incidence by race are lacking.

Due to the low incidence rate of some cancers, large data-linkage studies are required to monitor racial inequalities for many cancers in Canada. Our objective was to measure differences in site-specific cancer incidence across racial groups in a representative Canadian population to identify inequalities in cancer incidence. We also assessed whether socioeconomic status and immigration contribute to differences in cancer incidence by race.

The study sample includes members of the 2006 and 2011 Canadian Census Health and Environment Cohorts (CanCHEC), a set of population-based probabilistically linked datasets based on the long-form census and the National Household Survey (NHS) (13). The 2006 long-form census questionnaire was a mandatory questionnaire sent to approximately one in five Canadian households collecting information on relationships, languages, labor, income, education, housing, immigration, race, and ethnic origin (14). The long-form census was replaced with the NHS in 2011, a voluntary survey collecting the same information as the long-form census sent to approximately one in three Canadian households, with a response rate of 68.6% (15). Both surveys targeted the noninstitutionalized usual residents of Canada on Census Day (mid-May 2006 or 2011), including both permanent and nonpermanent residents. Interviewer-led data collection methods were used to oversample individuals living in remote areas and on Indian Reserves. Survey respondents were probabilistically linked using personal identifiers to the Derived Record Depository (DRD), a national dynamic relational database of individuals residing in Canada (16). The 2006 and 2011 CanCHEC cohorts include the 90.8% of the 2006 long-form census respondents and 96.7% of 2011 NHS respondents who could be linked to the DRD (13).

Cohort records were linked via the DRD up to December 31, 2015 to the Canadian Vital Statistics Death Database (CVSD) (17), the CCR (18), and the annual postal codes file (19). The CVSD collects dates and cause of death information annually from all provincial and territorial vital statistics registries on all deaths in Canada. The CCR is a population-based registry that includes information about each new primary cancer diagnosed in Canada. The International Agency for Research on Cancer rules were used for determining multiple primary cancers in the CCR (20). Cancers were grouped by site using Surveillance, Epidemiology, and End Results site groupings based on ICD-O-3 codes (Supplementary Table S1). For head and neck cancers, we used the definition used by the 2021 Canadian Cancer Statistics (21).

The annual postal code file collects the mailing addresses provided on income tax returns by tax filers. Postal codes were used to estimate changes in a person's place of residence for each year of follow-up. For individuals in the CCR or CVSD, we used the postal codes available in these registries instead of the annual postal code file for the years in which they were diagnosed with cancer or died. The geographical information related to each postal code was extracted using the Postal Code Conversion File Plus program version 7D (22). For years with missing postal code data due to missing tax records, we imputed the last known place of residence from the CCR, annual postal code file, or long-form census/NHS (in order of priority) forward to subsequent years.

We based our categorization of race on Canadian Institute of Health Information standards for race-based and Indigenous identity data health reporting (23). These standards distinguish between race and ethnicity, with race defined as a social construct used to categorize people based on perceived differences in physical appearance, and ethnicity defined as a multidimensional construct based on community belonging and a shared cultural group membership. For historical and legal reasons, race data is often collected by Canadian institutions using three constructs: Indigenous peoples, visible minority groups, and those who are neither. Indigenous peoples refer to three distinct population groups—First Nations people, Métis and Inuit—who are recognized in the Constitution Act. Visible minority groups are defined by the Employment Equity Act as persons, other than Indigenous peoples, who are non-Caucasian in race or non-white in color. Because much of the race data collected in Canada uses these definitions, in this paper we use the term race when referring to all racial groups and the term visible minority groups when referring specifically to racial groups who are not White and not Indigenous.

Racial identity and immigrant status were self-reported based on the Aboriginal group, visible minority group, immigrant status, and citizenship questions in the long-form census (2006) and the NHS (2011; refs. 24, 25). The visible minority group question was directed to respondents who did not identify as Indigenous, and included 11 response categories (White, South Asian, Chinese, Black, Filipino, Latin American, Arab, Southeast Asian, West Asian, Korean, and Japanese) and a write-in space. Respondents who reported belonging only to a single group were included in that corresponding racial group category. Respondents who selected multiple groups were classified in released census data using rules used by employment equity programs (24). Respondents who reported being both 'White' and 'Chinese', 'South Asian', 'Black', 'Filipino', 'Southeast Asian', 'Japanese', or 'Korean' were included in their corresponding visible minority group category. Respondents who reported being both 'White' and 'Latin American', 'Arab', or 'West Asian' were included in the 'White' category, as these are not considered to be visible minority groups by employment equity definitions and are grouped with 'White'. Respondents who reported belonging to multiple non-White visible minority groups were categorized as 'Multiple visible minority groups'. Those who provided a write-in response not classifiable into any other group were categorized as 'Other visible minority groups'. Canadian-born individuals (nonimmigrants) are those who indicated they were born in Canada; immigrants are those born outside of Canada who have been granted the right to live in Canada permanently by immigration authorities; and nonpermanent residents are those who are neither born in Canada nor immigrants (e.g., person with a temporary work/study permit).

The 2006 and 2011 CanCHECs were pooled together to increase the sample size and statistical power for the analysis. Respondents contributed person-time at risk and events from the year of their survey response (2006 or 2011) up to their year of death or December 31, 2015. Because the province of Québec has not reported new cancer cases to the CCR since 2010, we excluded person-time from Québec residents after 2010. To calculate person-time and events by age, we used each person's age at the midpoint of each calendar year (July 1) of follow-up.

We calculated age-standardized cancer incidence rates and age-standardized incidence rate ratios (ASIRR) by population group using the 2011 Canadian population age distribution as standard (26). Age-specific incidence rates were weighted by CanCHECs survey weights in order to be representative of the Canadian population living in private dwellings. The CanCHECs survey weights are calibrated to account for survey sampling probability, nonresponse rates, and DRD linkage probabilities (13). The 95% confidence intervals (CI) for age-standardized cancer incidence rates and ASIRR were calculated through bootstrapping, using the 2.5th and 97.5th percentile intervals of 500 bootstrap replicate weights developed for the CanCHECs. To validate sample representativeness, we compared cancer incidence rates in the CanCHECs with national cancer incidence rates reported during the same period; the results were highly similar (Supplementary Tables S2–S3). In secondary analyses, we also age-standardized using the 1960 World Population for comparisons with cancer incidence rates in the world region of origin of visible minority groups using GLOBOCAN 2020 data (RRID: SCR_012750; refs. 27, 28).

To assess whether incidence rate ratios between racial groups could be attributed to a healthy immigrant effect or differences in socioeconomic status, we used negative binomial regression models. Regression models used the logarithm of group person-years at risk as an offset. We included an interaction term between race and immigration status to calculate different rates for Canadian-born individuals compared with foreign-born immigrants and nonpermanent residents. The largest population group (White Canadian-born) was used as the reference level in regression models. Incidence rates were adjusted for 5-year age group, sex, calendar year, province of residence, rural/urban place of residence, education level, and quintiles of household after-tax income adjusted for household size. Due to the small numbers of events in some racial groups when assessing interactions, some populations were grouped together in regression analyses to improve model stability while maintaining meaningful distinctions: Chinese, Filipino, Southeast Asian, Korean, and Japanese were grouped as East/Southeast Asian; West Asian and Arab were grouped as Middle Eastern; and those in multiple visible minority groups were combined with the other visible minority groups category.

We obtained ethical approval from the McGill University Institutional Review Board for this analysis of secondary data, conducted in accordance with the Tri-Council Policy Statement: Ethical Conduct for Research Involving Humans (29). To protect respondent confidentiality, analysis outputs were vetted using rules developed by Statistics Canada, which include rounding of all counts to the nearest 5 and not disclosing descriptive statistics for groups with less than 5 events. In analyses where the number of cancer cases in a racial group did not meet the disclosure threshold, we either combined this group with the ‘Other visible minority group’ or suppressed outputs as indicated.

Restrictions apply to the availability of these data, which were used under license by Statistics Canada for this study. Eligible researchers can apply for data access through the Statistics Canada Research Data Centre program. GLOBOCAN data is publicly available online (28). Code used for the current analyses is available at the Borealis repository (30).

There were 5.9 million respondents in the 2006 CanCHEC and 6.5 million respondents in the 2011 CanCHEC (Table 1). The most common racial groups in both cohorts were White (75%–78%), Indigenous peoples (7%–8%), South Asian (4%–5%), and Chinese (3%–4%) populations. Immigrants constituted 18% to 20% of respondents and nonpermanent residents 1% of respondents. The proportion of immigrants and nonpermanent residents was substantially higher among visible minority groups than among White individuals and Indigenous peoples. The median time since immigration for immigrants was 18 years (interquartile range 8–36).

Age-standardized cancer incidence rates by race, sex, and immigration status for all cancer sites combined are presented in Table 2; Fig. 1. Age-standardized cancer incidence rates were lower in nearly all non-White racial groups than in the overall population, while White individuals and Indigenous peoples had a higher overall age-standardized cancer incidence rate (ASIRRs, 1.03–1.04). Overall, women had a lower age-standardized overall cancer incidence rate than men (ASIRR, 0.83; 95% CI, 0.82–0.83), but the female to male cancer ASIRR displayed substantial variation by racial group. Immigrants had substantially lower age-standardized overall cancer incidence rates than Canadian-born individuals (ASIRR, 0.83; 95% CI, 0.82–0.84). The ASIRR between immigrant and Canadian-born individuals also displayed substantial variation by racial group.

Site-specific age-standardized incidence rates and ASIRRs are presented in Tables 3 and 4, respectively. White individuals had higher age-standardized incidence rates than the overall population across most sites, with the notable exceptions of stomach, liver, and thyroid cancers, and multiple myelomas. With these cancers excepted, visible minority groups generally had lower or equivalent age-standardized cancer incidence rates compared with the overall population. Indigenous peoples had higher incidence rates for certain cancer sites (head and neck, stomach, colorectal, liver, pancreas, lung and bronchus, cervical, kidney and renal pelvis) but lower incidence rates for other cancer sites (melanoma, uterus, prostate, testis, bladder, brain and central nervous system, thyroid, Hodgkin Lymphoma, and leukemia) compared with the overall population.

Differences in most cancer incidence rates between racial groups persisted even after adjustment for age, sex, year, province, place of residence, education level, and household income in regression models (Table 5). Stratification by immigration status further revealed relationships with immigration which differed by site. For most cancer sites, the cancer incidence rates showed a trend of convergence, where Canadian-born members of racial groups had closer incidence rates to the Canadian-born White individuals than foreign-born members of racial groups. This was not the case for all cancer sites, however: stomach, cervical, thyroid cancer, and multiple myeloma incidence rates were in most cases equivalent between foreign-born and Canadian-born racial groups.

Age-standardized incidence rates by visible minority groups compared with GLOBOCAN 2020 incidence estimates by world regions of origin are provided in Supplementary Table S4 and Supplementary Fig. S1. For most cancer sites, incidence rates tended to be intermediate between the rates in White Canadians and the rates from their world region of origin. However, for head and neck, lung and bronchus, esophageal, pancreatic, and cervical cancers, the incidence rates in visible minority groups were markedly lower than in both White Canadians and their world region of origin. For thyroid cancer, the incidence rates in visible minority groups were strikingly higher than in both White Canadians and their world region of origin.

This study estimated cancer incidence rates by race using a very large representative population-based sample of the Canadian population linked to a cancer registry with national coverage. Because the CCR does not collect data on race, our results are to our knowledge the most nationally representative up-to-date estimates of cancer incidence rates stratified by race in Canada. We found that for most cancer sites, non-White and non-Indigenous racial groups had lower age-standardized cancer incidence rates than the overall population. Indigenous peoples had higher age-standardized incidence rates for some cancer sites but lower for others compared with the overall population. Differences in cancer rates between racial groups remained even after adjusting for differences in socioeconomic status. However, immigration status was an important modifier of cancer risk, suggesting that a healthy immigrant effect influences cancer risk by race in Canada.

Inequalities in cancer incidence are influenced by the social determinants of health (5, 31). Individuals who are more resourced are in general better able to adopt healthy lifestyles and avail themselves of preventive healthcare (32). Though Canada has a public healthcare system with universal health coverage available for all citizens and permanent residents, there is a socioeconomic gradient in cancer risk across multiple sites (33–36). Nevertheless, our analysis suggests that differences in household income, education, and urbanicity are not the fundamental causes of the observed differences in cancer incidence observed between racial groups, as incidence differences persisted despite adjustment for these variables. Systemic racism and enduring colonial influences on Indigenous peoples’ health contribute to health inequalities between racial groups (37). The observed patterns suggest that differences between racial groups may also be linked to other social determinants of health affecting lifestyles, culture, early life exposures, and selection factors for immigration.

Immigration was an important modifier of cancer risk. Most immigrants to Canada are economic immigrants selected based on their ability to contribute to labor market needs (38). Immigrants applying for permanent residency must undergo a medical examination. Applicants may be found inadmissible for residency if they present medical conditions deemed to potentially cause excessive demands on health and social services. There is therefore a strong selection for immigrants in better health than Canadian-born individuals, known as the healthy immigrant effect (39, 40).

We hypothesize that cancer rates would be intermediate between the world region of origin and those in Canadian-born individuals for cancer sites where adult lifestyles are important risk factors such as obesity, diet, and occupational exposures. We found that this was the case for many cancer sites such as colorectal, breast, uterus (corpus), prostate, kidney and renal pelvis, non-Hodgkin lymphoma, and leukemia. However, for many other sites the cancer incidence rate in racial groups was lower than in both Canadian-born White individuals and their world region of origin. This was the case for head and neck, esophagus, lung and bronchus, and cervical cancers. These cancers are strongly related to exposures adopted early in life prior to immigration and linked with socioeconomic status: tobacco consumption, alcohol drinking, and early life infections (41, 42). The low incidence rates for these cancers suggest there is a selection for immigrants in better health and from higher socioeconomic backgrounds from within their country of origin, who are less likely to be smokers (40) and heavy drinkers (43), and exposed to some infections. The high rates of cervical cancer in Indigenous women are attributable to both higher human papillomavirus (HPV) infection prevalence and lower screening coverage (44–47); however we are unaware of any studies comparing HPV prevalence by immigrant status or other racial groups in Canada. Immigrant women and some visible minority groups have lower cervical cancer screening rates than White and Canadian-born women (10, 48). This suggests that the lower risk of cervical cancer in many visible minority groups cannot be attributed to more screening, and is either due to a lower prevalence of HPV or of other cofactors leading to cervical cancer progression.

Stomach, liver, and thyroid stood out as the cancer sites where incidence rates were systematically higher across most racial groups than in White individuals. It is likely that for stomach and liver cancers, the higher incidence rates reflect childhood exposures to Helicobacter pylori and hepatitis B virus in the countries of origin of many immigrants (49, 50), as well as higher prevalences of H. pylori and hepatitis C virus in some Indigenous populations (50–52). Stomach cancer rates are known to be higher in Northern Canada, where higher proportions of Indigenous populations live (53). Remarkably, the rates of these cancers were also high in Canadian-born East Asian, Southeast Asian, and Middle Eastern individuals. It is possible that the persisting high rates of these cancers in Canadian-born members of these racial groups are due to vertical or familial transmission of these pathogens. While all provinces perform prenatal screening for hepatitis B to provide vaccination against hepatitis B to at-risk children, prenatal hepatitis B screening and follow-up has imperfect coverage (54). There is no routine screening for H. pylori in Canada. The risk of stomach cancer does not appear to decline with time since immigration in many Canadian immigrant groups (55). While H. pylori prevalence tends to decrease in successive generations of immigrants, it still often remains higher than the prevalence of the host country population (56). Persistence of stomach cancer risk across generations may also reflect conserved cultural differences in diet, such as consumption of foods high in salt (57).

Thyroid cancers are often diagnosed through incidental findings during examinations for other health problems. More incidental cancer diagnoses in the Canadian healthcare system likely account for the higher rates of this cancer in visible minority groups compared with their world regions of origin. However, it is not clear why rates are higher in visible minority groups than in White Canadians and Indigenous peoples. This finding contrasts with the United States, where thyroid cancer incidence rates are highest in White populations and lowest in Black populations due to racial disparities in access to healthcare (58). Visible minority groups have similar rates of contact with specialist physicians as White Canadians (59); it is nonetheless possible there could be more incidental findings of thyroid cancer due to the presence of other underlying conditions increasing health care utilization.

An important limitation of our analysis is that we are unable to account for emigration; it is therefore possible we may have underestimated cancer incidence rates due to the inclusion of person-time from cohort members who have left Canada. Cancer incidence might therefore be underestimated in groups with more nonpermanent residents, such as Korean and Latin American populations. Cancer incidence may also be underestimated due to the exclusion from CanCHECs of homeless or institutionalized populations living in nursing homes and penitentiaries, or due to the lower response rates to the voluntary 2011 NHS. However, we compared CanCHECs cancer incidence rates with the national incidence rates (which are not subject to these biases) and found very similar rates. This suggests that this underestimation is likely to be minimal overall. This study relied on self-report of race; there may be some misclassification due to individuals unwilling to disclose their race in the census. This may be more common for Indigenous peoples due to legacies of colonialism in Canada (60). We aimed to provide as disaggregated data as possible for racial categories, in alignment with calls for disaggregated health data by race (1, 4). However, this disaggregation came at the cost of lower statistical power for some cancer sites to detect meaningful differences. Even by pooling two CanCHECs with several years of follow-up, we were unable to meet confidentiality disclosure thresholds for rarer cancers in some racial groups. While we did not disaggregate data by Indigenous identity in this study, several studies have previously published on cancer outcomes in Indigenous peoples distinguishing between First Nations, Inuit, and Métis populations (11, 44, 61–65).

While previous studies have examined racial and ethnic differences in cancer rates in Canada, they have tended to be based in a single province or examined only the most incident cancer types (6, 7). A study of the 1991 CanCHEC established the existence of a healthy immigrant effect for cancer incidence rates by immigration status (8). It is often assumed that cancer rates in immigrants will trend towards those in their host country with time since immigration due to acculturation. However, that study and others (39) found that the convergence of cancer incidence rates and mortality rates between immigrants and nonimmigrants over years since arrival might be partly explained by changes in the countries of origin of immigrants over time, as more recent immigrants have come from world regions with lower cancer rates. These findings suggest that either immigrant lifestyles might not assimilate as much as previously believed, or that lifestyle changes later in adulthood may be less important for cancer risk than early life exposures. For example, the proportion of overweight immigrants does not greatly increase with time since immigration, and remains lower than in Canadian-born individuals even a decade after immigration (40). While there are few studies of dietary acculturation in Canada, some studies suggest that many immigrants maintain their traditional diet to a certain extent (40, 66). If culture and lifestyle are conserved across generations, this may explain why we found that cancer incidence did not fully converge between Canadian-born visible minority groups and Canadian-born White individuals.

A previous study of the 2006 CanCHEC estimated differences in overall cancer incidence rates by ethnicity and immigration status (9), but did not estimate site-specific cancer incidence rates as we have done. That study also examined the construct of ethnicity rather than race, using the ethnic origin census question rather than the Indigenous group and visible minority group questions. While there is construct overlap, a person's ethnic origin may not necessarily correspond with their race. We opted to use the constructs of visible minority group combined with Indigenous identity to capture race rather than ethnicity, and to align more closely with race-based and Indigenous identity data standards for measuring racial health inequities in Canada (23).

Cancer is a heterogenous group of diseases with different aetiologies and long latency. We focus the discussion above on large trends across sites; however, an understanding of the drivers of cancer inequalities requires a separate analysis by cancer site and over time to account for changes in risk after adoption of a new home country. Finally, the large differences we have observed between racial groups point to important potential gains in cancer prevention. The ComPARe study estimated that 33% to 37% of cancers among adults in Canada are attributable to preventable risk factors such as tobacco consumption (42). However, the differences we observed between racial groups are in some cases larger than the estimated preventable fractions of some cancer sites. This suggests that the preventable fraction of cancer from environment and lifestyle may still be underestimated. More studies of immigrant and racial group lifestyles and experiences may further elucidate opportunities for cancer prevention.

T. Malagón reports grants from Canadian Institutes of Health Research during the conduct of the study; and is a board member of the International Papillomavirus Society. S. Morais reports grants from Canadian Institutes of Health Research during the conduct of the study. M. El-Zein reports a patent for DNA methylation markers for early detection of cervical cancer pending. E.L. Franco reports grants from Canadian Institutes of Health Research during the conduct of the study; personal fees from Merck outside the submitted work; and fees from Elsevier, eLife Sciences, and Oxford University Press as Editor of Medical Journals. No disclosures were reported by the other author.

T. Malagón: Conceptualization, formal analysis, validation, investigation, methodology, writing–original draft, project administration. S. Morais: Validation, methodology, writing–review and editing. P. Tope: Validation, methodology, writing–review and editing. M. El-Zein: Project administration, writing–review and editing. E.L. Franco: Supervision, funding acquisition, project administration, writing–review and editing.

This study was funded by a Canadian Institutes of Health Research (CIHR; grant 179901) to E.L. Franco, and by a CIHR HIV/AIDS and STBBI Research Initiative, sponsored by the CIHR Institute of Infection and Immunity, Postdoctoral Fellowship Award (support to S. Morais; Funding Reference Number: 202110HIV-477526-93701).

Data Source: Statistics Canada, CanCHECs 2006 & 2011, 2006 long-form census, 2011 NHS, CVSD 2006–2015, and CCR 2006–2015. Reproduced and distributed on an "as is" basis with the permission of Statistics Canada. This does not constitute an endorsement by Statistics Canada of this product. The Postal Code^OM Conversion File Plus (7D) is based on data licensed from the Canada Post Corporation.

The publication costs of this article were defrayed in part by the payment of publication fees. Therefore, and solely to indicate this fact, this article is hereby marked “advertisement” in accordance with 18 USC section 1734.