Background
The number of individuals diagnosed annually with thyroid cancer (TC) is approximately 289,000 globally [
1], including approximately 6,300 Canadians [
2], 62,450 Americans [
3]. Furthermore, TC accounts for the most rapid rise in incidence of any cancer in Canada and the United States [
2,
3]. TC carries one of the lowest risks of disease-related mortality among malignancies [
2,
3]. Approximately three out of every four TC diagnoses are made in women, reflecting a marked female sex predilection [
3]. In Canada, there is significant geographic variability in TC incidence rates across the nation [
2,
4], with particularly elevated rates observed in urban centers [
5], and especially the province of Ontario [
2]. In contrast to many other malignancies, smoking is associated with a reduced risk of TC [
6]. An in-depth analysis of regional distribution of Canadian TC cases in population-dense urban centers, including representation of a variety of racial and immigrant groups across a spectrum of socioeconomic class, is of great interest, in unraveling current incidence trends of this malignancy.
There were approximately 26,600 Canadians [
2] and 221,200 Americans [
3] diagnosed with lung cancer (LC) in 2015. In Canada, lung cancer (LC) is currently the second most common malignancy diagnosed in women and the third most common in men, with a slightly lower age- standardized incidence rate in women (48 per 100,000 population) compared to men (58 per 100,000 population) [
2]. LC carries the highest cancer-related death of all cancers in Canada [
2] and the United States [
3]. LC incidence has been decreasing in men and has been relatively stable in Canadian women since 2006, with future projected decreases in women attributable to tobacco control, given that smoking is a well-established risk factor for this malignancy [
2]. There is significant regional variability of LC incidence rates across Canadian provinces [
2]; the highest age-standardized incidence rates of LC are estimated in Quebec, and the risk of LC is more than 40 % higher in both sexes in that province compared to the neighboring Ontario [
2]. Taken together, these data are suggestive of contrasting epidemiologic risk factors and disease behavior of TC and LC. As such, in this work, we examined the regional variability of respective TC and LC incidence among women in Toronto neighborhoods and explored the relationship of incidence patterns with socio-demographic variables (ethnicity, immigration, and household income). Metropolitan Toronto is Canada’s most populated urban center rendering it an ideal setting to examine disease-ethnic associations. This study was focused on women, given their known predilection to TC risk [
3]; whereas in contrast, the risk of LC may be slightly lower in women compared to men [
2]. Restricting our analysis to the female gender would also influence any potential impact of gender difference on the interpretation of results. The public health implications of our study are that it provides some insight into the regional incidence patterns, potential explanatory factors, and health resource utilization implications of these malignancies for the region of interest.
Discussion
In comparing the regional distribution of TC and LC cases among Toronto women, we observed that the neighbourhoods where the density of TC was the highest, were not overlapped with the areas where the density of LC cases was the highest. In comparing the association between distribution of ethnic background and cancer incidence, a higher regional density of Asian race was associated with a significantly increased age-adjusted density of TC cases, but significantly reduced relative risk of LC. However, a higher regional density of African race was not significantly associated with the age-density of either TC nor LC female cases. A higher regional density of recent immigrants was associated with a slight reduction in TC relative risk but slight increase in LC relative risk. In a national Canadian study, Carriere et al. reported an inverse gradient of overall cancer incidence rates and regional concentration of foreign-born individuals; however the opposite relationship was observed for thyroid cancer [
13]. Similarly, we have previously reported a positive association between proportion of immigrants and TC incidence among health regions within the province of Ontario [
14]. It is important to note that in this analysis, we defined immigrant status by recent immigration (within 5 years) and not on the basis of foreign birth without time restriction, which may account for some of the difference in our findings. Interestingly, Horn-Ross et al. reported that in an analysis of California data from the California Cancer Registry, for the period of time similar to our analysis (years 2003 to 2009) the average annual age-adjusted incidence rate for thyroid cancer per 100,000 population was relatively comparable in Non-hispanic Whites (12.03, 95 % confidence interval [CI] 11.75, 12.32) to that of Asian/Pacific Islanders born in the United States (12.40, 95 % CI, 10.91, 14.03), but it appeared slightly lower in foreign-born Asian Pacific Islanders (9.17, 95 % CI, 8.62, 9.76) [
15]. Thus, the existing North American reports on any association between immigration and thyroid cancer incidence appear conflicting. Higher socioeconomic status (reflected by regional median household income) was not significantly independently associated TC nor LC age-adjusted risk in this study, after adjustment for race, and immigration in our study. An inverse association between socioeconomic status and LC incidence has been previously reported [
16‐
18], although it appears to be attenuated by adjustment for smoking history [
19]. A positive association between socioeconomic status and TC risk has been previously reported in other studies [
5,
20‐
24], but was not observed in our study, which may in part, relate to differences in the method of analysis (spatial analysis adjusted for multiple variables in our study), classification of socioeconomic status, variability of incomes within studied regions, differences in healthcare delivery, gender differences of study populations (as our study was restricted to women), or other lifestyle or cultural factors. In spite of the observed contrasting epidemiologic associations with TC and LC risk in our study, some of the excess regional risk remained unexplained, as reflected by the residual risk plots generated by our analyses.
The incidence of TC in Toronto is known to be higher than that of the majority of other large Canadian cities, after adjustment for demographic and socio-economic variables [
25]. Furthermore, Corsten et al. have recently reported that the incidence of TC in the northern Greater Toronto Area (eg. Markham, Vaughn, and Richmond Hill) is significantly higher than in the Toronto city core [
25]. Although our method of analysis was different and we restricted our analysis to women, our study results generally corroborate those published recent findings suggesting increased TC relative risk in the northern Greater Toronto area. Individual-level analyses are now required to explore potential environmental, healthcare access, or other potential factors contributing to the disproportionately high risk of TC in the northern Greater Toronto Area.
Another finding of this study, was the observation of the relationships between regional race representation and cancer incidence. Specifically, we observed an inverse independent association of regional representation of Asian ancestry, with risk of TC (increased) and LC (reduced) among Toronto women. Jin et al. have recently analyzed data from the Surveillance, Epidemiology, and End Results [SEER] Program from 2009 to 2011 from eight states, and observed that in Asian American women, the risk of TC was comparable (21.5 per 100,000 population, 95 % confidence interval [CI] 20.8–22.2) to that of Non-hispanic white women (22.4, 95 % CI 22.1–22.7), whereas the risk of LC was reduced in Asian women (28.6, 95 % CI 27.8, 29.5) compared to Non-Hispanic white women (59.4, 59.0, 59.8) [
26]. With respect to the risk of papillary thyroid cancer (the most common type of TC), Aschebrook-Kilfoy et al., has previously reported that among American women in the SEER database, the risk of papillary thyroid cancer was greatest among Asian Americans and lowest among African Americans [
27]. Furthermore, Aschebrook-Kilfoy observed that among Asian Americans, regional papillary thyroid cancer incidence rates were highly variable, ranging from 5.3 per 100,000 population in the state of Connecticut to 9.5 per 100,000 in Iowa [
27]. Yet we did not observe any significant association of African race and TC nor LC risk in our study. Previous reports analyzing data from SEER, have suggested a reduced risk of TC in African American women compared to white women [
27,
28]. Our study highlights the importance of advancing our understanding of the relationship between race (and the genetics of race), the environment, lifestyle, and cancer susceptibility.
Some of the strengths of our study included the retrieval of epidemiologic and cancer incidence data from well-established databases. Furthermore, the number of LC and TC cases included in this study was relatively large. Another strength of this study was the use of Bayesian inference with the INLA package in the R statistical programming language; the INLA software has made fitting Geostatistical models to count data and point location data a practical and convenient option for studies of this kind, in contrast to earlier inference methodologies which were computationally intensive and required specialist knowledge (see the
Discussion in Diggle et al. [
29]). Some limitations of this study include the lack of details of individual data (such as individual smoking status, personal income, or race/ethnicity, or generational status), the relatively limited geographic scope of the analysis, the restriction of the analysis to women, and the lack of specific details relating to the cases of malignancy (eg. how diagnosed, disease stage, treatments, and outcomes). Also, it is possible that regional level variables such as ethnicity may be capturing other associated variables from neighbourhoods, rather than reflecting ethnicity itself. Furthermore, more spatially related factors could have been included to attempt to explain the spatial variation observed in the study, but our analyses were limited to some extent by resource and time limitations, as well as the availability of well-validated relevant potentially explanatory data.
Conclusion
In conclusion, our main finding was a remarkably contrasting incidence pattern for thyroid and lung cancer in Toronto women. This research highlights how contrasting spatial variation of incidence rates of different malignancies (such as thyroid and lung), may be relevant to explore in understanding the underlying mechanisms of susceptibility and causative factors associated with their incidence. Furthermore, the spatial analyses presented herein exemplify a cost-efficient means to account for important confounders in conducting exploratory causation research for human malignancies. In future research, information on social, genetic, and reporting effects should be taken into account in detailed spatial and individual-level analyses, to better inform our understanding of the incidence patterns and possible causative or preventative factors, in the development of thyroid and lung malignancies.
Acknowledgements
The authors would like to acknowledge assistance of Mrs. Coreen Marino in executing administrative tasks for this project. Map tiles are by Stamen Design under CC BY 3.0 (creativecommons.org/licenses/by/3.0) using data from OpenStreetMap.org available under the Open Database License (opendatacommons.org/licenses/odbl).