Geographic variation and factors associated with colorectal cancer incidence in Manitoba

Abstract

OBJECTIVES: Although individuals living in areas with lower household income have been shown to have higher rates of mortality from colorectal cancer (CRC), findings on the effect of income on CRC incidence in countries with universal health care have been inconsistent. There are limited data from Canada. We investigated the geographic variation and factors associated with CRC incidence in Manitoba, a central Canadian province. METHODS:The Manitoba Cancer Registry and Manitoba Health population registry were used to determine age-sex-standardized CRC incidence rates between 1985 and 2012, which were geocoded to 498 small geographic areas (SGAs). The 2001 Canadian Census was used to determine the sociodemographic characteristics of the SGAs. Bayesian spatial Poisson modelling was used to assess geographic variation and factors associated with CRC incidence.RESULTS: CRC incidence in SGAs ranged from 11 to 1026 per 100 000 population per year. Importantly, in the fully adjusted model there was no significant association between either average household income or proportion of recent immigrants in the SGAs and CRC incidence. Individuals living in urban areas had an overall lower CRC incidence (incidence rate ratio: 0.76; 95% credible interval: 0.58-0.98). CONCLUSIONS:In a province with a universal health care system, our study suggests there are no disparities in CRC incidence by socio-economic level of the areas of residence. Rural areas should be a focus of CRC reduction initiatives in Manitoba. Similar analysis in other jurisdictions should be performed to evaluate the effect of the characteristics of SGAs on CRC incidence in different settings and target some of the efforts to reduce CRC burden.KEY WORDS: Epidemiology; spatial analysis; variation in risk; disparities La traduction du résumé se trouve à la fin de l'article.

Area-Level Determinants in Colorectal Cancer Spatial Clustering Studies: A Systematic Review

“…Descriptive summary of included studies concerning study location, setting and study design are shown in Figure 2 . All eligible studies were conducted spanning five countries that include Canada [ 41 ], United States [ 11 , 32 , 42 ], France [ 43 ], Portugal [ 35 ], and Iran [ 33 , 34 , 36 , 37 , 38 , 39 ]. Comparing study locations based on the WHO regions, four studies [ 11 , 32 , 41 , 42 ] accounted for the Region of the Americas (AMR), two studies [ 35 , 43 ] from the European Region (EUR) and six studies [ 33 , 34 , 36 , 37 , 38 , 39 ] were performed in the Eastern Mediterranean Region (EMR).…”

“…The analyzed articles were published in the year 2016 to 2021. Seven of these quantitative studies resumed cross sectional study design [ 33 , 34 , 35 , 39 , 41 , 42 , 43 ], two studies were conducted using retrospective cohort [ 11 , 32 ], and three articles of ecological studies [ 36 , 37 , 38 ].…”

Current State of Geospatial Methodologic Approaches in Canadian Population Oncology Research

“…Cancer rates may be confounded by distributions of unknown or unmeasured factors related to cancer (85). Ecologic fallacy 24 Results should be interpreted only in the context of area of residence (42).…”

“…Only a single year of data was used for covariates, which may not be reflective of early (or latter) time periods (42,72). Self-reported measurements, recall bias and other information biases (excluding misclassification of exposure)…”

“…However, unstable risk estimates due to small case counts or population sizes were commonly cited in this review. Statistical approaches such as Bayesian hierarchical modeling may be used in these situations to stabilize risk estimates (41)(42)(43)(44)(45)(46)(47). Saint-Jacques and colleagues used the Bayesian Besag-York-Mollie (BYM) model to quantify the risk of bladder or kidney cancer in relation to arsenic exposure (41).…”