This is a descriptive analysis of trends in UADT cancer adult male mortality rates attributed to smoking among neighborhood income quintiles in urban Canada. The analysis reveals that between 1986 and 2001 UADT cancer mortality fell in all quintiles and the reductions were comparable in all quintiles. Within the quintiles, reduction in smoking-attributable proportions of UADT cancer deaths were more pronounced.
According to the Peto method, the prevalence of smoking in the study population was estimated indirectly from lung cancer death rates of the population
. This indirect method substitutes observed current exposure of smoking estimates with prevalence of smoking that is considered necessary for causing the current lung cancer mortality burden
. For most smoking-related outcomes, the current burden of disease is largely influenced by the past smoking exposure in the population
[28, 29]. The prevalence estimates calculated through this method avoids the potential error resulting from the lag time between population changes in smoking prevalence and the resulting change in disease outcome
Excess risks were arbitrarily halved to calculate smoking-attributable fractions conservatively as some of the deaths can be attributed to other risk factors, such as alcohol and Human Papilloma virus infection. As the methods used have been acknowledged to be crude, presentation of apparently precise numbers should not be taken to suggest otherwise
. The statistical significance of the observed trends in smoking-attributable mortality rates were also not assessed using any method like weighted regression analysis, as the motive was to look at the trends of these rates in general in different social strata of Canada. The major pattern is, however, clear that smoking-attributable UADT cancer mortality is reducing among all social strata of Canada. This is in consensus with steady declines in male smoking prevalence (15 years and above) in Canada over the last five decades; the rates reduced from 61% in 1965 to 20% in 2010
. The trends observed here are in agreement with a study done by Gupta et al. in Canada, and the US, which stated that the incidence of UADT cancers reduced between 1984 and 2001
. A possible explanation of this reduction can be tobacco control policies (for example significant increase in tobacco taxes in 1980’s and early 1990’s) which were implemented at that time period affecting the smoking prevalence. The results observed are also in consensus with Reid et al., who observed smoking prevalence among different social strata in Canada, 1999–2006, also revealed absolute reductions in daily smoking and cigarettes consumed per day in both the highest and the lowest social strata
For the analysis, the relative risk of smoking-attributable UADT cancer mortality was considered the same across all quintiles; however, there are many factors other than smoking that differ between quintiles
 and as smoking interacts with other risk factors
, the hazard for the individual smoker must also be expected to be different across various quintiles. However, Thun et al., for a US study, determined that smoking-attributable deaths reduced by just 1% per year after adjusting for other factors like education, occupation, race, alcohol consumption, and various dietary factors, in addition to age and sex
Because of lack of any large national representative mortality study, usage of relative risks based on CPS II study was another limitation; however, the mortality risks of CPS II study for various diseases have been quite acceptable in the Canadian context. Another limitation was the use of neighborhood level information, instead of family or individual, and applying to individuals, which forces consideration of the ecological fallacy. However, past studies have argued for the validity of using income quintiles as a proxy for individual socioeconomic status
Mackenback and Kunst, in 1997, presented a framework for measuring the magnitude of socio-economic inequalities in health, according to which, simple and straightforward measures are more useful in informing policy makers
. Relative Index of Inequality (RII), Slope Index of Inequality (SII), and Concentration Index on the other hand have a complex interpretation and can easily lead to misunderstandings
. Therefore, we used rate ratios and rate differences to depict social inequalities.
Although the methods of estimation used are indirect and have some limitations, the uncertainties inherent in these methods affect all social strata similarly; therefore, cannot account for overestimation of the differences observed between social strata in smoking-attributed mortality.