Alcohol consumption has a significant negative impact on public health. Alcohol accounted for far more deaths than the number of deaths prevented by alcohol for Canadians
0 to 64 years of age. This paper improves upon previous estimates of mortality by using new methodology to 1) account for the undercoverage of alcohol consumption in the CADUMS 2008, which represented only 34% of recorded alcohol consumption in Canada, 2) take into consideration the right-skewed distribution of alcohol consumption, and 3) calculate 95% CIs. Additionally, we used updated risk ratio estimates in this study, as well as estimates for mortality for new disease categories such as colon cancer, rectal cancer, and tuberculosis [8
]. Implementation of these new methods described by Rehm and colleagues [10
] resulted in a higher estimated mortality attributable to alcohol compared to previous methods due primarily to the additional disease categories included in our analysis. The inclusion of new disease categories and use of new methodology allows for a more accurate characterization of alcohol-attributable mortality, which can potentially lead to better public health policies aimed at reducing alcohol-attributable harms.
This paper's analysis has its limitations. For our study we did not include all aspects of harms caused by the drinking of others, such as due to motor vehicle accidents and workplace injuries, which recently have been shown to constitute a large proportion of the burden of injury attributable to alcohol, as no meta-analyses exist that describe an RR which can be used to calculate this burden [25
]. Due to the absence of available data on the incidence of certain events, such as the incidence of fetal loss during pregnancy attributable to alcohol consumption, our analysis was restricted to deaths and PYLLs and does not provide a complete picture of the effects of alcohol on all aspects of health. Additionally, we did not estimate the number of deaths for Canadians over 65 years of age since specified causes of death for people in this age group were considered to be unreliable. If we had included in our analysis the number of deaths for Canadians aged 65 years and older, the number of net deaths attributable to alcohol would have risen from 3,969 to 8,953 and the number of PYLLs would have risen from 134,555 to 172,255; however, as indicated above, primary information concerning the causes of death for Canadians over 65 years of age is considered to be unreliable.
For this study, we used cross-sectional estimates of the average volume of alcohol consumption, not taking into consideration individuals' past and current patterns of drinking as distinct from volume, with the exception of injury which incorporated binge drinking estimates. While restrictions on volume may not be important for certain diseases, patterns of drinking impact the risk of diseases such as ischemic heart disease and other cardiovascular diseases [4
]. In addition, not including past patterns of drinking is problematic for studying the role of alcohol in a person's developing malignant neoplasms as the time between exposure to a carcinogen and exhibiting symptoms of cancer can range from between 15 and 30 years [28
]. Liver cirrhosis at the individual level is mainly affected by long-term heavy drinking; however, immediate effects decreasing its incidence are seen at the country level after a short time when alcohol consumption decreases dramatically [29
]. Accordingly, public policies aimed at reducing alcohol-attributable harms will not only have an immediate impact on alcohol-related harms but will also have long-term benefits.
Limitations in the study design, sampling, and response rates of the CADUMS 2008 led to an undercoverage of the resulting estimates of alcohol consumption in Canada. For example, the CADUMS 2008 excluded people living in the Northwest Territories (which have a greater proportion of individuals suffering from alcohol abuse problems compared to the rest of Canada), and excluded homeless and institutionalized individuals, as well as those individuals without a home phone. The inherent problems with the CADUMS 2008 were adjusted for in our study by triangulating the survey data with the per capita
alcohol consumption estimate (according to the methods of Rehm and colleagues [10
The response rate for the CADUMS 2008 survey was relatively low, with an overall response rate of 36.5%. However, it has been previously demonstrated that higher response rates do not necessarily change the resulting distribution of alcohol consumption when the response rate varies between 30% - 60% and similar sampling methods are used [19
]. Moreover, since we triangulated survey data with per capita
consumption data for our study, underreporting of alcohol consumption should not be problematic [31
Undercoverage of alcohol may be present in both the survey alcohol exposure data and the studies used in the meta-analyses to calculate the RR functions. This undercoverage may lead to an overestimation of the alcohol RR functions for all diseases; however, the effects of these two instances of undercoverage differ. Only undercoverage from measurement error (response bias and construct validity) affects the RR functions. Undercoverage in exposure estimates derived from survey data arises from measurement error, non-response to the alcohol survey, and from the exclusion of certain populations from the survey. Since we do not know the extent of undercoverage caused by measurement error in the studies that were used in the meta-analyses, it is impossible to determine the coverage rate to which our survey data should be standardized in order to correct the RR functions. Since it is very likely that the undercoverage in epidemiological studies is much less than in population surveys [32
], we corrected the undercoverage of the CADUMS to 90% of the per capita
consumption in order to account for the overestimated RRs calculated from observational studies.
If we did not adjust for the undercoverage of the survey data, the number of deaths attributable to alcohol consumption would be 47.9% of the total number of deaths calculated after adjustment to 90% of per capita consumption, resulting in a calculation of 1,896 net deaths from alcohol. It should be noted, however, that adjustment of per capita consumption data across countries is necessary if the mortality figures are to be comparable across countries.
Our study is also limited by the RRs used to calculate the number of alcohol- attributable deaths and PYLLs in Canada for 2004. These limitations stem from the meta-analyses of the adjusted RRs [33
]. There are two potential problems with the RR functions that could affect the results of our study. First, there is an assumption that RRs for alcohol consumption are stable across countries. While this assumption is plausible for chronic diseases, which are mainly determined by biological relationships (as demonstrated by the low heterogeneity in the chronic disease RR meta-analyses), this assumption is less plausible for injuries where social influences will affect the RRs (e.g. traffic injury depends on highway safety standards, density of cars, etc.). As most of the RRs for injury are estimated based on data from high-income countries with similar environmental conditions to Canada, these RR estimates are valid for our study. Second, the use of adjusted RRs may overestimate the true RRs as observed by Flegal and others [34
]; however, the size of the potential bias has recently been shown to be relatively small [37
Regardless of our study's limitations and assumptions, alcohol is clearly a major contributor to mortality and PYLLs, and has a significant public health impact in Canada. In contrast to tobacco-related mortality, the average age of death resulting from alcohol consumption is much lower [2
], such that mortality due to alcohol consumption occurs primarily in adults of working age and results in a negative economic impact. This was also observed in our study, as alcohol was responsible for 7.7% of all deaths and 8.0% of all PYLLs for people between 0 and 64 years of age.
A monitoring system on alcohol consumption and alcohol-related harms is imperative as the burden of alcohol-attributable harms in Canada is unnecessary and could be reduced in a relatively short period of time if effective public health policies were implemented [38
]. These policies could include taxation, additional and/or stronger sanctions to combat impaired driving, and specific measures to decrease aggression and violence [39
]. Rehm and colleagues analyzed the effects of these interventions in Canada for 2002 and found that an increase in price of 25% would result in a decrease of 59 net deaths and more than 1,500 net PYLLs [42
]. By reducing in Canada the maximum acceptable blood alcohol content while operating a motor vehicle from 0.08% to 0.05%, Rehm and colleagues found that there would be 173 fewer net deaths and more than 7,000 fewer net PYLLs (a 4.1% reduction and a 5.0% reduction, respectively from the baseline scenario) [42
]. A real world example of government policy affecting mortality due to alcohol consumption was observed in Russia during the Gorbachev anti-alcohol campaign, where annual per capita
consumption of pure alcohol fell from 14.2 l in 1984 to 10.7 l in 1987. As a result of this decrease in alcohol consumption, all cause mortality rates for 40-44 year olds decreased by 39% for men and by 29% for women and alcohol-related mortality rates for 40-44 year olds decreased by 63% for men and by 63% for women between 1984 and 1987 [44
]. Each of the policy measures suggested above has been shown to be effective in jurisdictions outside of Canada which face similar epidemiological profiles of mortality due to alcohol consumption [39
]. For the above Canadian examples, the noted reductions in mortality represent a net estimate and, thus, incorporate potential increases in the incidence of diseases where alcohol exhibits a protective effect.
A comprehensive alcohol consumption and alcohol-attributable harms monitoring system in Canada would allow for evaluation of the effectiveness of public health policies and, once established, could be used to identify those areas where targeted alcohol policies would have the greatest impact. Finally, such a system would help to estimate alcohol-attributable harms in the future using time-trend analysis. Monitoring systems on the harms of alcohol should include the resulting social harms. In some regions it is possible that the social harms associated with alcohol consumption may exceed the burden of mortality.