Chronic disease conditions attributable to alcohol were identified through a review of epidemiological literature that found most assessments converge on the chronic disease categories causally related to alcohol (3,9). This study takes a conservative approach and includes only chronic disease categories with an established biological pathway, temporal order, consistent effects, and dose–response relationships (18). Table 1 gives an overview of chronic conditions identified. The meta-analyses reviewed to determine the nature of the effect of alcohol on chronic conditions in this study showed that the effect of alcohol consumption on chronic diseases can be detrimental or, in some cases, beneficial. The nature of the effect (detrimental or beneficial) of alcohol on chronic conditions depends on the level of alcohol consumed for some conditions, such as stroke (22) or ischemic heart disease (3,21). Lower levels of consumption have a beneficial effect; higher levels of consumption have a detrimental effect. For type 2 diabetes mellitus among men and cholelithiasis (i.e., gall bladder disease) among both sexes, the effect is beneficial for all drinking categories (Table 1). Either detrimental or beneficial effects may be based on confounding phenomena (23), which, in scientific literature, is discussed more for beneficial than for detrimental effects.

To measure level of alcohol consumption, we followed the approach of English et al (24) and used four drinking categories based on average volume of alcohol consumed (Table 2). The four levels are 1) abstainer or very light drinker, 0 to less than .25 grams of alcohol per day for men and women; 2) category I, .25 to less than 20 grams of alcohol per day for women and .25 to less than 40 grams for men; 3) category II, 20 to less than 40 grams of alcohol per day for women and 40 to less than 60 grams for men; and 4) category III, 40 grams of alcohol or more per day for women and 60 grams or more for men. Most meta-analyses have given relative risks based on these categories (9).

The prevalence data of different levels of alcohol consumption were collected between 2003 and 2004 through the Canadian Addiction Survey (CAS) (25) with a total sample size of 13,909 (n = 5,721 men and n = 8,188 women) and a 47% response rate. The CAS is a stratified (by region) and randomized (by household and respondent) telephone-based survey of the population aged 15 years and older initiated by the Canadian Centre on Substance Abuse (CCSA). This survey was used despite a relatively low response rate because it had the necessary alcohol consumption measures, a large sample size, and closest temporal proximity to the mortality data provided by Statistics Canada (27). It has previously been reported that higher response rates in surveys did not essentially change distribution results of alcohol consumption (28). 

Characteristics of the CAS sample were weighted to correspond to age and sex distribution of the Canadian population. Average volume of alcohol consumption was derived from a quantity–frequency measure and adjusted by adult per capita consumption (26). Quantity of alcohol was assessed in beverage-specific drinks per usual occasion (25). A per capita consumption adjustment was used because it is considered the most valid choice for alcohol consumption measurements for a population (28). Measurements based on usual quantity and frequency tend to underestimate alcohol consumption because respondents do not usually include heavy drinking occasions in their responses (29).

Mortality data, with underlying cause of death coded according to the International Classification of Diseases, 10th Revision (ICD-10), were obtained from Statistics Canada for 2002 (27). Comparisons of causes of death in the elderly have methodological problems; death certificates tend to be less valid for older individuals than for younger individuals because multiple causes of death are often involved (30,31). In this analysis, only data for those aged 69 years and younger are included because there is indication that relative risks, such as for ischemic heart disease and for substance-attributable mortality and morbidity, tend to converge with increasing age (26) and to result in overestimation of alcohol-attributable deaths caused or prevented in older age groups.

Alcohol-attributable fractions (AAFs) are generally defined as the proportion of a disease in a population that will disappear if alcohol is removed (32,33). AAFs for this study were calculated from alcohol usage proportions in Canada and disease- and sex-specific pooled relative risks from previous meta-analyses (Table 1) and by using the following formula:

where i is the category with usage (i = 1) or no alcohol (i = 0), RR(i) is the relative risk at exposure level i compared with no alcohol consumption, P(i) is the prevalence of the i^th category of alcohol consumption, and k is the highest drinking category (i.e., category III).

For each disease category, the sex- and age-group-specific AAFs were calculated as follows: the prevalence of alcohol consumption (exposure category) in the population was multiplied with the excess risk (RR-1) for the given level of alcohol consumption. The numerator in the formula represents the sum of all alcohol-attributable cases of a disease by exposure category within a given sex- and age-group. This sum is divided by all cases of a disease in the given sex- and age-group to derive a proportion that is attributable to the exposure (i.e., alcohol consumption). The counterfactual alternative in this conceptualization is no (or zero) consumption. To derive estimates of the number of deaths due to alcohol for a specific disease (by sex- and age-group), the AAFs were then multiplied by the number of deaths from that specific disease within the sex- and age-group.

We cite a meta-analysis for each condition (e.g., malignant neoplasms, type 2 diabetes mellitus, neuropsychiatric conditions, cardiovascular diseases, digestive diseases) (18) (Table 1) to derive information of relative risk for the diverse categories. Meta-analyses usually cited are from Gutjahr et al (19), a series of meta-analyses based on English et al (24). The Gutjahr et al meta-analyses incorporate literature published since publication of the English et al paper. For some disease categories, other meta-analyses were cited if they were more comprehensive than the Gutjahr et al paper. For stroke, we based relative risks on Reynolds et al (22) because, contrary to other authors, this work separates ischemic and hemorrhagic stroke. Alcohol usage has been shown to differentially influence both types of stroke (3). For hemorrhagic stroke, the effect of alcohol consumption, even at low levels, is detrimental for men (34); for women, the effect may be protective at low levels of drinking (categories I and II) (20), but the effect depends on the pattern of drinking (35). The effect of alcohol consumption on ischemic stroke is partly protective for low to moderate consumption levels (35), and the beneficial effect is more pronounced among women. For depression data, we used mental health surveys to estimate AAFs directly and took into consideration the rates of comorbidity and time of onset for alcohol use disorders and depression in comparison to other mental diseases comorbid with alcohol (3).

No alcohol consumption was used as a counterfactual scenario that was selected for the following reasons: 1) the counterfactual of zero consumption is the most widely used and allows comparisons with other studies; 2) the level of alcohol associated with lowest burden differs from country to country and is different for mortality and morbidity; 3) alcohol-attributable harm is related to average volume of alcohol consumption and consumption patterns; and 4) selection of any volume of consumption other than zero without considering the other dimensions would be arbitrary (36,37).

We calculated AAFs separately by sex and age (men and women 15–29 years, 30–44 years, 45–59 years, 60–69 years). To show detrimental and beneficial effects of alcohol, we applied AAFs to Canadian mortality data (27) to estimate the number of alcohol-attributed deaths by age and sex. A negative AAF, in this case, means that more deaths were prevented than caused by alcohol for the respective disease condition (Table 3).

We hypothesized that persons dying due to alcohol consumption would have lived longer if they had not consumed alcohol. The average extra time such individuals would have lived is known as residual life expectancy. If a man died of alcoholic liver cirrhoses at age 50 in Canada, he would have a residual life expectancy of 28.4 years (38). The sum of residual life expectancies for people dying from alcohol consumption is known as potential years of life lost (PYLL) due to alcohol. PYLL for each sex- and age-group can be estimated by interpolating the observed mean age at death and the standard life expectancies tables for each respective sex– and age–group. This study uses the World Health Organization (WHO) 2000 standard life expectancies table for Canada (38).

To calculate the mean ages within age intervals, we followed rules specified by the WHO Global Burden of Disease study (39). Canadian PYLL were calculated for each age group (15–29, 30–44, 45–59, and 60–69) by multiplying the number of deaths by the interpolated life expectancy for the observed mean age at death for the age interval. The upper age limit of 76.0 years for men and 81.5 years for women was used to approximate Canadian life expectancy at birth. PYLL were calculated per 100,000 population (Table 4).

Canadian fiscal year 2002 to 2003 hospital diagnoses data (ICD-10) were obtained from the Canadian Institute for Health Information (CIHI) Hospital Morbidity Database (HMDB) for both national and provincial levels (40). National data were available for seven provinces (Alberta, British Columbia, Newfoundland, Nova Scotia, Ontario, Prince Edward Island, Saskatchewan) and two territories (Northwest Territories and Yukon) and were provided for disease conditions, and sex- and age-groups in 5-year segments. Data for the provinces of Quebec, Manitoba, New Brunswick, and Nunavut (a northern territory) were not available according to ICD-10 classifications. Aggregated data for each condition were estimated based on total population and the combined data of available provinces and territories. The HMDB captures information on patients separated (through discharge or death) from acute-care facilities in Canada and provides national data on acute-care hospitalizations by diagnoses and procedures. Diagnoses were categorized based on most responsible diagnosis (MRD) of hospitalized patients. In cases where multiple diagnoses may be classified as most responsible, the diagnosis associated with the longest stay in a hospital was used. AAFs were then applied to hospital data to estimate alcohol-attributable hospital diagnoses for each treatment facility by age and sex.

This study estimates premature deaths and hospitalizations for Canadians aged 69 years and younger and provides an overview of the 2002 estimated volume of alcohol exposure in Canada by sex- and age-group. As expected, men consumed on average more than women, and alcohol consumption decreased with age.

Table 5 provides estimates of alcohol-attributable deaths — deaths caused or prevented by alcohol consumption. In Canada in 2002, there were 1631 net alcohol-attributable premature deaths estimated, 1155 deaths among men and 476 among women. These numbers were derived by multiplying AAFs with number of deaths for each category to produce numbers with decimals. As a result, there may be minor rounding errors after collapsing numbers across different categories.

The 1631 alcohol-attributable deaths constituted 2.4% of the deaths in Canada for people aged 69 years and younger. These were net figures, and the estimates of deaths prevented by alcohol have been taken into account. There were 2577 deaths (1906 men and 672 women) attributable to alcohol and 947 deaths (751 men and 195 women) prevented by alcohol consumption. Figures were calculated by using the epidemiological procedure previously described.

Moderate drinking (category I, less than 20g per day of pure alcohol for women and less than 40g per day for men) was associated with 25% of deaths caused by alcohol consumption and 85% of deaths prevented by alcohol consumption. In this group of moderate drinkers, the deaths prevented (n = 827) outnumbered deaths caused (n = 677) by alcohol use.

Among premature deaths caused by alcohol, malignant neoplasms accounted for 891 deaths (608 men and 283 women), and digestive diseases accounted for 881 deaths (663 men and 218 women) (Table 3). Among deaths prevented by alcohol for those aged 69 and younger, 788 prevented deaths were ascribed to ischemic heart disease (678 men and 110 women).

Figure 1 gives an overview of deaths prevented and caused by alcohol. Of the deaths caused by alcohol, including detrimental effects of alcohol on some cardiovascular diseases, cancer (34.6%) and digestive diseases (34.2%) had the highest proportions of deaths caused by alcohol.

In 2002, the PYLL rate for Canada for premature deaths due to alcohol was 196 per 100,000 for men and 92 per 100,000 for women aged 15 to 69 (Table 4). For every population of 100,000, there was a potential net loss of 196 years of life among men and 92 years of life among women as a result of premature death due to alcohol consumption. In total, 42,996 potential years of life were lost due to alcohol in Canada in 2002.

Table 6 provides estimates of alcohol-attributable diagnoses for those aged 69 and younger, and Table 7 shows AAFs applied to hospital data to estimate alcohol-attributable fractions and mean age at hospital diagnosis. In Canada for fiscal year 2002 to 2003, 91,970 alcohol-attributable diagnoses among hospital diagnoses from acute care facilities were estimated and accounted for 65,161 hospital separations for men and 26,809 for women. These numbers were derived by multiplying AAFs with number of diagnoses for each category to produce numbers with decimals. As a result, there may be minor rounding errors after collapsing numbers across categories.

Figure 2 gives an overview of alcohol-attributable chronic disease diagnoses. The number of diagnoses (91,970) is a net figure and includes estimates of hospital separations prevented by alcohol consumption. There were 124,621 acute-care hospital diagnoses (hospital separations for 89,022 men and 35,599 women) caused by alcohol and 32,651 hospital diagnoses (23,861 men and 8790 women) prevented by alcohol consumption. Among acute-care hospital diagnoses attributable to alcohol, neuropsychiatric conditions accounted for 49.9%, cardiovascular disease 33.7%, digestive diseases 13.0%, and cancer (malignant neoplasms) 3.4% of diagnoses (Table 6).