QUANTIFICATION OF ALCOHOL-RELATED MORTALITY IN SWEDEN

Harmeet Sjögren, Anders Eriksson*, Göran Broström1 and Kristin Ahlm

Department of Community Medicine and Rehabilitation, Forensic Medicine, Umeå University, P.O. Box 7642, SE-907 12 Umeå and
1 Department of Statistics, SE-901 87 Umeå University, Sweden

Received 10 September 1999; in revised form 30 May 2000; accepted 13 June 2000


    ABSTRACT
 TOP
 FOOTNOTES
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 REFERENCES
 
The main aim of the present study was to estimate total alcohol-related mortality in Sweden. For natural deaths, a meta-analysis carried out in Australia was updated to the end of March 1998, and pooled estimates of the relative risks were calculated for different diseases based on data from scientific studies that have been published in the international literature. The proportion of current alcohol drinkers from recent Swedish surveys, and the pooled relative risk estimates were used to estimate disease-specific alcohol-attributable fractions. Natural deaths ‘caused’ or ‘prevented’ by alcohol were estimated for the period 1992–1996. For unnatural deaths, all cases from 1992 through 1996 in Sweden were analysed (n = 23 132). Alcohol was regarded to attribute to the death: if the deceased was a ‘known alcoholic’; if the underlying or contributing cause of death was alcohol-related; if the deceased had an alcohol-related in-patient diagnosis during a 3-year period prior to death; if the case tested positive for blood alcohol. Person years of life lost/gained (<70 years) due to alcohol were also assessed. The assumptions underlying the attributable risk methods used to analyse alcohol-related mortality due to natural causes need to be borne in mind when interpreting the results on natural deaths. Moreover, the preventive effect of alcohol on coronary heart disease and stroke is still controversial. The findings of alcohol-related mortality due to unnatural causes were much more reliable. About 3.5% of deaths were attributed to alcohol; alcohol involvement was more than twice as common in deaths of males (4.8%) than in those of females (2.0%). About one-quarter of the deaths in those aged below 50 years were attributed to alcohol. In those (<=69 years, alcohol had a net harmful effect in that it ‘caused’ more deaths than it ‘prevented’; 7% of deaths were in net ‘caused’. More person years of life were in net lost than were saved by alcohol, suggesting that alcohol has a negative effect on premature mortality. We conclude that alcohol accounted for about 3.5% of deaths in all ages, and 25% of deaths in those aged below 50 years, and about 10% of person years of life lost in Sweden. The balance of harm and benefit does not weigh in favour of making a recommendation to the public to drink in order to prevent death.


    INTRODUCTION
 TOP
 FOOTNOTES
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 REFERENCES
 
Alcohol is responsible for considerable morbidity, mortality, and social problems. Public health problems associated with alcohol use and abuse include mortality from injuries associated with acute exposure to alcohol and mortality from diseases associated with long-term use/misuse of alcohol.

A number of attempts have been made to assess the proportion of deaths ‘caused’/‘prevented’ by alcohol in other countries. English et al. (1995) performed a meta-analysis of the published scientific literature and found that 5% of all deaths in Australia were attributable to high alcohol intake, relative to low alcohol intake (Holman and English, 1995Go). Scragg (1995) used data on the proportion of current alcohol drinkers from recent New Zealand surveys, and pooled relative risks derived from a meta-analysis by Holman et al. (1990) to calculate disease-specific population attributable risks for New Zealand. Using a similar approach, Single et al. (1999) in a recent study, estimated the mortality attributable to alcohol use in Canada.

Shultz et al. (1990) estimated alcohol attributable fractions for different diagnoses from a review of clinical research and found that 4.9% of the total number of deaths in the USA were related to alcohol. Other studies from the USA have also used alcohol attributable fractions derived by Shultz et al. (1990), to estimate alcohol-related mortality in different regions (Peterson et al., 1990Go; Stinson and DeBakey, 1992Go; Sutocky et al., 1993Go). A Spanish study also used alcohol attributable fractions derived by Shultz et al. (1990) and found that 6.1% of the deaths in Spain were related to alcohol consumption (Yañez et al., 1993Go). Similarly, 9% of the overall mortality in France was found to be attributable to alcohol (Pignon and Hill, 1991Go). Others have also estimated alcohol attributable fractions for different causes of death from reports in the literature (Parker et al., 1987Go; Gorsky et al., 1988Go). Harmful/protective effects of alcohol on global mortality have also been estimated by Murray and Lopez (1997).

Despite the fact that alcohol is the most commonly misused substance in Sweden, little is known about its impact on the total mortality in the country. Epidemiologic studies in Sweden have investigated the association between alcohol and mortality and morbidity in selected populations (Petersson et al., 1982Go; Adami et al., 1992Go; Romelsjö et al., 1993Go), and also the association between alcohol and specific diagnoses, such as stroke (Harmsen et al., 1990Go; Hansagi et al., 1995Go), ischaemic heart disease (Messner and Petersson, 1996Go), breast cancer (Holmberg et al., 1995Go), and gastric cancer (Hansson et al., 1994Go). Much has been done on the role of alcohol in specific populations in Sweden in traffic crashes (e.g. Björnstig et al., 1985; Öström et al., 1995; Sjögren et al., 1997) and other unnatural deaths (e.g. Ormstad et al., 1986; Örnehult and Eriksson, 1987; Berglund et al., 1987; Karlsson et al., 1988; Norström, 1988).

The aim of the present study was to estimate the proportion of deaths ‘caused’/‘prevented’ by alcohol and premature death due to all causes of death in Sweden. Alcohol involvement was assessed for different specific causes of natural and unnatural deaths (cause-specific mortality). As a comparison, alcohol involvement was also assessed for all-cause mortality (all-cause mortality), for ages 20–39 years based on risk estimates for a cohort of Swedish conscripts (Andréasson et al., 1991Go).


    MATERIALS AND METHODS
 TOP
 FOOTNOTES
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 REFERENCES
 
Natural deaths
The present study applied the methods used by English et al. (1995), who estimated the proportion of deaths in Australia ‘caused’ and ‘prevented’ by alcohol consumption (Holman and English, 1995Go; Holman et al., 1996Go). The literature (1987 up to May 1994) used in the meta-analysis by English et al. (1995), was updated up to March 1998. The pooled relative risks were estimated for different diseases, based on the relative risk estimates presented by English et al. (1995) together with relative risk data in new scientific studies that have been published in the international literature. The pooled relative risks for female breast cancer and epilepsy changed due to new studies since May 1994, compared to the meta-analysis by English et al. (1995). The proportion of current alcohol drinkers from recent Swedish surveys, and the pooled relative risk estimates, were used to estimate disease-specific alcohol-attributable fractions (AAFs) and alcohol-related mortality for different age groups (Sjögren et al., 2000aGo).

For ischaemic heart disease, pooled relative risks for different alcohol levels of consumption from a recent meta-analysis by Mäkelä et al. (1997) were used to estimate AAFs and alcohol-related mortality. For cerebrovascular disease, we used a stricter selection of studies than that used by English et al. (1995) in their meta-analysis, and estimated pooled relative risks for different levels of alcohol consumption.

The articles were reviewed carefully and those that provided data on a relationship between a disease and alcohol consumption were chosen. Suitable data on relative risk/odds ratio for a particular disease and alcohol consumption were extracted from the articles. The risk ratios adjusted for various confounding variables, such as sex, age or smoking, were preferred.

Pooling multiple study results to estimate the relative risk. We used precision-based weighting to obtain pooled estimates of relative risks as described by English et al. (1995). The pooled estimate of the relative risk from N studies using precision-based weighting is:

where RR is the pooled relative risk estimate, and RR i is the relative risk in published study, i, and Wi = 1/Var [ln(RR i)], i = 1..., N

An approximate 95% confidence interval around the pooled estimate is:

Prevalence of alcohol consumption in Sweden. The prevalence data of different levels of current alcohol consumption in the Swedish population aged >=16 years were estimated from TEMO (test house market opinion) surveys for 1992, 1993, 1994, and 1996. The TEMO national survey was a household survey on a sample of 4434 persons (2224 males, 2210 females). The selected persons were interviewed in their homes by professional interviewers using a structured questionnaire. The sample was randomly drawn from the whole of Sweden taking into consideration age, sex and region to avoid sample bias. It was estimated that there was a 25–30% drop-out rate. The crude data from the TEMO surveys were analysed by Håkan Leifman (Department of Sociology, Karolinska Institute, Stockholm, Sweden). The data with alcohol consumption levels used in the present study were provided by Håkan Leifman, and tabulations customized for our purpose were generated.

Derivation of AAFs for different diseases. In the present study, AAF is defined as the fraction of the disease in the population that would not have occurred if the effect associated with alcohol, which is the risk factor, was absent. Since alcohol may ‘cause’ or ‘prevent’ deaths, the AAF can be positive or negative. The AAFs were calculated from alcohol exposure prevalence in Sweden and the pooled relative risks for the diseases from the update of meta-analysis, as described by English et al. (1995).

If category 0 is the baseline category, Pi is the prevalence of i-th category, RRi is the relative risk in the corresponding category compared with the reference category, then the attributable fraction due to k categories of exposure is:

Mortality data. The age- and sex-specific number of deaths due to each disease included in the present study for Sweden between 1992 through 1996 were obtained from the Swedish Official Register of causes of death (Statistics Sweden, 1992Go, 1993Go; National Board of Health and Welfare, 1994Go, 1995Go, 1996Go).

Unnatural deaths
All unnatural death cases in Sweden (1992–1996) were included in the study sample (n = 23 132; male n = 14 740, female n = 8389 and three of unknown sex). The sample included medico-legally autopsied (n = 15 630), hospital-autopsied (n = 1144) and non-autopsied (n = 6358) cases.

Information on medico-legally autopsied cases was obtained from the Forensic Medicine Database (FMD) of the National Board of Forensic Medicine. This database includes information on age, sex, the underlying and contributing cause of death (coded according to ICD-9), external cause of death (ICD-9, E-code), circumstances associated with the death, date and place of death, hospitalization, hospital treatment, mode of death (accident, suicide, homicide, undetermined), type of autopsy, and a number of other variables. Unique code numbers of the cases in the FMD made it possible to link the data to the Forensic Toxicology Database of the National Board of Forensic Medicine.

The data for cases which did not undergo a medico-legal autopsy (hospital-autopsied and non-autopsied cases) were obtained from the Cause-of-Death Register (Centre for Epidemiology, National Board of Health and Welfare). The data included information on, e.g. the unique code number, sex, age, underlying and contributing causes of death (N-code), external causes of death (E-code), and type of autopsy where applicable.

The unique code numbers were also used to link our data to the in-patient register of the National Board of Health and Welfare, revealing all episodes of hospitalization, diagnoses for medical conditions treated and for other diagnoses that were of importance for treatment/hospitalization for a period of 3 years prior to death.

Cases with alcohol-related underlying and contributing causes of death (ARCD), and/or with alcohol-related in-patient diagnoses (ARID) were identified. The following diagnoses (ICD-9) were regarded to be related to alcohol: alcoholic psychosis (291A–291F, 291W, 291X); chronic alcoholism (303); alcohol abuse (305); alcoholic polyneuropathy (357F); alcoholic cardiomyopathy (425F); oesophageal varices (456A, 456B); alcoholic gastritis (535D); liver cirrhosis (571A–571D); pancreatitis (577A–577B); and alcohol intoxication (980A– 980C, 980W, 980X).

Alcohol was regarded to be associated with the death: if there was any indication in the information from the FMD that the deceased was a ‘known alcoholic’ or ever had any health or other problem associated with drinking; if they had ARCD; if they had ARID during a period of 3 years prior to death; if the case tested positive for blood alcohol (Sjögren et al., 2000aGo,bGo).

Potential years of life lost/gained
Potential years of life lost/gained up to the age of 69 years were estimated according to the method described by Hakulinen and Teppo (1976) which was also used by English et al. (1995).

All-cause mortality analysis
The role of alcohol for all-cause mortality between the ages of 20–39 years was worked out by using current alcohol consumption in Sweden together with relative risk of death for different levels of alcohol consumption estimated by Andréasson et al. (1991) in their nationwide study of a 20-year follow-up of 50 465 Swedish conscripts up to 1988. Current alcohol consumption in Sweden was estimated from recent Swedish surveys (see above). The methodology used for assessing AAFs by using the relative risk estimates (from the report by Andréasson et al., 1991) and the current Swedish alcohol consumption was as described above for natural deaths.


    RESULTS
 TOP
 FOOTNOTES
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 REFERENCES
 
Natural deaths
About 1200 natural deaths were estimated to be ‘caused’ by alcohol in Sweden annually, i.e. about 1.4% of all natural deaths (Table 1Go). The majority of deaths ‘caused’ by alcohol were due to conditions that are alcohol-related by definition (ARD) and due to malignant neoplasms. The most common cause of death in the group of ARD conditions was alcohol-dependence syndrome followed by alcoholic liver cirrhosis, and alcoholic liver damage. The most common causes of death due to malignant neoplasms associated with alcohol, were liver cancer followed by oesophageal cancer, and female breast cancer. Alcohol caused almost one-third of deaths due to laryngeal cancer and oesophageal cancer; these cancer forms have the highest risks with alcohol intake. The majority of deaths that can be hypothetically ‘prevented’ by alcohol are due to ischaemic heart disease and stroke. Alcohol involvement in malignant neoplasms was four times more common in males than in females. Deaths due to ARD conditions were four times more common in males than among females. Alcohol involvement in deaths due to gastrointestinal conditions was also almost twice as common in males than females.


View this table:
[in this window]
[in a new window]
 
Table 1. Natural deaths: total number of deaths (TD), alcohol-attributable fractions AAFs, deaths ‘caused’ (+) or ‘prevented’ (–) by alcohol consumption (alcohol-related deaths, AD) in males and females in Sweden from 1992 through 1996 (reference group is abstinence)
 
Alcohol ‘caused’ 3.4% of natural deaths below the age of 59 years, and in the age group above 60 years, almost 4.8% of natural deaths were ‘prevented’ by alcohol. However, in all ages, since the number of deaths ‘prevented’ by alcohol is much greater than the number ‘caused’ by alcohol, the net effect is that almost 4.1% of deaths are ‘prevented’ by alcohol.

Unnatural deaths
These are shown in Table 2Go. Just over 28% of the unnatural deaths were associated with alcohol; more than twice as common in deaths of males (35%) than of females (16%). When only autopsied cases or only blood-tested cases were taken as the denominators, 38% and 44% of the deaths were associated with alcohol, respectively. Alcohol involvement was also twice as common in intentional deaths (36%) than in unintentional deaths (18%). The intoxication group (78%) had the highest fraction of deaths that could be associated with alcohol, followed by the undetermined group (62%), homicide (49%), fire (41%), suicide (35%), asphyxia (29%), traffic (18%), and fall (9%) groups.


View this table:
[in this window]
[in a new window]
 
Table 2. Unnatural deaths: alcohol-related deaths (AD) and alcohol-attributable fractions (AAF) in males and females in Sweden from 1992 through 1996
 
Natural and unnatural deaths
About 3.5% of deaths were ‘caused’ by alcohol; alcohol involvement was more than twice as common in deaths of males (4.8%) than in those of females (2.0%) (Table 3Go). However, 5.2% of deaths were ‘prevented’ by alcohol. The net effect of alcohol was that 1.8% of deaths were ‘prevented’ (Tables 3Go).


View this table:
[in this window]
[in a new window]
 
Table 3. Natural and unnatural deaths: total number of deaths (TD) in Sweden (1992–1996), alcohol-attributable fraction (AAF), alcohol-related deaths (AD) in males and females in all age groups
 
When only premature mortality (i.e. <=69 years of age) is considered, about 10.9% of the deaths were ‘caused’ by alcohol and 3.6% of the deaths were ‘prevented’ by alcohol (Table 4Go). Thus alcohol had a net negative effect on premature mortality in that it ‘caused’ about 7% of the deaths. The net harmful effect was significantly greater in males (8.6%) than in females (4.7%) (Table 4Go).


View this table:
[in this window]
[in a new window]
 
Table 4. Natural and unnatural deaths: total number of deaths ‘caused’ (+) and ‘prevented’ (–) by alcohol by age and sex in Sweden in 1992–1996
 
In those aged below 50 years, alcohol ‘caused’ one-quarter of the deaths (Table 4Go). Alcohol involvement increased with age, to reach a peak in the age group 20–29 years and then decreased, and in the age group 70+ years, alcohol ‘caused’ only 1% of the deaths. The beneficial effect of alcohol became apparent in the age group 30–39 years and increased with age (Fig. 1Go). The net effect of alcohol was harmful up to the age of 59 years (Fig. 1Go).



View larger version (10K):
[in this window]
[in a new window]
 
Fig 1. Natural and unnatural deaths: fraction of deaths (alcohol-attributable fraction, AAF) ‘caused’ (positive AAF) or ‘prevented’ (negative AAF), and the net effect of alcohol vs age for all deaths in Sweden.

 
Person years of life lost/gained
More person years of life were lost due to conditions directly caused by alcohol, such as alcohol dependence, than saved by cardio/cerebrovascular diseases (Table 5Go). Almost 2.5 times more person years of life were net-lost due to unnatural deaths than natural deaths. Drinking alcohol saved person years of life due to ischaemic heart disease and stroke, in other words those who died due to ischaemic heart disease and stroke would have died earlier if all were abstainers. The net effect for all causes of mortality was that more person years of life were lost than were saved by alcohol. In males, more than four times more person years of life were net-lost, compared to females. Alcohol accounted for about 10% of person years of life lost due to all causes of deaths.


View this table:
[in this window]
[in a new window]
 
Table 5. Person years of life (PYL) lost (+) or saved (–) by alcohol consumption for males and females (<=69 years) in Sweden
 
All-cause mortality analysis
Alcohol was involved in a higher proportion of deaths of males, than of females, aged 20–39 years (Table 6Go). Alcohol involvement in the age group 20–29 years was higher than in the age group 30–39 years. About 16% of the deaths between 20–29 years were associated with alcohol (Table 6Go). When all-cause mortality analysis was compared to cause-specific mortality analysis, it was found that the latter analysis showed a higher alcohol involvement than the former analysis; this difference was more apparent in males, than in females (Table 6Go).


View this table:
[in this window]
[in a new window]
 
Table 6. Alcohol-related mortality in Sweden worked by two approaches: total deaths (TD) in Sweden (1992–1996), alcohol-attributable fraction (AAF), and alcohol-related deaths (AD) in males and females
 

    DISCUSSION
 TOP
 FOOTNOTES
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 REFERENCES
 
Alcohol as a causative/preventive factor
Terms such as deaths associated with alcohol, deaths attributed to alcohol (thereby AAF), role of alcohol, ‘alcohol involvement’, alcohol-related mortality, deaths ‘caused’ by alcohol, and deaths ‘prevented’ by alcohol, are used in the present report, and imply that alcohol is a component that is present in the events/diseases that lead to death. According to the theory of modified determinism by Rothman (1986), "a cause [or in the present study, a ‘protection’] is any act, event or state of nature which initiates or permits, alone or in conjunction with other causes, a sequence of events resulting in an effect". Alcohol is one of the components of a cluster of causal/protective factors that together are sufficient to produce a condition. The issue of interpretation of association, as an indication of causation, has been discussed by Holman and Armstrong (1992) and MacMahon (1992). This latter author argued that all differences in disease rates between users and non-users of alcohol do not necessarily have to be causal in nature. On the basis of the above discussion, taken together with the fact that the protective effect of alcohol on coronary heart disease and stroke is still controversial (Filmore, 2000Go), the terms ‘caused’ and ‘prevented’ are written with quotation marks in the present report.

Cause-specific alcohol-related mortality
The present study is unique, because it is the first to attempt to quantify alcohol-related mortality in Sweden. Quantification of alcohol-related mortality is difficult, especially the role of alcohol in diseases which are associated with long-term use/misuse of alcohol.

In general, comparison of our results with those from other studies is difficult due to differences in the categorization of deaths, in criteria used to define alcohol involvement, in survey periods, in selection of study populations, in the extent of blood-alcohol testing, in geographical differences in alcohol consumption, in drinking patterns in the population, and also in restrictions in availability of alcohol in the population.

The present study showed that 3.5% of deaths in all ages and one-tenth of deaths in those aged <=69 years were ‘caused’ by alcohol in Sweden. This is in line with recent studies showing that alcohol contributed to 3.4% of deaths in Canada (Single et al., 1999Go) and 3.3% of deaths in Italy (Cipriani et al., 1998Go). The study by Scragg (1995), which based its quantification of alcohol-related mortality on the meta-analysis by Holman et al. (1990), showed that 5.3% of all deaths in New Zealand were ‘caused’ by alcohol. Others have shown alcohol involvement in 3.3% of deaths in Minnesota, USA (Parker et al., 1987Go), 4% of deaths in New Hampshire, USA (Gorsky et al., 1988Go), 4.9–5% of deaths in the USA (Shultz et al., 1990Go, Stinson and DeBakey, 1992Go), 6.2% of deaths in California, USA (Sutocky et al., 1993Go), 4.5% of deaths in Wisconsin, USA (Peterson et al., 1990Go), 6.1% of deaths in Spain (Yañez et al., 1993Go), 9% of deaths in France (Pignon and Hill, 1991Go), and 6% of premature deaths in Canada (Ouellet et al., 1979Go). A more recent study, using the Finnish official death register, found that 6% of all deaths in Finland were related to alcohol (Mäkelä, 1998Go).

We found that 5.2% of the deaths in Sweden were ‘prevented’ by alcohol. This is slightly lower (6.9%) than that found by Scragg (1995) in the study from New Zealand, but higher than a recent study from Canada (Single et al., 1999Go) showing that 3.8% of deaths were ‘prevented’. In line with the study by Scragg (1995), we found that, in those aged <=69 years, alcohol drinking was estimated to have a net harmful effect, but in all ages, a net protective effect was found. The present findings are in line with estimates of alcohol-related global mortality, where it was found that, in established market economies, the protective effects of alcohol probably avert as many deaths as its harmful effects cause, and that a net harmful effect is apparent in those below 69 years of age (Murray and Lopez, 1997Go). Further support for the present finding of a net protective effect of alcohol on mortality is provided by a recent study by Single et al. (1999) showing that 0.4% of deaths in Canada are in net averted by low-level alcohol use.

Reliability of data
The reliability of the findings of alcohol involvement was greater for unnatural, than for natural, deaths. The assessment of the role of alcohol in unnatural deaths was based on a well-defined sample of all unnatural deaths from the whole of Sweden, of which 73% of the cases were autopsied and 57% were both autopsied and tested for blood alcohol. Furthermore, other well-defined criteria were used to identify unnatural deaths associated with alcohol (Sjögren et al., 2000bGo).

Limitations in the assessment of alcohol involvement in natural deaths. There are a number of limitations in the methodology used to estimate involvement of alcohol in natural deaths and these need to be borne in mind when evaluating the present results. Furthermore, it should be kept in mind that the relation between alcohol consumption and coronary heart disease and stroke is still controversial. The reliability of the results on natural deaths is dependent on three main factors: (1) the reliability of the data presented in the studies included in the meta-analysis that were used to estimate the pooled relative risks; (2) the reliability of the alcohol consumption data for Sweden that were used for calculation of the AAFs; (3) the reliability of the Cause of Death Register for Sweden used to assess alcohol-related mortality.

Reliability of data from studies included in the meta-analysis. A number of studies have shown a reduction in coronary heart disease mortality with moderate alcohol consumption (e.g. Doll et al., 1994; Goldberg et al., 1994; Rimm et al., 1996; Rehm et al., 1997; Thun et al., 1997). However, a recent study that adjusted for socioeconomic factors found that there was no clear relation between alcohol consumption and coronary heart disease mortality (Hart et al., 1999Go). As for stroke, the risk for haemorrhagic stroke is thought to increase with alcohol consumption (Stampfer et al., 1988Go; Iso et al., 1995Go), whereas the risk for ischaemic stroke has been shown to be reduced (Palomäki and Kaste, 1993Go), increased (Iso et al., 1995Go), or unaffected (Stampfer et al., 1988Go) for moderate drinkers. The risk of mortality from all types of stroke has shown variable associations with alcohol consumption: increased risk with amount of alcohol consumed (Doll et al., 1994Go; Iso et al., 1995Go; Hart et al., 1999Go), and a J-shaped relation (Gill et al., 1991Go; Iso et al., 1995Go). In view of these contradictory results for coronary heart disease and stroke, the present results on the preventive effects of alcohol are controversial.

A number of assumptions were made in the attributable risk methods used in the meta-analysis: findings from non-Swedish populations can be extrapolated to the Swedish population; relative risks did not vary with age; relative risks did not differ between males and females; relative risks for morbidity would also apply to mortality. The assumption that findings from non-Swedish populations can be extrapolated to the Swedish encompasses a large variety of factors such as patterns of alcohol consumption. However, it needs to be kept in mind that AAF for unnatural deaths were based on medico-legal autopsy cases from Sweden, and that 80% of the deaths ‘caused’ by alcohol were due to unnatural causes and conditions that are related to alcohol by definition such as alcohol dependence syndrome, suggesting that the reliability of the present results on harmful effects of alcohol is not as questionable as it may seem.

Reliability of alcohol consumption in Sweden. The other important factor is the reliability of the alcohol consumption data for Sweden used in the present study to calculate the AAFs. The present data were obtained from household surveys when randomly selected subjects from the whole of Sweden were visited at home by professional interviewers who used a structured questionnaire. The drop-out rate was estimated to be 25–30%. It is possible that the high alcohol consumers were underestimated in our data, as very high consumers may be unwilling to take part in the survey. There is also some evidence that heavy drinkers underestimate their alcohol intake more than light drinkers (Poikolainen, 1985Go). Thus if there is an underestimation of heavy drinkers, causes of mortality (for example, the neoplasms) that have an increasing risk with increasing alcohol consumption, are also underestimated in our results.

Reliability of the official Swedish mortality data. The official Swedish mortality statistics are based on the underlying cause of death indicated on death certificates issued for each deceased person. Ågren and Jakobsson (1987) studied the mortality and the cause of death in a cohort of male alcoholics in Stockholm and found that the underlying cause of death was incorrect in 21.8% of the cases. They found that, in the alcohol-related conditions, cardiomyopathy was underestimated as an underlying cause of death; in the official statistics 0.5% of the deaths were due to cardiomyopathy and after validation the proportion rose to 6%. Assuming that the underestimation of alcohol cardiomyopathy deaths in official Swedish mortality statistics also applies for the present period studied (1992–1996), the cardiomyopathy deaths due to alcohol would increase from 157 to 1884, which indicates that the deaths due to conditions directly ‘caused’ by alcohol as a group should be 57% higher (i.e. 4770 deaths; see Table 1Go) in the present study. Others in Sweden have also found that deaths due to conditions directly ‘caused’ by alcohol are underestimated (Petersson et al., 1982Go; Romelsjö et al., 1993Go).

Person years of life lost/gained
Person years of life lost is useful in that it gives higher weight to deaths at younger ages and thereby is an indicator of premature death. There are, however, statistical limitations when person years of life are estimated and this estimation must be taken to be just an approximation. More person years of life (below 70 years) were net lost than were saved by alcohol, suggesting that alcohol has a negative effect on premature mortality. Alcohol accounted for just over 10% of person years of life lost. This is somewhat higher than that found by Single et al. (1999) for Canada (6.1%).

All-cause mortality analysis
There are many reasons to be cautious about generalizing results from one country to another, particularly when analysing all-cause mortality. The death rate for a specific cause reflects the distribution of all risk factors in the population, not just alcohol, including e.g. specific genetic factors, diet, ethnicity, and smoking, which may themselves be related to alcohol consumption in different populations. In view of this, the present assessment of all-cause alcohol-related mortality for Sweden was based on a Swedish study on a nationwide cohort of conscripts. Since the conscripts were followed up between the ages 18 and 39 years, the all-cause alcohol-related mortality was estimated only for the age interval 20–39 years. It was found that 16% of the deaths were associated with alcohol.

There are both strengths and weaknesses in the two different approaches (cause-specific and all-cause) of estimating alcohol-related mortality. The strength in assessment of alcohol association in all-cause mortality is that the risk ratios were based on a nationwide cohort from Sweden. The weaknesses with this all-cause mortality approach are: (1) the cohort was followed up to the age of 39 years which limits the assessment to the age interval 18–39 years and the risks cannot be applied to older groups; (2) the cohort was followed between 1969 and 1988 and the spectrum of different causes of deaths and also the alcohol consumption patterns probably differed from that in the period investigated (1992–1996); (3) the risks were based on males only. The strength in cause-specific alcohol-related mortality was that reliable data for unnatural deaths for all of Sweden between 1992 and 1996 could be obtained. The weaknesses with the cause-specific approach for natural deaths are discussed above (reliability of data).

We found that the cause-specific mortality analysis showed a higher alcohol-related mortality than the all-cause mortality analysis. The difficulties and weaknesses in the two approaches probably explain this difference. However, keeping in mind the weaknesses in the two approaches, it is reasonable to assume that about 20% (range 16%–25%) of deaths between the ages of 20 and 39 years in Sweden today are associated with alcohol.

In conclusion, as far as premature mortality (<=69 years) is concerned, we found that deaths ‘caused’ by alcohol more than balance the deaths that may be expected to be ‘prevented’ by alcohol, and more person years of life are lost than averted due to alcohol, indicating that a public health recommendation that emphasizes the positive effects of alcohol and thereby promoting increase in per capita alcohol consumption, would do more harm than good in Sweden. This is in line with the general opinion of other scientists in this field (e.g. Marmot and Brunner, 1991; Addiction Research Foundation, 1993; World Health Organization, 1994; Anderson, 1995).


    FOOTNOTES
 TOP
 FOOTNOTES
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 REFERENCES
 
* Author to whom correspondence should be addressed. Back


    REFERENCES
 TOP
 FOOTNOTES
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 REFERENCES
 
Adami, H. O., McLaughlin, J. K., Hsing, A. W., Wolk, A., Ekbom, A., Holmberg, L. and Persson, I. (1992) Alcoholism and cancer risk: a population-based cohort study. Cancer Causes Control 3, 419–425.[ISI][Medline]

Addiction Research Foundation (1993) A Joint Policy based on the International Symposium on Moderate Drinking and Health. Addiction Research Foundation, Toronto.

Ågren, G. and Jakobsson, S. W. (1987) Validation of diagnoses on death certificates for male alcoholics in Stockholm. Forensic Science International 33, 231–241.[ISI][Medline]

Anderson, P. (1995) Alcohol consumption and all-cause mortality. Addiction 90, 481–484.[ISI][Medline]

Andréasson, S., Romelsjö, A. and Allebeck, P. (1991) Alcohol, social factors and mortality among young men. British Journal of Addiction 86, 877–887.[ISI][Medline]

Berglund, M., Krantz. P. and Lundqvist, G. (1987) Suicide in alcoholism. A prospective study of 55 cases with autopsy findings. Acta Psychiatrica Scandinavica 76, 381–385.[ISI][Medline]

Björnstig, U. L., Bylund, P. O., Lekander, T. and Brorsson, B. (1985) Motorcycle fatalities in Sweden. Acta Chirurgica Scandinavica 151, 577–581.[ISI][Medline]

Cipriani, F., Balzi, D., Sorso, B. and Buiatti, E. (1998) Alcohol-related mortality in Italy. Public Health 112, 183–188.[ISI][Medline]

Doll, R., Peto, R., Hall, E., Wheatley, K. and Gray, R. (1994) Mortality in relation to consumption of alcohol: 13 years' observations on male British doctors. British Medical Journal 309, 911–918.[Abstract/Free Full Text]

English, D. R., Holman, C. D. J., Milne, E., Winter, M. G., Hulse, G. K., Bower, C. I., Codde, J. P., Corti, B., Dawes, V., de Klerk, N., Knuiman, M. W., Kurinczuk, J. J., Lewin, G. F. and Ryan, G. A. (1995) The Quantification of Drug Caused Morbidity and Mortality in Australia. Commonwealth Department of Human Services and Health, Canberra.

Filmore, K. M. (2000) Is alcohol really good for the heart? Addiction 95, 173–174.[ISI][Medline]

Gill, J. S., Shipley, M. J., Tsementzis, S. A., Hornby, R. S., Gill, S. K., Hitchcock, E. R. and Beevers, D. G. (1991) Alcohol consumption — a risk factor for hemorrhagic and non-hemorrhagic stroke. American Journal of Medicine 90, 489–497.[ISI][Medline]

Goldberg, R. J., Burchfiel, C. M., Reed, D. M., Wergowske, G. and Chiu, D. (1994) A prospective study of the health effects of alcohol consumption in middle-aged and elderly men: The Honolulu Heart Program. Circulation 89, 651–659.[Abstract]

Gorsky, R. D., Schwartz, E. and Dennis, D. (1988) The mortality, morbidity, and economic costs of alcohol abuse in New Hampshire. Preventative Medicine 17, 736–745.[ISI][Medline]

Hakulinen, T. and Teppo, L. (1976) The increase in working years due to elimination of cancer as a cause of death. International Journal of Cancer 17, 429–435.[ISI]

Hansagi, H., Romelsjö, A., Gerhardsson de Verdier, M., Andréasson, S. and Leifman, H. (1995) Alcohol consumption and stroke mortality. 20-year follow-up of 15,077 men and women. Stroke 26, 1768– 1773.[Abstract/Free Full Text]

Hansson, L. E., Baron, J., Nygren, O., Bergström, R., Wolk, A. and Adami, H. O. (1994) Tobacco, alcohol and the risk of gastric cancer. A population-based case-control study in Sweden. International Journal of Cancer 57, 26–31.[ISI]

Harmsen, P., Rosengren, A., Tsipogianni, A. and Wilhelmsen, L. (1990) Risk factors for stroke in middle-aged men in Göteborg, Sweden. Stroke 21, 223–229.[Abstract]

Hart, C. L., Smith, G. D., Hole, D. J. and Hawthorne, V. M. (1999) Alcohol consumption and mortality from all causes, coronary heart disease, and stroke: results from a prospective cohort study of Scottish men with 21 years of follow up. British Medical Journal 318, 1725–1729.[Abstract/Free Full Text]

Holman, C. D. J. and Armstrong, B. K. (1992) The quantification of alcohol-caused morbidity and mortality in Australia: the authors respond [comment]. Medical Journal of Australia 157, 560–561.[ISI][Medline]

Holman, C. D. and English, D. R. (1995) An improved aetiologic fraction for alcohol-caused mortality. Australian Journal of Public Health 19, 138–141.[ISI][Medline]

Holman, C. D. J., Armstrong, B. K., Arias, L. N., Martin, C. A. et al. (1990) The quantification of drug caused morbidity and mortality in Australia, 1988. Commonwealth Department of Community Services and Health, Canberra.

Holman, C. D., English, D. R., Milne, E. and Winter, M. G. (1996) Meta-analysis of alcohol and all-cause mortality: a validation of NHMRC recommendations. Medical Journal of Australia 164, 141–145.[ISI][Medline]

Holmberg, L., Baron, J. A., Byers, T., Wolk, A., Ohlander, E. M., Zack, M. and Adami, H. O. (1995) Alcohol intake and breast cancer risk: effect of exposure from 15 years of age. Cancer Epidemiology and Biomarkers Preview 4, 843–847.[Abstract]

Iso, H., Kitamura, A., Shimamoto, T., Sankai, T., Naito, Y., Sato, S. et al. (1995) Alcohol intake and the risk of cardiovascular disease in middle-aged Japanese men. Stroke 26, 767–773.[Abstract/Free Full Text]

Karlsson, T., Ormstad, K. and Rajs, J. (1988) Patterns in sharp force fatalities — a comprehensive forensic medical study: Part 2. Suicidal sharp force injury in the Stockholm area 1972–1984. Journal of Forensic Science 33, 448–461.[ISI]

MacMahon, B. (1992) The quantification of alcohol-caused morbidity and mortality in Australia: a critique. Medical Journal of Australia 157, 557–560.[ISI][Medline]

Mäkelä, P. (1998) Alcohol-related mortality by age and sex and its impact on life expectancy. European Journal of Public Health 8, 43–51.[Abstract]

Mäkelä, P., Valkonen, T. and Poikolainen, K. (1997) Estimated numbers of deaths from coronary heart disease "caused" and "prevented" by alcohol: an example from Finland. Journal of Studies on Alcohol 58, 455–463.[ISI][Medline]

Marmot, M. and Brunner, E. (1991) Alcohol and cardiovascular disease: The status of the U shaped curve. British Medical Journal 303, 565–568.[ISI][Medline]

Messner, T. and Petersson, B. (1996) Alcohol consumption and ischemic heart disease mortality in Sweden. Scandinavian Journal of Social Medicine 24, 107–113.[ISI][Medline]

Murray, C. J. L. and Lopez, A. D. (1997) Global mortality, disability, and the contribution of risk factors: global burden of disease study. Lancet 349, 1436–1442.[ISI][Medline]

National Board of Health and Welfare (1994) Causes of death. Statistics — Health and Diseases. The National Board of Health and Welfare. Centre for Epidemiology. Official Statistics of Sweden, Stockholm, Sweden.

National Board of Health and Welfare (1995) Causes of death. Statistics — Health and Diseases. The National Board of Health and Welfare. Centre for Epidemiology. Official Statistics of Sweden, Stockholm, Sweden.

National Board of Health and Welfare (1996) Causes of death. Statistics — Health and Diseases. The National Board of Health and Welfare. Centre for Epidemiology. Official Statistics of Sweden, Stockholm, Sweden.

Norström, T. (1988) Alcohol and suicide in Scandinavia. British Journal of Addiction 83, 553–559.[ISI][Medline]

Ormstad, K., Karlsson, T., Enkler, L., Law, B. and Rajs, J. (1986) Patterns in sharp force fatalities — a comprehensive forensic medical study. Journal of Forensic Science 31, 529–542.[ISI]

Örnehult, L. and Eriksson, A. (1987) Accidental firearm fatalities during hunting. American Journal of Forensic Medicine and Pathology 8, 112–119.[ISI][Medline]

Öström, M., Sjögren, H. and Eriksson, A. (1995) Role of alcohol in traffic crashes involving women: passenger car fatalities in northern Sweden. Journal of Studies on Alcohol 56, 506–512.[ISI][Medline]

Ouellet, B. L., Romeder, J.-M. and Lance, J.-M.(1979) Premature mortality attributable to smoking and hazardous drinking in Canada. American Journal of Epidemiology 109, 451–463.[Abstract]

Palomäki, H. and Kaste, M. (1993) Regular light-to-moderate intake of alcohol and the risk of ischemic stroke. Is there a beneficial effect? Stroke 24, 1828–1832.[Abstract]

Parker, D. L., Shultz, J. M., Gertz, L., Berkelman, R. and Remington, P. L. (1987) The social and economic costs of alcohol abuse in Minnesota, 1983. American Journal of Public Health 77, 982–986.[Abstract]

Peterson, D. E., Akgulian, N. A. and Remington, P. L. (1990) Alcohol-related disease in Wisconsin, 1988, Wisconsin Medical Journal 89, 232–236.[Medline]

Petersson, B., Krantz, P., Kristensson, H., Trell, E. and Sternby, N. H. (1982) Alcohol-related death: A major contributor to mortality in urban middle-aged men. Lancet ii, 1088–1090.

Pignon, J. P. and Hill, C. (1991) Estimation of alcohol-related deaths in France in 1985. Gastroenterology: Clinical and Biological 15, 51–56.

Poikolainen, K. (1985) Underestimation of recalled alcohol intake in relation to actual consumption. British Journal of Addiction 80, 215–216.[ISI][Medline]

Rehm, J. T., Bondy, S. J., Sempos, C. T. and Vuong, C. V. (1997) Alcohol consumption and coronary heart disease morbidity and mortality. American Journal of Epidemiology 146, 495–501.[Abstract]

Rimm, E. B., Klatsky, A., Grobbee, D. and Stampfer, M. J. (1996) Review of moderate alcohol consumption and reduced risk of coronary heart disease: is the effect due to beer, wine, or spirits? British Medical Journal 312, 731–736.[Abstract/Free Full Text]

Romelsjö, A., Karlsson, G., Henningsohn, L. and Jakobsson, S. W. (1993) The prevalence of alcohol-related mortality in both sexes: variation between indicators, Stockholm, 1987. American Journal of Public Health 83, 838–844.[Abstract]

Rothman, K. J. (1986) Modern Epidemiology. Little, Brown. Boston, MA.

Scragg, R. (1995) A quantification of alcohol-related mortality in New Zealand. Australia and New Zealand Journal of Medicine 25, 5–11.[ISI]

Shultz, J. M., Rice, D. P. and Parker, D. L. (1990) Alcohol-related mortality and years of potential life lost — United States, 1987. Morbidity and Mortality Weekly Report 39, 173–178.[Medline]

Single, E., Robson, L., Rehm, J. and Xiaodi, X. (1999) Morbidity and mortality attributable to alcohol, tobacco, and illicit drug use in Canada. American Journal of Public Health 89, 385– 390.[Abstract]

Sjögren, H., Björnstig, U., Eriksson, A., Öhman, U. and Solarz, A. (1997) Drug and alcohol use among injured motor vehicle drivers in Sweden. Alcoholism: Clinical and Experimental Research 21, 968–973.[ISI][Medline]

Sjögren, H., Eriksson, A. and Ahlm, K. (2000a) Alcohol and unnatural deaths in Sweden: a medico-legal autopsy study. Journal of Studies in Alcohol 61, 507–514.[ISI][Medline]

Sjögren, H. Eriksson, A. and Ahlm, K. (2000b) Role of alcohol in unnatural deaths in Sweden: a study of all deaths in Sweden. Alcoholism: Clinical and Experimental Research 24, 1050–1056.[ISI][Medline]

Stampfer, M. J., Colditz, G. A., Willet, W. C., Speizer, F. E. and Hennekens, C. H. (1988) A prospective study of moderate alcohol consumption and the risk of coronary disease and stroke in women. New England Journal of Medicine 319, 267–273.[Abstract]

Statistics Sweden (1992) Causes of death. Official Statistics of Sweden. National Central Bureau of Statistics. Statistics Sweden, Stockholm, Sweden.

Statistics Sweden (1993) Causes of death. Official Statistics of Sweden. National Central Bureau of Statistics. Statistics Sweden, Stockholm, Sweden.

Stinson, F. S. and DeBakey, S. F. (1992) Alcohol-related mortality in the United States, 1979–1988. British Journal of Addiction 87, 777–783.[ISI][Medline]

Sutocky, J. W., Shultz, J. M. and Kizer, K. W. (1993) Alcohol-related mortality in California, 1980 to 1989. American Journal of Public Health 83, 817–823.[Abstract]

Thun, M. J., Peto, R., Lopez, A. D., Monaco, J. H., Henley, S. J., Heath, C. W. Jr and Doll, R. (1997) Alcohol consumption and mortality among middle-aged and elderly U.S. adults. New England Journal of Medicine 337, 1705–1714.[Abstract/Free Full Text]

World Health Organization (1994) Cardiovascular disease risk factors: new areas for research. WHO Technical Report Series 841. World Health Organization, Geneva.

Yañez, J. L., Del Rio, M. C. and Alvarez, F. J. (1993) Alcohol-related mortality in Spain. Alcoholism: Clinical and Experimental Research 17, 253–255.[ISI][Medline]





This Article
Abstract
FREE Full Text (PDF)
Alert me when this article is cited
Alert me if a correction is posted
Services
Email this article to a friend
Similar articles in this journal
Similar articles in ISI Web of Science
Similar articles in PubMed
Alert me to new issues of the journal
Add to My Personal Archive
Download to citation manager
Search for citing articles in:
ISI Web of Science (3)
Request Permissions
Google Scholar
Articles by Sjögren, H.
Articles by Ahlm, K.
PubMed
PubMed Citation
Articles by Sjögren, H.
Articles by Ahlm, K.