Methodological characteristics extracted included 1) the approaches used in the cost estimates, i.e. prevalence or incidence approach, 2) the methods used for estimating the cost of premature mortality, i.e. human capital or demographic approach, 3) the inclusion of the positive effects of alcohol drinking, i.e. using gross cost or net cost estimates, and 4) the discount rate used for adjusting future monetary values.

For each eligible study, the total estimated costs of alcohol and total cost estimates in term of Gross Domestic Product (GDP) or Gross National Product (GNP) were presented as originally published. To facilitate the comparison across different settings and years, total cost estimates were also presented in 2007 $US values by inflating the original cost to its 2007 value using country-specific GDP inflators21. The costs were then converted into US dollars using Purchasing Power Parities (PPP)21. To facilitate further comparison across studies in term of the magnitude of the total estimates relative to GDP, we calculated the total cost and direct cost as percentage of GDP (PPP) 2007 and cost per capita, using information from The World Economic Outlook Database by the International Monetary Fund (IMF)21.

In this study, cost components in each identified study were classified into three main categories, namely: direct costs, indirect costs, and intangible costs. Direct costs measure the value of resources used as a consequence of alcohol abuse. The direct cost in this study was further classified into 1) health care cost, 2) research and prevention costs, 3) costs of crime and law enforcement, 4) costs of property damage or loss, 5) administration costs, 6) costs of welfare assistance or social work. (These costs did not, however, include any welfare payments), 7) costs of alcohol beverage, and 8) other costs.

In contrast, indirect costs are those for which resources are lost without a direct payment actually being made. In this study, indirect costs have been classified into five categories as follows; 1) the cost of premature mortality, 2) the cost of reduced productivity, which includes both the cost of productivity loss due to absenteeism and that when the workforce comes to work (presenteeism), 3) the cost of incarceration, 4) the cost of loss of employment or early retirement, and 5) costs associated with crime i.e. time loss for victims due to crime.

The last cost category is referred to as intangible costs, which represented pain, suffering, and the deterioration of quality of life. This type of cost, when reduced or eliminated, does not yield resources that can be made available for other uses, and is less likely to be included in cost estimations22.

The type of alcohol attributed diseases included in the estimation can also be an important factor. In this study, diseases were classified according to the International Statistical Classification of Diseases and Related Health Problem 10^th Revision (ICD-10) into 10 groups as follows; A = Infectious diseases(A00-B99), B = Neoplasm(C00-D48), C = Endocrine diseases (E00-E99), D = Mental and behavioural disorders (F00-F99), E = Diseases of the nervous system(G00-G99), F = Diseases of the circulatory system (I00-I99), G = Diseases of the digestive system (K00-K93), H = Diseases of the skin (L00-L99), I = Conditions associated with pregnancy or child birth, and certain conditions in the perinatal period (O 00-O99, P00-P96), J = Injuries, poisoning, and other external causes (S00-T-98, V01-Y98).

The initial search strategy, conducted in August 2007, identified 318 potential relevant articles from MEDLINE databases. Of these, only 9 studies 8911121416181920 fulfilled the eligibility criteria, as indicated in table 1. Eleven additional eligible studies were identified through expert communication and bibliographies of reviewed studies46710131517232425 Upon completion of this strategy, 20 studies467891011121314151617181920232425 conducted in 12 developed countries (Australia, Canada, England and Wales, France, Germany, Japan, the Netherlands, New Zealand, Portugal, Scotland, Sweden and the United States) and 1 developing countries (Thailand) 25were included in the full review. It should be noted that the Australian study 6 contained cost estimations for 2 different years (1988 and 1992).

One non-English study with English summary and tables conducted in South Korea 5and one English article with only cost summary table conducted in Finland 26were also included in the partial review. Other 11 non-English studies conducted in Spain27, Italy28, Germany29, Argentina30, Chile31, Norway32, Switzerland33, France3435, Slovenia36, and Belgium37 were also identified by expert communications but were not included in this review. Search and retrieval processes are shown in figure 1.

The general characteristics of the studies included in this review have been summarized in table 2. It was found that all of the studies identified were prevalence-based studies using societal perspective. However, it should be noted that only external costs (costs that affect others than the consumers) were taken into account since private costs (costs that affect the consumers) were considered to be offset by benefit from the consumption. Nevertheless, a few studies 567included private cost (i.e. cost of alcoholic beverage) in their estimations.

Regarding the inclusion of the positive effects of alcohol drinking, only three studies 67 from Australia estimated the economic cost of alcohol in terms of net cost, while 17 studies 4567891013141516171920232425 calculated the gross cost. Two recent studies in Germany 12and Sweden 11 estimated total costs in term of both net cost and gross cost and found that the net cost estimation was approximately 12.5-14.5% lower than the gross cost estimation.

When the method employed in estimating the cost of premature mortality was reviewed, it was found that three studies, all conducted in Australia67, adopted the demographic approach, while the remaining 17 studies 458910111214151617181920232425employed the human capital approach. Of the 17 studies that employed the human capital approach 458910111214151617181920232425, the discount ranged from 3%1125 to10%8, with the most frequently cited value being 6%91014161718192023, complying with the 5 - 10% range recommended by WHO's guideline 22.

Table 3 gives an overview of the cost components included in all 22 reviewed studies. Both the direct costs and indirect costs were reported in all studies, while intangible costs were reported in only 5 46711.

As for direct costs, health care costs and the cost of crime and law enforcement were identified in all but 2 of the 22 studies, conducted in Korea 5and Germany12, in which the cost of crime and law enforcement was not taken into account. The cost of property damage or loss was taken into account in all studies except those from New Zealand8, Scotland19, Sweden 11and Germany12. The cost of research and prevention, and administrative costs were identified in 1291011121415161718192023 and 15 studies56101213141516171820232426, respectively. Very few studies included the cost of alcohol beverages 567or welfare assistance/social service costs 1114192426.

Regarding indirect costs, the costs of reduced productivity were calculated in all 22 studies. All of the studies, with one exception13, also incorporated the cost of premature mortality into their estimations. The cost of productivity loss due to incarceration and time loss for victims of crime was calculated in 1078910111617232426, and 7 studies4101116172324, respectively.

Only 5 studies included intangible costs in their estimation46711 Most of these studies employed the willingness-to-pay method38 in estimating intangible costs. The value of life was estimated at 2,000,000 Australian dollars (in 1996) in Australia7. Contrastingly, in the study conducted in England and Wales4, the intrinsic value lost for a fatal causality was estimated to be £783,000 (in 2002). It should be noted that in the Swedish study11 the intangible cost was measured in terms of Quality-Adjusted Life Years (QALYs) lost. In this study, the QALY lost due to mortality for consumers, friends, relatives and victims of crime was 121,791 and 145,356 QALY in net and gross estimation, respectively.

Diseases attributable to alcohol included in the estimation are displayed in table 4. The review showed some variations in the number of disease groups included in each study. As shown in table 4, only 5 studies included infectious diseases in their estimations, while neoplasm, mental disorders, diseases of the circulatory system, diseases of the digestive system, and injury, poisoning, and external causes were considered in all studies. Also, it should be noted that one study 13included only wholly alcohol-attributable diseases.

Cost estimations from each study are presented in table 5 in terms of total cost in the study year, cost as a percentage of GDP or GNP, and the share of the direct cost, indirect cost, and intangible cost, respectively. The share of the direct costs represented from 4% 25to 52% 13of the total cost, while the share of the indirect cost accounted for 23%7 to 96% 25 of the total cost. Where intangible cost was considered46711, its share ranged from 21%6 to 27%7 of the total cost. When comparing across cost categories, it was found that indirect costs represented the largest proportion of the total cost in 16 of the 22 studies 58910111214161718192023242526, as shown in table 5.

When looking at each of the direct cost components as shown in table 6, there was no general agreement across studies on what constitutes their largest share. For example, health care costs, which were included in all studies, made up 90.4% of the total direct cost in Japan14, but was only 6 to 7% of the total direct cost in Australia67. Besides the differences due to country-specific factors, this could also be explained by the fact that net cost instead of the gross cost was estimated in these studies67. Similarly, law enforcement and criminal justice costs accounted for 68 to 80% in New Zealand8, but less than 1% of the total direct cost in both Japan 14 and France9.

In the 18 studies 45679101314151617182023242526 where the cost of property damage or loss was taken into account, its share ranged from 0.3% 14 to 54%8. In addition, two studies 411 included the cost of stolen property in their estimations. In the Netherlands13, France9, and England and Wales4, where the costs associated with property damage accounted for the major share of the total direct cost, the estimates included the cost of victim support, the cost of medical expenses, production losses, and the cost of materials associated with road traffic accidents.

Where the cost of research and prevention and cost of administration were taken into account, its share was found to be less than 8% of the total direct cost. On the other hand, cost of welfare assistance or social work ranged widely from 0.04% in Japan14 to about 40% of the total direct cost in Sweden19. When the cost of alcohol beverage was taken into account 567, this component represented a sizable amount, ranging from 33%-63% to 75% of the total direct cost in Australia 67and South Korea 5, respectively. Regarding other direct costs, it was found that transfer costs including disability pensions, welfare payments, social security payments, and accident compensation were incorporated in several studies 8131426.

For the components of indirect cost, as shown in table 6, the cost associated with premature mortality played the largest part in contributing to the total indirect costs in just over half of the studies67911121819202526. In the remaining 9 studies458101416172324, the cost of reduced productivity accounted for the highest proportion of the total indirect cost. As for the cost of reduced productivity, seven studies 8131416171820 took into account both costs associated with absenteeism and situations where the workforce came to work, albeit with compromised productivity, with the impairment rate ranging from 20 to 25%. The cost of productivity losses due to incarceration and being a victim of crime was found to be relatively small, ranging from less than 1% to 17% of the total indirect cost.

Table 7 presents total costs in PPP US$ 2007, costs as % of GDP 2007, costs per capita, and total direct costs as % GDP in the 12 selected countries57891011121314152425. The studies selected in this table were the most recent for each country. As shown in the table, total cost as % GDP ranged from 0.6%13 to 5.44%8 while cost per capita ranged from 85.53 US$ PPP 24to 1,012.21 US$ PPP8. On the other hand, total direct cost as percentage of GDP ranged from 0.08%25 to 0.81%8. When only the 7 studies791014152425 in which 5 main costs components (health care cost, crime and law enforcement cost, property damage, premature mortality, and reduced productivity) were included, the total cost as % GDP ranged from 0.45% 24 to 2.11% 10 while cost per capita ranged from 85.53 US$ PPP 24to 850.86 US$ PPP10.