International Comparisons of Consumption

Kim Bloomfield, Dr.P.H., Tim Stockwell, Ph.D., Gerhard Gmel, Ph.D., and Nina Rehn, M.A. Pol.Sc.

International comparisons of alcohol consumption and its consequences can serve multiple purposes. For example, despite differences among countries in drinking cultures, drink sizes and strengths, and methods of measuring alcohol consumption, international survey research has provided a substantial amount of information on the rates of abstinence or current drinking, the frequency of drinking or , and the mean consumption among both adults and youths in many countries. Other studies using aggregate-level data have analyzed per capita alcohol consumption in various countries. These studies can be used to relate per capita consumption to certain alcohol-related outcomes and to evaluate changes of both consumption and different outcomes within a country or across countries over time. Some problems associated with international research, however, such as issues of comparability of surveys, still need to be resolved. KEY WORDS: international AODR (alcohol and other related) problems; international differences; cultural patterns of drinking; research quality; alcohol quantity–frequency methods; measure of AOD (alcohol and other drug) volume and strength; cross- sectional study; gender differences; AOD abstinence; aggregate-level statistical data

esearchers in numerous coun- determine how variations in social, tries have conducted analyses cultural, political, environmental, and KIM BLOOMFIELD, DR.P.H., is an associ- of alcohol consumption and genetic factors can influence drinking ate professor in the Department of Health R Promotion Research, University of general population surveys to ascertain behavior. For example, in the case of the level and consequences of alcohol research on gender differences in Southern Denmark, Esbjerg, Denmark. use. In recent years, investigators also alcohol use, international compar- TIM STOCKWELL, PH.D., is a professor have made attempts to compare drink- isons could help distinguish which and director of the National Drug Re- ing rates and other drinking variables differences in men’s and women’s drink- search Institute, Curtin University, Perth, across different countries. One reason ing behavior can be attributed to bio- Australia. for researching across national borders logical differences and which to socio- is the need for descriptive epidemiology cultural factors (Wilsnack et al. 2000). GERHARD GMEL, PH.D., is co-director (Room 1988). For example, national Epidemiologic research into the of research at the Swiss Institute for governments often want to know how underlying mechanisms (i.e., the the Prevention of Alcohol and Drug their countries measure up against oth- etiology) of any disorder generally Problems, Lausanne, Switzerland. ers in per capita consumption or in addresses two questions (Rose 1985): other comparative rankings of alcohol NINA REHN, M.A. POL. SC., is technical use. Another reason for comparative •What are the causes of individual officer, Management of Substance research is the desire to further theo- cases of the disorder (e.g., )? Dependence at the World Health Organ- retical knowledge; social scientists ization, Geneva, Switzerland. often employ comparative designs to • What factors, such as sociocultural develop or test theories. In the case of or political influences, contribute to The National Drug Research Institute alcohol research, comparisons among the incidence of the condition in an is funded by the Australian National different countries can help researchers entire population? Drug Strategy.

Vol. 27, No. 1, 2003 95 The type of question to be answered General Methodological More recent comparative research, determines the level at which Issues in International however, has found that, especially in researchers compare data in interna­ Survey Research , the previous wet/dry division tional research. To answer the first seems to be disappearing and a question, investigators would mainly homogenization of consumption rates use individual-level studies (e.g., deter- Differences in Drinking Cultures and beverage preferences is increasingly mine the level of alcohol consump­ Throughout the world, numerous evident. For example, both Allamani tion in individual drinkers within a different drinking cultures and atti­ and colleagues (2000) and Leifman country) to address within-popula­ tudes toward alcohol exist. A more (2001) have reported that con­ tion variability. To answer the second theoretical literature, which is some- sumption has decreased in the tradi­ question, it may be more useful to what separate from research compar­ tionally wet Mediterranean countries conduct aggregate-level studies that ing actual survey rates of alcohol con­ and that overall alcohol consumption determine overall alcohol consump­ sumption across countries, has been has increased in the northern Euro­ tion in a population (e.g., per capita devoted to describing these differing pean countries. Room and Mäkelä consumption) to model between- drinking cultures. This research has (2000) also have reconsidered the population variability. Such aggregate- had various historical traditions simple wet/dry dichotomy and have level analyses are particularly useful (Room 1988; Room and Mäkelä instead proposed a new typology that when some societal or environmental 2000), the most recent of which has considers a variety of drinking behav­ factors are constant or almost constant focused primarily on differences iors, such as the regularity of drinking within a population. For example, between North American and Euro­ and the extent of drunkenness. Such a Rose (1985) notes the hypothetical pean countries, dividing them into typology may better fit the distinctions example of a population in which groups with either high or low per in drinking cultures that are emerging every person smokes 20 cigarettes per capita alcohol consumption, or so- today. Nevertheless, the wet/dry day. An epidemiologist who uses dichotomy has represented a scale of called wet or dry cultures. These two individual-level data to study risk fac­ extremes on which to measure drink­ categories also are commonly associat­ tors for lung cancer in such a society ing cultures and around which a fair could identify factors that vary ed with various correlates, such as a amount of past research literature has between people with and without the history of temperance movements or a been organized. disease (e.g., genetic susceptibility) dominance of wine versus distilled Researchers must take differences but would not be able to identify spirits consumption (e.g., Peele 1997; in drinking cultures into account smoking as a cause of lung cancer. Levine 1992; Room 1982). when conducting international com­ Similarly, drinking patterns may show Traditionally, the wet/dry distinc­ parisons, in order to allow for a valid little variability within a drinking cul­ tion has been described as follows: measurement of alcohol consumption ture; accordingly, it may be more in any given country. For example, in valuable to study the effects of vari­ •In wet cultures, alcohol is integrated countries where drinking typically is ous drinking patterns by comparing into daily life and activities (e.g., is frequent and regular, a simple ques­ drinking patterns of different cul­ consumed with meals) and is widely tionnaire asking how often and how tures. For any research question, available and accessible. In these cul­ much people drink (i.e., a quantity– however, both aggregate-level and tures, abstinence rates are low, and frequency index) may correctly mea­ individual-level studies have advan­ wine is largely the beverage of pref­ sure most of the consumption. In tages as well as limitations (e.g., erence. European countries border­ countries where infrequent heavy Greenland and Robins 1994). ing the Mediterranean have tradi­ episodic drinking occurs, however, This article describes some of the tionally exemplified wet cultures. questions regarding occasions when methodological problems involved in higher quantities of alcohol are con­ measuring drinking rates across •In dry cultures, alcohol consump­ sumed are indispensable. Some data countries, such as differences in tion is not as common during collection instruments can be adjusted drinking cultures, drink sizes, and everyday activities (e.g., it is less fre­ to the local drinking culture to accu­ measurement instruments. It then quently a part of meals) and access rately measure alcohol consumption reviews the results of various types of to alcohol is more restricted. Ab­ and reflect the consumption patterns. studies that have examined drinking stinence is more common, but Because of the differences in drink­ rates across countries. Finally, the when drinking occurs it is more ing patterns, no single best instrument article discusses how such results likely to result in intoxication; exists for measuring consumption. should be interpreted given the limi­ moreover, wine consumption is less Nevertheless, a standard instrument tations of such studies, and it gives common. Examples of traditionally that is flexible enough to cover most some recommendations for improv­ dry cultures include the Scandi­ drinking patterns should be used in ing comparative international alcohol navian countries, the United States, comparative research to ensure that epidemiology for the future. and Canada. the data obtained from different

96 Alcohol Research & Health International Comparisons cultures are indeed comparable. In spirits, and fruit- or -based drinks could be more meaningful than the addition, the accuracy and compara­ are available for sale. number of glasses. bility across studies of alcohol con­ Another factor adding to the com­ Such local efforts to maximize the sumption measurements depend on plexity of accurately measuring alco­ validity of the most basic (conceptual) several potential sources of measure­ hol consumption is that consumption unit of measurement in alcohol surveys— ment error. These include response is conventionally expressed in grams the “drink”—can improve the interna­ rates among study participants, the of . One cannot ask respon­ tional comparability of alcohol surveys, mode of survey administration (e.g., dents how many grams of ethanol provided that other methodological face-to-face versus telephone inter- they consume, however, because peo­ considerations, such as sampling views), and the representativeness of ple generally do not think about their methodology, are also addressed. the sample. All of these factors can drinking in terms of alcohol content. With this approach, researchers in influence estimates of drinking vari­ Instead, researchers ask about the different countries may have to refer ables. A recent international confer­ number of “drinks,” “units,” “bottles,” to differently sized “standard drinks” ence focused on developing consensus or “cans” a person typically consumes, or drink containers because respon­ on questionnaire items for measuring depending on the national culture. dents most readily understand and alcohol consumption and alcohol- These everyday units, which respon­ report on these concepts. At the same related social harm in international dents recognize, contain varying time, however, investigators must comparisons. The recommendations amounts of ethanol, even within a have a valid estimate of the amount from this conference (Dawson and particular country. This problem is of ethanol present in each of these Room 2000), as well as the recently further compounded by the fact that drinks or containers to ensure compa­ when drinks are poured from a com­ published International Guide for rability of the data. For example, mon container (e.g., a bottle, cask, or Monitoring Alcohol Consumption and based on this information regarding can), the amounts poured will vary drink size and alcohol content, four Related Harm (World Health Organ­ greatly both among “home measures” ization [WHO] 2000), may help British pints of beer (16 g ethanol (Lemmens 1994) and among drinks each), five North American bottles of optimize comparability in interna­ served on licensed premises (Banwell tional alcohol surveys. beer (12–14 g ethanol each), and six 1999). For example, Kaskutas and Australian “middies” of regular beer The following sections demonstrate Graves (2000) found that respondents (10 g ethanol each) would be consid­ the complexity of assessing alcohol con­ in a sample of African American ered to contain approximately the sumption using two examples—drink women commonly poured themselves same quantity of ethanol (i.e., a total size and strength, and measurement “drinks” that contained up to six of approximately 60 g ethanol). instruments. times the amount defined as a standard Researchers must also scrutinize serving (i.e., 12 g alcohol). Similarly, assumptions about the usual ethanol Variation in Drink Sizes and Banwell (1999) reported that the aver- content (the strength) of major bever­ Strengths age serving for a glass of wine in ages. For example, a recent Australian Melbourne bars was 180 mL, not The basic problem for researchers study noted substantial differences in 100 mL as defined both in surveys conducting surveys of alcohol con­ the typical strengths of and dis­ and by health promoters in Australia. sumption is how to measure the tilled spirits across both time and The WHO (2000) recommends place in Australia during the 1990s amount of pure alcohol (chemically that researchers address the issue of known as ethanol) a respondent con­ (Catalano et al. 2001). Similar find­ assumed drink sizes in order to en­ ings were reported for wine in sumes, both on an individual drink­ hance comparability among alcohol ing occasion or day, and cumulatively Canada (Single and Giesbrecht surveys in different countries. This 1979). Furthermore, a study by the over a longer period. All researchers approach requires that investigators conducting surveys of alcohol con­ Finnish Foundation for Alcohol in each country conduct observation­ Studies (1977) documented changes sumption must make assumptions al studies to validate assumptions re­ about the serving size and alcohol in alcohol content as high as 6 per- garding typical serving sizes for each cent for cider and 8 percent for dis­ content of the drinks people say they major beverage in different settings. tilled spirits in different countries have consumed. In economically Similarly, assumptions about typical over time. developed countries, however, it is drink strengths and container sizes, becoming increasingly difficult to which ideally would be based on offi­ make such assumptions because of a cial sales data across time and place, Methods for Measuring bewildering array of beverages on the should be validated. In addition, Consumption market that come in containers of investigators could have respondents The most commonly used and simplest varying sizes and with varying strengths. indicate the number of beverage con­ measure of alcohol consumption is the For example, in Australia at least tainers or the portions of bottles con­ quantity–frequency index. With this 10,000 varieties of wine, distilled sumed in cases where those measures measure, respondents estimate how

Vol. 27, No. 1, 2003 97 often they drink and how much they consumption and may miss consump­ consumption data from each coun­ drink on a typical drinking occasion. tion by infrequent drinkers. try studied. One drawback of this approach is that Another method for preventing the respondents tend to ignore occa­ underreporting of typical alcohol The WHO (2000) recommends sional episodes of heavy consumption, consumption is to use the graduated- the graduated-frequency approach as which results in underestimates of true frequency method. With this approach, the method of choice for most purpos­ alcohol consumption (Gruenewald and respondents are asked to estimate es; however, investigators have experi­ Nephew 1994). However, if respon­ how frequently they drink different enced difficulties in setting several dents are asked only about consump­ daily quantities of alcohol—for exam­ comparable quantity thresholds across tion on recent occasions—a so-called ple, on how many days per week or different countries. Given local differ­ recent-recall approach—rather than month they have 0, 1–2, 3–4, drinks, ences in drink sizes and strengths, it is average consumption, they generally and so on. When researchers com­ possible that only one or two such report significantly higher amounts per pared this method with both the quantity thresholds, which should cor­ day (see Lemmens et al. 1992). quantity–frequency index and a respond to certain levels of risk of Various examples of recent-recall weekly-recall method, the graduated- experiencing adverse alcohol effects, approaches exist, such as drinking frequency method resulted in the are practical for use in international diaries or survey questions focusing highest estimates of alcohol consump­ comparative research. The WHO on very recent alcohol consumption. tion in a general population sample (2000) suggests that the thresholds for These approaches appear to produce from Ontario, Canada (Rehm et al. consumption that results in a medium relatively high estimates of total alcohol 1999). risk of acute adverse outcomes should consumption. Because the volume of Several other methodological issues be set at > 60 g of alcohol1 per day for alcohol reported in drinking surveys is also must be addressed in order to men and > 40 g of alcohol per day for normally only between 40 and 60 per- maximize the comparability of differ­ women. For consumption that results cent of the amounts known to be sold ent national alcohol surveys (WHO in a high risk of adverse consequences, in the relevant region, researchers 2000). Although the choice of methods the thresholds should be > 100 g of conducting drinking surveys generally must reflect the major research ques­ alcohol per day for men and > 60 g of assume that the higher the reported tions being addressed, the main policy alcohol per day for women. amounts are, the more accurately outcomes to be measured, and the The use of different consumption they reflect actual drinking behavior. resources available for the study, the thresholds for men and women—which Knibbe and Bloomfield (2001) com­ WHO recommends that investigators are based on gender differences in alcohol pared the ability of six European pay attention to the following issues: metabolism—is still controversial. One national surveys to account for national review concluded that different thresh­ levels of consumption as determined • The measures should reflect both olds are most appropriate when studying from sales data. The French survey, short-term and long-term physiological which contained only one question the overall volume consumed and the patterns of drinking. effects of alcohol (Graham et al. 1998). asking how much alcohol the respon­ However, according to that review differ­ dents had consumed the day before •Valid local estimates of typical units ent thresholds are less important for the the survey, achieved the highest esti­ study of behavioral gender differences in mate. In another analysis, the Australian of alcohol consumption should be available. the effects of alcohol consumption (e.g., Institute of Health and Welfare alcohol-related injuries), which are mod­ (Mathers et al. 1999) used a survey erated by the slower drinking pace fre­ that inquired about alcohol consump­ • The study should include a refer­ ence time period that matches other quently found in women, than for the tion only over the 3 days prior to the study of biomedical gender differences survey. When the data were weighted measures used for harms, such as alcohol-related workplace problems (e.g., liver disease). Conversely, a recent for the day of the week, the reported case control study found marked gender in the past 12 months. consumption accounted for over 70 differences in the risk of alcohol-related percent of the per capita alcohol con­ •Sampling methods should be injury at different thresholds of con­ sumption for that year. One plausible sumption. In that study, women at all designed to maximize the represen­ interpretation of these findings is that levels of alcohol consumption had a sub­ tativeness of samples. they indicate that poor recall is a stantially greater risk of injury than men major source of underestimation of (Stockwell et al. 2002). alcohol consumption. Nevertheless, • Whenever possible, researchers recent-recall approaches also have a should use multiple methods to compare consumption internation­ 1According to dietary guidelines in the United States, one major weakness in that they provide a (e.g., 12 ounces of beer, 5 ounces of valid picture only of recent behavior, ally, using recent-recall, graduated- wine, or 1.5 ounces of distilled spirits) contains approxi­ which may not necessarily reflect typical frequency, and per capita alcohol mately 12 grams of alcohol.

98 Alcohol Research & Health International Comparisons

International Survey search terms as “international com­ among men and women in 10 Research parisons” and “cross-cultural compar­ countries (Wilsnack et al. 2000). isons.” Although this list of studies Using comparable measures con­ Numerous investigators have con­ may not be exhaustive, it provides a structed from the data sets collected ducted studies of alcohol consumption general picture of the state of recent in the study countries, the group among adults and youth in a number comparative alcohol epidemiology. found that lifetime abstention for of countries. Such consumption data The studies reviewed here tend to both men and women was highest collected at the individual level (i.e., represent bounded comparisons—that in Israel (ages 18–40 years) and low­ through interviews and question­ is, comparisons among several rather est in the Czech Republic. naires) have several advantages over similar countries (Room 1988). Only aggregate-level data based on alcohol two of the studies (Wilsnack et al. sales statistics. For example, individual- •Hupkens and colleagues (1993) 2000; Fillmore et al. 1991) sought to analyzed data on beverage-specific level studies can gather information include a wider array of countries on drinking patterns, such as average drinking frequency from the from various regions (e.g., North Eurobarometer, an annual survey consumption, frequency of consump­ America, Australia, the Middle East, tion, and other variables, such as conducted in the EU to monitor and Europe). The other studies pri­ the attitudes of the population heavy episodic drinking. Moreover, marily included groups of member these studies allow for comparisons toward the EU and its policies. The states of the European Union (EU) or study reported that among the 12 among subgroups in the population Scandinavian countries. This selection (e.g., based on age, gender, and eth­ EU member states in 1988, Ireland reflects the present major funding had the highest rate of abstinence nicity), which are not possible with bodies for such research, such as the per capita consumption data based (timeframe not specified) and EU. In addition, although such a Denmark had the lowest rate. on sales. The following sections selection may appear limited, a focus review findings obtained in interna­ on relatively similar countries may tional surveys conducted among •In examining trends in alcohol use improve the comparability of the data. adults and youths. across as many as 15 European Despite considerable variation in countries, Simpura and Karlsson drinking cultures and research (2001) reported rates of abstinence Surveys of Adults approaches, European countries tend (definitions vary) for 1995. Among to have more in common in these Several international comparisons of women, the rate of abstention was drinking rates in adults based on gen­ respects than do, for example, African and European countries (Room 1988). highest in Portugal and lowest in eral population surveys have been Denmark. Among men, the absten­ conducted in the last decade or so. With most of these studies, ques­ tionnaires (including discrepancies in tion rate was highest in Italy and The studies presented here meet the lowest in Denmark. following criteria: (1) They are recent translations) and response rates varied cross-sectional general population to such an extent that the resulting measurement and nonresponse bias •In a study examining episodic heavy surveys. (2) They reported prevalence drinking with a common question­ rates of drinking behaviors, such as discourage direct comparison of con­ naire in four Scandinavian coun­ rates of abstinence or current drink­ sumption levels (e.g., Simpura et al. in tries, Mäkelä and colleagues (2001) ing (i.e., alcohol consumption in the press; Knibbe and Bloomfield 2001). past 12 months) and rates of heavy Thus, most studies also included other found that Sweden had the highest drinking, as well as mean alcohol measures (e.g., gender ratios) to exam­ 12-month abstinence rate for men, consumption and mean frequency of ine trends and relationships between Norway had the highest rate for drinking. Studies measuring alcohol drinking and other variables when women, and Denmark had the low­ dependence or alcohol problems are comparing data across countries. est rate for both genders. not discussed here. A comprehensive review of the literature to identify Abstinence. Several studies compared •Arecent EU project, funded par­ such studies was beyond the scope of abstinence rates among adults in various tially by the European Commission, this article. Rather, the authors chose countries. The countries included in conducted a survey of alcohol use, studies they were familiar with and the different studies vary, making an alcohol-related problems, and atti­ supplemented this information with overall analysis difficult. The studies tudes toward alcohol in six Euro­ the studies’ own literature and a con­ found the following results (see table 1): pean countries (Leifman 2002). firmatory search of the Alcohol and Surprisingly, among these countries, Alcohol Problems Science Database •In investigating gender differences abstinence in the last 12 months (ETOH), which is maintained by the in drinking, the International was lowest in Finland. Equally sur­ National Institute on Research Group on Gender and prising, the highest rate of absti­ and Alcoholism (NIAAA), using such Alcohol conducted a comparison nence was found in France.

Vol. 27, No. 1, 2003 99 Table 1 Summary of Results from Selected Comparative Studies of Drinking Rates Australia Austria Belgium Canada Czech Republic Denmark Estonia Finland France Germany Greece Ireland Israel Italy Luxemburg Netherlands Norway Portugal Russia Spain Sweden Switzerland UK USA Hupkens et al. 1993 (12 EU countries) Abstinence L H Frequency L HW HM

Ahlström et al. 2001 (7 European countries) Current drinking HM L HW L Frequency per month L H Mean consumption L H Heavy drinking L HM HW

Leifman et al. 2002 (6 EU countries) Abstinence L H Daily drinking H L Average volume L H Binge drinking L H

Wilsnack et al. 2000 (10 countries) Lifetime abstinence L H Current drinking H LW LM Frequency HM L HW Average volume LM H LW Heavy drinking HM L HW

Simpura & Karlsson 2001 (15 countries) Abstinence L HM HW Daily drinking L H

Mäkelä et al. 2001 (4 Scandinavian countries) Abstinence L HW HM Mean consumption H L Per occasion HW HM/LW LM 6+ frequency H LW LM

NOTE: Countries participating in the respective studies are shaded; countries with the highest (H) or lowest (L) levels and for men (M) or women (W) of each variable studied are indicated.

Overall, these studies indicate that countries appear to be declining. trend toward a homogenization of in recent years the Mediterranean Finally, Denmark appears as a special European drinking styles. countries had comparatively higher case in both Scandinavian and EU- rates of abstention than countries of wide comparisons as a country with Current Drinking. Two studies have central Europe. Furthermore, the low abstinence rates. These findings looked at current drinking rather than abstinence rates in Scandinavian agree with the recently observed abstinence in measuring the respon-

100 Alcohol Research & Health International Comparisons

dents’ drinking status. One study found the highest frequencies of Binge Drinking. Binge drinking (some- compared drinking patterns between drinking (i.e., number of drinking times called heavy episodic drinking) Germany and the United States. This occasions in a month) among Dutch has been included as a measure in analysis found that within a compara­ women and Czech men and the some recent comparative studies: ble age range and with fairly similar lowest frequencies among Estonian instruments, Germany had one-third women and men. •Wilsnack and colleagues (2000), in more current drinkers (i.e., people an analysis that included 10 coun­ who had consumed alcohol in the past Two studies examined the rates of tries, found that Canadian men and 12 months) than did the United States daily drinking in various countries. Swedish women had the highest per­ (Bloomfield et al. 2002). The other Leifman (2002) reported that when all centages of drinkers who had study was part of an EU concerted respondents (i.e., drinkers and non- engaged in heavy episodic drinking action that examined women’s alcohol drinkers) were included in the analysis, (the definitions of this term varied consumption and alcohol problems (as Italy had the highest and Finland the among the countries) in the last 12 well as gender differences in drinking lowest rates of daily drinking. Simpura months. Conversely, Israeli men and patterns) in nine European countries and Karlsson (2001) examined seven women had the lowest percentages of through a secondary analysis of survey countries regarding daily drinking; this heavy episodic drinking. (This analy­ data. In one of its analyses of basic analysis noted that wine countries had sis included current drinkers only.) drinking measures, Ahlström and col­ the highest rates, followed by beer- leagues (2001) found that current drinking countries and former spirits- •Studying data from four Scandi­ drinking rates (i.e., consumption in consuming countries. This order was navian countries, and considering the last 12 months) were highest among true for both men and women. data from all respondents (drinkers French men and Swedish women and Finally, in the two-country compari­ and nondrinkers), Mäkelä and col­ were lowest among Italian men and son of Bloomfield and colleagues leagues (2001) found that Danish women. These findings for current (2002), respondents (i.e., drinkers and men and women had the highest drinking (which is the inverse of nondrinkers) in Germany reported annual frequencies of consuming six abstinence) appear to fit the conclu­ almost twice as many drinking days as or more drinks on one occasion, sions drawn regarding abstinence rates did U.S. respondents. All of these and Norwegian men and Finnish as described in the preceding section. studies suggest that the main wine- women had the lowest frequencies. consuming (and wine-producing) Frequency of Drinking. Another aspect countries of Europe have the highest •In his study of six EU member of drinking behavior often measured in frequencies of drinking. states, Leifman (2002) reported that comparative studies is the frequency of The international Collaborative when both drinkers and non- drinking. Several studies have assessed Alcohol-Related Longitudinal Project drinkers were included in the analy­ this variable in international compar­ (e.g., Fillmore et al. 1991) conducted sis, people in the United Kingdom isons. These analyses, which included meta-analyses on 39 longitudinal data had the highest annual frequency of information from current drinkers only, sets3 on alcohol consumption from 15 heavy drinking (defined as drinking had the following results: countries. Its goal was not necessarily a bottle of wine or the equivalent on to examine differences in basic drink­ one occasion), and people in France • The study by Hupkens and col­ ing rates across countries, but to had the lowest. leagues (1993) involving 12 EU examine and predict drinking patterns member states found that Spain had and problems over the life course •In a comparison between the the highest frequency of drinking for (Fillmore et al. 1991). In one of its United States and Germany includ­ men, Italy had the highest frequency analyses examining the frequency of ing all respondents (Bloomfield et for women, and Ireland had the low­ drinking, the study took into account al. 2002), the number of days per est frequency for both genders (in national origin of the data sets. It month on which five or more general frequency categories2). grouped the studies into regions, drinks were consumed was almost including the United States, Canada twice as high in Germany as in the • Ahlström and colleagues (2001) deter- (including New Zealand), Europe, United States. mined that among nine European and the United Kingdom (including countries, France had the highest and Ireland). This comparison found that 2General frequency categories do not measure the actual Finland had the lowest frequency of in relation to the United States number of days when alcohol consumption occurs but drinking (i.e., number of drinking provide more general categories, such as “every day or (which was used as a reference), the almost every day,” “3 to 4 days per week,” or “never.” occasions within a month). frequency of consumption was highest in the United Kingdom, followed by 3In meta-analyses, data are pooled from several studies •Wilsnack and colleagues (2000), to allow researchers to draw overall conclusions. Longi­ the European countries and Canada tudinal studies follow the same respondents over an with their sample of 10 countries, (Johnstone et al. 1996). extended period of time (e.g., several years).

Vol. 27, No. 1, 2003 101 No consistent pattern in binge of drinking (e.g., relatively high absti­ tionnaire is voluntary). Thus, the drinking rates emerges from these nence rates in the Mediterranean response rates usually exceeded 85 studies. This lack of consistency countries and highest frequency of percent. The data presented here refer might be partially attributable to the consumption in the wine-producing to the most recent versions of both small number of studies and the vary­ countries of Europe). Measures of surveys. ing definitions of the behavior. binge drinking and mean consump­ tion levels, however, exhibited less Abstention Rates. The HBSC found Mean Alcohol Consumption. Several consistency. It is difficult to deter- that in all countries abstention from recent international studies also have mine whether these inconsistencies alcohol covaried highly between the compared mean levels of alcohol con­ stem from methodological problems sexes—that is, if the abstention rate sumption. In the study by Ahlström or from real changes in drinking was high among males, it usually was and colleagues (2001), Italian respon­ behaviors, which no longer fit the also high among females. In addition, dents reported the highest mean traditional typologies of drinking cul­ the countries maintained their rela­ monthly consumption and Finnish tures (e.g., the wet/dry dichotomy). tive positions in abstention rates in all respondents the lowest, when Such questions deserve increased age groups tested. However, the dif­ drinkers only were considered. attention as researchers develop meth­ ferences among countries in absten­ Leifman (2002), who included all odology and concepts of future studies. tion rates diminished among older respondents in the analysis, found the students. The HBSC findings also highest mean annual alcohol con­ Surveys of Youth suggested that with regard to absten­ sumption in the United Kingdom tion rates, the distinction between and the lowest consumption in Sweden. To date, researchers have conducted dry or wet drinking cultures either Mäkelä and colleagues (2001) reported two large-scale international youth did not apply to adolescents or devel­ that among the Scandinavian countries, surveys relating to alcohol use, the oped later in life. Thus, the HBSC Denmark had the highest and Norway Health Behavior of School-Aged found high abstention rates in proto­ had the lowest annual consumption Children (HBSC) survey and the typical wet countries, such as France in an analysis of all respondents. In European School Survey Project on and Switzerland, but also in Norway, the study comparing the United Alcohol and (ESPAD). The the United States, and Israel. Con­ States and Germany (Bloomfield et HBSC, conducted for the fourth time versely, in Denmark, Scotland, Wales, al. 2002), mean monthly alcohol in 1997–1998, included children ages England, the Czech Republic, and consumption among all respondents 11, 13, and 15 in 26 European coun­ Slovakia, abstention was uncommon was more than twice as high among tries, Canada, and the United States even at age 13. German respondents as among U.S. (Currie et al. 2000). The ESPAD sur­ veyed 15-year-olds from 30 European respondents. Finally, a somewhat Weekly Drinking. The HBSC survey countries for the second time in 1999 older study examined differences in also provided data on the percentages (Hibell et al. 2000). The main advan­ the sociodemographic correlates of of respondents who consume alcohol tage of both surveys is the use of a drinking patterns in the general pop­ at least once a week. Because these common methodology in all partici­ ulations of Switzerland, Germany, percentages included the abstainers, pating countries. For example, investi­ and the Netherlands (Knibbe and the authors of this article recalculated gators in the participating countries of Lemmens 1987). This study, which the percentages of weekly drinkers each survey used a common question­ analyzed all respondents, found that relative to the percentages of drinkers naire, employed a standardized sam­ Germans had the highest average to determine the proportion of week­ pling methodology to ensure that their weekly consumption in terms of stan­ ly drinkers among all drinkers in each samples were representative of that dard glasses, followed by the Swiss country. In general, these analyses country, surveyed the same age and Dutch respondents. demonstrate that as the percentage of As with binge drinking, these find­ group(s) across countries, and collected their data in the same year. Moreover, drinkers increased, so did the per­ ings do not indicate a clear pattern in centage of weekly drinkers (r = 0.4).4 mean consumption among the coun­ both projects required minimum sam­ However, one group of countries tries studied. This lack of consistency ple sizes as criteria for participation in diverged from this pattern by record­ could be accounted for by the varying the respective collaborative study, ing, among both males and females, methods of measurement used across guaranteeing a certain statistical high percentages of drinkers but low the studies. power, and thus precision, of the results. Finally, the surveys were con­ percentages of weekly drinkers. This Summary. Overall, the studies re- ducted in school, a procedure that group—which consisted of Poland, viewed in this section suggest that rel­ reduces nonresponse because usually Lithuania, Estonia, Latvia, Finland, atively consistent patterns appear to only those students who are absent on and Sweden—was the only group of exist across the reported studies with the day of the survey do not partici­ respect to abstinence and frequency pate (although responding to the ques­ 4Israel was an exception to this general observation.

102 Alcohol Research & Health International Comparisons geographically connected countries These drinking variables were Overall, the wet/dry dichotomy was exhibiting a pattern that deviated adjusted relative to the percentage of not evident among young people. from the patterns in other countries. past-year drinkers. Malta was exclud­ Although some countries exhibited Countries with relatively low per­ ed from the analysis as a country that such prototypical drinking behaviors, centages of both drinkers and weekly was different from all others and pro­ the overall picture of drinking among drinkers included Norway, Switzer­ duced extreme values. young people was highly diverse. land, and the United States. Con­ The analyses indicate that beer was versely, high percentages of both clearly the dominant beverage of drinkers and weekly drinkers were choice among the 15-year-olds in this Aggregate-Level Studies found in England, Wales, Scotland, sample of countries (see table 2). and Per Capita Denmark, and Greece. Again, no With the exception of one country— Consumption marked gender differences existed Hungary—the percentages of beer across the countries—if the percent- drinkers outnumbered the percent- Some cross-national comparisons of age of male weekly drinkers was ages of wine drinkers. Similarly, beer alcohol consumption use aggregate- high, so was the percentage among drinkers outnumbered drinkers of level data. The most widely analyzed females. Finally, the ranking of the distilled spirits in all but three coun­ variable in such studies is per capita countries with respect to weekly tries (i.e., Hungary, Norway, and alcohol consumption—the amount drinking remained stable among the Portugal). Moreover, initial findings of ethanol in liters per year that every different age groups tested, similar to indicated that when beer drinking adult consumes. To calculate annual what had been observed for the was common, so was consumption of adult per capita consumption, one abstention rates. These findings in distilled spirits. In fact, distilled spir­ sums up production and imports of young people again did not seem to its, rather than wine, were the second alcoholic beverages, subtracts exports fit the prototypical wet/dry classifica­ most common beverage of choice. of alcoholic beverages, and then tion of adult drinking in that nations Thus, only in Estonia, Italy, Latvia, divides by the number of adults with high alcohol consumption Lithuania, and Romania did more (often all people age 15 and older) in among youth included both a wine students drink wine (at least three the population. Ideally, the calcula­ country, Greece, and beer-drinking times or more during the past 30 tion would also consider informal countries, such as Denmark, Eng­ days) than distilled spirits. Overall, alcohol production, consumption by land, and Wales. wine consumption was not correlated residents outside the country, duty- with consumption of beer or distilled free consumption, consumption by Drinking Patterns. The ESPAD sur­ spirits in young people, and high foreign tourists in the country, vey, which included some countries wine consumption did not generally imported alcohol re-exported to not participating in the HBSC, indicate a distinct “drinking culture.” other countries, and any additional allowed for more detailed analyses of The percentages of binge drinkers stocks5 (WHO 2000); however, these drinking patterns. To this end, the also varied significantly across coun­ data are difficult to obtain. authors of this article used the follow­ tries (see table 2). The lowest percent- Studies comparing per capita con­ ing variables from the 1999 ESPAD ages of binge drinkers—approximately sumption may have advantages over report (Hibell et al. 2000): 10 percent of all drinkers—were studies comparing individual-level found in Lithuania, Greece, Slovak data, especially for cross-country Republic, Portugal, and Romania. comparisons. For example, aggregate •Frequency of consuming different Conversely, in Poland, Ireland, the types of alcoholic beverages data can easily (and inexpensively) United Kingdom, and many northern be obtained for many countries. countries, the percentages of students Furthermore, data often are available • Alcohol consumption three times or bingeing at least three times per more in the past 30 days for several years, permitting the month ranged between 20 and 40 comparison of trends. In addition, percent. Bingeing, however, showed per capita consumption data may •“Binge” drinking (i.e., consumption no significant association with the of five or more drinks in a row) three paint a more accurate picture of percentage of drinkers in a given overall consumption levels than times or more during the last 30 days country. Low-bingeing countries had surveys, which commonly result in both high (Portugal, Romania) and lower estimates of the total consumption •Frequency of any alcohol consump­ low (Greece, Slovak Republic) per­ tion in the past 12 months, with centages of abstainers. Similarly, in abstinence defined as no alcohol 5These additional stocks refer to alcoholic beverages high-bingeing countries, both low stored because they require aging (e.g., whiskey and consumption in the past 12 months (United Kingdom) and high (Iceland) cognac). Thus, these beverages may have been pro­ percentages of abstainers were found. duced in earlier years but are released for sale or export • Amount consumed on the last only after reaching maturation. However, the effect of A similar picture emerged with stockpiling on estimates of total alcohol consumption in drinking occasion. regard to frequency of drinking. most countries is likely to be small.

Vol. 27, No. 1, 2003 103 Table 2 Indicators of the Drinking Behavior of 15-Year-Olds Participating in the European School Survey Project on Alcohol and Drugs (ESPAD)

Percentage of 15-year-olds who:

Drank Drank Drank Consumption Average beer wine spirits on the last frequency Binged 3 or at least at least at least occasion of drinking Are more times in 3 times in 3 times in 3 times in (in centiliters of in past Country drinkers past 30 days* past month* past month* past month* pure ethanol)* 12 months* Bulgaria 82 13 33 17 27 6 10 Croatia 73 16 32 19 21 11 12 Cyprus 79 15 34 10 30 7 13 Czech Republic 94 18 43 19 30 9 20 Denmark 96 31 55 14 42 12 28 Estonia 89 16 28 15 9 8 12 Faroe Islands 75 20 33 7 31 13 15 Finland 86 21 20 6 10 14 13 Former Yugoslav Republic of Macedonia 57 16 28 21 25 11 9 France 77 16 32 16 30 9 12\ Greece 94 10 37 18 31 6 19 Greenland 81 31 56 6 26 12 11 Hungary 80 15 15 16 24 7 9 Iceland 69 25 25 6 19 16 10 Ireland 89 35 39 9 38 17 24 Italy 75 n/a 41 25 20 7 12\ Latvia 88 16 34 15 14 7 11 Lithuania 91 10 31 16 11 8 12 Norway 78 31 22 8 26 13 12 Poland 82 38 34 10 11 10 15 Portugal 74 8 24 5 27 10 11 Romania 79 6 25 18 8 6 11 Russia (Moscow) 87 18 46 11 16 9 16 Slovak Republic 90 9 23 22 22 8 13 Slovenia 83 30 33 27 27 10 13 Sweden 83 20 25 10 24 12 12 Ukraine 81 12 27 20 21 7 10 United Kingdom 91 33 41 18 35 16 22

* Recalculated based on percentages of current drinkers, by G. Gmel. NOTE: The ESPAD survey included 15-year-old male and female students. SOURCE: Hibell et al. 2000.

by a population.6 Finally, because age- and sex-specific consumption is that these analyses are prone to alcohol consumption also harms peo­ rates or the prevalence of certain biases (often known as ecologic bias) ple other than the drinkers them- drinking patterns. However, the same that are unique for such types of inves­ selves (see the article in this issue by per capita consumption may, in the­ tigations and are not found in individ­ Gmel and Rehm), aggregate-level ory, have a completely different ual-level studies (Morgenstern 1998).7 data may better capture the relation- impact on consequences, depending Furthermore, the recording of per ship between aggregate alcohol-related on factors such as the percentage of capita data depends on so many fac­ consequences and changes in the abstainers or typical consumption tors that their use as a standard for drinking patterns in a population patterns (e.g., more regular, moderate

(Norström and Skog 2001). consumption versus infrequent but 6Besides the previously mentioned problems of under- Aggregate-level data also have heavy drinking). reporting of respondents, these lower consumption estimates some drawbacks, however. For exam­ may also be attributable to a disproportionate underrepresen­ Another possible limitation of tation of certain groups of heavier drinkers in typical samples, ple, studies usually do not report aggregate-level studies of associations such as institutionalized or homeless people.

104 Alcohol Research & Health International Comparisons individual-level data is questionable. approach) instead of the original values patterns of drinking have indepen­ Sources of error may include illegal is believed to reduce confounding and dent effects on mortality. production, cross-border imports, therefore spurious associations (Nor­ By widening the spectrum of com­ consumption by tourists, and changes ström and Skog 2001). parative research to include developing in the age composition of a country One of the largest studies using countries, researchers can enhance over time (for more information on this approach in recent years was the understanding of the cultural impact these factors, see the special issue of European Comparative Alcohol on alcohol and related consequences Contemporary Drug Problems, Vol. 27, Study (ECAS) (Norström 2001a). because the cultural differences in No. 2; Summer 2000). This study was conducted in 14 EU alcohol consumption are more variable Despite these limitations, aggregate- countries, including traditional wine- on a worldwide scale than within one level studies can complement drinking countries in southern Europe or two continents. The WHO Global individual-level studies. Thus, the (e.g., Spain and France), beer-drink­ Alcohol Database, which includes combination of several data sources ing countries in central Europe (e.g., consumption estimates for 181 coun­ (e.g., aggregate-level studies, cross- the United Kingdom, Ireland, and tries, is currently the most comprehen­ sectional and longitudinal individual West Germany), and former spirits- sive source of per capita consumption studies, and laboratory studies) may drinking countries in northern Europe data. However, for some countries, strengthen the establishment of potential (e.g., Sweden, Norway, and Finland). the information on per capita con­ causal associations (see Pernanen Among its analyses, the ECAS com­ sumption originates from competing 2001). This strengthening effect may pared per capita consumption and sources and is based on either sales be even stronger if the evidence comes several outcomes in the participating figures or production data. The accu­ from different data sources rather than countries. For some outcomes, the racy of sales data generally depends from repeated analyses of the same data ECAS established the same associa­ on the accuracy of export and import source (Norström and Skog 2001). tion with alcohol consumption that data (WHO 1999). The most com­ Cross-cultural comparisons of per was found in individual-level studies prehensive source of information capita consumption can be used in (e.g., unintentional injuries, such as included in the Global Alcohol two major ways. First, investigators accidental falls). Surprisingly, many of Database is a data set published by the can compare per capita consumption the associations were stronger in Food and Agriculture Organization with certain alcohol-related outcomes northern countries that have lower (FAO) of the United Nations. It pro­ at a given point in time or as an aver- overall alcohol consumption than in vides estimates of per capita consump­ age over a range of time points. For southern countries, which generally tion based on production data and example, in a study involving 14 have higher alcohol consumption. includes not only beer, wine, and European countries, Ramstedt These observations indicate the distilled spirits but also several other (2001b) demonstrated that age- importance of different patterns of beverage categories, such as palm adjusted mortality rates for liver cir­ alcohol consumption. For example, it wine; maize, millet and sorghum beer; rhosis in both men and women is commonly assumed that although fruit wine; rice wine; rice-fermented increased with increasing per capita alcohol consumption is lower in the beverages; cider; grape must (i.e., consumption. Thus, although per northern countries, the drinking pat- pressed but unfermented grape ); capita consumption was available terns are different (e.g., people more only for the population as a whole, commonly drink to intoxication and 7A famous example of ecologic bias was reported by this variable correlated well with sex- the level of distilled spirits consump­ Durkheim (1951[1897]), who studied suicide rates in four specific rates of mortality. tion is higher). regions of Prussia and found that these rates increased with the proportion of Protestants in those regions. The Similar ecological analyses can be The relevance of drinking patterns aggregate-level data indicated that suicide rates among conducted with aggregated statistics is supported by a recent study by Protestants were nearly eight times greater than among non-Protestants. Individual-level analyses, however, from survey data—for example, to Gmel and colleagues (2001), who demonstrated that many suicides were committed by link mean consumption with the pro- modeled the effects of changes in non-Protestants, possibly because living as a religious portion of heavy drinkers or alcohol- aggregate consumption on mortality minority increases suicide risk. Accordingly, more non- Protestants might take their lives (and thus increase sui­ related consequences in a population across countries using indicators of cide rates) in regions with higher proportions of Pro­ (e.g., Colhoun et al. 1997). drinking patterns (e.g., drinking with testants. Thus, the Protestant area would have a contex­ Second, aggregate per capita con­ meals, frequency of drinking, drink­ tual effect on non-Protestant suicides. sumption data can be used to compare ing to intoxication, percentage of 8This study was part of a series of analyses of alcohol- 8 changes of both consumption and dif­ abstainers). The study showed that related consequences within the framework of the ferent outcomes (e.g., cirrhosis mortality, the more detrimental the general pat- Comparative Risk Assessment (CRA) of the Global Burden of Disease 2000 (study, which aims to estimate all-cause mortality, and traffic injuries) tern of drinking in a country, the the burden stemming from alcohol consumption on a within a country or across countries higher the impact of a change in worldwide scale, including developing countries (Rehm et al. in press). It was a pooled cross-sectional time series over time. Analyzing changes in con­ alcohol consumption on all-cause analysis. Pooling of data is particularly important in devel­ sumption or outcomes by using time mortality. This finding confirms that oping countries for which data often are available only for series (i.e., using a differencing in addition to volume of drinking, a few time points.

Vol. 27, No. 1, 2003 105 vermouth; and wheat-fermented bev­ recorded (e.g., because of revisions of behaviors. Simultaneously, in a small erages. These beverage categories are the diagnostic manuals, such as the random subset, each country could particularly important in many devel­ WHO’s International Classification apply the instrument that is assumed oping countries. of Diseases9). There is also evidence to be best for that country. Such a The second main source of data that alcohol-related diseases (e.g., design would allow researchers both included in the Global Alcohol Data- liver cirrhosis) may not be labeled as to compare countries with the same base, World Drink Trends (WDT) of such in some countries, possibly to measurement instrument and to the Dutch Distiller’s Association, is avoid stigmatization or financial explore whether using the “best” provided by the alcohol industry and penalties to the patients and their instrument for each country would is based mostly on sales and tax statis­ families (Cipriani et al. 2001). alter the conclusions. tics (Productschap voor Gedistilleerde Second, investigators could focus Dranken 2000). Although the WDT on associations between drinking is assumed to provide more accurate Recommendations for measures and drinking-related prob­ data than the FAO data set, it included Future Research lems rather than on comparisons of only 58 countries in 2000. It should rates of several drinking variables. be noted that the choice of data source Comparing alcohol consumption and Studies found that different epidemio­ may affect comparative research drinking patterns internationally is logical measurement instruments of because the correlation between FAO not an easy task. This review has dis­ alcohol consumption commonly and WDT estimates is not perfect cussed various sources of method­ result in comparable relative positions (r = 0.74 for those 45 countries in the ological problems that make such of drinkers (e.g., Feunekes et al. 1999; WHO Global Alcohol Database for comparisons so difficult, and the Rehm et al. 1999). For example, a which data are available from both comparative research presented here person who is considered one of the sources). reveals that the results of existing studies heaviest drinkers in a population Because of the limitations of the are only partly consistent. Consequently, using one instrument also will be con­ data sets and discrepancies between the question arises as to how researchers sidered one of the heaviest drinkers them, all per capita consumption data can best conduct comparative alcohol using other instruments. If researchers should be carefully checked over time research in the future. This section can identify stable associations among both within a country and across presents some suggestions for possible differing consumption measures and countries before one can draw any new directions. outcomes, they could conduct com­ conclusion from comparative research An obvious problem, which in the­ parative research of such associations, that uses data from different sources. ory could easily be solved, is to maxi­ regardless of the instrument used to However, an analysis of changes mize the comparability of surveys by measure alcohol consumption. between time points instead of original standardizing sample selection, survey Third, investigators could combine data (i.e., use of a differencing protocols, response rates, modes of findings from comparative research using approach for time series data) might survey administration, and socio­ different types of designs (e.g., general be one possibility to reduce problems demographic factors (e.g., the age population surveys and aggregate-level of comparative time series research range) of study samples. In practice, studies) to substantiate differences (although differenced time series models however, standardization is difficult among countries. also have potential pitfalls [e.g., see because studies have differing funding For quantitative alcohol epidemi­ Rehm and Gmel 2001]). If possible, and priority levels across countries. ology, survey designs should provide findings from aggregate-level analyses A related issue is the problem of valid as well as reliable quantitative should be corroborated with findings measuring alcohol consumption. assessments of alcohol consumption from individual-level analyses. The There is an inherent conflict between and its consequences. Therefore, ECAS (Norström 2001b) or the work efforts to improve comparability investigators should make greater within the Comparative Risk (through standardization) and efforts efforts to ensure that assumptions Assessment (Rehm et al. in press) are to improve validity (through use of about the alcohol content of “drinks” promising examples of such verification. measures best suited to a country’s in different countries are locally valid The comparability of analyses of drinking culture and typical drink and should pay attention to sampling per capita consumption may be sizes). Some possible solutions to this and questionnaire-design issues. To affected by more than the use of dif­ issue include the following. First, determine the success of such efforts ferent data sources, however. For future comparative research could researchers should compare survey example, Ramstedt (2001a) demon­ incorporate methodological studies results to actual alcohol consumption strated that time series analyses— using a split sample design. With when consumption can be reliably even within limited geographical such a design, all participating coun­ regions, such as Europe—might be tries could use a common instrument 9The International Classification of Diseases, published by biased by changes in the way alcohol- deemed to be the best available mea­ the WHO, provides diagnostic criteria for all medical con­ related outcomes are diagnosed and sure of the widest range of drinking ditions and allows the classification of those conditions.

106 Alcohol Research & Health International Comparisons estimated from sales, taxation, import, alcohol use and consequences on a use and its consequences. In addition, export, or production data (WHO cross-cultural level. The study a recent study of per capita alcohol 2000). The validity of comparisons of includes such diverse countries as consumption and liver cirrhosis consumption rates across regions or Mexico, Argentina, Sri Lanka, mortality in 14 European countries time periods will always be limited if Kazakhstan, France, Austria, Israel, suggested that the strength of the the survey results account for only a Hungary, and Japan as well as the relationship between drinking level low (or varying) proportion of actual United States. It employs a standard­ and mortality may vary by regional alcohol consumption. ized questionnaire in as many coun­ drinking pattern and may not be as Most of the comparative research tries as possible and has suggested dependent on total consumption level to date has focused on established guidelines for sampling and survey as previously thought (Ramstedt market economies (i.e., developed methods for incoming project part­ 2001b). This finding could be the countries). With these studies, differ­ ners. motivation for more detailed analy­ ences in findings may partly reflect Second, future international com­ ses. Thus, numerous possibilities exist simple “noise” as these countries are parative research should develop cri­ for further international comparative rather similar to one another com­ teria for considering the validity of studies that could promote the pared with developing countries. By comparisons of both aggregate-level methodological development of this widening the range of participating and individual data. Studies also research area. ■ countries, researchers may enhance should consider and combine data knowledge about the effects of drink­ across different sources in order to ing patterns. Such work is currently strengthen comparative analysis under way in different developing under less than optimal conditions. Acknowledgments countries (Demers et al. 2001). Using such approaches, a primary The authors express their apprecia­ In conclusion, researchers should task would be to determine how dif­ tion to Ms. Elisabeth Grisel-Staub for consider the following when creating ferent drinking patterns are related to her expertise in formatting the an agenda for future research. First, outcomes (e.g., mortality and mor­ manuscript and to Ms. Stephanie investigators need to tackle more bidity) and whether such differences Kramer for her stylistic improvements directly the major barriers to valid indicate drinking patterns that are to the text. Furthermore, the authors cross-cultural research on drinking, less harmful than others. Another are grateful to the two anonymous using and improving on previous task would be to identify effective attempts to address the major meth­ measures taken in various countries reviewers for their comprehensive and odological hurdles both for compara­ to prevent negative alcohol-related constructive suggestions. tive monitoring purposes and for consequences. This type of research social and epidemiological investiga­ involves not only cross-sectional studies References tion. The ECAS, which combines in each study country, but also the aggregate-level findings with general implementation of longitudinal AHLSTRÖM, S.; BLOOMFIELD, K.; AND KNIBBE, population surveys using a compara­ designs. Current international research, R.A. Gender differences in the drinking patterns ble methodology, is currently the best however, has remained at the stage of in nine European countries: Descriptive findings. 22(1):69–85, 2001. example of such efforts. In general, attempting to compare drinking rates the greater the geographical, cultural, and improve the needed methodology ALLAMANI, A.; VOLLER, F.; KUBICKA, L.; AND and economic diversity among partic­ and therefore has not yet reached the BLOOMFIELD, K. Drinking cultures and the posi­ tion of women in nine European countries. ipating countries, the greater the stage of addressing these more applied Substance Abuse 21(4):231–247, 2000. methodological difficulties. Com­ issues. parisons within Europe will be more Finally, future international com­ BANWELL, C. How many standard drinks are there in a glass of wine? Drug and Alcohol Review valid than, for example, comparisons parative research should continue to 18(1):99–101, 1999. between urban European areas and fulfill a concrete public health–related rural Africa. One example of an purpose through national and inter- BLOOMFIELD, K.; GREENFIELD, T.K.; KRAUS, L.; attempt to conduct valid comparisons national monitoring activities that AND AUGUSTIN, R. A comparison of drinking patterns and alcohol-related problems in the of diverse countries is a project called serve policy and political purposes for United States and Germany, 1995. Substance Use GENACIS (Gender, Alcohol and countries and international bodies. & Misuse 37:399–428, 2002. Culture: An International Study), a Additionally, international research multinational study funded by various should continue to test theories regard­ CATALANO, P.; CHIKRITZHS, T.; STOCKWELL, T.; ET AL. Trends in Per Capita Alcohol Consumption in sources including the U.S. National ing political, economic, cultural, and Australia, 1990/91–1998/99. NDRI Monograph 4. Institute on Alcohol Abuse and biological differences in alcohol use Perth, Western Australia: National Drug Research Alcoholism, the European Com­ and its consequences across countries. Institute, Curtin University of Technology, 2001. mission, and WHO.10 The study, For example, the GENACIS project which currently involves 26 countries, examines the influence of biology and 10In addition, many national governments are financing is investigating gender differences in culture on gender differences in alcohol their countries’ respective research contributions.

Vol. 27, No. 1, 2003 107 CIPRIANI, F.; LANDUCCI, S.; AND BLOOMFIELD, K. Drug Use Among Students in 30 European Coun­ NORSTRÖM, T., ed. Alcohol in Postwar Europe: Alcohol-related mortality in Europe: A tentative tries. Stockholm: Swedish Council for Infor­ Consumption, Drinking Patterns, Consequences analysis from the EU project “Alcohol consumption mation on Alcohol and Other Drugs, 2000. and Policy Responses in 15 European Countries. and alcohol problems among women in European Stockholm: National Institute of Public Health, HUPKENS, C.L.H.; KNIBBE, R.A.; AND DROP, countries.” Substance Abuse 22(1):55–68, 2001. European Commission 2001b. Available at: M.J. Alcohol consumption in the European http://www.fhi.se/pdf/ECAS_2.pdf COLHOUN, H.; BEN-SHLOMO, Y.; DONG, W.; ET Community: Uniformity and diversity in drink­ AL. Ecological analysis of collectivity of alcohol ing patterns. Addiction 88(10):1391–1404, 1993. NORSTRÖM, T., AND SKOG, O.-J. Alcohol and consumption in England: Importance of average mortality: Methodological and analytical issues in JOHNSTONE, B.M.; LEINO, E.V.; AGER, C.R.; ET drinker. British Medical Journal 314(7088):1164– aggregate analyses. Addiction 96(Suppl. 1):S5– 1168, 1997. AL. Determinants of life-course variation in the frequency of alcohol consumption: Meta-analysis S17, 2001. CURRIE, C.; HURRELMANN, K.; SETTERTOBULTE, of studies from the Collaborative Alcohol-Related PEELE, S. Utilizing culture and behaviour in epi­ W.; ET AL.; eds. Health and Health Behaviour Longitudinal Project. Journal of Studies on demiological models of alcohol consumption and Among Young People. Copenhagen: World Health Alcohol 57(5):494–506, 1996. consequences for Western nations. Alcohol and Organization (WHO), Regional Office for Europe, 2000. KASKUTAS, L.A., AND GRAVES, K.L. An alterna­ Alcoholism 32(1):51–64, 1997. tive to standard drinks as a measure of alcohol PERNANEN, K. What is meant by “alcohol-related” DAWSON, D.A., AND ROOM, R. Towards agree­ consumption. Journal of Substance Abuse 12(1-2): ment on ways to measure and report drinking 67–78, 2000. consequences? In: Klingemann, H., and Gmel, patterns and alcohol-related problems in adult G., eds. Mapping the Social Consequences of Alco­ general population surveys: The Skarpö Con­ KNIBBE, R.A., AND BLOOMFIELD, K. Alcohol con­ hol Consumption. Dordrecht, Netherlands: ference overview. Journal of Substance Abuse sumption estimates in surveys in Europe: Kluwer Academic Publishers, 2001. pp. 21–31. 12(1-2):1–21, 2000. Comparability and sensitivity for gender differ­ ences. Substance Abuse 22(1):23–38, 2001. Productschap voor Gedistilleerde Dranken. DEMERS, A.; ROOM, R.; AND BOURGAULT, C.; eds. World Drink Trends 2000. Henley-on-Thames, Survey of Drinking Patterns and Problems in Seven KNIBBE, R.A., AND LEMMENS, P.H. Korrelate des England: NTC Publications, 2000. Developing Countries. WHO/MSD/MSB/01.2. Gen­ Alkoholkonsum in der Schweiz, Deutschland eva: World Health Organization (WHO), 2001. und den Niederlanden. Drogalkohol 11:27–41, RAMSTEDT, M. Comparative Studies on Alcohol- 1987. Related Problems in Postwar Western Europe. DURKHEIM, E. Suicide: A Study in Sociology. New Stockholm: Department of Sociology, University York: Free Press, 1951 [1897]. LEIFMAN, H. Homogenisation in alcohol con­ of Stockholm, 2001a. sumption in the European Union. Nordic Studies FEUNEKES, G.I.J.; VAN’T VEER, P.; VAN STA­ on Alcohol and Drugs 18(Suppl. 1):15–30, 2001. RAMSTEDT, M. Per capita alcohol consumption VEREN, W.A.; AND KOK, F.J. Alcohol intake and liver cirrhosis mortality in 14 European coun­ LEIFMAN, H. A comparative analysis of drinking assessment: The sober facts. American Journal of tries. Addiction 96(Suppl. 1):S19–S33, 2001b. Epidemiology 150(1):105–112, 1999. patterns in six EU countries in the year 2000. Contemporary Drug Problems 29(3):501–548, REHM, J., AND GMEL, G. Aggregate time-series FILLMORE, K.M.; HARTKA, E.; JOHNSTONE, 2002. regression in the field of alcohol. Addiction 96(7): B.M.; ET AL. Preliminary results from a meta­ 945–954, 2001. analysis of drinking behavior in multiple longitu­ LEMMENS, P.H. The alcohol content of self- dinal studies. British Journal of Addiction 86(10): report and “standard” drinks. Addiction 89(5): REHM, J.; GREENFIELD, T.K.; WALSH, G.W.; ET 1203–1210, 1991. 593–601, 1994. AL. Assessment methods for alcohol consump­ tion, prevalence of high risk drinking and harm: Finnish Foundation for Alcohol Studies. Inter- LEMMENS, P.H.; TAN, E.S.; AND KNIBBE, R.A. national Statistics on Alcoholic Beverages: Produc­ Measuring quantity and frequency of drinking in A sensitivity analysis. International Journal of tion, Trade and Consumption 1950–1972. Forssa, a general population survey: A comparison of five Epidemiology 28(2):219–224, 1999. indices. Journal of Studies on Alcohol 53(5):476– Finland: Aurasen Kirjapaino, 1977. REHM, J.; ROOM, R.; MONTEIRO, M.G.; ET AL. 486, 1992. GMEL, G.; REHM, J.; AND FRICK, U. Methodo­ Alcohol as a risk factor for burden of disease. In: logical approaches to conducting pooled cross- LEVINE, H.G. Temperance cultures: Concern World Health Organization (WHO). Com­ sectional time series analysis: The example of the about alcohol problems in Nordic and English- parative Quantification of Health Risks: Global association between all-cause mortality and per speaking cultures. In: Lader, M.; Edwards, G.; and Regional Burden of Disease Due to Selected capita alcohol consumption for men in 15 Euro­ and Drummond, D.C.; eds. The Nature of Risk Factors. Geneva: WHO, in press. pean states. European Addiction Research 7(3): Alcohol and Drug-Related Problems. New York: 128–137, 2001. Oxford University Press, 1992. pp. 15–36. ROOM, R. Region and urbanization as factors in drinking practices and problems. In: Kissin, B., GRAHAM, K.; WILSNACK, R.W.; DAWSON, D.A.; MÄKELÄ, P.; FONAGER, K.; HIBELL, B.; ET AL. and Begleiter, H., eds. The Biology of Alcoholism, AND VOGELTANZ, N.D. Should alcohol consumption Episodic heavy drinking in four Nordic coun­ vol. 6: The Pathogenesis of Alcoholism: Psycho- measures be adjusted for gender differences? tries: A comparative survey. Addiction 96(11): social Factors. New York: Plenum Press, 1982. Addiction 93(8):1137–1147, 1998. 1575–1588, 2001. ROOM, R. Cross-cultural research in alcohol GREENLAND, S., AND ROBINS, J.M. Invited com­ MATHERS, C.; VOS, T.; AND STEVENSON, C. The studies: Research traditions and analytic issues. mentary: Ecologic studies—biases, misconcep­ Burden of Disease and Injury in Australia. Canberra: In: Harford, R., and Towle, L., eds. Cultural tions, and counterexamples. American Journal of Australian Institute of Health and Welfare, 1999. Influences and Drinking Patterns—A Focus on Epidemiology 139:747–760, 1994. MORGENSTERN, H. Ecologic studies. In: Hispanic and Japanese Populations. NIAAA GRUENEWALD, P.J., AND NEPHEW, T.M. Drink­ Rothman, K.J., and Greenland, S., eds. Modern Research Monograph 19. Rockville, MD: ing in California: Theoretical and empirical anal­ Epidemiology. Philadelphia: Lippincott-Raven National Institute on Alcohol Abuse and yses of alcohol consumption patterns. Addiction Publishers, 1998. pp. 459–480. Alcoholism, 1988. pp. 9–40. 89(6):707–723, 1994. NORSTRÖM, T., ed. Alcohol and mortality: The ROOM, R., AND MÄKELÄ, K. Typologies of the HIBELL, B.; ANDERSSON, B.; AHLSTRÖM, S.; ET post-war experience in the EU countries. Addiction cultural position of drinking. Journal of Studies AL. The 1999 ESPAD Report: Alcohol and Other 96(Suppl. 1, entire issue):2001a. on Alcohol 61(3):475–483, 2000.

108 Alcohol Research & Health International Comparisons

ROSE, G. Sick individuals and sick populations. European Countries. Stockholm: National Institute WILSNACK, R.W.; VOGELTANZ, N.D.; International Journal of Epidemiology 14(1): of Public Health, European Commission, 2001. WILSNACK, S.C.; ET AL. Gender differences in 32–38, 1985. alcohol consumption and adverse drinking conse­ SINGLE, E., AND GIESBRECHT, N. The 16 per cent quences: Cross-cultural patterns. Addiction SIMPURA, J., AND KARLSSON, T. Trends in drinking solution and other mysteries concerning the accura­ 95(2):251–265, 2000. patterns among adult population in 15 European cy of alcohol consumption estimates based on sales World Health Organization (WHO). Global countries, 1950 to 2000: A review. Nordic Studies data. British Journal of Addiction 74:165–173, 1979. Status Report on Alcohol. Geneva: WHO, Sub- on Alcohol and Drugs 18(Suppl. 1):31–53, 2001. stance Abuse Department, 1999. STOCKWELL, T.; MCLEOD, R.; STEVENS, M.; ET SIMPURA, J.; KARLSSON, T.; AND LEPPÄNEN, K. World Health Organization (WHO). Inter- AL. Alcohol consumption, setting, gender and European trends in drinking patterns and their national Guide for Monitoring Alcohol Consump­ socioeconomic background. In: Norström, T., ed. activity as predictors of injury: A population- tion and Related Harm. Geneva: WHO, Alcohol in Postwar Europe: Consumption, Drinking based case-control study. Journal of Studies on Department of Mental Health and Substance Patterns, Consequences and Policy Responses in 15 Alcohol 63(3):372–379, 2002. Dependence, 2000.

Vol. 27, No. 1, 2003 109