Background: To examine the association between age at drinking onset and subsequent binge drinking, and to examine whether there are differences in this association between four countries.
Methods: The data consisted of 68 539 women aged 18–47 years randomly selected from the general population in Denmark, Iceland, Norway and Sweden. Frequency of binge drinking, defined as consuming >6 U of alcohol at the same occasion once or more per month, and age at drinking debut were assessed through a questionnaire survey.
Results: Overall, 12–26% reported binge drinking once or more per month in the four countries. Median age for starting drinking was 16 years in all four countries. Women who started drinking at 14 years or younger were significantly more likely to binge drink than women who started drinking at 19 years or older with adjusted odds ratios of 2.9 (95% confidence intervals 2.3–3.7), 2.8 (2.1–3.6) and 2.6 (1.9–3.4) for binge drinking in Denmark, Iceland and Sweden, respectively. Among Norwegian women the association was stronger with an adjusted odds ratio at 4.4 (3.5–5.6). The association in all four countries was more pronounced in women younger than 30 years than in older women. Conclusion: In the four Nordic countries, there is a strong relation between age at drinking onset and later binge drinking. The strong relationship found in countries with such different alcohol cultures is most likely generalizable to other Western countries.
age at drinking onset
It is well established that binge drinking is strongly associated with physical and social harm both to drinkers and to the people around them,1 and it has been shown that there are both acute and long-term detrimental effects of drinking in binges vs. a steady intake.2 From a public health point of view, it is therefore important to investigate predicting factors of binge drinking. Influence from family and peers have been associated with binge drinking as have socio-demographical factors and personality.3 Many of the same factors have been associated with age at drinking onset.4,5 Therefore, it is likely that age at drinking onset is associated with binge drinking in adulthood either as a causal factor through mechanisms as higher tolerance to alcohol, as a mediating factor between e.g. peer and family influence and binge drinking or as a predicting, but non-causal, variable.Few cross-sectional studies and very small prospective studies have suggested a positive association between age at drinking onset and alcohol consumption in adulthood.6–8 Other studies have suggested a positive association between age at drinking onset and alcohol abuse and dependence in adulthood,9,10 frequency of drinking in young ages,11 and between alcohol consumption in adolescence and binge drinking in adulthood.12 Most of these studies were performed in North American states, in most of which the legal age of alcohol consumption is 21 years, and where the alcohol policies differ from most European countries. It is unsure whether the association between age at drinking onset and later alcohol consumption is universal or dependent on nationality. Denmark is known to have, both presently and in the past, a more liberal alcohol policy than e.g. Iceland, Norway, and Sweden, both with regard to legal age of purchasing and drinking alcohol and to availability and legislation.13 Furthermore, the drinking cultures differ from each other across the countries: in Denmark a higher frequency of drinking alcohol and especially of binge drinking is reported than in Norway and Sweden.14,15 It is therefore plausible that an association between age at drinking onset and binge drinking differs as well. In this study, we look at the association between age at drinking onset and later alcohol consumption in Denmark, Iceland, Norway and Sweden, separately, to see whether differences in the association across nationalities exist or not.
The aim of the present large cross-sectional study is therefore to assess whether age at drinking onset is associated with binge drinking in adulthood controlled for possible confounders, and whether there are differences in this association between the Nordic countries.
From November 2004 to June 2005 a total of 22 199 Danish women (participation rate 81.4%), 15 051 Icelandic women (54.6%), 16 604 Norwegian women (68.0%) and 15 713 Swedish women (63.3%) from 18- to 45-years-old responded to a questionnaire about life stile, health and illness as a part of a study of associations between life stile and different types of cancer in women. The women were randomly chosen from the Central Population Register in each country, where each citizen has a unique personal identification number. The data collection have been described in details elsewhere.16
In the analysis, 81 women were excluded due to incomplete questionnaires or discrepancy between their personal identification number and year of birth reported. Furthermore, women with missing values on alcohol drinking debut (n = 526) or on binge drinking (n = 421) were excluded, leaving 68 539 women to be included in the study. In the analyses, women who had never been drinking (n = 3331) were also excluded due to the impossibility to have a positive outcome (i.e. binge drinking), thus in total 65 211 women were included in the analyses.
Binge drinking was defined as drinking >6 U of alcohol at the same occasion. We asked about the normal number of episodes of binge drinking per month: never, less than once, one to three times, four to eight times, nine times or more. If binge drinking once or more times per month the subjects were categorized as binge drinkers, and the remainder was categorized as non-binge drinkers.
The subjects were also questioned about age at drinking onset with the question ‘how old were you when you started to drink alcohol’ and with following possible answers: have never drunk, ≤12 years, 13–14, 15–16, 17–18, 19–20 or ≥21 years. The analyses were conducted with the following four groups: ≤14, 15–16, 17–18 and ≥19 years.
Alcohol intake was assessed through questions about the average number of drinks per month of ordinary beer, strong beer, red or white wine and hard liquor or fortified wine. The reported amounts were used to calculate the average intake per week with adjustment for differences in alcohol volume in the respective consumptions in the four countries so that 1 U of alcohol contained ∼12 g of ethanol.
Following covariates, chosen on a priori knowledge,3,5 were included in the analyses to control for possible confounding and were all based on self-reported data: Age at survey (grouped as <25, 25–29, 30–34, 35–39, ≥40 years), smoking habits (every day, at least once per week, less than once per week, ex-smoker, never smoker), age at smoking initiation (never smoked, ≤12, 13–14, 15–16, 17–18, ≥19 years), self-rated health (excellent, very good, good, fair, poor), marital status (married/cohabitant, single, separated/divorced, widow), educational level (<13, 13–16, >16 years), age at first intercourse (never, <14, 15–16, 17–18, ≥19 years) and number of sexual partners (0, 1–2, 3–4, 5–6, 7–8, 9–10, 11–20, >20 years).
Using logistic regression, adjusted odds ratios for binge drinking were estimated, according to age at drinking onset controlling for possible confounders (SAS package version 9.1). Women who started drinking at 19 years of age or later were chosen as the reference group for alcohol debut [odds ratio (OR) = 1]. The analyses were performed in three steps: one with adjustment for age, one with adjustment for age and the background variables: educational level, age at first intercourse and age at initiation of smoking. In the last analysis, adjustment was made for all of the covariates. Missing values of the covariates were not included in the analyses.
To assess whether there were differences between the four countries, all the analyses were stratified on nationality. Furthermore, an analysis stratified on age when responding were conducted. The two covariates were tested for interaction with the age at drinking onset with a cut-off level for significance at P = 0.10. In the last analyses the age at onset was collapsed into larger categories.
The study was approved in each country by the Data Protection Board and the Scientific Ethical Committee.
The median age among the 68 539 women at drinking onset was 16 years in all four countries (table 1).The proportion of the subjects who reported binge drinking once or more per month was highest among the Danish and Norwegian women with 25.8 and 24.3%, respectively, whereas it applied only to 14.9 and 11.7% in Iceland and Sweden. The Danish women were also those with the highest alcohol intake per week with a median intake of 2.5 U compared with 1.2, 1.7 and 1.4 U/week for Iceland, Norway and Sweden, respectively.
Only a small fraction of the subjects in all four countries had never been drinking alcohol, and for the Danish women only a small fraction (4.9%) started drinking at 19 years of age or later. In all four countries a higher age at drinking onset was associated with a smaller amount of alcohol intake both concerning alcohol intake per week, binge drinking and exceeding the alcohol limit of 14 U/week, which is the recommended limit of alcohol intake for women in the four countries. Furthermore, it was associated with less smoking and a higher age at onset of smoking and sexual activity. Concerning self-rated health, it was lowest among women who had never been drinking in Denmark, Norway and Sweden, whereas in Iceland the lowest self-rated health was seen among those who started drinking at 14 years old or younger. The highest fraction of subjects with a good self-rated health was seen for those who started drinking at 19 years or older in Iceland and Norway, at 17–18 years old in Sweden and at 15–16 years old in Denmark.
Among women who started to drink at age 14 years or younger there were strong significantly increased age-adjusted odds ratios for subsequent binge drinking of 3.6 (95% CI 2.9–4.6), 3.8 (2.9–4.8) and 4.5 (3.5–5.8) compared with an age at drinking onset at 19 years or older (ref, OR = 1) for Denmark, Iceland and Sweden, respectively, whereas Norway had an age-adjusted OR at 5.8 (4.7–7.2) for an age at drinking onset at 14 years or younger compared with over 19 years (table 2). Adjustment for educational level, age at first intercourse, smoking start, smoking status, self-rated health, marital status and number of sexual partners attenuated the relation between drinking debut and binge drinking a little, but still there was a highly significant association in all four countries (P < 0.001). After adjustment, the Danish, Icelandic and Swedish women who started drinking at an age of 14 years or younger had an OR for binge drinking of 2.9 (2.3–3.7), 2.8 (2.1–3.6) and 2.6 (1.9–3.4), respectively, compared with those who started drinking as 19 years old or older. For Norwegian women the corresponding odds ratio was 4.4 (3.5–5.6). When testing for differences between the four countries in the association between age at drinking onset and binge drinking a significant interaction was found (P < 0.001).
↵b: Adjusted for age, educational level, age at first intercourse and age at smoking start
↵c: Adjusted for the variables listed above, smoking status, self-rated health, marital status and number of sexual partners
The odds ratios for binge drinking when age at drinking onset was under 14 years compared with ≥19 years were higher for women younger than 30 years than for those over 30 years in all four countries (table 3). A test for interaction between age at drinking onset and the current age showed a significant interaction in Denmark (P = 0.004) and Sweden (P = 0.001), but not in Iceland and Norway (P > 0.1). The country specific differences in odds ratios diminished among the respondents over 30 years, but still an interaction between nationality and age at drinking onset was found (P = 0.07).
↵a: Adjusted for age, educational level, age at first intercourse, age at smoking start, smoking status, self-rated health, marital status and number of sexual partners
↵b: Significant interaction between age at alcohol debut and age
The present study shows that a young age at drinking onset is strongly associated with binge drinking in adulthood both in Denmark, Iceland, Sweden and Norway. We found a much stronger association in Norway than in the other three countries. In all four countries, the association between age at drinking onset and subsequent binge drinking tended to be strongest for the youngest women, and an interaction was found for current age and age at drinking onset in Denmark and Sweden, but not in Iceland and Norway.
The potential association between age at drinking onset and risk of later binge drinking has until now only been addressed in few and fairly small population based studies. Two previous studies from Finland and Norway with respectively 331 adults and 465 adolescents support our results,7,8 as does another study from North America with 2631 persons where the peak blood alcohol concentration was measured as an indicator of binge drinking.6 Other studies found an association between age at drinking onset and alcohol dependence and misuse,9,10 which also indicates that a young age at drinking onset is associated with later excess drinking including binge drinking.The finding of stronger associations between the age at drinking onset and binge drinking for younger ages than for older ages is also in line with previous studies of the association between alcohol consumption in adolescence and in adulthood.4,6,12 However, in contrast to these studies, we still found a significant association in Denmark, Iceland and Norway among the women older than 30 years in contrast to these studies. An explanation for this discrepancy may include the fact that other studies were based on much smaller study populations, or that they used different measures for alcohol intake (i.e. alcohol intake in adolescence, alcohol intake in adulthood and peak blood alcohol concentration, respectively). A decrease in the correlations between age at drinking onset and subsequent binge drinking with increasing age may however be expected, as variables more proximal in time, in general, are more likely to be more correlated than distal ones, especially when the variables measures related behaviours as in this case.6 Moreover, it is expected that women in their 20s could be more affected by factors during the maturational process than those in the late 30s and 40s. Nonetheless, the decreases could also be due to cohort differences because of the cross-sectional study design. However, the findings from the longitudinal studies by Labouvie and Jefferis, where no relationship between age at drinking onset and subsequent drinking, and amount of drinking in adolescence and subsequent binge drinking, respectively, was found for women in older ages,4,12 speak against this explanation as the only one.The cross-sectional design of our study implies that inference regarding causality cannot be made. However, several factors have been suggested to explain the relationship, both as a non-causal and as a causal relationship.17 Both parental and peer influence have been found to be associated with age at drinking onset and binge drinking, and it is suggested that they are underlying factors for the relation.3,5 Hawkins et al.18 found that age at drinking onset mediated the relation between parental and peer influence and alcohol consumption, and therefore suggested a causal relationship. Other mechanisms between age at drinking onset and subsequent binge drinking could be a greater tolerance and habituation of alcohol when started drinking in a young age compared to those who started drinking in older ages.
Some have suggested a non-causal association with a more direct relationship between other underlying factors and the alcohol consumption, hence the age at drinking onset is only seen as an indicator for alcohol consumption.6,17 Possible limitations in this study are a lack of information on possible confounders such as family and peer influence and socioeconomic position; factors that may be associated with binge drinking and age at drinking onset.3,5 Furthermore, information on binge drinking and age at drinking onset was collected via self-administered questionnaires. The validity of this method has been assessed in the Danish part of the MONICA project.19 Here, information on alcohol intake obtained by frequency questionnaire was compared with information on alcohol intake obtained by dietary interview. A close agreement between the two information sources was observed. Hence, the self-reported alcohol intake must be seen as a valid measure of the real alcohol intake.The validation of self-reports of age at drinking debut has been investigated by Labouvie et al.4 They found that a later age at drinking onset was reported when the same subjects were asked at 18 years old than at 15 years. However, the women in the present study are all older than 18 years, so the study of Labouvie et al. cannot directly be applied to our results. Nevertheless, there will still be a risk of this kind of misclassification of the age at drinking onset in the study. By analysing the data in two age groups, we tried to reduce the consequences of the misclassification, but it may still have an impact on the results. However, traditional recall bias are unlikely to occur, as it is difficult to imagine that binge drinkers will classify themselves with a younger or respectively older age at drinking onset than non-binge drinkers. Hence, the possible misclassification is non-differential and may affect the estimates toward the null value. Further, although a binge-drinking pattern later in life may influence the recall of ones drinking onset, the likelihood of any bias in this context seems small. Another possible limitation could be the risk of inherent truncation bias,20 in that older subjects have the possibility of reporting older ages at onset than the younger age groups.Strengths of our study include the large study population that is sampled with a representative sampling method and with fairly high response rates, which help reduce potential selection bias. Moreover, the study is based on four different nationalities with very different drinking cultures and policies in Europe,14,15 which all together make the results generalizable to other Western countries. Furthermore, the age range of the respondents is wide, making it possible to assess differences in the relationship between age at drinking onset and subsequent binge drinking with increasing age.
However, there is still a need for a prospective study to be conducted with a larger sample than previous ones and for a longer sampling period so any eventual recall bias and inherent truncation bias can be reduced. Furthermore, there is a need for studies of the mechanisms or underlying factors behind the relationship.
Our findings suggest that age at drinking onset is an indicator for subsequent binge drinking. The association may be causal, which would imply that a delay in age at drinking onset would cause a reduction in binge drinking. Meanwhile, if the association is non-causal, the age at onset can be used as an indicator for high-risk groups on binge drinking, and an intervention could be addressed directly to those who started to drink in a young age. In that binge drinking is seen to have a lot of harmful effects both on long and short term it is advisable to have a greater focus on the age at drinking onset, either by delaying the onset or by addressing interventions on people with a young age at drinking onset, and hereby reduce binge drinking in adulthood.
In all Nordic countries, there is a strong relationship between age of drinking onset and later binge drinking. The strong relation found in countries with such different alcohol cultures is most likely to be generalizable to other Western countries.
The Danish Ministry of Health and Prevention (government department); Merck & Co., Inc. (research grant EPO 8014.016).
Conflicts of interests: None declared.
The main findings of the study is a strong positive relation between age at drinking onset and subsequent binge drinking for women aged 18–47 years in four Nordic countries.Due to the different alcohol policies and different alcohol culture in the four countries, the relationship is most likely generalizable to other Western countries.
By delaying the drinking onset or by addressing binge-drinking interventions on people with a young age at drinking onset, binge drinking in adulthood might be reduced.
The authors would like to thank Erik Dasbach from Merck Research Laboratories, Merck & Co., USA, who contributed to the data collection.