Interventions delivered in primary and secondary schools to prevent and/or reduce alcohol use by young people under 18 years old

Transcription

1 Interventions delivered in primary and secondary schools to prevent and/or reduce alcohol use by young people under 18 years old Epidemiology review and development of an epidemiological model Dr Tom Jefferson 1, Lisa Jones 2, Professor Mark Bellis 2 1 Cochrane Vaccines Field, Anguillara Sabazia, Rome, Italy; 2 National Collaborating Centre for Drug Prevention, Centre for Public Health, Liverpool John Moores University, Liverpool, UK April 2007

2 CONTRIBUTION OF AUTHORS Dr Tom Jefferson Lisa Jones Professor Mark Bellis Undertook study selection, data extraction and synthesis of the data for the review of mortality and morbidity. Developed and populated the evidence-based disease incidence model. Contributed to the writing of the report. Undertook study selection, data extraction and synthesis for the sections of the report on prevalence and the relationship between adolescent and adult drinking, and adolescent and acute outcomes. Development of the conceptual model and checking data entry. Overall responsibility for the layout and content of the review. Responsible for overall management of the project. Commented on all drafts of the report. ACKNOWLEDGEMENTS We wish to thank Karen Hughes, Michela Morleo and Dr Harry Sumnall from the Centre for Public Health for their comments on the draft versions of this report. 2

3 CONTENTS 1 Introduction Methods Search strategy Inclusion criteria Methods of analysis Prevalence of alcohol use Recommended sensible drinking guidelines Overview of surveys available General population surveys in young people General population surveys in adults Surveys of hazardous/harmful and alcohol dependent drinkers Discussion Change in alcohol-related behaviour between adolescence and adulthood Results of literature review Overview of studies which examined changes in drinking behaviour between adolescence and adulthood Overview of studies which examined the effects of age of drinking onset on adult drinking behaviour Discussion Adolescent drinking and acute outcomes Results of the literature review Overview of studies that examined the association between adolescent alcohol consumption and acute outcomes Discussion Mortality and morbidity associated with alcohol consumption Results of the literature review Overview of studies which examined mortality and morbidity associated with alcohol consumption Evidence-based disease incidence model Sensitivity analysis Discussion Conceptual model Adolescent drinking model Adult drinking model Transition from adolescent to adult patterns of drinking References

4 LIST OF APPENDICES Appendix 1. Search strategy Appendix 2. Formulae used to convert RR and OR into RD Appendix 3. Data extraction tables Appendix 4. Data for disease burden model LIST OF TABLES Table 1. Characteristics of surveys... 9 Table 2. Percentage of year olds who had ever had an alcoholic drink, by age and sex in 2005 (Becker et al., 2006) Table 3. Maximum daily amount drunk last week by age and sex in 2005 (Goddard 2006) Table 4. Mean number of units per week in 2004 (Lader & Goddard 2006) Table 5. Odds ratios (95% CI) for binge drinking at different ages in adulthood according to drinking levels in adolescence (Jefferis et al. 2005) Table 6. Estimated frequency of time varying adolescent measures and their association with alcohol dependence in frequent alcohol users at age 20 years (Bonomo et al. 2004) Table 7. Mean adult alcohol use (at age 36/42) by age of onset of drinking and gender (Pitkänen et al. 2005) Table 8. Adjusted odds ratios and 95% confidence intervals for involvement in physical fights, having been injured in physical fights and injuring others in physical fights by drinking characteristics among youth drink alcohol (Swahn et al., 2004) Table 9. Number of alcohol-related expected cases in the 2005 birth cohort by outcome Table 10. Number of expected cases in 2005 birth cohort at different levels of daily alcohol consumption Table 11. Number of expected cases in 2005 birth cohort using RR estimates from Corrao et al., Table 12. Number of expected cases of lung cancer in 2005 birth cohort using alternative estimates from Freudenheim et al. (2005) Table 13. Studies that examined change in alcohol-related behaviour between adolescence and adulthood Table 14. Studies that examined an association between adolescent drinking and acute outcomes Table 15. Causality evidence from systematic reviews and meta-analyses Table 16. Causality evidence from primary studies Table 17. Calculation of expected cases in 2005 birth cohort

5 Table 18. Sensitivity analysis (1): Calculation of expected cases in 2005 birth cohort at different levels of daily alcohol consumption Table 19. Sensitivity analysis (2): Calculation of expected cases in 2005 birth cohort using RR estimates from Corrao et al., Table 20. Sensitivity analysis (3): Calculation of expected cases of lung cancer in 2005 birth cohort using alternative RR estimates

6 1 INTRODUCTION The consumption of alcohol has both health and social consequences. In addition to the range of negative health outcomes associated with chronic alcohol use, alcohol contributes to traumatic outcomes through violence and injury. There is increasing evidence that in addition to the volume of alcohol consumed, patterns of drinking may be relevant for the negative health outcomes (WHO Global Status Report on Alcohol 2004). An association between alcohol consumption and many types of disease and injury has been established (Single et al., 2001; see Box 1). Box 1. Examples of disease and injury associated with alcohol consumption Alcoholic psychosis Alcohol dependence Alcohol abuse Alcoholic polyneuropathy Alcoholic cardiomiopathy Alcoholic gastritis Alcoholic liver cirrhosis Ethanol toxicity Other alcoholic poisonings Lip cancer Oral cancer Pharyngeal cancer Oesophangeal cancer Colon cancer Rectal cancer Hepatic cancer Pancreatic cancer Laryngeal cancer Breast cancer Pellagra Hypertension Ischaemic heart disease Cardiac dysrhythmias Heart failure Stroke Oesophangeal varices Gastro-eosophangeal haemorrage Cholelithiasis Acute pancreatitis Low birthweight Road injuries Fall injuries Fire injuries Drowning Aspiration Machine injuries Suicide Assault Child abuse Adapted from Single et al The purpose of this epidemiological review is to explore the prevalence and patterns of alcohol consumption in the general population, explore the relationship between adolescent drinking behaviours and drinking in adulthood, and to determine the mortality and morbidity risk associated with alcohol use. The review is divided into five main sections: Section (3) examines prevalence data from national surveys which have examined drinking in the general population; Section (4) reviews the literature which has examined the relationship between adolescent and adult drinking patterns; Section (5) reviews studies which have examined the association between adolescent drinking behaviour and a range of acute outcomes; Section (6) reports on the development of an evidence-based disease incidence model which examined the association between the consumption of alcohol and a range of acute and chronic events in adults; and Section (7) presents a conceptual framework for the consumption of alcohol and its consequences. The review is intended as the first step in the development of a conceptual framework. This framework will be used to form the structure for an economic model that will examine the cost- 6

7 effectiveness of interventions delivered in primary and secondary schools to prevent and/or reduce alcohol use by young people under 18 years old. 2 METHODS 2.1 Search strategy English language studies published since 1990 were identified by searching major medical databases; principally MEDLINE, EMBASE and CINAHL. The search strategy, developed by the Centre for Reviews and Dissemination at the University of York, is presented in Appendix 1. In addition, reference lists of retrieved studies, key documents (e.g. documents related to the Alcohol Harm Reduction Strategy for England, ACMD Pathways to Problems) and relevant websites (e.g. Department of Health) were searched. 2.2 Inclusion criteria National surveys that examined the prevalence of alcohol consumption in the UK or England were included. International surveys were not sought, as good quality surveys with a national focus were judged to be available for both young people and adults. Studies that reported changes in drinking behaviours between adolescence and adulthood, or an association between alcohol consumption and the risk of acute and chronic outcomes were included if they met the following criteria: English language Populations of children, young people or adults from developed countries 1 Published from 1990 onwards Initially, study inclusion was considered according to the hierarchy of evidence (Guyatt et al., 1995; Guyatt et al., 2000). We prioritised the inclusion of systematic reviews and/or meta-analyses, but if these types of studies were unavailable or did not report sufficient data, then the inclusion of casecontrol or cohort studies was considered against the inclusion criteria described above. When evidence was still unavailable then the inclusion of cross-sectional surveys, case series/reports and expert opinion was considered. 2.3 Methods of analysis All included studies were extracted following the tabular description contained in the Cochrane Review Manager software (Version 4.2 for Windows), under the following headings: methods, participants, results and notes. The results of studies that examined the prevalence of drinking, change in alcohol-related behaviours, or which examined acute outcomes associated with alcohol 1 Studies were included if they were conducted in Europe, North America, Australia or New Zealand. Geographical restrictions were introduced in order to assess the effects of drinking only where drinking patterns were likely to be similar to those of the UK population. 7

8 consumption in adolescence are presented in a narrative overview. Studies which examined mortality and morbidity associated with adult alcohol consumption were used to develop the disease incidence model. Development of the evidence-based disease incidence model Whenever possible we used evidence from systematic reviews, augmented and updated with evidence from studies published after the review searches were carried out. Measures of the association between exposure to alcohol and outcome, reported as odds ratios (ORs), relative risks (RRs) and β coefficients 2, were extracted from the included studies. The β coefficients were converted to RR estimates by taking the exponential of the β coefficient. The relationship between the RR and β is explained by the following equations: ln RR = βx Exp (βx) = RR where x is equal to unit of alcohol consumption (g/day) To determine the difference in the level of risk between populations exposed and unexposed to alcohol, estimates were converted into risk difference (RD). The latest available incidence data for the general population of England 3 (or sometimes England and Wales) (see Appendix 2 for statistical methods used) was used as a proxy to determine the level of risk in unexposed populations. If incidence in the general population was presented by gender, this was converted to an all gender rate. Only those outcomes, for which the recent literature has provided consistent evidence for a causal relationship on alcohol-related pathogenesis (Rehm 2003a and b), were incorporated into the birth cohort model. The pathology or condition-specific RDs calculated in the causality model were then applied to the England and Wales 2005 birth cohort (n= 645,835). This allowed for the definition of an evidence-based disease incidence and progression model based on the origin and progression of alcohol-related pathologies in the general population (the birth cohort model ). Data relating to the alleged protective effects of alcohol were not incorporated into the model given the weakness and uncertainty of the evidence (Corrao et al., 2000). Whenever possible, data from the most conservative levels of exposure to alcohol were incorporated in the model. This was due to the complexity of the dose-response gradient, the observed heterogeneity and variable methodological quality, and differences in disease coding and definition in the identified studies. 2 The β coefficient expresses the variation of the natural logarithm of the relative risk. 3 Incidence data were extracted from Hospital Episode Statistics data for 2005/2006 and cancer incidence data were extracted from Cancer Statistics Registrations for

9 3 PREVALENCE OF ALCOHOL USE 3.1 Recommended sensible drinking guidelines In 1995, the Government published new sensible drinking guidelines based on daily consumption of alcohol. They recommend a maximum intake of 3-4 units of alcohol per day for men and 2-3 units of alcohol per day for women. Similar guidelines do not exist for children or young people s drinking. 3.2 Overview of surveys available Prevalence of alcohol use in the general population was drawn from a number of national surveys. Drinking behaviour among young people is measured in an annual survey of secondary school children in England. In addition, the European School Survey Project on Alcohol and Other Drugs (ESPAD), a Europe-wide survey provides estimates of the prevalence of drinking in young people aged years old. Adult drinking in the general population is collected via the General Household Survey, a continuous survey of UK adults aged 16 and over. In addition, the Office for National Statistics (ONS) Omnibus survey has surveyed adults behaviour and knowledge, most recently in The Alcohol Needs Assessment Research Report (ANARP) was commissioned by the Department of Health to measure the gap between the demand for and provision of specialist alcohol treatment services in England. As part of the research available data was collected on the prevalence of alcohol use disorders 4 in England. Table 1 reports the main features of the surveys used. Table 1. Characteristics of national surveys of alcohol use Survey name Sample Follow-up Drug use, smoking and Survey of 9,202 secondary school children in England aged 11 to The survey has drinking among young 15. Both schools and pupils randomly selected. been carried out people in England in 2005 annually since (Becker et al., 2006) Core questions cover: (1) smoking status, and the number of cigarettes smoked in the last week; (2) drinking status, and alcoholic drinks consumed in the last week; and (3) awareness of individual drugs, which drugs pupils have been offered, whether pupils ever tried each drug, and when drugs were last used European School Survey 2,068 UK students born in 1987 (aged at the time of data Repeated every 4 Project on Alcohol and collection) were randomly sampled from participating schools. years. Other Drugs Report 2003 (ESPAD) (Hibell et al., Core questions cover tobacco use, alcohol consumption, 2004) drunkenness, binge drinking and illicit drug use. 4 The World Health Organisation (WHO) defines three levels of alcohol use disorder: hazardous, harmful and dependent. Hazardous drinkers are those who drink above sensible levels but who are not yet experiencing significant alcoholrelated problems. Harmful drinkers are those who drink above sensible levels and are experiencing damage to their health and show evidence of alcohol-related harm. Alcohol dependence is characterised by psychological dependence referring to the experience of impaired control over drinking. In more severe cases, alcohol dependent drinkers may experience physical dependence. 9

10 General Household Survey (GHS) 2005: Smoking and drinking among adults, 2005 (Goddard 2006) ONS Omnibus survey (Lader and Goddard 2006) Alcohol Needs Assessment Research Report (ANARP) (Drummond et al., 2005) Survey of people living in private households in the UK. For the 9 months covered by the survey, sample size of 12,802 households and 30,069 adults aged 16 and over. The GHS collects data on a wide range of topics. The sample population consisted of 3,528 adults aged 16 or over residing throughout Great Britain. Questions about drinking were included in the Omnibus survey at the request of the Department of Health. The ANARP is an audit of treatment, giving an assessment of the levels of alcohol misuse and the availability of alcohol treatment services throughout England. As part of the project, surveys conducted by the ONS were re-analysed and the number of hazardous, harmful and alcohol dependent drinkers determined. Continuous survey, data reported annually. Annual survey One-off 3.3 General population surveys in young people The most recent data indicates that the overall prevalence of drinking increases sharply between the ages of 11 and 15 (Becker et al., 2006). In 2005, 22% of 11 year olds had drunk alcohol compared to 86% of 15 year olds (see Table 2). Twenty-two percent of year olds had drunk alcohol in the past week. The proportion of pupils who reported drinking in the past week also increased by age, from 3% of 11 year olds to 46% of 15 year olds. The mean alcohol consumption of year olds who had drunk in the last week was 10.5 units; year olds reported drinking an average 8.2 units and 15 year olds reported an average of 11.8 units. Table 2. Percentage of year olds who had ever had an alcoholic drink, by age and sex in 2005 (Becker et al., 2006) Age Total % % % % % % Boys Girls Total Data from the ESPAD 2003 survey (Hibell et al., 2004) found than 91% of year olds in the UK reported drinking alcohol in the past year; 34% reported drinking 20 times or more in the past year. Seventeen percent of year olds reported drinking alcohol 10 times or more during the last 30 days. The average consumption of alcohol on the last drinking occasion was estimated at 10.2 cl. Seventy-five percent reported that they had been drunk at least once; 27% reported having been drunk 20 times or more in their lifetime. Drunkenness in the last year was also measured, 68% reported one occasion of drunkenness in the last year and 24% reported being drunk 10 times or more during the last year. Forty-six percent reported being drunk in the last 30 days with 23% having been drunk three times or more. Thirty-six percent reported having been drunk at age 13 or younger. Binge drinking was defined as having five or more drinks in a row. Fifty four percent of year olds in the UK met the criteria for binge drinking in the past 30 days, and 27% reported binge drinking three times or more during the last 30 days. In the UK, 61% of respondents had consumed at least one glass of beer or wine by the age of

11 3.4 General population surveys in adults Data from the General Household Survey 2005 (Goddard 2006) indicated that men were more likely than women to have had a drink in the previous week; 72% of men and 57% of women reported having had a drink on at least one day during the previous week. Among men, those aged years and those aged 65 years and over were least likely to report drinking alcohol during the previous week. Men reported drinking on more days of the week than women; 22% of men compared to 13% of women reported drinking on at least five of the seven preceding days. Compared to women, men were more likely to have exceeded the recommended daily intake on at least one day during the previous week (35% of men vs. 20% of women). Men were also twice as likely to have drunk heavily 5 (19% of men vs. 8% of women). Respondents aged were significantly more likely than other age groups to report exceeding the recommended number of daily units on at least one day (see Table 3). The same pattern was evident with reports of heavy drinking (see Table 3). Table 3. Maximum daily amount drunk last week by age and sex in 2005 (Goddard 2006) Max daily amount Age and over Total % % % % % Men Drank nothing last week Up to 4 units More than 4, up to 8 units More than 8 units Women Drank nothing last week Up to 3 units More than 3, up to 6 units More than 6 units All persons Drank nothing last week Up to 4/3 units More than 4/3, up to 8/6 units More than 8/6 units Data collected in the ONS Omnibus survey (Lader & Goddard 2006) found that on average, adults reported drinking 11.1 units a week (see Table 4). Men reported drinking more than women (16.0 units of alcohol a week vs. 7.0 units per week, respectively). Table 4. Mean number of units (SE) per week in 2006 (Lader & Goddard 2006) Total units Men Women Total Total (2.64) (1.11) (1.51) (1.10) (0.76) (1.05) (0.85) (0.38) (0.41) (0.36) 5 Defined as drinking more than eight units on one day for men and more than six units for women 11

12 3.5 Surveys of hazardous/harmful and alcohol dependent drinkers Data from ANARP (Drummond et al., 2005), based on a reanalysis of surveys conducted by the ONS, indicated that approximately 38% of men and 16% of women aged between 16 and 64 have an alcohol use disorder (26% overall). Of these, 32% of men and 15% of women can be classified as hazardous or harmful alcohol users (23% overall). Twenty one percent of men and 9% of women are binge drinkers. The overall prevalence of alcohol dependence was found to be 3.6%, with 6% of men and 2% of women meeting this criteria. The study found that there was a decline in all alcohol use disorders with age. Younger people showed a higher prevalence of both hazardous/harmful drinking and alcohol dependence than older people (see Figure 1). Figure 1. Percentage of people in England with an alcohol use disorder by age (reproduced from Drummond et al., 2005) In addition, black and minority ethnic groups were found to have a lower prevalence of hazardous/harmful alcohol use, but share a similar prevalence of alcohol dependence with the white population (see Figure 2). 12

13 Figure 2. Percentage of people in England with an alcohol use disorder by ethnicity (reproduced from Drummond et al., 2005) 3.6 Discussion Data from national surveys of drinking behaviour in young people indicate that by the age of years, the vast majority of young people have tried their first alcoholic drink. In addition, at age 15 nearly half of young people are consuming alcohol on a weekly basis and at levels similar to adults. Data from the ESPAD survey found that nearly a quarter of young people aged get drunk regularly and that binge drinking is prevalent. Binge drinking habits continue into young adulthood, more than a third of year olds report drinking over the sensible drinking daily limits. In addition, a quarter of young adults may be classified as having an alcohol use disorder. 13

14 4 CHANGE IN ALCOHOL-RELATED BEHAVIOUR BETWEEN ADOLESCENCE AND ADULTHOOD 4.1 Results of literature review Five studies were identified which examined changes in drinking behaviour between adolescence and adulthood. Three studies examined patterns of drinking in adolescence and their impact on adult drinking behaviours and two studies examined how age of onset of alcohol use affected later indications of problem drinking. Characteristics of these studies are presented in Appendix Overview of studies which examined changes in drinking behaviour between adolescence and adulthood Jefferis et al. (2005) assessed continuities in binge drinking across adulthood and the association between adolescent drinking level and adult binge drinking. The results suggested that women, who rarely or never drank at age 16, were less likely than light drinkers (0-2 units) to binge drink in adulthood. However, male light drinkers were no more likely than non-drinkers to binge drink as adults. Drinking 3 to 6 units in the past week at 16 years compared to 0 to 2 units increased the odds of adult binge drinking at each adult age in men and at 33 and 42 years in women. The heaviest drinkers at 16 years ( 7 units) were significantly more likely to binge drinking at each age in adulthood for men and at 42 years for women. Further analysis showed that the effects of adolescent drinking on binge drinking were similar across ages 23, 33 and 42 for men, whereas for women there was a significant difference, there was a stronger effect of adolescent drinking on adult binge drinking at age 42 than at ages 23 or 33. Table 5 presents odds ratios for drinking at different ages in adulthood according to drinking levels in adolescence. Table 5. Odds ratios (95% CI) for binge drinking at different ages in adulthood according to drinking levels in adolescence (Jefferis et al. 2005) Adolescent drinking a Adult binge Age (years) Rarely/never drink 0-2 units 3-6 units 7 units drinking Men n= (0.89, 1.25) (1.13, 1.70) 2.07 (1.71, 2.51) n= (0.71, 1.03) (1.00, 1.58) 1.65 (1.33, 2.04) n= (0.85, 1.21) (1.02, 1.59) 1.64 (1.33, 2.08) Women n= (0.55, 0.77) (0.93, 1.45) 1.43 (0.94, 2.12) n= (0.60, 0.90) (1.01, 1.74) 0.96 (0.55, 1.67) n= (0.83, 1.21) (1.02, 1.78) 2.88 (1.85, 4.48) a units of alcohol consumed in previous week McCarty et al. (2004) tested the hypothesis that late adolescent drinking behaviour was associated with harmful and binge drinking in adulthood. Harmful drinking at ages 17 to 20 was associated with an increased risk of harmful drinking at ages 30 to 31 for men [RR 2.71 (95% CI: 1.63, 4.48)], but did not reach significance for women [RR 1.43 (95% CI: 0.83, 2.46)]. Binge drinking at ages 17 14

15 to 20 also increased the risk of binge drinking at ages 30 to 31 for both men [RR 2.34 (95% CI: 1.81, 3.04)] and women [RR 3.38 (95% CI: 2.38, 4.78)]. The aim of the study by Bonomo et al. (2004) was to determine whether adolescent alcohol use predisposed participants to alcohol dependence in young adulthood. The authors found that frequent drinking and binge drinking in adolescence both showed strong associations with alcohol dependence in young adulthood. Participants who reported recurrent frequent drinking in adolescence had significantly increased odds for later dependence [OR 8.1 (95% CI: 4.2, 16)], as did those who reported recurrent binge drinking [OR 6.7 (95% CI: 3.6, 12)]. In addition, the likelihood of alcohol dependence increased with persistence of frequent drinking through adolescence [OR for frequent drinking at one wave: 2.0 (95% CI: 1.0, 4.3); OR for frequent drinking at multiple waves: 3.1 (95% CI: 1.2, 7.7)]. Table 6 shows the estimated frequency of time varying adolescent measures and their association with alcohol dependence in frequent alcohol users at age 20 years. Table 6. Estimated frequency of time varying adolescent measures and their association with alcohol dependence in frequent alcohol users at age 20 years (Bonomo et al. 2004). Adolescent measure: waves 1-6 Frequent drinking Binge drinking Category Estimated frequency Alcohol dependence at age 20 years n 95% CI OR 95% CI None , One wave , , 8.4 More than one wave 88 68, , 16 None , One wave , , 6.7 More than one wave , , 12 Wells et al. (2004) investigated the pattern of drinking in mid-adolescence and subsequent outcomes across the major domains of life in late adolescence and early adulthood. Data were used for 953 respondents to the Christchurch Health and Development Study, assessed up to 25 years. The authors found that four latent classes were required to describe patterns of drinking at age 16. Examination of these patterns showed a progression across the four classes from those in class 1 who had not consumed any alcohol in the past 3 months (24% of the sample) through to alcohol abusers in latent class 4 who drank often, consumed large amounts and reported a number of alcohol-related problems (9% of the sample). Alcohol consumption and alcohol dependence in the periods years and years all showed a strong linear trend that increased with latent class (p<0.002). That is, worse outcomes were more common for adolescents who drank more at age 16. After controlling for background and correlates, drinking behaviours consistently related to drinking at age 16 over both age periods (16-21 years and years) were drinking at least weekly, amount per last occasion (past year), largest amount on a single occasion (past year), and alcohol dependence. 15

16 4.3 Overview of studies which examined the effects of age of drinking onset on adult drinking behaviour Warner and White (2003) examined the relationship between age of first use, context of alcohol initiation and problem drinking. The mean age of drinking onset in the sample was 10.7 years. First drinking experience at a family gathering was reported by 73% of the sample (mean age of initiation 8.6 years; 81% reported drinking before age 11). For those who initiated drinking outside of the family, the mean age of onset was 14.2 years, with 18% reporting onset of drinking under the age of 11. Compared to participants who initiated drinking at older ages and at a family gathering, both early initiates who first drank at a family gathering and early initiates who first drank outside a family gathering had significantly higher odds of developing problems associated with alcohol use [OR 2.86 (95% CI: 1.36, 6.00); OR 8.32 (95% CI: 2.28, 30.41), respectively]. When there was a relatively faster transition from first drinking in a family context to drinking outside (<5 years), the odds of problem drinking were significantly greater than when five or more years had elapsed [OR 2.54 (95% CI: 1.45, 4.42)]. The aim of the study by Pitkänen et al. (2005) was to investigate the relationship between age of onset of drinking and several indicators of alcohol use. The mean age of onset of drinking among participants in the study was 15.5 years (range years, SD = 2.4). Two percent had begun drinking at age Participants who began drinking at 13 years scored significantly higher on all indicators of adult use of alcohol (frequency of drinking, binge drinking and alcoholism screening tests) than the oldest group ( 18 years), and year olds with the exception of women s binge drinking. Table 7 shows the means and standard deviations of the sample for four indicators of adult alcohol use according to age of drinking onset (annual frequency of drinking, binge drinking during the past 12 months, CAGE score and Mm-MAST score). Table 7. Mean adult alcohol use (at age 36/42) by age of onset of drinking and gender (Pitkänen et al. 2005). Males n=162 Females n=146 Indicator Onset age group Mean SD Mean SD Annual frequency of 13 years or less drinking ( years days/year) years Binge drinking (0-5 occasions) CAGE (0-8 points) Mm-MAST (0-9 points) 18 years or more years or less years years years or more years or less years years years or more years or less years years years or more

17 4.4 Discussion The studies identified have shown that people who binge drink in adolescence are more likely to be binge drinkers as adults. In addition, frequent drinking and binge drinking have been shown to increase the risk of developing alcohol dependence in young adulthood. Two studies indicated that early onset of alcohol drinking can increase the risk of later dependence and that people who start drinking early are more likely to drink frequently and to binge drink. 17

18 5 ADOLESCENT DRINKING AND ACUTE OUTCOMES 5.1 Results of the literature review Five primary studies were identified that examined the association between adolescent alcohol consumption and acute outcomes. These studies examined a variety of outcomes included sexual risk behaviours, risk of fighting and injuries, and juvenile offending. 5.2 Overview of studies that examined the association between adolescent alcohol consumption and acute outcomes Sexual behaviour Stueve and O Donnell (2005) examined the relationship between early alcohol use and subsequent alcohol and sexual risk behaviours in a sample of 1,034 Black and Hispanic adolescents. Around a quarter of the sample (26%) reported lifetime use of alcohol use at age years compared to 63% at age years. Early initiation of alcohol use was significantly associated with recent alcohol use [adjusted OR 2.40 (95% CI 1.75, 3.29)], binge drinking in the past month [adjusted OR 1.87 (95% CI 1.25, 2.80)], getting drunk or high in the past year [adjusted OR 2.01 (95% CI 1.23, 3.27)], having more than two lifetime sexual partners [adjusted OR 1.54 (95% CI 1.10, 2.26)] and lifetime pregnancy [adjusted OR 1.73 (95% CI 1.10, 2.70)]. However, there was no association between early initiation of alcohol use and sexual initiation or having sexual intercourse within the past 3 months. Sen (2002) investigated the association between alcohol use and adolescent sexual activity using data from 8,984 respondents of the National Longitudinal Study of Youth The results of the model indicated a strong association between alcohol use and sexual activity. Alcohol use was associated with 20 27% (20 37%) increase in the probability of sexual intercourse and an 11 16% (10 17%) increase in the probability of intercourse without contraception for adolescent girls (boys in brackets). Heavy drinking had weak effects on sexual intercourse. Fergusson and Lynskey (1996) examined the associations between alcohol misuse and measures of early onset sexual activity and sexual risk taking behaviours. Data from 953 children aged years were collected as part of the Christchurch Health and Development Study, a 16-year longitudinal study of a birth cohort of children born during mid Both boys and girls who reported misusing alcohol 6 (n=79) had higher rates of sexual intercourse, were more likely to report multiple (three or more) partners, and reported higher rates of unprotected intercourse than those who did not misuse alcohol. After adjustment for common and correlated risk factors there were 6 Using data gathered on the frequency of drinking, amounts consumed and alcohol-related problems in the last year, the authors identified a proportion of the sample that engaged in frequent, heavy or problem drinking. 18

19 small to moderate associations between alcohol misuse and early onset sexual activity [boys: OR 2.9 (95% CI 1.4, 6.0); girls: OR 6.2 (95% CI 1.6, 23.4)], and alcohol misuse and unprotected intercourse [boys: OR 6.9 (95% CI 2.5, 18.9); girls: OR 4.5 (95% CI 1.7, 11.9)]. There was no significant relationship between alcohol misuse and multiple partnerships [boys: OR 1.3 (95% CI 0.4, 4.1); girls: OR 1.9 (95% CI 0.7, 5.5)]. Further analyses of the same cohort by Wells et al. (2004) found that drinking at 16 was significantly related to the number of sexual partners at ages years and years. The number of sexual partners, sexually transmitted infection and pregnancy increased linearly with increasing consumption of alcohol at 16. Violence and injuries Swahn et al. (2004) examined the associations between specific alcohol use measures and physical fighting, injuries received and injuries inflicted on others whilst fighting. Data were examined for 8,885 adolescents aged years who responded to the National Longitudinal Study of Adolescent Health and who reported consuming at least one alcoholic drink in the past year. Fully adjusted multivariate logistic regression models with all alcohol variables included showed that adolescent drinkers who reported frequent drinking (9-30 days/month), binge drinking, problem drinking or peer drinking were more likely to be involved in all three violence and injury outcomes (fighting in past 12 months, injured in fight in past 12 months, injured others in past 12 months) than drinkers who did not report these patterns (see Table 8). Table 8. Adjusted odds ratios and 95% confidence intervals for involvement in physical fights, having been injured in physical fights and injuring others in physical fights by drinking characteristics among youth drink alcohol (Swahn et al., 2004) Measures Drinking frequency = 2-8 days/month Drinking frequency = 9-30 days/month Binge drinking frequency = 2-30 days/month Fighting, past year Injured in a fight, past year Injured others, past year 1.02 (0.84, 1.22) 0.87 (0.64, 1.18) 1.07 (0.89, 1.29) 1.41 (1.05, 1.87) 1.96 (1.31, 2.95) 1.55 (1.17, 2.04) 1.35 (1.12, 1.62) 1.85 (1.35, 2.54) 1.32 (1.06, 1.65) Problem drinking 1.51 (1.26, 1.81) 1.44 (1.08, 1.92) 1.47 (1.19, 1.83) 1-3 peers drink 1.34 (1.12, 1.61) 1.60 (1.22, 2.10) 1.47 (1.18, 1.84) Offending Fergusson et al. (1996) examined the associations between alcohol misuse and juvenile offending at ages years in a cohort of 953 respondents from the Christchurch Health and Development Study. After adjusting for covariate factors, the authors found that there was a small but significant association between alcohol misuse and violent offending [adjusted OR 3.2 (95% CI 1.4, 7.6)] but not property offences [adjusted OR 1.4 (95% CI 0.6, 3.3)]. Fergusson and Horwood (2000) further analyzed the effects of alcohol use on crime in this cohort using a fixed effects regression model. 19

20 Analysis of the association between alcohol abuse and crime rates showed that in all cases there was evidence of statistically significant increases (p<0.01) in rates of violent and property crimes with increases in symptoms of alcohol abuse. After controlling for confounding factors and observed time dynamics, there were still significant (p <0.001) associations between alcohol abuse and crime. The incidence rate ratios showed that a one-symptom increase in alcohol abuse was associated with a 1.15 times increase in the rate of violent crime and a 1.10 times increase in the rate of property offending. Further analyses of the same cohort by Wells et al. (2004) found that drinking behaviours at 16 was also significantly associated with the number of violent offences committed between ages years (but not ages years). 5.3 Discussion The studies identified provide evidence of the association between alcohol consumption during adolescence and a range of acute outcomes. Three studies highlighted positive associations between alcohol consumption and sexual activity, with heavy drinkers more likely to initiate sex earlier and have sex without using contraception. In addition, one study found that early onset of drinking was associated with having more sexual partners and pregnancy. One study demonstrated that frequent and binge drinkers were more likely to be involved in fights, and more likely to be injured as a result of fighting. A second study showed that alcohol misuse may be significantly associated with violent offending. 20

21 6 MORTALITY AND MORBIDITY ASSOCIATED WITH ALCOHOL CONSUMPTION 6.1 Results of the literature review A total of 43 studies (4 meta-analyses and 39 primary studies) were retrieved and met the inclusion criteria. An additional 16 studies were included and data extracted, but subsequently excluded. Salient characteristics of the studies included in the causality model are presented in Appendix 3. Spreadsheets for the causality model and birth cohort models are shown in Appendix Overview of studies which examined mortality and morbidity associated with alcohol consumption The four systematic reviews (Bagnardi et al., 2001; Corrao et al., 1999; 2000; 2004) identified were linked publications and examined the dose-response relationship between exposure to alcohol and a range of outcomes including various cancers, stroke, cirrhosis of the liver, chronic pancreatitis and suicide. Of the 39 primary studies, 13 studies examined the association between exposure to alcohol and the risk of various cancers including lung, head and neck cancer, prostate cancer, bladder cancer, breast cancer, oral cancer, adenocarcinoma of the small intestine and colon/rectal cancer. The remaining studies examined a wide range of outcomes and their association with alcohol consumption including injuries, risky sexual behaviour, depression, suicide, haemorrhagic stroke, all-cause mortality, coronary heart disease and crime. 6.3 Evidence-based disease incidence model A total of 12 studies (3 systematic reviews and 9 primary studies) were used to develop the causality model and generate risk difference estimates. Whenever possible, if the literature yielded more than one estimate of association we used the more conservative one. As described in the methods section the estimates of the association were converted to RD and applied to the birth cohort model. This allowed us to calculate the hypothetical number of alcohol-related events per outcome that would be expected to occur in the 2005 birth cohort for England and Wales 7 (n=645,835). These numbers are presented in Table 9. Where the number of cases is minus this indicates a potential protective effect of alcohol consumption. Table 9. Number of alcohol-related expected cases in the 2005 birth cohort by outcome Outcome Consumption level (g/day) Expected numbers in 2005 birth cohort (95% CI) Cancers of the lip, oral cavity & pharynx 1 1 (1, 1) Oesophageal cancer 1 1 (1,2) 7 (accessed 21/01/07) 21

22 Cancer of the larynx 1 0 Breast cancer (females) 1 5 (3, 6) Lung cancer >15 1,584 (512, 4094) Adenocarcinoma of the small intestine >25 17 (3, 49) Stomach cancer 25 6 (3, 9) Colon cancer (16, 49) Liver cancer 25 3 (2, 4) Ovarian cancer 25 7 (0, 32) Essential hypertension 1 1 (1, 2) Haemorrhagic stroke 1 6 (4, 9) Subarachnoid haemorrhage (SAH) <150 a 11 (-22, 66) Cirrhosis of the liver 1 11 (8, 14) Chronic pancreatitis 1 5 (2, 7) Suicide 1 1 (-1, 3) Hip fracture >0-165 (-256, -58) Injury >0 13,666 (9,483, 18,762) Depression >0 b 117 (46, 198) High risk intercourse Intoxication c 2,066 (1,105, 5,189) Unwanted pregnancy Bingeing d 432 (131, 543) Violent crime attributable to alcohol e Sexual offences attributable to alcohol - 98 e All recorded crime attributable to alcohol e a per week; b alcohol use in early 20s.; c one or more episodes of drunkenness a month; d consumption of 5 or more alcoholic drinks on 1 occasion; e NWPHO estimated using data from the Home Office and Office for National Statistics 6.4 Sensitivity analysis As the β coefficient represented the change in the ln RR per unit change in the value of x (1 g/day), we explored the risk of exposure at different levels of consumption of alcohol in a sensitivity analysis for the following outcomes: cancers of the lip, oral cavity and pharynx; oesophageal cancer, cancer of the larynx, breast cancer, essential hypertension, haemorrhagic stroke, cirrhosis of the liver; chronic pancreatitis; and suicide. The hypothetical number of alcohol-related events per outcome that would be expected to occur in the 2005 birth cohort at increasing levels of daily alcohol consumption are shown in Table 10. Alcohol consumption at 25 g/day corresponds to approximately two drinks per day. Table 10. Number of expected cases in 2005 birth cohort at different levels of daily alcohol consumption Pathology Expected cases in 2005 birth cohort 25 g/day 50 g/day 100 g/day Cancers of the lip, oral cavity & pharynx Oesophageal cancer Cancer of the larynx Breast cancer ,806 Essential hypertension Haemorrhagic stroke ,735 Cirrhosis of the liver 399 1,188 5,847 Chronic pancreatitis

23 Suicide We also carried out a sensitivity analysis on the estimates from the causality model using RR estimates for 25 g/day alcohol intake reported in Corrao et al. (2004) for the following outcomes: oesophageal cancer, cancer of the larynx, breast cancer, essential hypertension, haemorrhagic stroke, cirrhosis of the liver and chronic pancreatitis. The sensitivity analysis found that the estimation of the number of expected cases in the birth cohort were sensitive to the RR estimates used (see Table 11). For example, breast cancer cases ranged between 141 cases estimated using the converted β coefficient for 25 mg/day (Corrao et al., 1999) to 114 cases using the RR estimate of 1.25 (Corrao et al., 2004). Table 11. Number of expected cases in 2005 birth cohort using RR estimates from Corrao et al., Pathology Expected cases in 2005 birth cohort (95% CI) 25 g/day 50 g/day 100 g/day Cancers of the lip, oral cavity & pharynx 49 (43, 54) 119 (105, 135) 308 (269, 353) Oesophageal cancer 31 (29, 34) 75 (68, 80) 208 (188, 230) Cancer of the larynx 10 (8, 11) 23 (20, 26) 64 (54, 75) Breast cancer 114 (91, 132) 251 (201, 305) 643 (488, 820) Essential hypertension 50 (38, 62) 121 (90, 157) 366 (248, 525) Haemorrhagic stroke 64 (-10, 165) 275 (154, 430) 1,243 (789, 1,877) Cirrhosis of the liver 773 (696, 850) 2,494 (2,177, 2,848) 10,383 (8,650, 12,446) Chronic pancreatitis 141 (66, 223) 322 (141, 562) 905 (339, 1,897) Freudenheim et al. (2005) examined the risk of lung cancer associated with alcohol consumption by conducting a pooled analysis of seven cohort studies. In the pooled analysis there was weak evidence for a positive association between alcohol consumption and lung cancer risk. Men and women who consumed 30 g alcohol/day were at a slightly greater risk for lung cancer than those consuming 0 g alcohol/day [men: OR 1.21 (95% CI: 0.91, 1.61); women: OR 1.16 (95% CI: 0.94, 1.43)]. However, in males classified as never smoking, there was a strong positive association [RR for consumption of 15 g/day compared to 0 g/day: 6.38 (95% CI: 2.74, 14.9)]. The primary version of the disease burden model resulted in a large number of hypothetical cases in the birth cohort model, and because of the uncertainty surrounding the estimates of lung cancer risk we undertook a sensitivity analysis using the pooled estimates adjusted for smoking status at different levels of alcohol consumption for males and females separately (see Table 12). Table 12. Number of expected cases of lung cancer in 2005 birth cohort using alternative estimates from Freudenheim et al. (2005) Expected cases in 2005 birth cohort Alcohol consumption (g/day) Males Females >0 to <5-27 (-55, 6) -23 (-72, -20) 23

24 5 to <15 0 (-30, 38) -20 (-69, -7) 15 to <30-32 (-72, 19) -8 (-56, 28) (-17, 116) 17 (-13, 93) 6.5 Discussion When looking at this apparently large and complex body of data, readers should bear in mind the following issues. In particular, the main part of the model is centred on the work of Corrao et al. (1999, 2000, and 2004), who collected, evaluated and synthesised data from a large population of non-randomised comparative studies. Although we have no reason to criticise this work, several requests for the list of included and excluded studies in the reviews have gone unanswered. As a consequence we only have a partial view of the work and of the types of studies that were excluded from the review. In addition, Corrao et al s statements about the variable quality of included studies and the high likelihood of publication bias does not increase our confidence. Additionally the reason for multiple analyses and publications of the same data set are not clear (the searches in the Corrao group of publications appear to be the same and with no list of included studies there is no possibility of checking). Single studies included in the model were similarly of variable quality (the extreme case being a case-control study in which numerators and denominators of cases and their controls had been swapped around). Given the design, variable quality, heterogeneity and geographical spread of the studies the possibility of a sizeable effect of some known and mostly unknown confounders cannot be ruled out. Although we have tried to report adjusted measures of association (e.g. ORs) whenever available, we do not believe this is any sure protection against the play of confounders. We found no guidance in the literature on the building of models similar to ours. Although powerful and as evidence-based as possible, several features of the model must be taken into account. Baseline probability figures incorporated in the model are assumed to be the same as incidence. Whereas for short-lasting pathologies or conditions (such as some of the effects of violent crime) this should provide a robust estimate, for pathologies with longer natural histories such as most forms of cancer, their use may lead to an underestimation of the likely number of cases in our birth cohort. As accurate estimates of prevalence are difficult to obtain, we justify our choice with the high quality of incidence estimates based on cancer registrations. 24

25 7 CONCEPTUAL MODEL The relationship between alcohol consumption and health and social outcomes is complex and multidimensional (Rehm et al., 2003a). Therefore, in the development of the conceptual model (see Figure 3), only the main causal pathways have been illustrated. The model is in two stages. The first stage models the main drinking patterns identified in adolescence and the second model patterns of adult drinking. 7.1 Adolescent drinking model In the adolescent drinking model, following onset of drinking young people may become light drinkers with progression to drinking to intoxication and finally heavy drinking (characterised by binge drinking). The review of the literature found evidence that drinking to intoxication and heavy drinking may increase the risk of a range of acute outcomes including risky sexual behaviour (unprotected sex, sex with multiple partners), injuries and crime as shown in the model. 7.2 Adult drinking model In the adult model, adults can be light to moderate drinkers or progress through the three levels of alcohol use disorder as defined by the WHO (hazardous, harmful or dependent drinking). All levels of alcohol consumption are associated to some degree with the chronic diseases listed through the toxic and beneficial biochemical effects of alcohol. Acute health and social outcomes are associated with the three alcohol abuse disorders. 7.3 Transition from adolescent to adult patterns of drinking This review has found evidence to show that people who binge drink in adolescence are more likely to be binge drinkers as adults. In addition, frequent drinking and binge drinking have been shown to increase the risk of developing alcohol dependence in young adulthood. Therefore, in the transition to adult drinking, adolescents may enter at any of the four stages, but at which is dependent on the patterns of drinking they have exhibited through adolescence. 25