Early Predictors of Adult Drinking: A Birth Cohort Study

Abstract

Few studies have explored early predictors of problem drinking in youth, and fewer still have simultaneously considered the role of biologic, familial, and intrapersonal factors. The present study explored early life course and later life course predictors of alcohol abuse and dependence in young adulthood. Data were taken from a cohort of 2,551 mothers and their children recruited as part of the longitudinal Mater University Study of Pregnancy and its outcomes (MUSP) carried out in Brisbane, Australia, from 1981 to 1984. Data were collected prenatally and then postnatally at 6 months and at 5, 14, and 21 years. A range of biologic, familial, and intrapersonal factors was considered. A series of logistic regression models with inverse probability weighting was used to explore pathways to problem drinking from adolescence to early adulthood. For males and females, no association was found between either birth factors or childhood factors and a lifetime diagnosis of alcohol disorders at age 21 years. Externalizing symptoms and maternal factors at age 14 years were significantly associated with alcohol problems. For youth aged 14 years, maternal moderate alcohol consumption accounted for the highest percentage of attributable risk among those exposed. Results show that exposure to maternal drinking in adolescence is a strong risk factor for the development of alcohol problems in early adulthood.

Despite a wealth of knowledge on the nature of alcohol-related pathologies in Western populations, developmental pathways to alcohol-related disorders remain unclear. Many studies have followed high-risk or specific samples of “alcoholics” (1, 2) and have reported on the most extreme and severe sequelae of these disorders, leaving unanswered questions about the factors and mechanisms underpinning more common alcohol problems in the general population. This focus has resulted in a considerable body of research oriented around a specific group of problem drinkers—namely, middle-aged, middle-class males—and has left other population subgroups relatively understudied (1–3). More recently, public health concern has grown in response to the health burden associated with harmful drinking among young people, resulting in increased research into pathways to problem drinking, particularly among adolescents (4, 5). Recent evidence suggests that a range of familial, individual, social, and environmental factors is important in predicting early onset of drinking in adolescence, with peer models in particular strongly promoting drinking initiation in early adolescence (6, 7).

Little is known about the influence of early life factors on problem drinking in early adulthood. However, since David Barker (8) argued that events occurring around the time of birth can influence patterns of morbidity and mortality in the general population, there has been increased interest in the role of biologic factors in the development of physical and mental illnesses. With regard to mental health, small but robust associations have been found between obstetric complications and schizophrenia, and evidence from longitudinal studies points to an association between low birth weight and depressive symptoms in adulthood (9, 10). Another potentially important factor in the postnatal period is early infant temperament, which is typically regarded as a marker for certain biologic characteristics (11). Longitudinal evidence for the role of these factors in the development of alcohol problems is limited, with the exception of a recent Danish study that found an association between low birth weight and alcohol problems in a high-risk sample of men (12). To date, however, these findings are not known to have been replicated in a population study.

Childhood is another important period in the developmental pathway to many diseases in adulthood. Problematic behavior, parenting styles, and familial attitudes during this time appear to be associated with a range of difficulties in early adulthood (13, 14). Among the few papers published on early risk factors for alcohol abuse, findings from the Seattle Social Development Project point to permissive parental monitoring and externalizing and internalizing behaviors at age 10 years as important predictors of alcohol abuse and dependence at age 21 years (15).

Research interest in the role of adolescence in the development of alcohol problems is growing. In the Pittsburgh Youth Study, early signs of psychopathology, including conduct disorders, attention deficit/hyperactivity disorders, and depression at age 13 years, were all strongly associated with alcohol abuse in late adolescence (16). Another prospective population study found dysphoria strongly associated with later heavy drinking, with a bidirectional association evident for females early in adolescence (17). Findings from the Dunedin birth cohort study also support the view that personality traits are important predictors of persistent heavy drinking among young adults (13).

Understanding the developmental pathways to alcohol problems in the young population is important so that appropriate public health interventions can be tailored to identified risk factors at each point in the child's development. At present, however, the role of early social, familial, and biologic influences on the development of alcohol problems in early adulthood is poorly understood. Accordingly, we used longitudinal data collected over a 21-year period to explore the role of a range of factors at three periods in a child's development (birth, childhood, and adolescence) and their association with the development of alcohol-related disorders from adolescence to age 21 years.

MATERIALS AND METHODS

Data were taken from the Mater University Study of Pregnancy and its outcomes (MUSP), a prospective study of women and babies who received antenatal care at a major public hospital in Brisbane, Australia, between 1981 and 1984. The sample consists of 7,223 women (and their offspring) who have been followed up prospectively for 21 years. Data were collected at the first antenatal clinic visit, 3–5 days after birth, 6 months after birth, and 5, 14, and 21 years after birth. The study population and recruitment methods have been described extensively elsewhere (18, 19). Our cohort consisted of a selected subgroup of 2,551 youth (1,248 (48.9 percent) male and 1,303 (51.1 percent) female) who completed the lifetime version of the Composite International Diagnostic Interview–computerized version (CIDI-Auto) at the 21-year follow-up phase of the study (20). This group represents 35.3 percent of the 7,223 offspring in the original birth cohort.

In this paper, we assess the contribution of biologic, environmental, familial, and individual factors at three different stages of development: in utero/birth and infancy, early childhood (5 years), and adolescence (14 years). Written informed consent from the mother was obtained at all data collection phases. Written informed consent was also obtained from the young adult at the 21-year follow-up phase of the study.

Data collection

Data were drawn from five phases of the study: the obstetric records collected at birth, maternal self-report at 6 months, maternal and child self-reports at 5 and at 14 years, and the measure of alcohol disorders at the 21-year follow-up.

Outcome variable: Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition (DSM-IV) criteria of alcohol disorders

At the 21-year phase of the study, we used the CIDI-Auto (20) to assess the association of earlier and later factors with the development of alcohol disorders from adolescence to age 21 years, according to DSM-IV diagnostic criteria (21). Only a small percentage of youth (6.1 percent) reported a diagnosis of alcohol dependence. This small number made it difficult to meaningfully separate alcohol abuse and dependence in the analysis, particularly for the female group, where only 2.5 percent (n = 63) reported symptoms consistent with alcohol dependence. We therefore used DSM-IV criteria to assess a lifetime diagnosis of both alcohol abuse and dependence. Because use of lifetime diagnosis may have included cases who reported a disorder before the age of 14 years, we excluded, in sensitivity analyses, 38 youth (1.9 percent of the complete case analysis sample) who reported having a drink as often as once a week before or while taking part in the 14-year follow-up. Results from the analyses excluding these 38 cases did not vary substantively from those presented here.

Birth and early life period

Biologic predictors were taken from obstetric records at birth and included gestational age, birth weight, Apgar scores at 1 and 5 minutes, and transfer to intensive care. At 6 months, the child's temperament was assessed through use of a seven-item index developed for this study. Mothers were asked how often their baby had the following problems: colic, sleeplessness, vomiting, diarrhea or constipation, feeding problems, skin problems, and overactivity. Each item was scored on a 5-point scale (0–3 symptoms = “easy behavior”; 4 or more symptoms = “difficult behavior”). The internal consistency of the scale was acceptable (Cronbach's alpha = 0.65).

Maternal and individual predictors in early childhood

Maternal child-rearing practices were assessed at the 5-year follow-up by using two measures: a monitoring-style index and a physical punishment index. Mothers were asked at what age they would allow their children to 1) go to the movies, 2) go on holidays by themselves, 3) travel alone on a bus, 4) stay at home alone, and 5) drink alcohol. The index exhibited acceptable internal consistency (Cronbach's alpha = 0.61), and scores were reclassified as follows: some control, no control, and strict control. Secondly, mothers were asked in which of the following situations they would smack their child “sometimes/always /never”: 1) refuses to clean up his or her room, 2) takes something belonging to another child and punches that child, 3) makes fun of a crippled person, 4) touches a hot stove, and 5) breaks something indoors after being told to play outdoors. The index exhibited acceptable internal consistency (Cronbach's alpha = 0.61).

The internalizing and externalizing scales of a modified form of the Child Behaviour Checklist were administered to mothers when the child was age 5 years to assess early signs of behavior problems (22). Use of this scale in the MUSP study has been explained elsewhere (23).

Familial and individual influences at 14 years

Data on maternal alcohol consumption were obtained at the 14-year follow-up with two items assessing frequency (from never to daily) and quantity (from 0 to ≥7 standard drinks) of consumption. These data were used to classify the women into four categories (0 = abstainers; 1 = <1 drink a week; 2 = at least 1 drink a week; 3 = ≥1 drinks a day). Mothers also reported their current level of tobacco use at the 14-year follow-up by recalling how many cigarettes they smoked over a 7-day period. Smoking status was categorized as nonsmoker, 1–19 cigarettes per day, and 20 or more cigarettes per day.

Maternal depression and anxiety were assessed at the 14-year follow-up by using the Delusions-Symptoms-States Inventory (24). This inventory contains two seven-item subscales measuring depression and anxiety that have been found to correlate strongly with other scales of depression, including the Beck Depression Inventory (25). In this study, maternal symptoms of depression and anxiety were defined as the reporting of four or more of the seven symptoms in the Delusions-Symptoms-States Inventory depression and anxiety subscales.

At the 14-year follow-up, behavior problems were assessed by using the internalizing and externalizing symptom subscales of the Youth Self Report version of the Child Behaviour Checklist (26). Validation studies with both clinical and population samples have reported factor structures and reliability estimates consistent with Achenbach data (27). The use of the Youth Self Report in the MUSP study has been extensively described in previous papers, which also report on the good validity and internal consistency of the scales (23). Scale scores were dichotomized into two categories, with a 10 percent cutoff for “cases.”

Data analysis

Prospective analyses predicting alcohol consumption were conducted by using logistic regression. The initial models reflected each of three developmental periods under study: birth and the early life period (up to 6 months), childhood (5 years), and adolescence (14 years). All variables were adjusted for the effect of other variables in the model and for maternal age and education at baseline and marital status at birth, 5 years, and 14 years.

Variables that met statistical significance at the p < 0.10 level in these initial models were included in two cumulative models for males and females, considering all time periods, with predictors adjusted for each other and for maternal demographic confounders. Finally, we calculated the attributable risk of developing an alcohol abuse or dependence disorder by age 21 years for each of the significant risk factors in the final, adjusted analyses (28).

Attrition

We used inverse probability weighting with robust estimates for standard errors to account for those lost to follow-up. We explored all sociodemographic, lifestyle, and biologic variables available at baseline. Thirty variables were included in this exploratory logistic regression model to determine whether those subjects remaining in the study significantly differed from those lost to follow-up. Measures that predicted loss to follow-up at 21 years differed by gender and included maternal education, depression and anxiety, smoking and drinking habits, parity, and late attendance to the first antenatal visit. In further analyses, we stratified by gender and retained in the two models those variables that were significant for each group. We then estimated inverse probability weighting by fitting these measures in two logistic regression models, one for the female sample and another for the male sample. As suggested by Hogan et al. (29), we assessed lack of fit by using Hosmer and Lemeshow's deciles of risk statistic. For our models, this statistic was 7.48 (df = 8, p = 0.28) for females and 2.27 (df = 5, p = 0.81) for males, indicating a very satisfactory fit. The distribution of most weights' values was between 2 and 3, with a minimum value of 0.20 for females and 0.25 for males, indicating stable weights (29). When weighting adjustments were included in the analyses, results did not change substantively from the unweighted analyses presented here.

RESULTS

Of the 2,551 youth who completed the CIDI-Auto (20) at age 21 years, 25.1 percent (n = 640) met the criteria for a lifetime diagnosis of alcohol abuse and/or dependence. Of these, 69.7 percent (n = 446) were male and 30.3 percent (n = 194) were female. This prevalence is high; however, it reflects lifetime rather than current diagnoses and, in most cases, indicates a diagnosis of alcohol abuse (19 percent) rather than alcohol dependence (6 percent). Young adulthood is the period in which excessive drinking is most prevalent in Australia (30), with the most recent household survey finding that, of those aged 20–29 years, 14.2 percent reported at least weekly consumption of alcohol in a pattern putting them at risk of short-term harm, and 14.7 percent reported consuming alcohol in a pattern risking long-term harm (31). Among Australians, the average age at which males first consume a full glass of alcohol is 16.2 years; for females, it is 17.6 years (31).

Table 1 shows, separately for males and females, unadjusted and adjusted associations between a variety of potential birth and early-life risk factors—such as gestational age, birth weight, Apgar scores at 1 and 5 minutes, transfer to intensive care, and early difficult behavior—and alcohol disorders from adolescence to age 21 years. We found no associations between alcohol disorders and the biologic and obstetric factors considered, except that, for females, difficult behavior at age 6 months increased the odds of developing an alcohol disorder by age 21 years, whereas transfer to intensive care at birth decreased these odds (p for interaction test for a gender difference = 0.16 for temperament and = 0.10 for intensive care).

There was no significant association between early signs of mental health/behavioral problems at age 5 years and the development of alcohol problems by age 21 years (table 2). However, males exposed to a stricter parental monitoring style at age 5 years had a decreased risk of developing alcohol problems compared with boys who were given more freedom. For females, no childhood factors remained significant after confounders were included in the model.

Table 3 shows unadjusted and adjusted odds ratios for the occurrence of alcohol problems by age 21 years as a function of a child's externalizing problems and of maternal anxiety, depression, alcohol problems, and smoking when the child was age 14 years. Externalizing problems and maternal depression, smoking, and drinking at the 14-year follow-up were associated with alcohol disorders in youth, with maternal consumption of one or more drinks a day putting males at a twofold risk of developing alcohol problems by young adulthood.

Table 4 shows, for males, associations between alcohol disorders by age 21 years and exposure to distal and proximate risk factors, and related attributable risk expressed as percentages. A child's externalizing symptoms and maternal depression reported at the 14-year follow-up increased the odds of developing an alcohol disorder by age 21 years, as did both moderate and daily maternal drinking reported at the 14-year follow-up. Maternal drinking alone accounted for over 21 percent of the attributable risk of alcohol abuse and dependence in boys exposed. Compared with that for abstainers, the lifetime risk of alcohol abuse was higher for boys whose mothers were in either the “at least one glass a week” or the “≥1 drink a day” group (table 4).

Table 5 shows, for females, influential predictors of alcohol disorders throughout the life course. The strongest predictors of developing an alcohol disorder by age 21 years for females were self-reported externalizing symptoms and daily maternal drinking at age 14 years (table 5), which together accounted for over 15 percent of the attributable risk in the female sample.

DISCUSSION

The aim of our study was to examine the association between a range of early life factors and a lifetime diagnosis of alcohol disorders reported at age 21 years. In particular, we were interested in exploring the specific contribution of three sensitive periods—birth, childhood, and early adolescence—to establish which factors and time periods predict the development of alcohol problems from adolescence to early adulthood.

Ours is the first longitudinal population study known to examine a possible relation between birth factors and alcohol disorders in early adulthood. An association between these factors and alcohol problems would suggest that problem drinking, as well as other mental and physical illnesses (9), may have its origins very early in life. In our study, we found no evidence of a biologic effect on the drinking patterns of young adults. This finding contrasts with those from a recent Danish study that suggested an etiologic association between birth weight and a lifetime diagnosis of alcohol abuse and dependence at age 30 years (12). We conclude that biologic effects are not associated with alcohol problems in a general population sample, although we accept that they might be markers of fetal distress among specific populations at risk, such as those represented in the Danish cohort study.

Our study provides some support for the view that individual traits in early adolescence, such as the child's externalizing symptoms, increase the odds of having an alcohol disorder by age 21 years (16, 17, 32, 33), although we did not find an association between internalizing symptoms in childhood and the development of problem drinking (7, 13, 32, 34). Findings from this study point to maternal factors, including maternal depression and drinking when the child is age 14 years, as strong contributors to the development of problem drinking from adolescence to age 21 years. Those born to mothers who reported depressive symptoms and/or drinking more than one glass of alcohol a day at the 14-year follow-up had a twofold risk of reporting drinking problems at the 21-year follow-up. More importantly, adolescent boys exposed to moderate maternal alcohol consumption still had an increased risk of alcohol disorders compared with boys of abstinent mothers.

Overall, the risk factors considered in this study explained about 37 percent of the attributable risk of developing alcohol problems between ages 14 and 21 years in males, and 15 percent in females. Because males represented 70 percent of those reporting alcohol disorders, these risk factors explained about 36 percent of the variance in alcohol disorders reported in the full sample. Among males, exposure to daily maternal drinking explained 7 percent of this attributable risk; exposure to more moderate maternal drinking explained almost 14 percent.

To our knowledge, this is the first longitudinal study to highlight the extent of the risk associated with exposure to moderate maternal drinking in early adolescence. Although difficult to interpret, this finding points to a plausible sequential pathway from abstinence in the early teens to alcohol disorders in early adulthood. Because Australian women drink less than men (35), the drinking patterns of the mothers in this study may well reflect exposure to similar or more excessive paternal drinking and possibly similar patterns of consumption in the child's broader social environment. This pattern is likely to reinforce adolescents' positive perceptions of alcohol consumption as a “normal” and socially acceptable behavior, devoid of the negative connotations attached to excessive drinking.

More importantly, moderate maternal alcohol consumption in the family context may facilitate greater access when teenage drinking begins. Easy availability, coupled with pressure from current aggressive marketing of alcohol products and a traditional acceptance of teenage drinking within Australian society (36), may result in increased risk of excessive alcohol consumption. Indeed, in this population sample, one quarter of young adults and more than one third of males exhibited symptoms consistent with alcohol abuse and (less often) dependence over the period ranging from adolescence to early adulthood. The effect of moderate maternal drinking was evident for males only, who are likely to start drinking earlier and in greater quantities and who may therefore be more vulnerable than girls to parental modeling influences (37). Studies with the capacity to assess the effect of moderate drinking in both parents, means of access to alcohol in adolescence, and other associated factors are needed to replicate our findings and to more clearly explicate the pathway proposed here.

Results of this study should be viewed in the context of some limitations. Despite the fact that the study used comprehensive health and psychosocial assessments at each time period, we were unable to consider a number of potentially important factors, such as genetic predisposition, adverse childhood experiences, societal influences such as peer group modeling, and paternal mental health and alcohol consumption. Our data did not lend themselves to assessing the impact of early substance use behavior on alcohol disorders in early adulthood. For example, at the 14-year follow-up, no participants reported drinking as often as daily, and only 1.9 percent (n = 38) reported drinking one glass of alcohol as often as once a week. The small number of adolescents reporting regular drinking and smoking at the 14-year follow-up made empirical assessment of the impact of these factors impossible. In addition, our assessment of maternal alcohol consumption does not accurately match current standardized screening measures, which include consideration of quantity, frequency, and binge drinking over a specific time period (38). However, we were able to classify alcohol consumption into categories comparable to those used by similar prospective studies (39). Furthermore, the lifetime diagnosis of alcohol abuse and dependence we used for assessing the development of alcohol disorders in youth may be sensitive to recall bias. Ideally, we would need measures collected at short, regular intervals because alcohol consumption during the developmental years is likely to change rapidly. From this perspective, the MUSP study is limited because measurements are available at only lengthy, irregular intervals (i.e., birth, 6 months, 5 years, and 14 years).

It is also worth noting the loss to follow-up in our cohort. If the risk factors and poor outcomes considered here were less prevalent among those lost to follow-up, our results would overestimate the association between the identified risk factors and alcohol consumption at age 21 years. However, because we typically find that those lost to follow-up are more likely to be anxious/depressed, to be smokers and drinkers, and to be of lower socioeconomic status (19), this possibility seems unlikely. Indeed, it is more likely that those lost to follow-up would exhibit higher rates of alcohol consumption at age 21 years and therefore that the associations presented here would be a conservative estimate (19). Although we are aware that all statistical simulations have limitations in accounting for missing data, attaching inverse probability weighting to subjects included in the analyses is an acceptable method to restore representation of those lost to follow-up (29, 40). The fact that we found little difference between the weighted and nonweighted results suggests that attrition is unlikely to have substantively biased our findings in either direction.

Despite these limitations, a key strength of this study is that it was longitudinal. In addition, it is one of the few studies able to assess both alcohol problems, according to DSM-IV criteria for alcohol disorders, and the contribution of biologic, individual, and family factors to the etiology of drinking problems in early adulthood. Results further our understanding of the interactions between early psychopathology and family factors in the etiology of alcohol problems in young adults.

Even though an overall decrease in alcohol consumption is occurring in the general population, it is steadily increasing among young people (30, 36, 37). Our study directs attention to normative influences, entrenched in the social structure of a fast-changing, modern society, as strong contributors to the development of heavy drinking behaviors that we found in our young adult sample (36). Public health interventions will need to raise parents' awareness of the risks associated with easy availability of alcohol in the family context during early adolescence if we are to reduce the high rates of alcohol misuse currently experienced among young people. An awareness of the risks associated with maternal depression may also facilitate primary interventions targeted at youth who may go on to develop problematic patterns of alcohol consumption in adolescence.

The National Health and Medical Research Council (NHMRC), Queensland Health, Queensland Treasury, the Centre for Accident Research and Road Safety–Queensland (CARRS-Q), the Australian Institute of Criminology (AIC), and the Telstra Foundation funded this project.

Dr. Alati is funded by an NHMRC Public Health Fellowship grant (301298). Dr. Mamun is funded by an NHMRC Capacity Building grant (252834).

Drs. R. Alati, J. M. Najman, and S. A. Kinner developed the study aim and design with advice from Drs. M. O'Callaghan, W. Bor, and G. M. Williams. Drs. A. A. Mamun and G. M. Williams advised on statistical methods and analysis. Drs. J. M. Najman, M. O'Callaghan, G. M. Williams, and W. Bor set up, and are responsible for, the conceptual development and continued management of the MUSP study. Dr. R. Alati wrote the first draft of the paper. All authors contributed to the final version.

The authors thank the MUSP team, the Mater Misericordiae Hospital, and the Schools of Social Science, Population Health, and Medicine (University of Queensland) for their support.

Special thanks to the MUSP 21-year follow-up team who collected the 21-year follow-up data: Rosemary Aird, Stacey Allerton, Ruth Armstrong, Samantha Batchelor, Pauline Bonnici, Rachael Bor, Emma Brown, Justine Butcher, Fiona Cameron, Narelle Constantine, Sophie Gudgeon, Jatinder Kaur, Jane Maclean, Amanda Margerison, Kobie Mulligan, Kelly Quinlan, Marie Seeman, and Jennifer Winn.

Conflict of interest: none declared.

References