Introduction
Adolescent drinking is associated with developing non-communicable diseases (Patton et al. Reference Patton, Coffey, Cappa, Currie, Riley, Gore, Degenhardt, Richardson, Astone, Sangowawa, Mokdad and Ferguson2012; Swendsen et al. Reference Swendsen, Burstein, Case, Conway, Dierker, He and Merikangas2012), and being the leading risk factor for disability-adjusted life years lost in 10- to 24-year-olds (Gore et al. Reference Gore, Bloem, Patton, Ferguson, Joseph, Coffey, Sawyer and Mathers2011). Surprisingly then, parents are a major source of alcohol consumed by children. Approximately a third of adolescents in US, European and Australian studies (Hearst et al. Reference Hearst, Fulkerson, Maldonado-Molina, Perry and Komro2007; White & Bariola, Reference White and Bariola2012; Fuller, Reference Fuller2013) report parental supply. Reasons cited for parental supply include: parental desire to ‘socialize’ children into responsible drinking, ‘inoculating’ them from heavy consumption (Donovan & Molina, Reference Donovan and Molina2008; Gilligan et al. Reference Gilligan, Kypri and Lubman2012; Jackson et al. Reference Jackson, Ennett, Dickinson and Bowling2012); parental concern that peers’ parents will provide alcohol anyway (Gilligan et al. Reference Gilligan, Kypri and Lubman2012; Jackson et al. Reference Jackson, Ennett, Dickinson and Bowling2012; Wadolowski et al. Reference Wadolowski, Hutchinson, Bruno, Aiken, Najman, Kypri, Slade, McBride and Mattick2016); and supply at cultural/religious, or celebratory events (Gilligan et al. Reference Gilligan, Kypri and Lubman2012; Jackson et al. Reference Jackson, Ennett, Dickinson and Bowling2012). Yet, the quality of evidence on whether parental supply is associated with more, or less, adolescent drinking is compromised by design and analytic limitations.
While a recent review concluded that parental supply was ‘generally associated’ with increased child drinking (Kaynak et al. Reference Kaynak, Winters, Cacciola, Kirby and Arria2014), that view was qualified by study design limitations. Ignoring the cross-sectional studies – of little assistance in understanding the impact of parental supply – seven cohorts showed associations between parental supply and drinking (Kaynak et al. Reference Kaynak, Winters, Cacciola, Kirby and Arria2014). Yet, assessments were limited to one (Shortt et al. Reference Shortt, Hutchinson, Chapman and Toumbourou2007; Livingston et al. Reference Livingston, Testa, Hoffman and Windle2010) or two follow-up occasions (Jackson et al. Reference Jackson, Henriksen and Dickinson1999; Komro et al. Reference Komro, Maldonado-Molina, Tobler, Bonds and Muller2007; Van Der |Vorst et al. Reference Van Der Vorst, Engels and Burk2010; Danielsson et al. Reference Danielsson, Romelsjö and Tengström2011; McMorris et al. Reference McMorris, Catalano, Kim, Toumbourou and Hemphill2011), in quite late adolescence (18–19 years of age; Livingston et al. Reference Livingston, Testa, Hoffman and Windle2010) or in early adolescence [10–12 (Jackson et al. Reference Jackson, Henriksen and Dickinson1999), 12–13 (Shortt et al. Reference Shortt, Hutchinson, Chapman and Toumbourou2007) or 12–14 years of age (Komro et al. Reference Komro, Maldonado-Molina, Tobler, Bonds and Muller2007; McMorris et al. Reference McMorris, Catalano, Kim, Toumbourou and Hemphill2011)], hampering conclusions about both the development and the sequelae of parental supply. One study did span 13–15 years of age (Jackson et al. Reference Jackson, Henriksen and Dickinson1999), and a key study assessed 13–16 years of age (Van Der Vorst et al. Reference Van Der Vorst, Engels and Burk2010). Yet, these two studies left unmeasured known predictors of adolescent drinking and potential confounders, including: parental/familial alcohol problems (Jackson et al. Reference Jackson, Henriksen and Dickinson1999; Van Der Vorst et al. Reference Van Der Vorst, Engels and Burk2010), parental drinking (Jackson et al. Reference Jackson, Henriksen and Dickinson1999), child externalizing and other substance use (Van Der Vorst et al. Reference Van Der Vorst, Engels and Burk2010) and child age (Jackson et al. Reference Jackson, Henriksen and Dickinson1999; Van Der Vorst et al. Reference Van Der Vorst, Engels and Burk2010). Incomplete control for demographics and risk factors was noted in all seven cohorts (Kaynak et al. Reference Kaynak, Winters, Cacciola, Kirby and Arria2014).
In this last regard, several other parental (Barnes et al. Reference Barnes, Reifman, Farrell and Dintcheff2000), familial (Nash et al. Reference Nash, McQueen and Bray2005), child (Zernicke et al. Reference Zernicke, Cantrell, Finn and Lucas2010) and peer (Borsari & Carey, Reference Borsari and Carey2001) predictors, suggested by parenting and child socialization research and theory (Oetting & Donnermeyer, Reference Oetting and Donnermeyer1998; Ajzen, Reference Ajzen2002; Real & Rimal, Reference Real and Rimal2007), are associated with adolescent drinking (Fisher et al. Reference Fisher, Miles, Austin, Camargo and Colditz2007; Alati et al. Reference Alati, Maloney, Hutchinson, Najman, Mattick, Bor and Williams2010; Donovan & Molina, Reference Donovan and Molina2011; Swendsen et al. Reference Swendsen, Burstein, Case, Conway, Dierker, He and Merikangas2012; Kuperman et al. Reference Kuperman, Chan, Kramer, Wetherill, Bucholz, Dick, Hesselbrock, Porjesz, Rangaswamy and Schuckit2013; Rossow et al. Reference Rossow, Keating, Felix and McCambridge2016). Factors such as parental/familial drinking/problems (Donovan & Molina, Reference Donovan and Molina2011), family intactness (Donovan & Molina, Reference Donovan and Molina2011; Kuperman et al. Reference Kuperman, Chan, Kramer, Wetherill, Bucholz, Dick, Hesselbrock, Porjesz, Rangaswamy and Schuckit2013), monitoring of child activities (DiClemente et al. Reference Diclemente, Wingood, Crosby, Sionean, Cobb, Harrington, Davies, Hook and Oh2001), child externalizing/internalizing (Crum et al. Reference Crum, Green, Storr, Chan, Ialongo, Stuart and Anthony2008) and peer substance use (Ary et al. Reference Ary, Duncan, Duncan and Hops1999) are often unmeasured. Incomplete adjustment does not constitute strong analysis, and creates uncertainty about relationships (McCambridge et al. Reference McCambridge, McAlaney and Rowe2011). In addition, when researching associations of parental supply with child drinking, there is also a need to control for alcohol supplied from non-parental sources (other adults, friends, siblings, self-purchased, etc.), referred to hereafter as ‘other supply’. Supply of alcohol from non-parental sources has been associated with greater drinking and negative outcomes than parental supply (Foley et al. Reference Foley, Altman, Durant and Wolfson2004; Bellis et al. Reference Bellis, Hughes, Morleo, Tocque, Hughes, Allen, Harrison and Fe-Rodriguez2007; Dietze & Livingston, Reference Dietze and Livingston2010), raising the potential for differential harms. However, no prospective study has reported how these different sources are associated with adolescent drinking. Overall, the piecemeal picture limits confident conclusions from existing research.
We investigated the associations between exposure to parental supply and other supply of sips or whole drinks of alcohol across four adolescent years (approximately 12–15 years of age), and two outcomes –consumption of whole drink(s) (drinking) and binge drinking (bingeing; consuming more than four standard drinks on a single occasion, a pattern of drinking that puts the drinker at increased risk of harm; National Health and Medical Research Council, 2009) – using longitudinal mixed-model analytic methods, unadjusted and adjusted for the full range of relevant covariates. Guided by the literature and pre-specified aims (Aiken et al. Reference Aiken, Wadolowski, Bruno, Najman, Kypri, Slade, Hutchinson, McBride and Mattick2015), we hypothesized that while parental supply, other supply, and the covariates would have significant unadjusted associations with these two outcomes, the adjusted analyses would show parental supply of alcohol to have the greatest odds of drinking and bingeing, given concerns that such provision signals parental permissiveness (Kaynak et al. Reference Kaynak, Winters, Cacciola, Kirby and Arria2014).
Method
Design
This longitudinal cohort study is registered (ClinicalTrials.gov: NCT02280551). Institutional review board (IRB), and STROBE (Strengthening the Reporting of Observational Studies in Epidemiology) statement reporting requirements are met (von Elm et al. Reference Von Elm, Altman, Egger, Pocock, Gøtzsche and Vandenbroucke2007).
Sample recruitment and characteristics
In 2010–2011 a cohort of adolescents and parents was recruited from grade 7 classes in Sydney, Hobart and Perth; detailed methods are described elsewhere (Aiken et al. Reference Aiken, Wadolowski, Bruno, Najman, Kypri, Slade, Hutchinson, McBride and Mattick2015). As in similar research (Jackson et al. Reference Jackson, Ennett, Dickinson and Bowling2012), schools made recruitment materials available to students (either by mail or face-to-face at school), but had no other role. Study information packs were distributed to grade 7 students at participating schools. Parents could submit a form indicating interest in consenting to be in the study, and were eligible to participate if: (a) the child was enrolled in grade 7; and (b) signed parental consent was provided. Of 2017 parents expressing interest in their child participating, 90 were ineligible, resulting in 1927 adolescents in the cohort (see Fig. 1 for details of eligibility and retention); 1910 adolescents provided baseline data, with high retention thereafter (>85%). Adolescents were a mean of 12.9 years of age at baseline, and similar to the Australian population on important demographic measures: 44.9% of adolescents were female (48.7% in the population of 12- to 13-year-olds); there were 2.6 adolescents per household (1.9 in the population); 79.6% of the adolescents lived in two-parent households (81.0% in the population); 73.8% of parents were Australian-born (72.3% in the adult population); 73.4% of parents had post-high school education (67% in the adult population); 81.2% of parents were employed (80.0% of males and 65% of females in the adult population); median household income was in the same range as in the population (Aiken et al. Reference Aiken, Wadolowski, Bruno, Najman, Kypri, Slade, Hutchinson, McBride and Mattick2015).
Time and study flowchart of recruitment and assessment of the cohort (n = 1927).
Rates of alcohol use in the adolescent cohort, and of parental drinking and supply, were also similar to those in population surveys. National school student samples (past-12-month alcohol use) and our sample (past 12-month alcohol use) had similar drinking: 21.3% of 12-year-olds drank in 2011 v. 19.8% of our sample that year; 32.1% of 13-year-olds drank in 2012 v. 32.9% of our sample that year (White & Bariola, Reference White and Bariola2012). School data for 2013 and 2014 are not available, but other national survey data showed a secular decline of approximately 10% by 2013 in 12- to 17-year-olds ever consuming alcohol (Pennay et al. Reference Pennay, Livingston and MacLean2015). Applying this secular trend to the school survey data: the national rate of 14-year-olds drinking of 45.9% should decrease to about 35% by 2013 (compared with the rate of 37.3% of 14-year-olds in the cohort by 2013); and from 60.2 to about 50% (compared with our rate of 47.9% of 15-year-olds in the cohort by 2014). Parental supply was reported by 34.7% of 15-year-olds in the cohort (Table 2), compared with 34.9% of 12- to 15-year-olds in population surveys (White & Bariola, Reference White and Bariola2012). Parental frequency of drinking (2.6% daily, 48.6% weekly, 38.3% less than weekly, 10.5% not in the past 12 months) (Aiken et al. Reference Aiken, Wadolowski, Bruno, Najman, Kypri, Slade, Hutchinson, McBride and Mattick2015) was similar to Australian adult population use (7.7% daily, 41.7% weekly, 33.0% less than weekly, and 7.7% not in the past 12 months), indicating regular alcohol involvement (Australian Institute of Health and Welfare, 2011).
Measures
Outcome variables [whole standard drink(s) and binge drinking]
Primary outcomes were: (a) past 12-month consumption of whole standard drink(s) (drinking); (b) binge drinking. A binary variable (no/yes) was constructed to indicate drinking one or more whole standard drink(s) (10 g of alcohol), and consumption of more than four standard drinks (no/yes) on any single occasion (Bush et al. Reference Bush, Kivlahan, McDonell, Fihn and Bradley1998) was coded into a binary variable: binge drinking. Current Australian guidelines recommend drinking no more than four standard drinks on a single occasion to reduce the risk of alcohol-related injury; hence binge drinking was coded as consumption as more than four standard drinks on one occasion (National Health and Medical Research Council, 2009). Secondary analyses were conducted of typical quantity consumed on a drinking occasion, and number of drinks consumed in the year.
Exposure variables
Parental supply
Children were asked about who supplied them alcohol (sips or whole drinks) in the past 12 months, including mother, father, other adults, friends, siblings and self-supply. From this, a dichotomous exposure variable was coded indicating those who had received supply of alcohol from parental supply, and those who had not. For a secondary planned dose–response analysis, an exposure variable was coded into a measure of the number of years in which parental supply occurred (0, 1, 2, or 3) (see Statistical analysis).
Other supply
Another exposure variable of ‘other supply’ included supply from other adults, friends, siblings or self-supply, compared with adolescents reporting no supply from these sources. Parental supply and other supply were not mutually exclusive, and as supply could be derived from both sources, each source was controlled for separately in analyses.
Covariates
Covariates identified from the literature as associated with adolescent drinking were measured annually. These variables are fully described in the online Supplementary material, and included: parental factors [alcohol use (Donovan & Molina, Reference Donovan and Molina2011; Swendsen et al. Reference Swendsen, Burstein, Case, Conway, Dierker, He and Merikangas2012), alcohol accessible at home without parental knowledge (Swendsen et al. Reference Swendsen, Burstein, Case, Conway, Dierker, He and Merikangas2012), alcohol-specific rules (Van Der Vorst et al. Reference Van Der Vorst, Engels, Meeus, Dekovic and Van Leeuwe2005; Van Der Vorst et al. Reference Van Der Vorst, Engels, Deković, Meeus and Vermulst2007), monitoring (Swendsen et al. Reference Swendsen, Burstein, Case, Conway, Dierker, He and Merikangas2012), responsiveness/demandingness/consistency (Alati et al. Reference Alati, Maloney, Hutchinson, Najman, Mattick, Bor and Williams2010; Donovan & Molina, Reference Donovan and Molina2011), religiosity (Donovan & Molina, Reference Donovan and Molina2011)]; family factors [one- or two-parent household (Alati et al. Reference Alati, Maloney, Hutchinson, Najman, Mattick, Bor and Williams2010), family conflict/positive relations (Ary et al. Reference Ary, Duncan, Duncan and Hops1999), family alcohol problems (Kuperman et al. Reference Kuperman, Chan, Kramer, Wetherill, Bucholz, Dick, Hesselbrock, Porjesz, Rangaswamy and Schuckit2013), older siblings (Fisher et al. Reference Fisher, Miles, Austin, Camargo and Colditz2007)]; child factors [sex, age in years, or part thereof, at time of survey completion (Fisher et al. Reference Fisher, Miles, Austin, Camargo and Colditz2007; Swendsen et al. Reference Swendsen, Burstein, Case, Conway, Dierker, He and Merikangas2012), money to purchase alcohol (Swendsen et al. Reference Swendsen, Burstein, Case, Conway, Dierker, He and Merikangas2012), tobacco use (Kuperman et al. Reference Kuperman, Chan, Kramer, Wetherill, Bucholz, Dick, Hesselbrock, Porjesz, Rangaswamy and Schuckit2013), externalizing (Swendsen et al. Reference Swendsen, Burstein, Case, Conway, Dierker, He and Merikangas2012; Kuperman et al. Reference Kuperman, Chan, Kramer, Wetherill, Bucholz, Dick, Hesselbrock, Porjesz, Rangaswamy and Schuckit2013), internalizing (Crum et al. Reference Crum, Green, Storr, Chan, Ialongo, Stuart and Anthony2008; Kuperman et al. Reference Kuperman, Chan, Kramer, Wetherill, Bucholz, Dick, Hesselbrock, Porjesz, Rangaswamy and Schuckit2013), problems socializing (Achenbach, Reference Achenbach1991)]; and peer factors [peer substance use, and peer disapproval of alcohol/tobacco use (Fisher et al. Reference Fisher, Miles, Austin, Camargo and Colditz2007; Swendsen et al. Reference Swendsen, Burstein, Case, Conway, Dierker, He and Merikangas2012; Kuperman et al. Reference Kuperman, Chan, Kramer, Wetherill, Bucholz, Dick, Hesselbrock, Porjesz, Rangaswamy and Schuckit2013)]. Participants completed annual paper or online questionnaires, forwarded separately to adolescents and parents to minimize reporting biases. Data from 4 years (waves 1 to 4) are included in this study.
Statistical analysis
Primary
Logistic regressions (random intercept mixed-effects model controlling for within-respondent and within-school clustering/correlation, and time) determined the relationship between parental supply and other supply and two outcomes: (a) drinking whole standard drinks; and (b) binge drinking, controlling for each source of supply and covariates and collinearity. We analysed 4 years of data, with parental supply (yes/no) from each year used to model the two outcomes in the following years, while controlling for covariates (fully adjusted model). The analysis included paired time periods (period 1 = waves 1–2; period 2 = waves 2–3; period 3 = waves 3–4), with exposures and covariates from each year predicting the outcomes in all the later years. Variables were included in adjusted analyses if they showed unadjusted significance at an α of 0.05 (five covariates were dropped).
Secondary
Fixed-effect logistic regressions assessed for a dose–response between number of years in which parental supply occurred (0, 1, 2, or all 3 initial years) and drinking outcomes in the 4th year. Analyses were also conducted using multiple imputation for missing data (online Supplementary Table S1). Sensitivity analyses tested robustness of the dose–response relationship under three scenarios: (a) complete case analysis (no imputation); (b) all participants with missing outcomes coded as drinkers/bingers; and (c) all participants with missing outcomes coded as non-drinkers. Analyses used Stata 14.1 (Stata Corporation, 2012). Number of drinks consumed in the year and single occasion drinking rates were also examined using negative binomial regression (random intercept mixed-effects). Finally, potential moderating effects between parental supply and four other variables on the primary outcomes were investigated through analysis of interaction terms between parental supply and: child externalizing, family history of alcohol problems, peer substance use, and peer disapproval of substance use.
Results
Sample drinking behaviour over time
There was a steady increase in consumption of drinks and bingeing across the waves (Table 1). Table 2 provides the sources and mean number of drinks/bingeing consumed by adolescent drinkers, separately for adolescents reporting any parental supply, and those reporting any other supply. Mean consumption was similar across supply source. The main sources of ‘other supply’ were other adults and peers, and the frequency of supply from each ‘other supply’ source is presented in online Supplementary Table S2.
Adolescent drinking and binge drinking rates and percentages at each measurement wave a
a Students were a mean age of 12.9 years at wave 1.
b Number of respondents with valid data.
c Number of participants reporting the behaviour at each measurement wave.
Source(s) of supply of alcohol and rates of adolescent drinking whole standard drinks and binge drinking by wave a
s.d., Standard deviation; p.a., per annum.
a Bingeing is defined as >4 standard drinks on an occasion.
b The categories of ‘parental supply’ and ‘other supply’ are not mutually exclusive; adolescents could derive supply from more than one of these two sources – the analyses of the impacts of parental supply controlled for other supply (and the other covariates) when estimating odds ratios/associations and vice versa, so that the impact of each type of supply can be estimated independent of and controlling for each type of supply and each covariate.
Unadjusted analyses
Unadjusted, parental supply was strongly associated with both drinking whole standard drinks [odds ratio (OR) 5.71, 95% confidence interval (CI) 4.41–7.39] and bingeing (OR 4.66, 95% CI 3.48–6.24) (Table 3). Of the 30 variables analysed, the same six showed no significant unadjusted associations with drinking whole drinks, or bingeing: having older siblings; parent is Australian-born; parent education; parent employment; home access to alcohol; and family history of alcohol problems. These variables were not included in fully adjusted models.
ORs and 95% CIs for associations of source of supply at current wave, and subsequent wave drinking of whole beverages and binge drinking
OR, Odds ratio; CI, confidence interval; s.d., standard deviation; YSR, Youth Self Report.
a Results of random intercept mixed-effects logistic regression models. Data included as paired waves, with covariates from one wave and outcome from subsequent wave. Significant random effects were observed for the multivariate logistic mixed model of drinking whole drinks (school intercept s.d. = 0.37, individual intercept s.d. = 1.54, p < 0.001) and binge drinking (school intercept s.d. = 0.38, individual intercept s.d. = 1.45, p < 0.001).
b Models controlled for other covariates (see online Supplementary Tables S3 and S4): parental average alcohol use; home access to and availability of alcohol; parental alcohol specific rules; parental monitoring; authoritative parenting: demandingness; authoritative parenting: responsiveness; parenting consistency; parental religiosity at baseline; parent born in Australia; two-parent household; family conflict; family positive relations; relative socio-economic disadvantage of area of residence; household income; child has money to buy alcohol; smoking; YSR: externalizing; YSR: anxious/depressed; YSR: withdrawn/depressed; YSR: social problems; age; sex; peer use of alcohol and/or tobacco; peer disapproval of alcohol and/or tobacco use.
c Included here was alcohol supply from: a family friend or relative or adult; a brother/sister; friends; or alcohol received as part of religious service; or alcohol where ‘I got it myself’.
Adjusted analyses
Adjusting for other variables, we still found significant associations between parental supply and drinking whole drinks, but not with bingeing. Parental supply at any wave was associated with doubled odds of drinking at subsequent waves (OR 1.80, 95% CI 1.33–2.45), but was not associated with bingeing (OR 1.12, 95% CI 0.80–1.55). In contrast, other supply of alcohol from non-parental sources was associated with significant increased risk of both outcomes, unadjusted and adjusted. While supply from other sources was similarly associated with a doubling in the odds of drinking whole drinks (OR 2.53, 95% CI 1.86–3.45), it was associated with a more than tripling in the odds of binge drinking (OR 3.51, 95% CI 2.53–4.87) (online Supplementary Tables S3 and S4). Consistent with this result, adolescents who received alcohol only from other supply sources drank significantly more drinks than adolescents who were supplied alcohol only by their parents (and who were not supplied by other supply sources) (F = 30.80, p < 0.001).
Drinking and bingeing were also associated with time, although in different ways. The odds of consuming whole drinks increased by around half in time period 2 (OR 1.45, 95% CI 1.05–2.01), before increasing markedly in time period 3 (OR 4.75, 95% CI 3.32–6.78) (online Supplementary Table S3). In contrast, binge drinking remained steady in time period 2 (OR 1.15, 95% CI 0.79–1.67), but increased significantly in time period 3 (OR 3.13, 95% CI 2.13–4.60) (online Supplementary Table S4).
Secondary analyses showed odds increasing with number of years of parental supply, consistent with a dose–response relationship, although the 95% CIs overlapped. There was an association (p = 0.003) between the number of waves of parental supply and drinking in wave 4; those supplied alcohol even once by parents in any of waves 1–3 were more likely to consume whole drinks in wave 4 (OR 1.48, 95% CI 1.03–2.12), and odds of consumption increased among those who were supplied alcohol in 2 years during waves 1–3 (OR 1.92, 95% CI 1.29–2.85), increasing further if supply occurred in all waves 1–3 (OR 2.15, 95% CI 1.21–3.82). Secondary analysis of number of waves of parental supply showed no dose–response relationship for bingeing (online Supplementary Table S5). Finally, the secondary analyses showed similar results for covariates as primary analyses, suggesting robust relationships.
Sensitivity analyses
Analysis of the number of drinks consumed in the year showed a similar pattern to analysis of drinking whole drinks. Parental supply of alcohol was associated with a fourfold increase in the incidence rate of number of drinks [incidence rate ratio (IRR) 3.68, 95% CI 2.30–5.90] with a similar increase associated with other supply (IRR 2.67, 95% CI 1.63–4.35) (online Supplementary Table S6). Number of drinks also showed a strong increase in the incidence rate over time, doubling in time period 2 (IRR 2.35, 95% CI 1.47–3.74), before a dramatic increase in time period 3 (IRR 10.70, 95% CI 6.52–17.56). The sensitivity analysis of number of drinks consumed on a single typical drinking occasion (online Supplementary Table S7) supports the results of the model of binge drinking. Adolescents supplied alcohol by parents drank lower numbers of drinks on a typical occasion (IRR 0.89; 95% CI 0.79–0.99) than those adolescents supplied only by other sources (i.e. other supply). The sensitivity analyses of the dose–response relationship found that this relationship was maintained in complete case analysis. When more extreme assumptions were made, either assuming missing outcome data were all non-drinkers or all drinkers, the significant association with parental supply remained, although the pattern of ORs was less consistent with a dose–response relationship (online Supplementary Tables S8–S10). Two additional post-hoc sensitivity analyses were also conducted. First, to separate out the influence of sources of ‘other supply’ on drinking behaviour, we conducted post-hoc analyses of the two primary outcomes with each source of ‘other supply’ entered separately (i.e. other adults, siblings, peers, religious service, and self-supply). The only significant ‘other supply’ sources predictive of drinking and/or bingeing were self-supply (OR 2.62, 95% CI 1.71–4.03 and OR 3.16, 95% CI 2.11–4.74, respectively) and peer-supply (OR 3.91, 95% CI 2.16–7.08 and OR 2.64, 95% CI 1.56–4.49). Second, to eliminate the possibility that results were influenced by less common living arrangements, we re-conducted the primary analysis limiting the sample to adolescents who reported living in a two-parent mixed-sex (mother/father) household. A breakdown of adolescent-rearing environment for each wave is presented in online Supplementary Tables S11 and S12. The sensitivity analysis showed no substantive differences to the primary analysis.
Moderating variables
Analysis of potential moderating effects showed two significant interactions (of child externalizing and peer substance use) with parental supply on the two outcomes. For both outcomes, the effect of parental supply was stronger with lower externalizing scores, with the effect declining as externalizing score increased (online Supplementary Fig. S1). That is, if a child is higher on externalizing, parental supply is less important to that child's drinking – externalizers seem more likely to seek out alcohol whether their parents supply or not. On the other hand, the interaction of parental supply and peer substance use showed a reversal in the direction of effect, with parental supply increasing the odds of drinking when low peer substance use was observed, but decreasing the odds when peer substance use was higher (online Supplementary Fig. S2). Family history of alcohol problems, and peer disapproval did not show any significant interaction effect.
Discussion
This first longitudinal study of associations between parental supply and other supply of alcohol and adolescent drinking, conducted over 4 years in a large representative cohort with excellent retention, comprehensively adjusted for known covariates (Kaynak et al. Reference Kaynak, Winters, Cacciola, Kirby and Arria2014). Before adjustment, parental supply of alcohol was associated with five-fold increased unadjusted odds of consuming whole drinks, consistent with earlier studies (Jackson et al. Reference Jackson, Henriksen and Dickinson1999; Komro et al. Reference Komro, Maldonado-Molina, Tobler, Bonds and Muller2007; Shortt et al. Reference Shortt, Hutchinson, Chapman and Toumbourou2007; Livingston et al. Reference Livingston, Testa, Hoffman and Windle2010; Van Der Vorst et al. Reference Van Der Vorst, Engels and Burk2010; Danielsson et al. Reference Danielsson, Romelsjö and Tengström2011; McMorris et al. Reference McMorris, Catalano, Kim, Toumbourou and Hemphill2011). Also, there were increased unadjusted odds of bingeing among adolescents supplied alcohol by parents. In adjusted longitudinal mixed-model analyses which: (a) controlled for familial/adolescent/peer characteristics in each wave being interrelated; (b) assessed parental supply controlling for growth in drinking rates over time; and (c) included variables with significant unadjusted associations, parental supply was still associated with a doubling of the odds of drinking, but no longer associated with bingeing. Our results are consistent with the results of studies that consistently find increased drinking, but not necessarily increased heavy episodic drinking (Kaynak et al. Reference Kaynak, Winters, Cacciola, Kirby and Arria2014). We also found adolescents supplied alcohol by parents drank less on a typical drinking occasion than adolescents supplied alcohol from other sources (and not from their parents), as some others also report (Foley et al. Reference Foley, Altman, Durant and Wolfson2004; Bellis et al. Reference Bellis, Hughes, Morleo, Tocque, Hughes, Allen, Harrison and Fe-Rodriguez2007; Dietze & Livingston, Reference Dietze and Livingston2010), possibly due to the supervisory nature where there is parental presence. In this last regard, it is likely that the drinking context matters, as do child perceptions of parental permissiveness, issues which we are pursuing in this cohort. There was no evidence of over-adjustment or collinearity affecting these results.
As expected from other research (Alati et al. Reference Alati, Najman, Kinner, Mamun, Williams, O'Callaghan and Bor2005; Fisher et al. Reference Fisher, Miles, Austin, Camargo and Colditz2007; Donovan & Molina, Reference Donovan and Molina2011; Kuperman et al. Reference Kuperman, Chan, Kramer, Wetherill, Bucholz, Dick, Hesselbrock, Porjesz, Rangaswamy and Schuckit2013), we also detected some variables associated with lower odds of drinking, specifically: parental monitoring; responsive and consistent parenting; religiosity; child social problems; and peer disapproval of substance use (online Supplementary Table S3). Increased odds of drinking were observed for: peer substance use and child externalizing. Turning to bingeing, odds were decreased by: alcohol-specific rules; responsive and consistent parenting; religiosity; and peer disapproval of substance use. Externalizing was associated with increased odds of bingeing, as reported elsewhere (Swendsen et al. Reference Swendsen, Burstein, Case, Conway, Dierker, He and Merikangas2012; Kuperman et al. Reference Kuperman, Chan, Kramer, Wetherill, Bucholz, Dick, Hesselbrock, Porjesz, Rangaswamy and Schuckit2013).
Our results also showed a dose–response relationship: adolescents supplied alcohol by parents in the one, two, or in all three initial years were increasingly more likely to drink whole beverages in year 4. Sensitivity analyses showed that this relationship was robust when only complete cases were analysed, and also when missing responses were classified as non-drinkers. Interestingly, being supplied alcohol from other sources increased the odds of both drinking and bingeing. It thus seems that while both parental supply and other supply were equally related to drinking, only other supply – but not parental supply – was more likely to be associated with bingeing. Findings from these mixed-effect regressions were replicated in fixed-effect logistic regression analyses, indicating robust results, confirming relationships reported in cross-sectional studies (Foley et al. Reference Foley, Altman, Durant and Wolfson2004; Bellis et al. Reference Bellis, Hughes, Morleo, Tocque, Hughes, Allen, Harrison and Fe-Rodriguez2007; Dietze & Livingston, Reference Dietze and Livingston2010). Notably, given no other cohort studies control for other supply, associations of parental supply with drinking in those studies may be inflated. This is an important point, as causality is often inferred via associations.
Our study has limitations. First, IRB approval allowed information about the study to be available through classrooms, but families had to opt in, so the self-selected sample restricts generalizations to the population, and reporting biases may underestimate population rates of alcohol consumption (Stockwell et al. Reference Stockwell, Donath, Cooper-Stanbury, Chikritzhs, Catalano and Mateo2004). However, our child and parent drinking rates were very similar to those in population surveys (Australian Institute of Health and Welfare, 2011; White & Bariola, Reference White and Bariola2012;Wadolowski et al. Reference Wadolowski, Bruno, Aiken, Stone, Najman, Kypri, Slade, Hutchinson, McBride and Mattick2015), and the sample closely matches the Australian population in sex, age, household composition and socio-economic status. Additionally with regard to bias, we aimed to assess associations between exposures and outcomes, where population representativeness is less crucial. These observations suggest selective responding and reporting may not have substantially affected the alcohol estimates and associations herein. Second, the legal age to purchase alcohol in Australia is 18 years (parents may legally provide alcohol to children), so generalizations to other cultures should be considered carefully, although there is evidence of some cross-country similarities (McMorris et al. Reference McMorris, Catalano, Kim, Toumbourou and Hemphill2011). Third, the cohort is too young to show development of alcohol-use disorders (Hingson et al. Reference Hingson, Heeren and Winter2006), so we cannot assess long-term harms (Gore et al. Reference Gore, Bloem, Patton, Ferguson, Joseph, Coffey, Sawyer and Mathers2011; Patton et al. Reference Patton, Coffey, Cappa, Currie, Riley, Gore, Degenhardt, Richardson, Astone, Sangowawa, Mokdad and Ferguson2012). Fourth, we have not explored associations between either amounts (sips v. drinks) or settings (supervised v. unsupervised) of parental supply and adolescent consumption. Such analyses between drinking and the amounts/frequency and settings of parental supply, or indeed the specific sources of other supply (Samek et al. Reference Samek, McGue, Keyes and Iacono2015), are beyond our scope here, but need to be undertaken. Fifth, we do not address population trends over time (Keyes et al. Reference Keyes, Schulenberg, O'Malley, Johnston, Bachman, Li and Hasin2012), advertising or environmental/regulatory factors (Moreno et al. Reference Moreno, Furtner and Rivara2011; Tanski et al. Reference Tanski, McClure, Li, Jackson, Morgenstern, Li and Sargent2015).
While we report associations, the findings do strongly suggest the potential for parental supply to accelerate adolescent alcohol consumption, compared with no supply, consistent with conclusions by others (Kaynak et al. Reference Kaynak, Winters, Cacciola, Kirby and Arria2014). This possibility is concerning, given adolescent vulnerabilities to the effects of alcohol (Brown et al. Reference Brown, McGue, Maggs, Schulenberg, Hingson, Swartzwelder, Martin, Chung, Tapert, Sher, Winters, Lowman and Murphy2008), and associations with adverse adult outcomes (Patton et al. Reference Patton, Coffey, Cappa, Currie, Riley, Gore, Degenhardt, Richardson, Astone, Sangowawa, Mokdad and Ferguson2012). However, while adolescents supplied alcohol by parents were more likely to drink than those not supplied, parental supply was not associated with increased bingeing. This finding was supported by less alcohol being consumed on a typical drinking occasion by adolescents who have been supplied by parents, compared with those accessing alcohol from other sources (controlling for each source in the analyses). Also, in contrast, and somewhat surprisingly, other supply was associated with increased odds of bingeing after adjustment.
There are several possible explanations for the pattern of findings. Given that parental supply increases the odds of drinking, but not of bingeing, parental supply may have a protective effect, possibly due to the supervised nature of the supply. However, our view is that such a conclusion is premature at this time. These results should not be taken to suggest that parental supply is somehow protective of bingeing in the longer term. In fact, parents may be accelerating children into drinking alcohol, and laying down the potential for later harms (Shaffer et al. Reference Shaffer, Fisher, Lucas, Dulcan and Schwab-Stone2000; Hingson et al. Reference Hingson, Heeren and Winter2006; Jackson et al. Reference Jackson, Ennett, Dickinson and Bowling2012; Patton et al. Reference Patton, Coffey, Cappa, Currie, Riley, Gore, Degenhardt, Richardson, Astone, Sangowawa, Mokdad and Ferguson2012). It is possible that parents supplying alcohol to their children are setting up a pattern of drinking; higher frequency but lower volume than those supplied by others. Parental supply of alcohol may ultimately signify permissiveness to children, and while only other supply was associated with bingeing in this study, the patterns of use may alter as these children grow older. Further investigation of the patterns of use in adolescence and in early adulthood is needed. The significant interaction effect of child externalizing behaviour further adds to the complexity of the findings; depending on the children themselves, the impact of parental supply of alcohol is likely to be different. While these parents may be supplying alcohol as an attempt to moderate their child's drinking, or because they believe their child will receive alcohol from their peers anyway, as we have found in this cohort (Wadolowski et al. Reference Wadolowski, Hutchinson, Bruno, Aiken, Najman, Kypri, Slade, McBride and Mattick2016), it seems that children high on externalizing are less influenced by parental supply and seem likely to obtain alcohol whether their parents supply or not.
This is clearly a highly complex area and one that has significant relevance to public and mental health professionals and policymakers, and parents, as understanding these relationships can alter professional opinion, and parental and child behaviour. The results have international relevance as alcohol use increases with the growth in wealth of developing economies whose populations embrace Western habits (Jiang et al. Reference Jiang, Room and Hao2015). Given that the findings here are limited to drinking between early and mid-teen years, further study of the cohort is essential to understand the longer-term relationships between parental supply and early adult alcohol use and bingeing.
The questions are clear. Does parental supply protect in the short-term and set up patterns of moderate (non-binge) drinking into adulthood, or lead to heavier drinking when adolescents are of a legal age to drink? This is the issue that is central to our work – does parental supply harm or help? A precautionary principle remains appropriate for parents to ensure that early drinking does not compromise child well-being (Furtner & Rivara, Reference Furtner and Rivara2011). Follow-up of this cohort into adulthood will help clarify impacts of early drinking on more mature consumption patterns.
Supplementary material
The supplementary material for this article can be found at http://dx.doi.org/10.1017/S0033291716002373
Acknowledgements
This research was funded by: a 2010–2014 Australian Research Council Discovery Project Grant (DP:1096668) to R.P.M., J.N., K.K., T.S. and D.H., supplemented by funds from two Australian Rotary Health Mental Health Research Grants to R.P.M., M.W., A.A., J.N., K.K., T.S., D.H. and R.B.; an Australian Rotary Health Whitcroft Family PhD Scholarship and an Australian Postgraduate Award to M.W.; National Health and Medical Research Council Principal Research Fellowship Grants to R.P.M. (APP1045318) and K.K. (GNT0188568 and APP1041867); National Health and Medical Research Council Project Grants to R.P.M. for a Longitudinal Cohorts Research Consortium (GNT1009381 and GNT1064893); a Research Innovation Grant from the Australian Foundation for Alcohol Research and Education to R.P.M., J.N., K.K., T.S., D.H., R.B. and M.W.; the National Drug and Alcohol Research Centre, UNSW Australia, which is supported by an Australian Government Substance Misuse Prevention and Service Improvements Grant.
We have no financial relationships relevant to this article to disclose.
R.P.M., J.N., K.K., T.S. and D.H. conceptualized the study and, with M.W, A.A., R.B. and N.M., designed the data collection methods. M.W., A.A. and R.P.M. acquired the data. P.J.C. and R.P.M. conducted the data analyses, and R.P.M. and P.J.C. drafted the manuscript. All authors provided substantial contributions to the interpretation of the results, and all critically revised the manuscript for important intellectual content, and approved the final manuscript as submitted. All authors have agreed to be accountable for all aspects of the work.
Declaration of Interest
None.
