High rates of substance use (alcohol, tobacco and cannabis) in the perinatal period, in pregnancy and in the first year following childbirth, are of significant clinical and public health concern globally (Louw, Reference Louw2018; Rodriguez & Smith, Reference Rodriguez and Smith2019). It is well established that high maternal perinatal substance use poses significant risks to offspring health and development (Conner, Carter, Tuuli, MacOnes, & Cahill, Reference Conner, Carter, Tuuli, MacOnes and Cahill2015; Crume, Reference Crume2019; El Marroun et al., Reference El Marroun, Tiemeier, Steegers, Jaddoe, Hofman, Verhulst and Huizink2009; Floyd, Decouflé, & Hungerford, Reference Floyd, Decouflé and Hungerford1999; Philips et al., Reference Philips, Santos, Trasande, Aurrekoetxea, Barros, von Berg and Jaddoe2020). In utero exposure to alcohol, tobacco and cannabis are associated with restricted foetal growth and preterm birth and can cause harms in early pregnancy even prior to pregnancy recognition (Conner et al., Reference Conner, Carter, Tuuli, MacOnes and Cahill2015; Crume, Reference Crume2019; El Marroun et al., Reference El Marroun, Tiemeier, Steegers, Jaddoe, Hofman, Verhulst and Huizink2009; Floyd et al., Reference Floyd, Decouflé and Hungerford1999; Philips et al., Reference Philips, Santos, Trasande, Aurrekoetxea, Barros, von Berg and Jaddoe2020). Substance use in the first year postpartum is further associated with depressive symptoms in mothers, infant exposure through breastfeeding and second hand smoke and other disruptions to parental care (Chapman & Wu, Reference Chapman and Wu2013; Forray, Reference Forray2016; Herrmann, King, & Weitzman, Reference Herrmann, King and Weitzman2008; Ko et al., Reference Ko, Tong, Bombard, Hayes, Davy and Perham-Hester2018; Polcaro & Vettraino, Reference Polcaro and Vettraino2020). In the longer term, parent substance use also increases risks of future substance use in offspring (Crume, Reference Crume2019; De Genna, Goldschmidt, Richardson, Cornelius, & Day, Reference De Genna, Goldschmidt, Richardson, Cornelius and Day2018; Kerr, Capaldi, Pears, & Owen, Reference Kerr, Capaldi, Pears and Owen2012). For these reasons, guidelines in Australia and other countries recommend abstinence of substance use in pregnancy (Australian Government Department of Health, 2019; O'Keeffe et al., Reference O'Keeffe, Kearney, McCarthy, Khashan, Greene, North and Kenny2015).
However, in high-income countries including Australia, prevalence of perinatal substance use is high. Within general population samples, it is estimated that 30–60% of mothers drink alcohol during pregnancy (McCormack et al., Reference McCormack, Hutchinson, Burns, Wilson, Elliott, Allsop and Mattick2017; Muggli et al., Reference Muggli, O'Leary, Donath, Orsini, Forster, Anderson and Halliday2016), 11–14% smoke tobacco cigarettes (Crume, Reference Crume2019; Kurti et al., Reference Kurti, Redner, Lopez, Keith, Villanti, Stanton and Higgins2017) and approximately 4% use cannabis (el Marroun et al., Reference el Marroun, Tiemeier, Jaddoe, Hofman, Mackenbach, Steegers and Huizink2008; Ko et al., Reference Ko, Tong, Bombard, Hayes, Davy and Perham-Hester2018), with cannabis use likely to rise with recent policy changes to decriminalisation in many countries (Metz & Stickrath, Reference Metz and Stickrath2019; Young-Wolff et al., Reference Young-Wolff, Sarovar, Tucker, Conway, Alexeeff, Weisner and Goler2019). Furthermore, while most women decrease or cease substance use upon pregnancy recognition (typically after the fourth week of gestation) (Floyd et al., Reference Floyd, Decouflé and Hungerford1999), a high proportion continue use after pregnancy recognition (e.g. approximately 30% continue to drink alcohol at low levels) (Hutchinson, Moore, Breen, Burns, & Mattick, Reference Hutchinson, Moore, Breen, Burns and Mattick2013; Muggli et al., Reference Muggli, O'Leary, Donath, Orsini, Forster, Anderson and Halliday2016). Following childbirth, many women recommence use in the postpartum period (Borschmann et al., Reference Borschmann, Becker, Spry, Youssef, Olsson, Hutchinson and Patton2019; De Genna et al., Reference De Genna, Goldschmidt, Richardson, Cornelius and Day2018; Hutchinson et al., Reference Hutchinson, Moore, Breen, Burns and Mattick2013). These trends make it pertinent to more precisely define who is most at risk of substance use in the perinatal period and when it is most effective to intervene.
From a life course perspective, most substance use problems have their origins in adolescence and young adulthood, with earlier onset, as well as heavier and more frequent use, associated with worse outcomes (Kessler et al., Reference Kessler, Berglund, Demler, Jin, Merikangas and Walters2005; Luciana & Feldstein Ewing, Reference Luciana and Feldstein Ewing2015; McGorry, Purcell, Goldstone, & Amminger, Reference McGorry, Purcell, Goldstone and Amminger2011; Palmer et al., Reference Palmer, Young, Hopfer, Corley, Stallings, Crowley and Hewitt2009). Consistent with this, there is emerging evidence from retrospective studies that preconception substance use may be an important determinant of substance use in pregnancy (Skagerstróm, Chang, & Nilsen, Reference Skagerstróm, Chang and Nilsen2011; Young-Wolff et al., Reference Young-Wolff, Sarovar, Tucker, Conway, Alexeeff, Weisner and Goler2019). This has potentially significant public health implications; however, prospective multigenerational cohort studies remain rare, which means that much remains unknown about when and for whom, preconception interventions to prevent perinatal substance use might be most effective.
Here we bring together data from two long-running Australian intergenerational cohort studies to examine the extent to which maternal binge drinking, tobacco use and cannabis use in adolescence and young adulthood predict use of these substances in pregnancy and 1 year postpartum for women of the peak reproductive age in Australia (ages 29–35 years). Specifically, we sought to address three objectives: (1) to describe the natural history of substance use from adolescence to young adulthood and into pregnancy and motherhood; (2) to examine the extent to which maternal preconception use predicts continued use in pregnancy; and (3) to examine the extent to which maternal preconception use predicts use by the end of the first year of life.
Method
Sample and procedure
Data were pooled from two Australian intergenerational cohort studies within the Australia and New Zealand Intergenerational Cohort Consortium (Olsson et al., Reference Olsson, Spry, Letcher, McAnally, Thomson, Macdonald and Patton2020). The Australian Temperament Project (ATP), established in 1983, began as a representative sample of 2443 infants and their families recruited from maternal and child health centres within the state of Victoria, Australia. Sociodemographic and child development data were collected from parents (Generation 1) and self-reported by children after age 11 (Generation 2) via mail surveys. Data were collected approximately every 2 years across childhood and adolescence, then again at ages 21, 24 and 28 years of age (Vassallo & Sanson, Reference Vassallo and Sanson2013). From 2012 to 2018, ATP participants were followed up via email and telephone to identify pregnancies from 29 to 35 years of age. Women who reported a pregnancy were eligible to participate in the ATP Generation 3 Study (ATPG3). Telephone or web-based surveys, including questions on substance use, were administered to prospective mothers in trimester 3 of pregnancy and 1 year postpartum.
The Victorian Adolescent Health Cohort Study (VAHCS), established in 1992, was based on a representative sampling framework with a high initial response of 1943 mid-secondary school students recruited via a two-stage cluster sampling design from schools also in Victoria, Australia. Data were collected every 6 months in adolescence until a mean age of 17.4 years-old and again at ages 21, 24, 29, 35 and 42 years-old, via telephone-based structured interviews. From 2006 to 2013, VAHCS participants were followed up via email and telephone to identify pregnancies between ages 29 and 35 years. Women who reported a pregnancy were eligible to participate in the Victorian Intergenerational Health Cohort Study (VIHCS) (Spry et al., Reference Spry, Olsson, Hearps, Aarsman, Carlin, Howard and Patton2020). Telephone and web-based surveys were administered to prospective mothers in trimester 3 of pregnancy and 1 year postpartum.
In both ATPG3 and VIHCS, multiple pregnancies during the study period were included in the analytic sample and accounted for using clustered analyses. Miscarried and terminated pregnancies were not included. Ethics approval for VAHCS and VIHCS and perinatal data collection for the ATP were obtained from the Royal Children's Hospital Human Research Ethics Committee. Prior ATP waves were approved by human research ethics committees at the University of Melbourne, the Australian Institute of Family Studies and/or the Royal Children's Hospital, Melbourne.
Measures
Preconception substance use exposures
In the ATP, preconception alcohol and cannabis use were measured at two waves in adolescence (15–16 years; 17–18 years) and three waves in young adulthood (ages 19–20 years; 23–24 years; and 27–28 years). Preconception tobacco use was measured at one additional time point from ages 13–14 years. At each wave, participants were asked how many times they had used substances in the past 30 days. For alcohol and tobacco, participants were also asked how much they consumed per occasion. Binge drinking was indicated by having consumed at least five drinks in quick succession (in adolescence) and by having consumed at least five drinks in 1 day (in young adulthood). In VAHCS, alcohol, tobacco and cannabis use were measured at six waves in adolescence (Waves 1–6; mean age 14.9–17.4 years) and three waves in young adulthood (Waves 7–9; mean ages 21, 24 and 29 years). At each wave, participants reported on frequency of cannabis use (adolescence: within the past 6 months; young adulthood: past 12 months). Alcohol and tobacco use were assessed at each wave using a retrospective diary. For tobacco use, this was each of the past 7 days prior to the survey. For alcohol use, this was each of the past 7 days prior to the survey in adolescence and four of the past 7 days (Friday, Saturday, Sunday and the most recent weekday when drinking occurred) in young adulthood.
Problematic Preconception Substance Use. We classified problematic levels of substance use in the preconception period as binge-drinking alcohol weekly or more frequently (5 or more drinks in a drinking session) (Patton, Reference Patton2004), smoking tobacco cigarettes weekly or more frequently, and using cannabis weekly or more frequently. This was derived in the ATP data as reporting any binge drinking, tobacco use and cannabis use on 4 or more days in the past month and in the VAHCS data as substance use at least 1 day in the past week (binge drinking and tobacco use) and cannabis use weekly or more frequently in the past 6 months (adolescence) or 12 months (young adulthood).
Persistent Problematic Preconception Substance Use. We then created a combined preconception substance use continuity variable for each substance based on the presence of frequent binge drinking, tobacco, or cannabis use at ⩾1 adolescent wave and frequent binge drinking, tobacco, or cannabis use at ⩾1 young adult wave. These dichotomised variables at each preconception period were then grouped into four levels of problematic substance use continuity: ‘none,’ ‘adolescent only,’ ‘young adult only,’ and ‘both adolescent and young adult.’
Maternal perinatal substance use outcomes
ATPG3 and VIHCS mothers reported on their perinatal substance use once in trimester three of pregnancy and once at 1 year post-partum. In trimester three of pregnancy, mothers reported how often they drank any alcohol, smoked cigarettes and smoked or used marijuana or any other cannabis product immediately prior to pregnancy awareness (i.e. ‘Just before finding out you were pregnant’), during pregnancy after pregnancy awareness (‘Shortly after finding out you were pregnant’) and currently (‘Now, [in 3rd trimester]’). Responses to ‘Shortly after finding out you were pregnant’ and ‘Now (in third trimester)’ were averaged to estimate substance use post-awareness of pregnancy. At 1 year postpartum, mothers reported their substance use ‘In the past month [30 days].’ In total, three perinatal periods were examined: (1) Pregnancy pre-awareness, (2) pregnancy post-awareness and (3) 1 year postpartum.
Potential confounders
We controlled for potential preconception confounders associated with both preconception substance use problems and perinatal substance use. Dichotomised variables were parental education [completed post-secondary school qualifications (ATP) or secondary school (VAHCS) prior to participant age 18 years v. did not complete post-secondary qualifications or secondary school), parental separation (not separated/divorced/widowed prior to participant age 18 years v. separated/divorced/widowed) and parent ethnicity (not from an ethnic minority group v. ethnic minority). Adolescent antisocial behaviours were defined as three or more problem behaviours in the past month between the ages of 13 and 18 years. Adolescent mental health problems were assessed as anxious or depressive symptoms before age 18 years. More detail on these adolescent measures are provided in the study protocol (Olsson et al., Reference Olsson, Spry, Letcher, McAnally, Thomson, Macdonald and Patton2020). Post-conception, we also controlled for whether or not mothers were expecting their first child, to adjust for any differences in participants' perinatal substance use associated with first parenthood experience. We also controlled for study cohort (ATPG3 or VIHCS).
Analyses
Data from the two cohorts were pooled by harmonising measures across studies, comparing frequencies in each sample (Table 1), and then combining samples to be able to analyse low prevalence outcomes (Hutchinson et al., Reference Hutchinson, Silins, Mattick, Patton, Fergusson, Hayatbakhsh and Horwood2015; Olsson et al., Reference Olsson, Spry, Letcher, McAnally, Thomson, Macdonald and Patton2020). Models were run separately for each exposure−outcome relationship (e.g. alcohol to alcohol) adjusting for the other substance use histories, potential confounding factors, and an indicator of study cohort. Each substance was modelled as a multivariable outcome defined across three discrete time periods: just prior to pregnancy awareness, pregnancy post-awareness, and 1 year postpartum.
Table 1. Frequency of preconception substance use and potential confounders, by study cohort
ATPG3, Australian Temperament Project Generation 3; VIHCS, Victorian Intergenerational Health Cohort Study.
Frequency estimates were calculated from imputed percentage estimates and total number of participants. Parental low education was measured in ATPG3 as non-completion of post-secondary school qualifications and in VIHCS as non-completion of secondary school.
Generalised estimating equations (GEE), with logit links, were used to estimate associations between preconception substance use and perinatal substance use to account for clustering of pregnancies and for repeated outcome measures in mothers. To examine associations between preconception substance use history and maternal substance use at each perinatal time period, simple slopes were calculated based on an interaction between substance use history and a variable denoting maternal substance use at the three time periods (pregnancy pre-awareness, pregnancy post-awareness, 1 year postpartum).
Missing data ranged from 0% to 37% in ATPG3 (mean = 18%; pregnancy pre-awareness = 36%, pregnancy post-awareness = 37%, 1 year postpartum = 8%) and from 0% to 32% in VIHCS (mean = 14%; pregnancy pre-awareness = 32%, pregnancy post-awareness = 32%, 1 year postpartum = 7%). To handle missing data, we imputed 50 complete datasets, separately by cohort based on a multivariate normal model (Bodner, Reference Bodner2008). All analytic variables were included in the imputation models. This included binary preconception problematic substance use variables defined as weekly or more frequent binge drinking/tobacco/cannabis use, in adolescence and young adulthood. Persistence variables were derived post-imputation. We additionally included low education (< year 12) and divorce/separation by young adulthood as auxiliary variables. Binary variables were imputed as continuous variables and then back transformed with adaptive rounding following imputation. Estimates were obtained by averaging results across the 50 imputed datasets with inferences under multiple imputation obtained using Rubin's rules (Rubin, Reference Rubin2004). In the analyses, varying omitted terms across imputations due to low cell sizes were handled by removing problematic imputations (i.e. those which had reduced samples in regression models due to omitted terms) as recommended by Stata (Stata Corp., 2019). Due to low cell sizes, 18% and 22% of imputed samples were dropped for the tobacco and cannabis models, respectively. Sensitivity analyses were conducted using all imputations by (1) collapsing none and adolescent only users (online Supplementary Material A), and (2) removing adolescent only tobacco users from all analyses (online Supplementary Material B). The interpretation of findings and strength of findings was consistent with sensitivity analyses using all imputations. All analyses were conducted in Stata release 16 (Stata Corp., 2019).
Results
In 1983, 1168 of 2443 recruited ATP participants were female. In 2012, on commencement of ATPG3 (age 29, 2012), 851 women remained in the study. During the pregnancy identification period (ages 29–35), 460 women reported pregnancies and were eligible for the study. Of those eligible, 400 women participated. Baseline demographic characteristics of those screened for ATPG3 broadly matched those of the original cohort (ATP), except that women screened for the study were less likely to have parents born outside of Australia (online Supplementary Material C). There were no demographic differences between women eligible and participating in ATPG3. An additional seven mothers had not participated in at least two of three data collection periods (i.e. adolescence, young adulthood, or the perinatal period) and were excluded from the analysis. The final analytic ATPG3 sample included 689 pregnancies to 393 women.
In 1992 1000 of 1943 recruited VAHCS participants were female. In 2006, at commencement of VIHCS (age 29, 2006), 885 women remained in the study. During the pregnancy identification period (ages 29–35) 465 women reported pregnancies were eligible for the study. Of those eligible, 398 participated. The final analytic VIHCS sample included 609 pregnancies to 398 women. Baseline substance use and demographic characteristics of those screened for, identified as eligible for, and participating in VIHCS, broadly matched those of the original adolescent cohort (VAHCS) (Spry et al., Reference Spry, Olsson, Hearps, Aarsman, Carlin, Howard and Patton2020) (online Supplementary Material D). A sampling flow diagram of both studies is provided in online Supplementary Material E.
Frequency of preconception substance use and potential confounders in ATPG3 and VIHCS are presented in Table 1. Demographics and substance use rates between the two cohorts were similar and were combined into one sample to be able to examine associations with low frequency perinatal outcomes (see Table 1). Sixty-three per cent [95% confidence interval (CI) 59–66] of the combined sample reported weekly or more frequent binge drinking at one or more preconception waves in adolescence or young adulthood. Forty-six per cent (95% CI 43–50) reported weekly or more frequent tobacco use, and 17% (95% CI 14–20) reported a preconception history of weekly or more frequent cannabis use, at one or more preconception waves in adolescence or young adulthood.
The proportion of the sample reporting substance use for any pregnancies at each perinatal assessment period is shown in Table 2. In the combined sample, during pregnancy just before finding out they were pregnant, 76% of mothers reported any alcohol use, 14% reported tobacco use, and 4% reported cannabis use. Subsequent analysis found that the majority of mothers reporting perinatal alcohol use (78%) stopped drinking after pregnancy awareness. Forty-six per cent and 63% of mothers stopped smoking tobacco and using cannabis (respectively) after pregnancy awareness.
Table 2. Perinatal substance use outcomes, by study cohort
ATPG3, Australian Temperament Project Generation 3; VIHCS, Victorian Intergenerational Health Cohort Study.
Frequency estimates were calculated from imputed percentage estimates and total number of participants.
The frequencies of perinatal substance use outcomes by preconception substance use history are presented in online Supplementary Material F. Of pregnancies in which mothers reported perinatal alcohol use, 65% (95% CI 61–68) were preceded by any history of weekly binge drinking [persistent use across adolescence and young adulthood: 24% (95% CI 21–26); use in young adulthood only: 34% (95% CI 31–37); use in adolescence only: 7% (95% CI 5–9)]. Of pregnancies in which mothers reported perinatal tobacco use, 90% (95% CI 85–95) were preceded by any history of weekly tobacco use [persistent: 69% (95% CI 62–76); young adulthood only: 17% (95% CI 12–23); adolescence only: 4% (95% CI 1–6)]. Of pregnancies in which mothers reported perinatal cannabis use 59% (95% CI 45–74) were preceded by any history of weekly cannabis use [persistent: 32% (95% CI 20–45); young adulthood only: 21% (95% CI 9–32); adolescence only: 7% (95% CI −1 to 14)]. Table 3 presents the odds ratios of preconception binge drinking, tobacco, and cannabis use histories and perinatal substance use outcomes adjusted for family demographics and adolescent risk factors. Estimated marginal probabilities of perinatal substance use for each preconception history and substance type are presented in Fig. 1.
Fig. 1. Estimated marginal proportions of perinatal alcohol, tobacco, and cannabis use by preconception substance use history. Preconception history groups: (1) None: No binge drinking/tobacco/cannabis use in adolescence or young adulthood; (2) Adolescence only: Weekly or more frequent binge drinking/tobacco/cannabis use limited to adolescence (ages 13–18 years); (3) Young adulthood only: Weekly or more frequent binge drinking/tobacco/cannabis use limited to young adulthood (ages 19–29 years); (4) Persistent: Weekly or more frequent binge drinking/tobacco/cannabis use across adolescence and young adulthood.
Table 3. Odds ratios of mothers' preconception substance use and substance use pregnancy pre-awareness, pregnancy post-awareness and 1 year postpartum
Note. Adjusted for study cohort, childhood parental separation, childhood parental low education, parent ethnicity, adolescent mental health problems, adolescent delinquency, first child. Tobacco model required dropping nine imputed samples (18%) and the cannabis model required dropping 11 imputed samples (22%).
Adolescent limited substance use
The odds of any alcohol use during pregnancy, post-awareness, were doubled for mothers with histories of frequent binge drinking limited to adolescence, compared to those without histories of frequent binge drinking in adolescence or young adulthood. The odds of tobacco use at 1 year postpartum were also higher for mothers with histories of frequent tobacco use limited to adolescence, compared to those without histories of frequent preconception tobacco use. Despite combining data from two cohorts, it was difficult to precisely estimate the association between adolescent limited preconception cannabis use and perinatal cannabis use due to low sample size resulting in wide confidence intervals.
Young adult limited substance use
Preconception substance use limited to young adulthood was associated with heightened use of every substance type at each perinatal assessment period. Specifically, the odds of reporting any alcohol use prior to pregnancy awareness, during pregnancy post-awareness, and at 1 year postpartum were doubled for mothers with histories of frequent binge drinking in young adulthood compared to those without histories of frequent preconception binge drinking. The odds of reporting any tobacco use across the three perinatal assessment periods were much higher for mothers with histories of frequent tobacco use in young adulthood compared to those without histories of frequent preconception tobacco use. The odds of reporting any cannabis use were again higher across the three perinatal assessment periods for mothers with histories of frequent cannabis use in young adulthood compared to those without histories of frequent preconception cannabis use.
Persistent adolescent and young adult substance use
There was a suggestion that persistent frequent substance use across both adolescence and young adulthood was associated with the highest odds of perinatal substance use for nearly all substance types and perinatal assessment periods, although confidence intervals were wide and overlapping for these estimates. Specifically, the odds of reporting any alcohol use just prior to pregnancy awareness and 1 year postpartum were doubled in those with persisting preconception binge drinking compared to those with no such histories. This increased to tripled post-pregnancy-awareness. The odds of perinatal tobacco use across the three perinatal assessment periods were higher again in those with persistent histories of tobacco use compared to those without. The odds of perinatal cannabis use were higher yet again across the three perinatal assessment periods for those with persistent histories of cannabis use compared to those without.
Discussion
Using rare intergenerational data from two Australian cohort studies, these analyses suggest that maternal alcohol, tobacco, and cannabis use during pregnancy, both prior to and following pregnancy awareness, as well as at 1 year postpartum, are likely to have their origins well prior to becoming a parent. This clarifies findings from prior reviews and retrospective research (Skagerstróm et al., Reference Skagerstróm, Chang and Nilsen2011; Young-Wolff et al., Reference Young-Wolff, Sarovar, Tucker, Conway, Alexeeff, Weisner and Goler2019), by showing that preconception risks extend back to adolescence. There was a suggestion that early frequent substance use beginning in adolescence and persisting into young adulthood was the strongest predictor of perinatal alcohol, tobacco, and cannabis use, although confidence intervals were wide for tobacco and cannabis use outcomes. Frequent substance use limited to young adulthood was also consistently associated with higher odds of perinatal substance use. Women with these two preconception substance use histories also accounted for the largest proportion of subsequent substance use in the perinatal period. Of pregnancies in which women reported any alcohol, tobacco, or cannabis use, 58%, 86%, and 53% of pregnancies, respectively, were preceded by a preconception history of weekly or more frequent binge drinking or tobacco/cannabis use across both adolescence and young adulthood, or in young adulthood only. Additionally, whilst mothers generally reduced their substance use following pregnancy recognition, mothers with a history of persistent and young-adult-limited frequent substance use had consistently higher odds of substance use during pregnancy post-awareness compared to mothers without these histories, indicating a high-risk group for early intervention prior to pregnancy. Frequent preconception substance use limited to adolescence had small associations with future substance use in the case of adolescent binge drinking, which was associated with higher odds of drinking during pregnancy post-awareness, and adolescent tobacco use, which was associated with higher odds of tobacco use at 1 year postpartum. Having controlled for adolescent behaviour and sociodemographic background, this may reflect to some extent that substance use that starts and stops in adolescence represents experimentation (Wells, McGee, Baxter, Agnew, & Kokaua, Reference Wells, McGee, Baxter, Agnew and Kokaua2009). That said, early experimentation often increases alcohol use and other risks from adolescence to young adulthood (Yuen et al., Reference Yuen, Chan, Bruno, Clare, Mattick, Aiken and Peacock2020). Overall, our findings suggest that interventions to prevent longer-term patterns of frequent substance use may be well positioned to begin in adolescence, before potential sustained problem use starts.
In this study we also found that patterns of perinatal substance use varied considerably by substance type. In close to eight in 10 pregnancies (76%), mothers reported some alcohol use before pregnancy awareness. Drinking during pregnancy substantially reduced after pregnancy recognition (18%), but then rebounded to pre-pregnancy levels (77%) 1 year postpartum. This reduction in drinking after becoming aware of the pregnancy and dramatic resurgence postpartum is consistent with other prospective studies (Borschmann et al., Reference Borschmann, Becker, Spry, Youssef, Olsson, Hutchinson and Patton2019; Hutchinson et al., Reference Hutchinson, Moore, Breen, Burns and Mattick2013). Although women with preconception histories of persistent or young-adult-limited frequent binge drinking had higher odds of perinatal alcohol use, 35% of pregnancies in which any drinking occurred were not preceded by earlier frequent use, suggesting that alcohol use in early pregnancy and postpartum are common regardless of a woman's preconception drinking history.
Although perinatal tobacco and cannabis use were lower than alcohol use, it was nonetheless concerning that in over one in ten pregnancies (14%) mothers reported smoking tobacco cigarettes before pregnancy awareness. The continuity between preconception use and perinatal use was also strong, with 90% of pregnancies in which mothers reported tobacco use preceded by a preconception history of weekly or more frequent smoking. Smoking during pregnancy somewhat reduced after pregnancy recognition (9%), but increased again at 1 year postpartum (11%). This pattern of quitting smoking and relapsing across the perinatal period is similar to that reported by other perinatal substance use studies, and shows that patterns of perinatal smoking have not changed substantially in the last decade despite prevention efforts (Crume, Reference Crume2019; Kurti et al., Reference Kurti, Redner, Lopez, Keith, Villanti, Stanton and Higgins2017). This is especially so for mothers who spontaneously quit smoking during pregnancy (Crume, Reference Crume2019), and points to the addictive nature of tobacco smoking. This finding is of significant clinical and public health concern because of the well documented adverse effects of passive smoking on child development (Herrmann et al., Reference Herrmann, King and Weitzman2008).
A somewhat different story was evident for cannabis use. Prevalence was lower than for tobacco, with mothers reporting cannabis use in approximately one in 20 pregnancies (4%) before pregnancy awareness, reducing to one in 50 (2%) after pregnancy awareness and remaining stable (2%) at 1 year postpartum. Lack of postpartum relapse is contrary to trends reported in prior perinatal studies of cannabis use (el Marroun et al., Reference el Marroun, Tiemeier, Jaddoe, Hofman, Mackenbach, Steegers and Huizink2008; Ko et al., Reference Ko, Tong, Bombard, Hayes, Davy and Perham-Hester2018). However, differences in study design may account for this. For example, previous studies have focused on younger mothers (<24 years) who, based on population trends, are more likely to use cannabis (Ko et al., Reference Ko, Tong, Bombard, Hayes, Davy and Perham-Hester2018), whereas our data are from mothers in the normative range for births in Australia (29–35 years). Postpartum rebound in tobacco, but not cannabis use, suggests differences in the addictive potential of the two substances and a need for more intensive intervention around smoking to reduce risk for postpartum use.
Strengths and limitations
A strength of this study was the integration of rare multigenerational data from two longitudinal population-based cohorts. In doing so we were able to maximise sample size and power for outcomes with a relatively low prevalence. Our multi-wave, prospective design also minimised recall bias of adolescent and young adulthood substance use and included a broad range of relevant potential confounding variables. Limitations of our study included the use of self-reported substance use, which could have underestimated prevalence rates and attenuated associations. In the third trimester pregnancy, mothers retrospectively reported their substance use just before finding out they were pregnant and shortly thereafter, which also may have introduced inaccurate reporting. Previous research has shown up to 97% concordance between women's self-reported substance use during pregnancy and biological samples (Hutchinson et al., Reference Hutchinson, Wilson, Allsop, Elliott, Najman, Burns and Ward2018), although most research and clinical practice supports the use of self-report screening in conjunction with biological measures for optimal substance use identification during pregnancy (Burns, Gray, & Smith, Reference Burns, Gray and Smith2010; Polak, Kelpin, & Terplan, Reference Polak, Kelpin and Terplan2019). Perinatal data were collected for pregnancies identified in women aged between 29 and 35 years, the peak period for births in high-income countries such as Australia. Because pregnancies at younger ages were excluded, generalisability of the findings may be reduced and associations attenuated between adolescent-limited substance use and perinatal use. Excluding miscarried or terminated pregnancies also may have introduced bias.
Data from the ATP/ATPG3 and VAHCS/VICHS study cohorts were collected over close to four and three decades, respectively, resulting in sample attrition and non-response. Notably, a higher proportion of participants were missing at the third trimester assessment in both study cohorts, due to the logistical challenges of identifying all pregnancies to cohort participants before birth. We addressed missing data with multiple imputation. Although we found minimal differences in demographics between participants who remained in the studies compared to those initially recruited (women in ATPG3 were less likely to have parents born outside of Australia), it is possible that those retained in the studies differed on unmeasured characteristics, as well as in their levels of perinatal substance use. Furthermore, measures of preconception and perinatal substance use were not identical across the ATP/ATPG3 and VAHCS/VIHCS studies. Variable derivations (e.g. weekly binge drinking) were created based on the closest possible alignment between studies, which in some cases involved long assessment periods. These derivations could have introduced misclassification in some cases, if patterns of substance use changed dramatically throughout the assessment period (e.g. up to 1 year). There were some differences between cohorts in demographic and substance use rates, although findings were similar across the two cohorts. Further, even within the pooled sample, the low prevalence rates of perinatal substance use outcomes resulted in low precision of odds ratio estimates. Wide confidence intervals limited interpretations of effect sizes, however the results consistently indicated elevated risk associated with preconception substance use. Finally, our study did not include measures of substance use for participants' partners (e.g. fathers of the infants), which has been identified as an important predictor of maternal substance use in pregnancy as well as a risk factor for infant offspring behaviour and future substance use (el Marroun et al., Reference el Marroun, Tiemeier, Jaddoe, Hofman, Mackenbach, Steegers and Huizink2008; Kerr et al., Reference Kerr, Capaldi, Pears and Owen2012). Future research should investigate the extent to which fathers' substance use histories persist into parenthood and how these are related to offspring outcomes.
Implications and conclusions
From a population health perspective, this study suggests that persistent frequent preconception substance use from adolescence to young adulthood plays a significant role in influencing substance use in pregnancy and across the first year of offspring life. This highlights the significance of the preconception years as a key period for investment in early intervention. Specifically, identifying and addressing common patterns that predict potential substance use harms through universal monitoring systems may have profound benefits. Universal approaches may be particularly beneficial for alcohol where use is highly prevalent in the population; even small reductions in alcohol use across many cases could result in significant population health improvements (Rose, Reference Rose1985; Stockwell et al., Reference Stockwell, Toumbourou, Letcher, Smart, Sanson and Bond2004). Similar principles apply to smoking prevention, where use is less prevalent but highly addictive. Furthermore, targeted interventions may be warranted for sustained patterns of frequent alcohol, tobacco, or cannabis use throughout adolescence and young adulthood, both to reduce harm to the health of the individual as well as to the health of potential future offspring. Importantly, interventions should prioritise the wellbeing of individuals irrespective of their health behaviours to avoid stigmatising individuals and encourage engagement with family planning and prenatal care (Roberts, Thompson, & Taylor, Reference Roberts, Thompson and Taylor2021). Supports should also be extended to partners of pregnant mothers who influence substance use habits and, in the case of tobacco and cannabis smoke, increase the risk of passive exposure within the household (el Marroun et al., Reference el Marroun, Tiemeier, Jaddoe, Hofman, Mackenbach, Steegers and Huizink2008; Schneider, Huy, Schütz, & Diehl, Reference Schneider, Huy, Schütz and Diehl2010).
Effective public health responses are likely to be multifaceted, focusing both on the reduction of substance use and greater planning around pregnancies. There is much opportunity to invest in school-based programmes, pre-emptively addressing substance misuse at the point of initiation at a population-level, in addition to addressing current substance misuse through targeted school-based programmes (Conrod et al., Reference Conrod, O'Leary-Barrett, Newton, Topper, Castellanos-Ryan, MacKie and Girard2013). Public health campaigns focusing on substance use in the family planning context are also examples of (late) preconception intervention. These could complement existing prenatal screening and intervention programmes which often miss a critical period before pregnancy recognition when in utero exposures may have already occurred (Louw, Reference Louw2018). The pause in substance use during pregnancy may offer a further point of post conception intervention to maintain reduced substance use patterns during the childrearing years (Chapman & Wu, Reference Chapman and Wu2013; Crume, Reference Crume2019; Skagerstróm et al., Reference Skagerstróm, Chang and Nilsen2011). Research examining predictors of substance use continuity and discontinuity (e.g. mental health, delinquency, social support) would likewise help inform prevention and intervention efforts. Inclusion of indicators of young adulthood and early-onset persistent substance use into population monitoring activities would also provide valuable information on when and where increased supports and services may benefit populations and ultimately reduce the prevalence of substance use in adulthood including parenthood. Studies comparing the natural history of preconception substance use in different policy settings (e.g. where cannabis is decriminalised) may also provide further insights into effective population-level strategies to address problematic drinking, tobacco, and cannabis use from an early age.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/S0033291721003925
Acknowledgements
The ATP, Australian Temperament Project Generation 3 (ATPG3), Victorian Adolescent Health Cohort Study (VAHCS), and Victorian Intergenerational Health Cohort Study (VIHCS) are located at The Royal Children's Hospital Melbourne. ATP and ATPG3 are a collaboration between Deakin University, The University of Melbourne, The Murdoch Children's Research Institute, The Australian Institute of Family Studies, The University of New South Wales, The University of Otago (NZ), and the Royal Children's Hospital. We acknowledge all collaborators who have contributed to the ATP, especially Professors Ann Sanson, Frank Oberklaid, Dr Diana Smart, and the late Professor Margot Prior. We would also like to sincerely thank the participating families in all studies for their time and invaluable contributions, as well as all collaborators and the study teams involved in the data collection and management. Further information on these studies is available at https://www.melbournechildrens.com/atp/ and https://lifecourse.melbournechildrens.com/cohorts/2000-stories/.
Financial support
Funding for this work was supported by grants for CAO from the Australian Research Council [DP130101459; DP160103160; DP180102447] and the National Health and Medical Research Council of Australia (NHMRC) [APP1082406]. GCP was supported by the NHMRC [APP1117873]. KCT was supported by funding from the Canadian Institutes of Health Research [MFE158141] and Michael Smith Foundation for Health Research [RT-2020-0550]. Additionally, this work was supported by the NHMRC (APP1008273, APP1063091, APP437015), Australian Rotary Health, Colonial Foundation, Perpetual Trustees, Financial Markets Foundation for Children (Australia), Royal Children’s Hospital Foundation and the Murdoch Children’s Research Institute. Research at the Murdoch Children’s Research Institute is supported by the Victorian Government’s Operational Infrastructure Program.
Conflict of interest
The authors have no conflicts of interest to declare.
Ethical standards
The authors assert that all procedures contributing to this work comply with the ethical standards of the relevant national and institutional committees on human experimentation and with the Helsinki Declaration of 1975, as revised in 2008.
