Article Text

Download PDFPDF

Smoking cessation in pregnancy and the risk of child behavioural problems: a longitudinal prospective cohort study
  1. Monique Robinson1,3,
  2. Neil J McLean2,
  3. Wendy H Oddy1,
  4. Eugen Mattes1,
  5. Max Bulsara1,4,
  6. Jianghong Li3,5,6,
  7. Stephen R Zubrick3,
  8. Fiona J Stanley1,
  9. John P Newnham7
  1. 1Telethon Institute for Child Health Research, Centre for Child Health Research, The University of Western Australia, Perth, Australia
  2. 2School of Psychology, The University of Western Australia, Perth, Australia
  3. 3Telethon Institute for Child Health Research, Curtin Health Innovation Research Institute, Centre for Developmental Health, Curtin University of Technology, Perth, Australia
  4. 4Institute of Health and Rehabilitation Research, University of Notre Dame, Fremantle, Australia
  5. 5School of Public Health, Curtin University of Technology, Perth, Australia
  6. 6Centre for International Health, Curtin University of Technology, Perth, Australia
  7. 7School of Women's and Infants' Health, The University of Western Australia at King Edward Memorial Hospital, Perth, Australia
  1. Correspondence to Monique Robinson, Telethon Institute for Child Health Research, PO Box 855, West Perth, WA 6872, Australia; moniquer{at}


Background The aim of this study was to examine the influence of smoking in pregnancy on child and adolescent behavioural development, in comparison with mothers who ceased smoking in the first 18 weeks of pregnancy and with those who never smoked, in a large prospective pregnancy cohort.

Methods The Western Australian Pregnancy Cohort (Raine) Study provided comprehensive data from 2900 pregnancies. Smoking was assessed at 18 weeks gestation, and children were followed up at ages 1, 2, 3, 5, 8, 10 and 14 years. The Child Behaviour Checklist (CBCL) was used to measure problem child behaviour with continuous z-scores and clinical cut points at ages 2, 5, 8, 10 and 14 years. Potential confounders included maternal and family sociodemographic characteristics and alcohol exposure.

Results After adjusting for confounders, children of light smokers who quit smoking by 18 weeks gestation had significantly lower CBCL total z-scores, indicative of better behaviour, than children of women who never smoked, children of heavy smokers who quit and continuing smokers. Maternal smoking during pregnancy resulted in higher CBCL total, internalising and externalising scores and a higher risk of clinically meaningful behaviour problems in children from ages 2 to 14.

Conclusion The maternal decision not to quit smoking, or the inability to quit smoking, during pregnancy appears to be a particularly strong marker for poor behavioural outcomes in children. There is a need for a greater understanding of the psychosocial characteristics associated with the decision and ability to quit smoking in pregnancy.

  • Smoking
  • pregnancy
  • behaviour
  • mental health
  • prenatal risk
  • Raine study
  • child development
  • mental health DI
  • smoking and pregnancy
View Full Text

Statistics from


The physical health risks associated with smoking in pregnancy have been known for decades and prenatal smoking is considered one of the most important preventable causes of poor child health.1 2 Smoking in pregnancy is associated with a number of adverse clinical consequences for offspring, including low birth weight, respiratory problems and early childhood overweight.3–5 Even a few cigarettes per day may have consequences for the developing child.6 However, smoking cessation during pregnancy, particularly during the first two trimesters, can alleviate some of this risk.7 8

Fetal exposure to maternal prenatal smoking has also been linked to behavioural problems later in life.9 Numerous studies have found that maternal smoking in pregnancy increases the child's risk for externalising problems, such as conduct disorder and antisocial behaviour,10–12 and internalising problems, such as depression and anxiety.11 13 Prenatal smoking is also linked to a higher risk of specific psychopathology such as Attention Deficit Hyperactivity Disorder (ADHD) diagnosis.14–16

However, there has been controversy surrounding these findings. Previous research has not always controlled for factors that could moderate any associations, such as changes in smoking patterns and maternal characteristics (such as self-efficacy) linked to the decision to smoke or quit smoking during pregnancy.17 18 In addition, the difficulty of separating the specific effects of smoking during pregnancy from other prenatal and postnatal risks strongly correlated with both smoking behaviour and child behaviour, such as socioeconomic disadvantage and maternal stress, has potentially confounded previous findings.2 19

Although previous research has shown that smoking cessation can reverse some of the adverse physical effects of smoking during pregnancy, such as intrauterine growth restriction,7 no study of which the authors are aware has examined the influence of smoking cessation in pregnancy on long-term behavioural outcomes. The present study aimed to compare the behavioural outcomes of children born to mothers who continue to smoke during pregnancy with those born to mothers who quit smoking within the first 18 weeks of pregnancy and with those born to mothers who never smoked. Using a prospective pregnancy cohort followed to 14 years, the aim was to control for confounding factors that may have limited the generalisability of previous research in this area and examine factors that may influence the decision to quit smoking in pregnancy.


Study design

The Western Australian Pregnancy Cohort (Raine) Study is a prospective pregnancy cohort study of 2868 live births followed to 14 years of age.


Women were recruited between May 1989 and November 1991 (N=2900) through the public antenatal clinic at King Edward Memorial Hospital (KEMH) and nearby private clinics in Perth, Western Australia. Comprehensive data regarding social and demographic characteristics were collected at 18 and at 34 weeks gestation. Data were collected at birth, including physiological and clinical information, and the study children and their families provided sociodemographic and behavioural data at 1, 2, 3, 5, 8, 10 and 14 years of age.


Complete details of enrolment methods have been published elsewhere.20 Briefly, to be eligible, women were required to have a pregnancy between 16 and 20 weeks gestation (average 18 weeks), sufficient English language skills, an expectation to deliver at KEMH, and an intention to reside in Western Australia to allow for future follow-up of their child. Informed consent to participate in the study was obtained from the mother of each child at enrolment and at each subsequent follow-up. The Human Ethics Committees at KEMH and/or Princess Margaret Hospital for Children approved the protocols for the study.

Loss to follow-up

All cohort studies have problems from loss to follow-up, although the present study achieved a reasonably low rate of attrition. Young mothers, single mothers and those who experienced high levels of stress were less likely to remain in the study in the early years.21 Given that the initial cohort over-represented socially disadvantaged women, this loss to follow-up could have reduced the bias and potentially increased the study generalisability.

Outcome variables: child behaviour

The Child Behaviour Checklist (CBCL), an empirically validated measure of child behaviour by parent report, was used to measure child and adolescent behaviour. The CBCL for Ages 2–3 (CBCL/2–3), with 99 items, was used at the 2 year follow-up and the CBCL for Ages 4–18 (CBCL/4–18), with 118 items, was administered at the 5, 8, 10 and 14 year follow-ups.22 23 The CBCL demonstrated good test-retest reliability, good sensitivity (83% overall) and reasonable specificity (67% overall) to a clinical psychiatric diagnosis in a Western Australian clinical calibration.24

Both CBCL instruments produce a raw score that was transformed into three summary z-scores for A. total behaviour, B. internalising behaviour (withdrawal, somatic complaints and anxious/depressed behaviour) and C. externalising behaviour (delinquent and aggressive behaviour), for each of the follow-up years. The z-scores were used as continuous scores in the analysis, with higher scores associated with more disturbed emotions and behaviours. The raw scores were also converted into T-scores for total, internalising and externalising behaviour.23 The recommended clinical cut-off scores (T≥60) were applied to the CBCL T-scores, to obtain three binary variables indicative of clinically significant total, internalising and externalising morbidities.23

Predictor variables: prenatal smoking

A four-category variable representing smoking continuation and smoking cessation by 18 weeks gestation was created based on classifications used in existing literature.17 25 The categories were: A. never smoked, B. prior light smokers (1–10 cigarettes daily) who quit in the first 18 weeks of pregnancy, C. prior heavy smokers (11+ cigarettes daily) who quit in the first 18 weeks of pregnancy and D. prior smokers who were still smoking at least one cigarette per day at 18 weeks gestation. The classification for continuing smokers was deliberately broad as smoking patterns fluctuate substantially in pregnancy and social desirability factors can limit the validity of self-report quantitative data.18 The data showed that the number of cigarettes smoked per day at 34 weeks gestation was strongly correlated with the number of cigarettes smoked at 18 weeks gestation (r=0.866, p<0.001) indicating that few participants quit smoking post-18 weeks gestation; therefore, smoking cessation after 18 weeks gestation was not assessed in this study.

Control variables

The control variables included numerous prenatal and perinatal factors known to have some effect on prenatal smoking and behavioural outcomes in children. These variables included maternal sociodemographic information measured at 18 weeks gestation such as maternal age, maternal education, family income and the presence of the biological father in the family home. The maternal experience of stressful events in pregnancy was measured at 18 and 34 weeks gestation and added together to produce a continuous variable representing the total number of events experienced.9 Alcohol intake at 18 weeks gestation was also included in the analyses.16 Variables from the perinatal and postnatal periods were also included: birth weight (grams, centred at the mean), gestational age (weeks, centred at the mean) and breast feeding duration (months), and the General Functioning Scale (GFS) from the McMaster Family Assessment Device (FAD) was applied to measure family functioning at each follow-up except at age 2 (when data were not available).26

Other variables

In order to assess differences in maternal self-efficacy that could enhance the interpretation of results in regards to smoking cessation, the mean total scores for the modified Cowen Perceived Self-Efficacy Scale from the 3 year follow-up were examined.27 This scale was adapted from the original Cowen scale28 and consists of 22 items relating to how sure the respondent is that they would be able to manage when faced with a number of everyday situations. Each question is answered using a five-point Likert scale ranging from 1 (‘not at all sure’) to 5 (‘very sure’). The Cronbach's α coefficient was calculated at 0.91, showing excellent internal reliability.

Statistical analyses

The aim of the present study was to compare the behavioural outcomes for the children of mothers who continued to smoke or quit smoking with those who never smoked. Therefore, a linear regression model with a random intercept (random effects model) was used to examine the ability of the predictor variable to effect changes on the continuous CBCL z-score and generalised estimating equations (GEE; a random effect logistic regression model) to assess whether such changes in score reflected clinically meaningful differences in child behavioural problems. First, cross-tabulations were used to assess the relationships between the outcome and predictor variables. The predictor variable was then included in a univariate random effects model, to account for repeated observations of the same individuals over time, and analysed using continuous CBCL z-scores for total, internalising and externalising behaviour at each year as outcomes. This analysis was followed by the inclusion of all the control variables (maternal age and education, maternal experience of stress events, total family income, alcohol intake, presence of the biological father in the family home, birth weight, gestational age, breast feeding duration and family functioning) into the model for multivariable analysis. The predictor variable was then entered into a GEE model, without the control variables, to estimate the risk of child behavioural problems over time. An unstructured working correlation matrix specification was used (which provided the best goodness of fit). This analysis was then repeated following the inclusion of all the control variables in the model. Two-way interaction effects were tested between the predictor variable and control variables but there were no significant results, and therefore interactions were not included in either model. SPSS 15.0 was used for the analyses.


Frequency characteristics for the predictor, outcome and control variables are presented in table 1. Forty-one per cent of mothers reported never having smoked, whereas another 32% quit smoking upon becoming pregnant. The remaining 27% of mothers smoked at least one cigarette per day prior to their pregnancy and continued smoking at least that amount during their pregnancy. The percentage of study children who scored above the clinical cut point for CBCL morbidity (T-score≥60) by maternal smoking status is presented in table 2. Total behavioural morbidity ranged from a low of 11.5% at the 2-year follow-up to a high of 21.3% at the 5-year follow-up, with 14% above the clinical cut point for total behaviour at the 14-year follow-up and 12.6% and 15.5% above the clinical cut points for internalising and externalising morbidity respectively at 14 years of age.

Table 1

Frequency characteristics for predictor, outcome and control variables

Table 2

Behavioural morbidity (CBCL T-score≥60) at each follow-up by total sample and smoking status

The reference category of mothers who ‘never smoked’ was compared with prior light smokers who quit by 18 weeks gestation, with prior heavy smokers who quit by 18 weeks gestation and with mothers who smoked before and during pregnancy in a linear regression model with a random intercept (table 3). Those smokers who continued to smoke during pregnancy had children with significantly higher CBCL z-scores for total, internalising and externalising behaviour in both the univariate and adjusted analyses. Unexpectedly, in the adjusted analysis, light smokers who quit smoking upon becoming pregnant had significantly lower scores for total behaviour across the 14 years of follow-up.

Table 3

Random effects model showing relationship between smoking and smoking cessation and CBCL z-scores at each age

Using the same reference category of mothers who never smoked, a generalised estimating equation model was applied to determine the clinical significance of the changes in CBCL score found in the random effects model analyses (table 4). Smoking before and during pregnancy showed a significant relationship with total, internalising and externalising behavioural morbidity over 14 years in both the univariate and multivariable analyses, and these ORs remained high following adjustment for confounding variables. Being a light or heavy smoker prior to pregnancy and then quitting before 18 weeks gestation was not significantly associated with a reduction in clinically meaningful T-scores over 14 years.

Table 4

Generalised estimating equation (GEE) model‡ showing relationship between smoking and smoking cessation and CBCL morbidity (T≥60) at each age

Finally, the maternal self-efficacy scores for the four groups were examined using one-way ANOVA to assess differences in mean scores across groups (figure 1). The prior light smokers who quit smoking prior to 18 weeks gestation had the highest self-efficacy of all groups, and significantly higher self-efficacy than the mothers who continued to smoke during pregnancy in post hoc analysis using Tukey's HSD (p=0.013).

Figure 1

Differences in mean maternal self-efficacy score at the 3-year follow-up by pregnancy smoking status (significant difference at p=0.013 between groups marked with asterisk).


These findings have important public health implications. Although continuing to smoke was associated with an increased risk of behavioural problems, the long-term negative consequences for child and adolescent behaviour could be avoided for smoking mothers who quit early in pregnancy. The results suggest that the maternal ability to quit smoking early in pregnancy is a marker for the successful behavioural development of their children.

Mothers who quit smoking during pregnancy have already been found to have babies with easier temperaments compared with those born to mothers who never smoked.17 The present findings observing a small yet significant positive change in CBCL z-score support those results and potentially extend them past infancy to 14 years of age, although it is noted that the observed movement in score was a subtle change. Light smokers who quit in the early stages of pregnancy had higher levels of self-efficacy at the 3-year follow-up than those who continued to smoke. High self-efficacy is a strong predictor of positive parenting practices.29 Self-efficacy and addiction theories also suggest that those who can moderate their substance use may be better adjusted and mentally healthier than both abstainers and heavy users.30 31 It could be that not only did light smokers keep their use of cigarettes under control prior to pregnancy, but they also had the self-efficacy to quit soon after becoming pregnant. It is plausible that success in quitting smoking in pregnancy could have subsequently increased maternal self-efficacy as well. As with any change in behaviour, factors influencing the decision and the ability to quit smoking are complex. Quitting may be influenced by maternal knowledge about the effects of smoking on the health of her unborn child. Previous research has shown that smoking cessation in pregnancy is motivated more by a maternal desire to protect the baby, rather than a long-term intention to quit the habit.17 Conversely, continuing to smoke during pregnancy is linked with a decreased likelihood of engaging in other positive antenatal behaviours, such as consuming a good quality diet.32

Based on previous findings,9 and like other cohort studies, a significant increase in the risk of a child developing behavioural problems if the mother smokes during pregnancy was observed.10–12 Stronger findings were also observed for externalising behaviour as compared to internalising behaviour for children whose mothers smoked throughout pregnancy.11 12 Previous research has found dose-response effects; for example, a greater than fourfold increase in conduct disorder in boys from mothers who smoked ten or more cigarettes per day during pregnancy.33 All smokers were included irrespective of their level of smoking in the present analysis (given smoking patterns fluctuate substantially during pregnancy), which may have led to the smaller, yet still significant, overall effect size.18

In the present study, mothers who continued to smoke during pregnancy experienced a high level of psychosocial adversity. They were more likely to be teenage mothers, to be less educated, to not be living with the baby's father, to be consuming alcohol daily, to experience more stressful events in pregnancy, and to have a low family income (table 1). There is also a growing body of work suggesting that mothers who smoke during pregnancy experience increased levels of domestic violence34 and mental health problems.35 Given that the loss to follow-up over-represented socially disadvantaged women, the associations between maternal smoking during pregnancy and adverse behavioural outcomes in children may be even stronger had it been possible to include those socially disadvantaged women in the analysis.

The study design enabled investigation of the influence of prenatal smoking and smoking cessation at multiple periods of behavioural development while controlling for many sociodemographic characteristics. The failure to control for these characteristics in previous research has limited the conclusions that can be drawn.36 37 The majority of studies on smoking exposure and child behavioural and health outcomes have not considered the developmental trajectory, and many have ignored the influence of maternal psychosocial characteristics in favour of a biological pathway.17 The use of the CBCL as a measure of behaviour was also a strength of this study, as it is a well-validated instrument and shows good internal consistency in the diagnosis of child psychopathology.38 39 Furthermore, the CBCL was administered at each follow-up giving the opportunity for in-depth longitudinal analyses.

Previous research has found that self-reported smoking data, although potentially underestimated, are reliable when compared with cotinine levels.2 40 41 However, self-reported information on smoking cessation is often less reliable, and this potentially has an effect on the present findings with regards to smoking cessation and amount smoked prior to pregnancy.41 In addition, it was not possible to determine with precision the exact nature of the timing and pattern of smoking cessation for those mothers who previously smoked, but who quit smoking prior to 18 weeks gestation.

It is important for any study on prenatal smoking and child development to note that smoking could be a proximal risk factor reflecting the larger influence of socioeconomic status (SES) on mental health outcomes, as smoking behaviours are closely linked with sociodemographic factors, including low income and maternal characteristics.1 17 Although a number of variables reflecting SES were included in the present analysis (eg, family income, maternal education, maternal stress and maternal age), it was not possible to control for every conceivable factor in what is a complex sociodemographic framework. However, of the sociodemographic influences, maternal education is particularly linked with the decision to smoke during pregnancy and was therefore an important covariate in the present analysis.42


The effects of smoking during pregnancy on child behavioural development are likely to occur via both biological and psychosocial pathways.43 There are more than 4000 chemically active compounds in cigarette smoke, many of which are candidates to alter fetal brain development.3 A dose-dependent relationship between increased smoking and fetal growth reduction has consistently been observed, but knowledge as to how this effect could be translated into behavioural development is incomplete. Alternatively, the nicotine in cigarettes is known to stimulate the sympathomimetic system,3 and this may indicate a potential role for the stress axis in this response. Like previous studies, which have observed a relationship between stress experience and behavioural development, it is possible that the smoking-induced stimulation of the sympathomimetic system prepares the fetus for a world that the mother perceives as dangerous and hence behavioural problems may reflect the fetal adaptation to a stressful world.44 It is also possible that the physical health risks for the fetus associated with smoking during pregnancy may have some interaction with behavioural development. For example, extremely low birth weight has been associated previously with higher CBCL total scores in children up to 10 years of age.45 In addition to the biological pathways, this study indicates a role for maternal psychosocial processes in child and adolescent behavioural development.17 The differences in maternal self-efficacy at age 3 suggest that self-efficacy may be one such psychosocial construct that influences the maternal decision to quit smoking in the first 18 weeks of pregnancy and may also have an influence on later parenting practices.29 Additionally, mothers who smoke during pregnancy are more likely to neglect and/or abuse their children.46

Implications of these findings

The present results suggest that if mothers can quit smoking early in their pregnancy they can avoid exposing their child to the higher risk of behavioural problems observed for the children of continuing smokers. Although smoking interventions in the clinical setting have generally been successful, smoking interventions in pregnancy have met with limited success.2 47 The majority of women who smoke will quit during the first trimester of pregnancy, as has been shown, with up to 95% of women who were smoking at 18 weeks gestation still smoking at 34 weeks.48 Given that pregnant women are usually concerned about fetal health and well-being, it should be an ideal time to quit smoking. However, research suggests that success depends not just on a cost-benefit analysis but also on the individual's perceived likelihood (ie, self-efficacy) of successfully making the change,49 and clinical trials have shown a weak effect of smoking cessation implemented in routine prenatal care.50 For the clinician, if a mother is still smoking in mid-pregnancy it is a marker for poorer behavioural outcomes for her children and it could warrant a more detailed examination of her mental health, substance use and broader social circumstances. However, it remains unclear whether a referral to specialist agencies increases the chances of quitting.50 Mothers who continue to smoke in pregnancy appear to warrant special attention rather than simplistic office-based approaches for changing behaviour.47 We run the risk of blaming mothers for failing to stop, rather than finding more effective solutions outside the clinical setting in their communities or even at a macro-social level with welfare policies.51 Smoking cessation interventions for pregnancy should be designed to address the maternal characteristics that strengthen a mother's capacity to quit.


The positive finding to come out of this study is that children of mothers who quit smoking within the first 18 weeks of pregnancy appear to avoid the negative consequences associated with continued smoking throughout pregnancy, which presents as a particularly strong marker for poor behavioural outcomes in childhood and adolescence. It may be that maternal characteristics influencing the decision and ability to quit smoking also underpin effective parenting. The failure rate of smoking cessation programmes in pregnancy and the adverse psychosocial experience of mothers who continue to smoke raise the question of which interventions beyond cessation programmes will best help these mothers during their pregnancies and while raising their families.

What is already known on this subject

Smoking during pregnancy is associated with poor behavioural outcomes in offspring; however, recent research suggests that not only does smoking cessation during pregnancy alter these outcomes, but also that the children of mothers who are able to quit smoking may be ‘easier’ infants than those born to mothers who never smoked.

What this study adds

  • In a prospective pregnancy cohort followed to 14 years, it has been shown that quitting smoking during pregnancy was associated with a small positive change in behavioural scores through to adolescence, whereas continuing to smoke increased the risk of behavioural problems in children.

  • The decision and ability to quit smoking in pregnancy are, therefore, important determinants of child behavioural development and should be a focus for early intervention policies in mental health.


We are extremely grateful to all the families who took part in this study and the whole Raine Study team, which includes data collectors, cohort managers, data managers, clerical staff, research scientists and volunteers.


View Abstract


  • Funding The Western Australian Pregnancy Cohort (Raine) Study is funded by the Raine Medical Research Foundation at The University of Western Australia, the National Health and Medical Research Council of Australia (NHMRC), the Telstra Foundation, the Western Australian Health Promotion Foundation, and the Australian Rotary Health Research Fund. We would also like to acknowledge the Telethon Institute for Child Health Research and the NHMRC Program Grant which supported the 14-year follow-up (Stanley et al ID 003209).

  • Competing interests None.

  • Ethical approval The Human Ethics Committees at KEMH and/or Princess Margaret Hospital for Children approved the protocols for the study.

  • Provenance and peer review Not commissioned; externally peer reviewed.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.