Fergusson DM, Woodward LJ, Horwood LJ. Maternal Smoking During Pregnancy and Psychiatric Adjustment in Late Adolescence. Arch Gen Psychiatry. 1998;55(8):721-727. doi:10.1001/archpsyc.55.8.721
Copyright 1998 American Medical Association. All Rights Reserved. Applicable FARS/DFARS Restrictions Apply to Government Use.1998
This study assessed the extent to which exposure to maternal smoking during pregnancy was associated with increased risks of psychiatric symptoms in late adolescence (adolescents aged 16-18 years) when due allowance was made for confounding or selection factors associated with maternal smoking during pregnancy.
Data were gathered during an 18-year longitudinal study of a birth cohort of 1265 children born in New Zealand. The measures collected included (1) maternal smoking during pregnancy; (2) assessments of psychiatric problems (conduct disorder, major depression, and anxiety and substance use disorders) at age 16 to 18 years; and (3) measures of potentially confounding social, family, and parental factors.
Children exposed to maternal smoking during pregnancy had higher psychiatric symptom rates for conduct disorder, alcohol abuse, substance abuse, and depression. Those children whose mothers smoked at least 1 pack of cigarettes per day during their pregnancy had symptom rates that were between 1.4 and 2.5 (median, 2.0) times higher than the children of nonsmokers. Smoking during pregnancy was also associated with a series of adverse or disadvantageous factors that included (1) socioeconomic disadvantage, (2) impaired child-rearing behaviors, and (3) parental and family problems. After adjustment for these confounding and selection factors, smoking during pregnancy was significantly associated with an increased rate of conduct disorder symptoms in late adolescence (P<.001). This effect was more pronounced for male than female adolescents.
This study suggests that maternal smoking during pregnancy may contribute to childrens' risk of later externalizing problems. There is a need to further explore the moderating effect of the sex of the child and to clarify the underlying pathophysiological features of this relationship.
THERE HAS been increasing concern about the effects of maternal smoking during pregnancy on children. Research indicates that maternal smoking during pregnancy is associated with increased risks of miscarriage,1,2 reduced birth weight,3,4 compromised perinatal status,4- 7 and reduced intelligence in children.8- 10 More recently, numerous studies have suggested that children whose mothers smoke during pregnancy are at an increased risk of developing later childhood externalizing problems, including attention deficit,11,12 oppositional defiant behavior, or conduct disorder.13- 16 Furthermore, recent evidence also indicates that the adverse effects of maternal smoking during pregnancy may persist into adolescence.14
While there has been increasing research into the relationship between exposure to maternal smoking during pregnancy and later psychological adjustment in childhood and adolescence, several features of this association merit further examination. The first concerns the extent to which associations between maternal smoking and adjustment are specific to externalizing behaviors or, conversely, the extent to which smoking during pregnancy is associated with generalized increases in risks of disorder. To date, to our knowledge, research in this area has focused on the relationship between maternal smoking during pregnancy and externalizing behavior problems,11- 16 while virtually no attention has been given to the extent to which smoking during pregnancy may be related to other child and adolescent outcomes, including depression, generalized anxiety, and substance misuse.
A second issue in this area concerns the extent to which statistical associations between maternal smoking during pregnancy and later adjustment reflect the following: (1) cause-and-effect associations in which, by various routes, exposure to cigarette smoking in utero influences childrens' longer-term susceptibility to adjustment difficulties; and (2) noncausal, or spurious, associations between smoking during pregnancy and child psychiatric adjustment that arise because of the social background, behavioral characteristics, and child-rearing practices of mothers who elect to smoke during pregnancy.16 To resolve this issue requires that observed associations between maternal smoking during pregnancy and subsequent psychological adjustment in children be shown to persist even after selection factors known or suspected to be associated with maternal smoking during pregnancy have been taken into account.
This article reports the results of an 18-year longitudinal study of the relationships between maternal smoking during pregnancy and psychiatric outcomes in a birth cohort of more than 1000 children born in New Zealand. This study examined the extent to which consistent dose-response relationships could be found between the extent of maternal smoking during pregnancy and rates of psychiatric symptoms in late adolescence (adolescents aged 16-18 years) and the extent to which associations between smoking during pregnancy and rates of psychiatric symptoms persisted when due allowance was made for a range of confounding or selection factors that were associated with smoking during pregnancy and later risks of disorder.
The data were collected during the course of the Christchurch Health and Development Study. The Christchurch Health and Development Study is a longitudinal study of a birth cohort of 1265 children born in the Christchurch, New Zealand, urban region during mid-1977. These children have been studied at birth, at age 4 months, at annual intervals to age 16 years, and again at age 18 years. The analyses reported in this article were based on a sample of 1022 children for whom there was complete data on the maternal smoking and all mental health variables measured at age 18 years. This sample represented 80.8% of the initial birth cohort and 92.1% of all cohort members still alive and resident in New Zealand at the age of 18 years. Losses to follow-up arose because of outmigration from New Zealand (56.3%), refusal to participate in the research (35.4%), and mortality (8.3%). A further 3 (0.2%) of the subjects were omitted because of missing data for maternal smoking during pregnancy variable.
To examine the effects of sample losses on the representativeness of the sample, comparisons were made between the 1022 children included in the analyses and the excluded 243 cohort members on a range of social background measures collected at birth. This analysis suggested that losses to follow-up were not associated with maternal age, ethnicity, family size, or sex. However, there were small but statistically detectable (P<.01) tendencies for this sample to underrepresent children from families with a lower socioeconomic status. Although these results suggest some bias in the sample, it is unlikely that this bias will materially influence the results reported in this article, since previous efforts to correct for nonrandom sample loss in the cohort have shown these effects to be negligible.17,18
At birth, mothers were questioned about their smoking habits during each pregnancy trimester. Because the number of cigarettes smoked per day across the 3 trimesters was highly correlated (r>0.85), a composite measure of the mean number of cigarettes smoked per day throughout the pregnancy was created. Based on this score, mothers were classified into 1 of 4 groups: (1) nonsmoking mothers, (2) mothers who smoked an average of between 1 and 9 cigarettes per day, (3) mothers who smoked an average of between 10 and 19 cigarettes per day, and (4) mothers who smoked an average of 20 or more cigarettes per day.
At age 18 years, sample members were questioned about their substance use and experience of mental health problems during the previous 2 years using a questionnaire based on the Composite International Diagnostic Interview19 and the Self-Report Delinquency Inventory.20 With the use of this information, DSM-IV21 criteria were used to construct a series of substance abuse and psychiatric diagnoses for each sample member. These diagnoses included conduct disorder, alcohol abuse or dependence, nicotine dependence, cannabis and other illicit substance abuse or dependence, generalized anxiety, and major depression.
In a preliminary examination of the data, analyses were conducted using DSM-IV diagnoses. This analysis suggested that maternal smoking during pregnancy was unrelated to disorder risk after control for selection factors. To test the robustness of this conclusion, the data were reanalyzed using scale measures of the number of DSM-IV symptoms reported by each subject. This analysis revealed the presence of significant associations between maternal smoking during pregnancy and psychiatric outcome even after control for the effects of selection factors (P<.05). The differences between the results for categorical and dimensional analyses suggested that a dimensional scoring approach led to greater predictive precision. This is consistent with a previous analysis of the predictive validity of categorical and dimensional scaling approaches to the measurement of psychiatric disorder within this cohort.22 Therefore, for this article, the variabilities in symptom levels were analyzed as dimensional variables reflecting the number of DSM-IV symptoms reported. The reliability of each symptom scale was assessed using the Cronbach α, which suggested that the measures used were of moderate to high reliability (conduct disorder, α=.72; alcohol abuse or dependence, α=.84; nicotine dependence, α=.85; illicit substance abuse or dependence, α=.89; generalized anxiety, α=.97; and major depression, α=.92).
To control for selection factors associated with maternal smoking during pregnancy, a range of social background, pregnancy, and family factors was selected for inclusion in the study. These factors were chosen because of their statistical associations with maternal smoking during pregnancy and their availability in the study database.
Measures of maternal educational level, maternal age, and family socioeconomic status were obtained at the time of the sample member's birth. Education was coded on a 3-point scale reflecting the highest academic qualification obtained (1, no formal qualifications; 2, high school qualifications; and 3, tertiary level qualifications). Family socioeconomic status at the time of the child's birth was assessed using the Elley and Irving 23 scale of socioeconomic status for New Zealand. This scale classifies families into 6 classes based on paternal occupation.
Measures of mother's pregnancy planning, alcohol consumption, and illicit drug use during pregnancy were also collected at the birth assessment. Alcohol use was coded as the mean number of drinks consumed per week across the 3 pregnancy trimesters. Illicit drug use was coded if mothers reported at any time having used an illicit drug (cannabis, hallucinogens, stimulants, sedatives, or opiates) without a physician's prescription during their pregnancy.
Four measures of maternal child-rearing practices were used in this analysis. First, the maternal emotional responsiveness subscale of the HOME Inventory was included as a measure of the frequency with which mothers were observed to make positive emotional responses to their 3-year-old child.24,25 Second, at age 18 years, sample members were questioned about the extent to which each parent used physical punishment to discipline them during their childhood (from birth to age 16 years). A composite measure of exposure to physical punishment was created by taking the highest level of punishment experienced by the sample member.26 This classification ranged from 1 (parent[s] never used physical punishment) to 4 (at least one parent used physical punishment too often or too severely). Third, at age 18 years, sample members were questioned about their exposure to childhood sexual abuse. Children were classified into 4 groups, ranging from 1 (no exposure to childhood sexual abuse) to 4 (exposure to childhood sexual abuse involving completed or attempted oral, anal, or vaginal intercourse).27
Five measures of parental behavior and family functioning were identified: parental separation, parental conflict, parental history of alcohol problems, parental history of criminal offending, and parental illicit drug use. Children were coded as having experienced a parental separation if their parents had separated or divorced within their first 5 years of life. The quality of parents' marital relations was coded annually using the following 3 items: (1) whether the parents had engaged in prolonged arguments during the last 12 months, (2) whether the child's mother had reported being assaulted by her spouse in the last 12 months, and (3) whether the child's mother had reported experiencing sexual difficulties in the last 12 months. These items were then combined to produce a scale measure of the extent to which the child was exposed to parental conflict in the period from birth to age 5 years.28 When the children were 15 years old, parents were asked about their history of criminal offending and alcohol abuse during the last 15 years. Finally, when the children were 11 years old, parents were questioned about their use of cannabis and other illicit drugs.
The analysis was conducted in 4 stages:
The associations between maternal smoking during pregnancy and DSM-IV symptoms were tabulated and tested for statistical significance using 1-way analysis of variance and F tests of linear and nonlinear association (Table 1).
The associations between maternal smoking during pregnancy and a range of confounding and selection factors were then assessed using χ2 tests of independence (Table 2).
The associations between smoking during pregnancy and psychiatric symptom rates were then adjusted for confounding and selection by fitting linear regression models to the data. Model fitting was conducted using methods of forward and backward variable elimination to identify the most parsimonious model. The adjusted dose-response functions were then computed using the parameters of the regression model to estimate the adjusted relationship between symptom rates and smoking during pregnancy (Table 3). An account of this method is provided by Lee.29
The effects of the sex of the child on the adjusted relationship between maternal smoking during pregnancy and subsequent psychiatric well-being were examined using a 2-way analysis of covariance (Table 4).
Table 1 shows the relationships between maternal smoking during pregnancy and the rates of conduct disorder, substance abuse, anxiety, and depressive symptoms. For all comparisons except anxiety, there was evidence of significant linear associations between maternal smoking during pregnancy and symptom rates in late adolescence (P<.05). There were no significant nonlinear associations. Children whose mothers smoked more than 1 pack of cigarettes per day during pregnancy had rates of conduct disorder symptoms, nicotine dependence, substance abuse symptoms, and depressive symptoms that were between 1.4 and 2.5 (median, 2.0) times higher than the children of mothers who did not smoke during pregnancy.
Table 2 shows the relationships between maternal smoking during pregnancy and a range of measures of maternal social background, antenatal history, child-rearing practices, and parental and family characteristics. For ease of data display, all measures have been presented in dichotomous form. The results revealed relatively strong associations between mothers' social background and their tendency to smoke during pregnancy. Mothers who smoked during pregnancy, and particularly those who smoked more than half a pack (10 cigarettes) per day, were younger, less well educated, and of a lower socioeconomic status compared with nonsmoking mothers. Compared with nonsmoking mothers, mothers who smoked during their pregnancy were less likely to have planned their pregnancy and were also more likely to have consumed alcohol and used illicit drugs while pregnant. For early child-rearing practices, women who smoked during pregnancy were rated as being significantly less nurturing towards their children at age 3 years. Children whose mothers smoked during pregnancy had greater exposure to child physical abuse and child sexual abuse compared with children whose mothers did not smoke during their pregnancy. Finally, mothers who smoked during pregnancy were more likely to have experienced a separation or divorce and reported higher rates of alcohol abuse or dependence, illicit drug use, and criminal offending compared with nonsmoking mothers. These findings clearly raise the possibility that the elevated rates of disorder found among the adolescent children whose mothers smoked during pregnancy could reflect the social circumstances, personal characteristics, and child-rearing practices of women who elected to smoke during their pregnancy rather than the direct causal effect of smoking on childrens' later adjustment.
Table 3 shows the relationships between maternal smoking during pregnancy and later psychiatric symptoms after adjustment for selection factors associated with maternal smoking. As noted in the "Statistical Methods" subsection of the "Subjects and Methods" section, these adjustments were obtained using multiple linear regression. For each outcome, Table 3 shows (1) the covariate-adjusted dose-response relationship between maternal smoking during pregnancy and rates of symptoms, (2) the regression coefficient and SE for the smoking during pregnancy factor, (3) the significance of the association, and (4) the significant covariates in the regression adjustment equations.
Table 3 shows that after controlling for confounding and selection processes, maternal smoking during pregnancy was not significantly related to symptoms of alcohol abuse or dependence, nicotine dependence, illicit drug abuse or dependence, major depression, or generalized anxiety. However, even after adjustment for confounding, there was still a clear and significant relationship between conduct disorder symptoms and maternal smoking during pregnancy. After covariate adjustment, children whose mothers smoked 1 pack of cigarettes or more per day during their pregnancy had rates of conduct disorder symptoms that were twice as high as the rate for the children of nonsmokers.
An exploration of the sensitivity of the results in Table 3 to the order and number of control factors included in the analysis indicated that, providing the significant covariates identified in Table 3 were included in the analysis, inclusion or exclusion of further covariates did not materially influence the estimates of the adjusted dose-response functions or the significance level of the associations.
To examine possible sex differences in the effects of maternal smoking during pregnancy on conduct disorder symptoms, the analysis was extended to an analysis of covariance model in which maternal smoking during pregnancy and sex were included as factors. This analysis produced evidence of a clear smoking during pregnancy × sex interaction (F3,936=7.18, P<.001) for conduct problems. This interaction is illustrated in Table 4, which shows the relationship between maternal smoking during pregnancy and conduct problems after adjustment for selection factors for male and female adolescents. While rates of conduct difficulties increase with increasing levels of maternal smoking during pregnancy for male and female adolescents, the rate of increase in conduct disorder symptoms is more marked for male than for female adolescents. These results are consistent with the view that male adolescents may be more susceptible than female adolescents to the effects of maternal smoking during pregnancy.
This study has examined the relationship between maternal smoking during pregnancy and psychiatric symptoms in late adolescence within a birth cohort of more than 1000 children studied to the age of 18 years. The major findings of the study and their implications are reviewed.
There was evidence of pervasive and general associations between maternal smoking during pregnancy and later adolescent adjustment, with children whose mothers smoked during pregnancy having elevated rates of symptoms of conduct disorder, alcohol abuse or dependence, nicotine dependence, illicit drug abuse or dependence, anxiety, and depression. Furthermore, there was evidence of consistent dose-response relationships in which increasing exposure to cigarette smoke during pregnancy was associated with steadily increasing rates of psychiatric symptoms.
However, subsequent analyses showed that maternal smoking during pregnancy was correlated with a range of social background, pregnancy, parental, and family functioning factors that could also have led to spurious correlations between maternal smoking during pregnancy and adolescent adjustment. The children of women who smoked during pregnancy were exposed to higher levels of social disadvantage, parental deviance, and family dysfunction throughout childhood compared with the children of women who did not smoke during pregnancy. Control for these social and contextual factors explained almost all of the associations between maternal smoking during pregnancy and self-reported psychiatric symptoms, including anxiety, depression, nicotine dependence, alcohol misuse, and illicit drug abuse or dependence. However, even after controlling for potential confounding and selection factors, maternal smoking during pregnancy remained significantly associated with the number of conduct disorder symptoms reported by sample members at age 18 years (P<.001). Children whose mothers smoked 1 pack of cigarettes or more per day during their pregnancy had mean rates of conduct disorder symptoms that were twice as high as those found among children born to mothers who did not smoke during their pregnancy.
These findings add to growing evidence that suggests possible causal links between exposure to maternal smoking during pregnancy and increased rates of externalizing symptoms in later life.11- 16 In support of this causal hypothesis, the following may be noted.
A growing number of studies have reported associations between maternal smoking during pregnancy and measures of externalizing symptoms during childhood. In all cases, these associations have been shown to persist after control for confounding and selection processes.
As illustrated in this study, there is evidence of consistent dose-response relationships in which increasing exposure to maternal smoking during pregnancy is associated with increased rates of externalizing symptoms.
This study also suggests that this association is specific to conduct problems and that similar associations do not exist between maternal smoking during pregnancy and other psychiatric symptom patterns.
Additionally, longitudinal evidence from this study suggests that the association is present during mid-childhood 13 and late adolescence.
Collectively, these considerations suggest that the evidence on maternal smoking during pregnancy and later externalizing behaviors meets many of the criteria (consistent dose-response relationship, replication, resilience to control for confounding, and specificity of association) that have been proposed as bases for causal inference in nonexperimental studies.30
While the available evidence clearly suggests a possible causal relationship between maternal smoking during pregnancy and later externalizing behaviors, there are still grounds for uncertainty about this conclusion. It remains possible that this association reflects uncontrolled confounding factors that are related to smoking during pregnancy and later externalizing symptom levels. One source of uncontrolled confounding comes from the possible effects of common genetic factors that are associated with maternal smoking during pregnancy and later externalizing behaviors. A further difficulty in this area has been that of developing a biologically plausible account of the physiological and developmental bases of any causal relationship between exposure to pregnancy smoking and later externalizing behavior. Several researchers have pointed to possible biological processes and mechanisms that may explain this association.12,14,15 These have included fetal hypoxia, changes in serotonin uptake, changes in dopaminergic systems, and changes in DNA and RNA synthesis in the brain. However, these explanations remain highly speculative. The uncertainties about the adequacy of control of confounding factors and the absence of compelling evidence about the mechanisms that might lead to causal links between maternal smoking during pregnancy and later outcomes continue to pose threats to the validity of causal inferences in this area.16 Finally, the results of this study may suggest that the effects of maternal smoking during pregnancy on externalizing behavior are modified by the child's sex, with the evidence suggesting that while boys show increases in externalizing symptoms with increasing exposure to maternal pregnancy smoking, the same may not be true for girls.
Accepted for publication March 19, 1998.
This study was supported by grants from the Health Research Council of New Zealand, and the National Child Health Research Foundation, Auckland; the Canterbury Medical Research Foundation, Christchurch, New Zealand; and the New Zealand Lottery Grants Board, Wellington.
Corresponding author: David M. Fergusson, PhD, Christchurch Health and Development Study, Department of Psychological Medicine, Christchurch School of Medicine, PO Box 4345, Christchurch, New Zealand (e-mail: firstname.lastname@example.org).