The literature on maternal prenatal smoking and externalizing behavior problems is clearly separated in two developmental periods. The bulk of the current literature on maternal prenatal smoking and externalizing behavior problems covers either childhood and adolescence or the preschool years. However, the rates of common childhood psychiatric disorders such as conduct disorder (CD) and attention deficit hyperactivity disorder (ADHD) and the patterns of comorbidity among them in early childhood are similar to those seen in later childhood (Egger & Angold,
2006). Limitations in current diagnostic criteria for early childhood psychopathology have resulted in a research focus on specific behaviors such as physical aggression (PA) and hyperactivity-impulsivity (HI) rather than clinical disorders in this age group. Both physical aggression and hyperactivity during early childhood appear to be typical precursors of full-blown CD and ADHD during the school years and beyond (Séguin, Nagin, Assaad, & Tremblay,
2004). A substantial number of studies show continuity between early and late childhood externalizing behavior problems (e.g. Campbell, Breaux, Ewing, & Szumowski,
1986; Keenan & Wakschlag,
2000) or have identified consistent developmental trajectories for PA and HI that start as early as age 1 1/2 years (e.g. Côté, Vaillancourt, LeBlanc, Nagin, & Tremblay,
2006; NICHD Early Child Care Research Network,
2004; Romano, Tremblay, Farhat, & Côté,
2006; Shaw, Lacourse, & Nagin,
2005).
In reviewing childhood and adolescence studies, maternal prenatal smoking has consistently been associated with CD- and ADHD- symptoms (for reviews, see Linnet et al.,
2003; Wakschlag, Pickett, Cook, Benowitz, & Leventhal,
2002). The associations have been observed in clinical samples (e.g. Mick, Biederman, Faraone, Sayer, & Kleinman,
2002; Milberger, Biederman, Faraone, Chen, & Jones,
1996), ‘at-risk’ samples (e.g. Wakschlag & Hans,
2002) and large population-based samples (e.g. Braun, Kahn, Froehlich, Auinger, & Lanphaer,
2006; Kotimaa et al.,
2003). Whereas these studies generally controlled for familial psychopathology and numerous environmental risk factors, there are also indications from studies using (large) twin samples that maternal prenatal smoking predicts children’s CD- and ADHD-symptoms beyond the effects of heritable risk (e.g. Button, Thapar, & McGuffin,
2005; Maughan, Taylor, Caspi, & Moffitt,
2004; Silberg et al.,
2003; Thapar et al.,
2003). Studies showing associations between maternal prenatal smoking and CD-symptoms and delinquency during adolescence and adulthood (Brennan, Grekin, Mortensen, & Mednick, 2002; Fergusson, Woodward, & Horwood,
1998; Räsänen et al.,
1999) and studies showing behavior problems in children of mothers who smoked during pregnancy at different time points (e.g. Maughan et al.,
2004; Wakschlag & Hans,
2002; Wakschlag, Pickett, Kasza, & Loeber,
2006) suggest long-lasting effects.
Although there are indications that maternal prenatal smoking is more strongly or even exclusively related to CD symptoms (Wakschlag & Hans,
2002; Wakschlag, Leventhal, Pine, Pickett, & Carter,
2006; Wakschlag, Pickett et al.,
2006), only few studies tested for specificity of the relations between maternal prenatal smoking and CD- or ADHD-symptoms (Button et al.,
2005; Mick et al.,
2002; Thapar et al.,
2003). This is an important matter considering the high comorbidity of these behavior problems (e.g. Jensen, Martin, & Cantwell,
1997). Mick et al. (
2002) found a robust link with ADHD-symptoms after controlling for CD-symptoms in a clinical population. In that study, children aged 6 to 17 years with ADHD and non-ADHD controls were compared after statistical control for CD. In contrast, Thapar et al.’s (
2003) and Button et al.’s (
2005) population-based twin study using the Greater Manchester Twin Register did not select on either CD- or ADHD-symptoms. Whereas Thapar et al. showed that prenatal maternal smoking predicted a unique proportion of the variance in ADHD-symptoms after control for CD-symptoms in children between 5 and 16 years, Button et al. (
2005) tested a more complex model showing unique variance for both CD- and ADHD-symptoms in children 5 to 18 years of age. Although the aforementioned studies used some form of statistical control for the symptoms associated with the ‘other’ disorder, only one study to date examined a relation between prenatal maternal smoking and co-occurring behavior problems (Wakschlag, Pickett et al.,
2006). In that study maternal prenatal smoking predicted the co-occurrence of Oppositional Defiant Disorder (ODD) and ADHD when high levels of both were contrasted with absence of both. This suggests that maternal prenatal smoking, to the extent that it is a causal factor, may affect behavior much more seriously than initially shown in specificity studies. However, such a conclusion may be premature because that analysis did not inform us if maternal prenatal smoking was related specifically to the co-occurrence of both behavior problems or whether it was driven by its association with either behavior problem because we do not know if the combined group differed from the ODD-only or ADHD-only groups. Thus whether maternal prenatal smoking is related to the co-occurrence of behavior problems remains an open question in this literature.
In contrast to childhood and adolescence studies, there is a more limited number of early childhood studies and they tend to focus on symptoms rather than diagnoses. Tremblay et al. (
2004) reported an association with early childhood trajectories of PA across time, and Romano et al. (
2006) found an association between maternal prenatal smoking and hyperactive symptoms from age 2 to 7, but both studies focused on one externalizing behavior problem. Other studies focused on multiple behaviors. Orlebeke, Knol, and Verhulst (
1997) and Williams et al. (
1998), using the Child Behavior Checklist (CBCL) in large cohort studies of 3-year-old and 4 to 6-year old children, respectively, found associations with externalizing behavior problems (aggressive, overactive, oppositional) but not with internalizing behavior problems (withdrawn, anxious, depressed). Day, Richardson, Goldschmidt, and Cornelius (
2000) reported significant associations of third trimester exposure with scores on each of the subscales of the Toddler Behavior Checklist (Physical Aggression, Oppositional Behavior, Immaturity, and Emotionality) and Activity level assessed with the Routh Activity Scale. Wakschlag, Leventhal et al. (
2006) noted a persistent association between maternal prenatal smoking and disruptive behaviors during early childhood. None of these multiple behavior studies examined co-occurrence. Thus there also remains a need in the early childhood literature to examine whether maternal prenatal smoking is specifically associated with the co-occurrence of problem behaviors.
From a developmental perspective, early onset of externalizing behavior has been associated with the poorest outcomes (Brame, Nagin, & Tremblay,
2001; Lacourse et al.,
2006; Moffitt,
1993; Woodward, Fergusson, & Horwood,
2002). In addition, co-occurring externalizing behaviors during middle childhood have been associated with the poorest behavioral outcomes during adolescence and adulthood (Lacourse et al.,
2006; Lahey, McBurnett, & Loeber,
2000; Séguin et al.,
2004). However, co-occurring behavior problems may be etiologically different from individual behavior problems, with their own social, sociodemographic and biological precursors (Waschbusch,
2002). Yet, we know very little about the prenatal markers of co-occurring behavior problems in general. Given that maternal prenatal smoking has been clearly associated with several single externalizing problems across development, in the present study, we test the hypothesis that maternal prenatal smoking is associated with co-occurring externalizing problems. Specifically we predict that maternal prenatal smoking will be associated to the co-occurrence of PA and HI in contrast with the absence of both, PA-only, and HI-only in a large early childhood sample.
Discussion
The results of this study of early childhood show that maternal prenatal smoking was associated with PA but not with HI in covariate regression analyses when PA and HI were examined separately. When PA and HI were combined, a contrast between the PA-only group and the remainder of the sample, which was neither high on PA or HI, fell short of significance. However, maternal prenatal smoking predicted co-occurring elevated levels of PA and HI. Therefore the key finding of this study is that maternal prenatal smoking predicted high PA, but largely when it is co-occurring with high HI. Had we just examined PA and HI separately, as has often been done in studies of specificity, we would have missed the fact that HI is also sensitive to maternal prenatal smoking, even after controlling for several confounding variables, but only when it is combined with PA.
The first implication of these findings is that maternal prenatal smoking may be associated with the most severe forms of PA and HI, i.e. their combination, in early childhood. Second, the fact that maternal prenatal smoking predicts co-occurring PA and HI may be considered an important finding in light of evidence indicating that multiple co-occurring behavior problems at an early age are associated with a higher risk of persistent antisocial problems (Brame et al.,
2001; Broidy et al.,
2003; Nagin & Tremblay,
1999; Lacourse et al.,
2006; Lahey et al.,
2000; Moffitt,
1993; Moffitt, Caspi, Dickson, Silva, & Stanton,
1996), as well as with several other functional impairments (Waschbusch,
2002; Séguin et al.,
2004). Finally, the risk factors identified for these combined behavior problems, are largely modifiable.
Our study replicates the reports from Orlebeke et al. (
1997), Williams et al. (
1998), Day et al. (
2000), and Tremblay et al. (
2004), based largely on early childhood behavior problems. However, we extend these findings because none of those early childhood studies had tested for the prediction of specific behavior problems or for the specificity of their combination. Tremblay et al. (
2004) covered the same developmental period and used the exact same PA scale as we did here on another sample. We both found that the association between maternal prenatal smoking and PA was not explained by control variables. However, our results for HI appear to contrast slightly with those of Romano et al. (
2006) who reported that the association between maternal prenatal smoking and hyperactivity was not explained by control variables. There are a number of key differences between the current study and Romano et al. (
2006). Besides using a different sample, we included impulsivity along with hyperactivity items, and focused exclusively on behavior during early childhood. Thus, in the current study the effects of maternal prenatal smoking on HI were largely explained by control variables. It is possible that control variables have a greater effect on the impulsivity component of HI than on the hyperactivity component. We note that our results appear to be consistent with those of Wakschlag, Leventhal et al. (
2006), who also failed to find a relation of maternal prenatal smoking but with ADHD, which includes impulsivity. Because of our goal for a strict test for specificity of effects on PA, it was more conservative to have a combination of impulsivity and hyperactivity items in our scale. Alternately, by having information extending beyond the early childhood period to estimate trajectories Romano et al. (
2006) may have more accurately identified children for whom maternal prenatal smoking matters beyond the effects of control variables. Despite this limitation to early childhood, and despite the fact that we cannot claim to have accounted for all possible control variables (examples include exposure to environmental smoking postnatally and forms of parental psychopathology other than maternal depression and parental history of antisocial behavior such as parental ADHD), this study replicates and extends all previous studies by providing an empirical basis for how the association of prenatal smoking to both PA and HI may manifest itself. It is now much more clear that studies need to examine co-occurrence of behavior problems.
There is a leap in drawing parallels between early childhood symptoms to DSM-based disorder, i.e., from PA to CD and from HI to ADHD. CD is characterized by aggression to people and/or animals, deceitfulness or theft, vandalism and serious rule violations whereas ADHD is characterized by hyperactivity, behavior disinhibition (or impulsivity), and inattention and distractibility (American Psychiatric Association,
1994). It is not yet clear to what extent behavior problems during the early childhood years map onto ADHD and CD symptoms during later childhood and (antisocial) behavior problems during adolescence and adulthood, although we have shown that childhood PA trajectories and hyperactivity trajectories were respectively related to CD and ADHD measured in adolescence (Séguin et al.,
2004). But manifestation of PA and HI symptoms are often not sufficient to warrant diagnoses. We also note that behavior problems during the early childhood years tend to be more common and typically decline over time (Côté et al.,
2006; Romano et al.,
2006; Bongers, Koot, Van Der Ende, & Verhulst,
2003).
A number of pioneering studies showed that behavior problems later in life (ranging from school-age to adulthood) could already be identified during early childhood (e.g. Campbell et al.,
1986; Keenan & Wakschlag,
2000), but only recently a number of studies have started to show, particularly with respect to HI, a significant consistency in the likelihood to display (very) high levels from age 1 1/2 years onwards (e.g. Romano et al.,
2006; Shaw et al.,
2005). With respect to PA, there is a larger group of very young children showing rather high levels, but most of them will use less PA from early school age onwards (e.g. Côté et al.,
2006; NICHD Early Child Care Research Network,
2004; Shaw et al.,
2005). Such general trends might also underlie differences in number of groups identified by the trajectory methodology in samples with different age ranges. Whereas our three PA-groups were consistent with the three groups identified by some authors (e.g. Côté et al.,
2006) covering early and middle childhood, other authors identified more than three groups for this period (NICHD Early Child Care Research Network,
2004; Shaw et al.,
2005). However, most children will not yet have finished learning alternative strategies (i.e. have undergone ‘socialization of aggression,’ Tremblay,
2003) between 17 and 42 months, so the expected ‘desisting’ patterns of PA will not yet be evident. One thing the abovementioned studies have in common is that they show a degree of consistency between early and middle childhood conduct problems, i.e. children with chronic PA during middle childhood generally had high PA levels during early childhood as well. This, in turn, emphasizes the importance of searching for predictors of co-occurring early childhood behavior problems: children most at risk of subsequent behavior problems might be those who display both high PA and high HI at this stage.
The combination of conduct problems and hyperactivity-impulsivity-attention problems has been distinguished from their pure forms at a number of levels (Waschbusch,
2002). Although research has again mainly focused on older children with diagnosed disorders rather than the combination of behaviors such as PA and HI (but see Séguin, Arseneault, Boulerice, Harden, & Tremblay,
2002; Séguin et al.,
2004), evidence has been provided for a number of etiological and developmental differences between the comorbid and individual conditions. For example, greater behavioral and autonomic nervous system reactivity in response to provocation has been shown in the comorbid compared to the separate conditions (Waschbusch,
2002). Children with CD+ADHD were also shown to have lower baseline sympathetic arousal than children with the separate disorders (specifically those with ADHD only; Herpertz et al.,
2001). There is also evidence for differences in patterns of brain activity between CD+ADHD and the separate conditions (Banaschewski et al.,
2003) and for specific heritable risk of CD+ADHD (Dick, Viken, Kaprio, Pulkkinen, & Rose,
2005; Thapar, Harrington, & McGuffin,
2001). Several studies have shown that specific family psychosocial characteristics (e.g. parent-child conflict; parental psychopathology) specifically predict CD+ADHD or predict CD+ADHD more strongly than the separate conditions (e.g. Burt, Krueger, McGue, & Iacono,
2003; Pfiffner, McBurnett, Rathouz, & Judice,
2005). Thus, although extrapolation to CD and ADHD must be done with caution, the results of the present study support an etiological basis for the combined behavior problems in an early childhood sample.
Finally, in addition to prenatal maternal smoking, we found that maternal hostile- reactive parenting was also associated with co-occurring PA and HI beyond its effects on the separate behaviors. Hostile-reactive parenting and similar ‘negative’ parenting behaviors have been associated with children’s PA and HI levels in earlier studies (e.g. Côté et al.,
2006; Romano et al.,
2006; Shaw et al.,
2005; Tremblay et al.,
2004), Negative parenting has also been related to the failure of training programs aimed at improving child conduct problems (Webster-Stratton, Reid, & Hammond,
2001). This association could be seen as resulting from the impact of maternal behavior on their child’s behavior. However, they may also reflect, at least in part, parent’s reactions to their child’s behavior. For instance, when faced with conduct problems, mothers of preschoolers at risk for ADHD and Oppositional Defiant Disorder tend to resort to more negative parenting strategies (Cunningham & Boyle,
2002). Furthermore, mothers hostile-reactive parenting toward their 5 month-old infants has been found to be partly driven by their infant’s difficult temperament (Boivin et al.,
2005).
Our finding that maternal hostile-reactive parenting is associated with an increased risk of co-occurring PA and HI may thus reflect the increased challenge of parenting when a child displays multiple behavior problems (cf. Seipp & Johnston,
2005). This pattern could lead to the establishment of a family coercive process and the further learning of anti-social behaviors by the child. However, the state of evidence precludes any definite conclusions at this point, and there is a clear need to better document the early dynamics of negative parenting and externalizing problems in early childhood.
To the extent that they may be involved in causation, the good news is that prenatal maternal smoking and hostile-reactive parenting could be significantly attenuated in prevention and intervention programs. The fact that the effects of smoking during pregnancy on both PA and PA+HI were robust, added to the evidence that PA+HI increases the risk for continued serious behavior problems, emphasizes the importance of smoking cessation during pregnancy (for a review on effectiveness of cessation programs, see Lumley, Oliver, Chamberlain, & Oakley,
2004). Considering the high prevalence of prenatal smoking (around 25% in Western countries), it will be important to further refine the identification of families most at risk of having children with a prepotency to develop co-occurring PA and HI. Our results further confirm that hostile-reactive parenting should be targeted in preventive intervention aimed at children with early multiple behavior problems. However, because parenting interventions often are the least effective in multiple-risk families (for a review, see Hutchings & Lane,
2005), including those characterized by prenatal smoking (Vuijk, van Lier, Huizink, Verhulst, & Crijnen,
2006), cessation programs targeting the most vulnerable families should be emphasized, intensified, and given appropriate means to succeed.