We found evidence for a developmental model suggesting both direct and indirect effects of prenatal exposure to cocaine and other substances on behavior problems in childhood. The indirect effects show a sequence of connected behavioral alterations starting in the neonatal period that lead to later behavior problems. Prenatal substance exposure predicted higher infant reactivity and stress at 1 month, which led to a more difficult temperament at 4 months. In turn, difficult temperament was associated with more behavior problems at 3 and 7 years. These indirect effects of prenatal substance exposure were observed in the presence of the direct effects of prenatal substance exposure and effects of SES on these behavioral outcomes. In other words, the indirect effects remained after controlling for the direct effects of prenatal substance exposure and SES. This suggests multiple pathways, both direct and indirect, from prenatal substance exposure to behavioral outcomes in childhood, and that the effects are cumulative.
The direct paths may be thought of as teratogenic effects and supports previous findings relating prenatal substance exposure to caregiver report of behavior problems in school-age children.5, 21, 22
Our findings suggest these may be “true” teratogenic effects because they remained when indirect effects were also included. On the other hand, half of the direct path coefficients failed to reach statistical significance in the SEMs (), suggesting these effects are less robust than when tested with more traditional regression analysis. These findings also provide an alternative complimentary model for studying unique (direct) effects of prenatal substance exposure. In addition, both direct and indirect effects were found in some models indicating they explain additive portions of the variance.
Our finding that SES mediates the effects of prenatal substance exposure on childhood behavioral problems is consistent with previous findings.23–25
Clearly, this is not a “causal” model; prenatal substance exposure does not “cause” low SES. Rather, SES is a proxy for postnatal environmental factors associated with substance use during pregnancy related to childhood behavior problems. Other factors, including quality of the home environment,26, 27
and parenting stress29
associated with the caregiving environment of mothers who used substances during pregnancy were also examined, but these models did not meet statistical criteria or were weaker than models with SES.
We included teacher and caregiver report of behavior problems. In previous work, prenatal cocaine exposure has been related to behavior problems using teacher report.30, 31
In contrast, we averaged teacher and caregiver report because the combination of two independent reports of behavior in different settings might provide a more complete assessment than either report alone. This may explain the higher total variance explained in the models for the combined caregiver and teacher report than for the separate report models.
Our results demonstrate a logical sequence of cascading effects on developmental processes leading to behavior problems in children with exposure to cocaine and other substances. This pathway could explain some of the behavioral origins of neurobehavioral disinhibition in later childhood related to adolescent substance use.9–11
Neurobehavioral disinhibition includes many of the behavioral dimensions reminiscent of the behaviors measured in our study, including emotional liability, irritability, difficult temperament in infancy, and externalizing and internalizing behavior problems. In older children, these behaviors are subsumed under the broader categories of dysregulated emotion and behavior undercontrol. Neurobehavioral disinhibition is thought to be due to dysfunction of the prefrontal cortex and also includes deficits in executive function that were not measured in the current study. We need to continue to follow the children in our study to determine if this developmental pathway extends to the profile of neurobehavioral disinhibition and later substance use in adolescence. Nonetheless, we demonstrated that some precursors of neurobehavioral disinhibition have behavioral echoes in early infancy.
The limitation of SEM is that alternative models could be developed that fit the data as well as or better than the model developed in this study. Although we examined alternative models (e.g., using measures of the caregiving environment), it is possible that other caregiving environment measures would also result in adequate models. The model we tested was theoretically based, and we acknowledge it could be modified. In addition, we did not measure genetic influences that are potentially involved in the development of behavior problems.
Three of the four measures of child behavior were based on caregiver or teacher report and could have been strengthened with the addition of more objective measures. The path coefficients between the 3- and 7-year CBCL scores were stronger when caregiver report was used at both ages than when caregiver report was related to teacher report (), suggesting possible reporter bias. An alternative explanation is that the observations of caregivers and teachers reflect the different contexts in which they interact with the children, supporting the decision to average caregiver and teacher scores.
Prenatal substance exposure was a single factor based on the presence or absence of each substance examined, including nicotine, alcohol, marijuana, and cocaine. We tested models that included estimates for each individual substance and the amount of exposure to each substance; however, these models did not have adequate goodness of fit statistics. The alternative would have been to use cocaine only and acknowledge the presence of other substances, but we thought it was more accurate to include all substances in a single model. It is interesting that the path coefficients for the four substances in the latent substance factor () are similar in size, suggesting the contribution of each substance to the factor is similar.
Early identification and prevention of behavior problems is in line with recent recommendations by the American Academy of Pediatrics32
and would address an important public health need. Our findings suggest it is possible to identify infants at 1 month who are on a path leading to behavior problems in childhood. Intervention studies could be developed to determine if altering behaviors measured by the NNNS (i.e., reducing arousal and stress) could prevent the development or reduce the magnitude of later child behavior problems. Our findings are optimistic because although we may not be able to treat the children who show direct effects of prenatal substance exposure on childhood behavior problems, we were able to identify children who left behavioral tracks in infancy that may be amenable to treatment. In the long term, altering the developmental trajectory of these children could address more severe conduct problems and substance use disorders in adolescence.