Overall, 2.06% of children surveyed met DSM-IV
criteria for CD in the past 12 months, equivalent to 560,000 U.S. children 8–15 years of age. This estimate is consistent with previous prevalence estimates that range from < 1.0% to 16.0% (Lahey et al. 2000
; Loeber et al. 2000
; Maughan et al. 2004
). Our analyses confirm prior observations that prenatal tobacco smoke exposure is associated with disruptive behavior disorders in children. We also found increases in the number of CD symptoms among children exposed to postnatal ETS. Finally, we found that lead exposure, measured using blood lead levels, was associated with increased odds of CD and increased CD symptom count in the past year.
Our results are the first to use a DSM-IV
–based instrument to assess conduct problems in a nationally representative sample of U.S. children. Previous work evaluated small to moderately sized case–control sets or prospective cohorts using behavior scales such as the CBCL (Needleman et al. 1996
; Wasserman et al. 2001
), self- or parent report of delinquent behavior (Dietrich et al. 2001
; Needleman et al. 1996
), or adjudicated case status (Needleman et al. 2002
). Estimates of the prevalence of CD may differ across studies because of variations in diagnostic instrument, informant, time period for assessing psychiatric status, and source population. As noted by Lahey et al. (1999)
, small changes in the diagnostic instrument can produce large changes in the prevalence. When criteria of the 3rd, revised edition of the DSM
were used, the prevalence of CD in three U.S.-based samples ranged from 1.2% to 16.0%, depending on the age and sex of the children. When DSM-IV
criteria were used, the prevalence of CD has been reported to be 1.3% for girls and 3.9% for boys (Loeber et al. 2000
). We did not find differences in the prevalence of CD diagnosis between boys (2.24%) and girls (1.86%) to be as large as previously reported (Loeber et al. 2000
). This may be a result of using parents as the informants of CD symptoms.
In this sample, children with prenatal tobacco smoke exposure had elevated odds of meeting DSM-IV
CD criteria, which is consistent with previous reports (Fergusson et al. 1998
; Wakschlag et al. 1997
; Weissman et al. 1999
). Using DSM-IV
criteria, Fergusson et al. (1998)
reported a 1.4- to 2.5-fold increase in CD symptom rate among children whose mothers smoked more than one pack of cigarettes per day during pregnancy, which is consistent with the increase in CD symptoms we observed among children exposed to prenatal tobacco smoke. However, unlike our study, Fergusson et al. (1998)
did not find an association between maternal smoking during pregnancy and CD diagnosis after controlling for confounding. The difference in our results may be attributable to differences in the confounders that were controlled for. Fergusson et al. (1998)
controlled for illicit drug and alcohol use during pregnancy, child-rearing practices, and family functioning. Our reported effect estimates may have been attenuated had we been able to adjust for these other confounders.
Children with increasing cotinine levels had increased odds of meeting DSM-IV
CD criteria and an increased prevalence of CD symptoms. This association was not a result of increased serum cotinine levels among actively smoking children because we excluded all children with serum cotinine levels indicative of active smoking (≥ 10 ng/mL). Our result is consistent with previous prospective cohort studies that have found similar increases in behavior problems (Weitzman et al. 1992
) and CD (Fergusson et al. 1993
) associated with postnatal ETS exposure. Fergusson et al. (1998)
reported that postnatal ETS exposure was associated with increased CD symptoms at 8, 10, and 12 years of age, using parent informants. Weitzman et al. (1992)
reported increased scores on the Behavior Problem Index of the CBCL among children whose mothers smoked only after pregnancy. Our study is the first to use an objective biomarker of tobacco smoke exposure to examine the association between postnatal ETS exposure and severe behavior problems among children.
Our results indicate that a substantial proportion of children are exposed to ETS outside of the home, leading to elevated serum cotinine levels. Previous studies have reported similar findings (Boyaci et al. 2006
; Cornelius et al. 2003
). Future studies would be well advised to use cotinine as a measure of ETS exposure in children given the high likelihood of exposure misclassification.
Children with blood lead levels ≥ 1.5 μg/dL had a 8.64-fold increased odds of having met DSM-IV
CD criteria in the past year compared with children with levels from 0.2 to 0.7 μg/dL. Our findings, which are consistent with prior research showing an increased risk of delinquency and criminality among children with higher bone or blood lead levels (Dietrich et al. 2001
; Needleman et al. 1996
; Wasserman et al. 2001
), provide evidence that contemporary children with considerably lower levels of lead exposure than those in previous studies remain at increased risk for CD. However, the results of our logistic regression models were very imprecise because of the small number of cases.
The relationship between environmental toxicant exposure and disruptive behavior disorders is not surprising given the wealth of animal literature showing adverse effects of nicotine and lead exposure on behavior (Ernst et al. 2001
; Lidsky and Schneider 2003
). It has been hypothesized that tobacco smoke exposure elicits its neurotoxic effects through two mechanisms: a
) fetal hypoxia as a result of carbon monoxide exposure and b
) the direct interaction of nicotine with the developing brain (Wakschlag et al. 2002
). Exposure to lead has been observed to cause changes in neurotransmitter concentrations and neurotransmitter receptor density (Lidsky and Schneider 2003
). Nicotine interacts with nicotinic acetylcholine receptors, which are present in the developing fetal brain. These receptors are involved in the modulation of neurotransmitters such as dopamine, serotonin, and γ-amino butyric acid. Thus, prenatal nicotine exposure may produce secondary effects through these other systems (Ernst et al. 2001
). Slotkin et al. (1987)
proposed that nicotine exposure is associated with regional abnormalities in cell number and macromolecule content in rats. In addition, he found that prenatal exposure to nicotine results in a premature switch from cell replication to cell differentiation. Animal models using rats and rhesus monkeys have shown that lead-exposed animals exhibit deficits in discrimination reversal, spatial delayed alternation, and fixed interval tasks (Rice 1996
). These deficits indicate impairment in the animals’ ability to inhibit inappropriate responses, temporally organize behavior, and learn from the consequences of previous actions. Recent work by Nigg et al. (2008)
among children with attention-deficit/hyperactivity disorder suggests that behavioral problems may be mediated by child IQ or poor cognitive control.
This study has several limitations that should be considered when interpreting our results. First, the cross-sectional nature of the data makes it difficult to infer causal relationships. The results of our study are consistent with previous birth cohorts that prospectively collected exposure information (Dietrich et al. 2001
; Fergusson et al. 1998
; Weitzman et al. 1992
). Concurrent blood lead levels may not be the optimal biomarker of a child’s risk for lead-associated behavior problems if lead induces neurotoxic effects during early development. However, recent studies indicate that concurrent blood lead level is a stronger predictor of lead-associated IQ decrements and behavior problems than is blood lead measured during early childhood (Chen et al. 2005
; Lanphear et al. 2005
). Another potential source of bias in cross-sectional data is exposure misclassification. Mothers of children with behavior problems may be more likely to recall gestational intake of potentially harmful substances, such as tobacco, owing to a drive to identify a cause of their child’s disorder. On the other hand, mothers may fail to report prenatal and postnatal tobacco smoke exposure because of social stigma (i.e., social desirability bias) or tobacco smoke exposures outside of the home. We minimized the possibility that postnatal ETS exposures were misclassified by using cotinine as a marker of exposure. Prior research indicates that mothers can accurately recall gestation intake of tobacco with sensitivities and specificities in the range of 0.81–0.86 and 0.94–0.97, respectively (Jacobson et al. 2002
; Tomeo et al. 1999
Another limitation to the NHANES data is that prenatal tobacco smoke exposure was collected only as a dichotomous variable, and we were unable to examine its relationship with CD in more than two categories. This would result in our effect estimate being biased toward the null if the effect of prenatal tobacco smoke exposure on CD diagnosis is greater at higher levels of prenatal tobacco consumption.
Although we were able to adjust for some confounders, we were unable to adjust for maternal education, family functioning, care-giving environment, parenting practices, prenatal alcohol use, and parental psychopathology. These factors tend to be associated with greater exposure to environmental toxicants and greater risk for CD. NHANES does collect data on maternal education and prena-tal alcohol use, but these data were not available in the publicuse NHANES data files. We did attempt to control for many confounders (or their proxies) in the relationship between environmental toxicants and CD, including race/ethnicity, maternal age at child’s birth, and socioeconomic status (PIR). Still, it is unlikely that these confounders would have been strong enough to eliminate our observed associations, given the previous literature showing robust effects even after controlling for numerous confounders.
Finally, the small number of exposed cases in our logistic regression model created imprecise estimates of the effect of lead exposure and limited our ability to adequately assess for an interaction between prenatal tobacco smoke exposure and blood lead levels. Still, the consistency of our results using the CD symptom counts suggests that exposure to environmental toxins may result in a shift of the symptom count distribution that would result in an increased number of CD-diagnosed children.
The reported prevalence of CD may be an underestimate of the true national prevalence because we relied on parents as the informants of CD symptoms (Loeber et al. 2000
; Shaffer et al. 2000
). Many disruptive or delinquent behaviors in children are not recognized by their parents. Although maternal and child report of CD symptoms are correlated (Burt et al. 2005
), some studies found that child informants were twice as likely to meet the diagnostic criteria for CD compared with parent or caregiver informants (Ezpeleta et al. 1997
This study confirms the previously observed associations between prenatal tobacco smoke exposure and CD. In addition, this study provides support that elevated blood lead levels are a risk factor for CD. Future research should be directed at confirming this observation in prospective birth cohorts, preferably using serial biomarkers of prenatal tobacco smoke exposure and environmental lead exposure. Despite dramatic reductions in children’s exposures to tobacco smoke and environmental lead, these results suggest that millions of contemporary children may be exposed to levels of these toxicants sufficient to increase the risk for persistent, disruptive, and even violent behavior problems.