3.1. Sample characteristics
We enrolled 1388 maternal-child dyads in the follow-up phase of MLS. 1028 (74.1%) had both the TRF and CBCL completed at least one time from 7 to 11 years of age. compares the mother-child dyads with and without completed TRF and CBCL. Of those children with evaluation of behavior problems, 79.9% were black compared to 67.2% of those with no assessment (p<0.001). The proportions of PCE, high PCE and opiate exposure were similar between those with and without behavior assessments. Mothers in each group did not differ as to age at the child's birth and in the use of other drugs during pregnancy as to the average number of cigarettes/d, absolute alcohol (oz/d) and number of joints/d of marijuana. Mean birth weight and gestational age did not differ between those children with or without behavioral assessment. A higher percentage of children were discharged from the nursery to other than their biological mother among those without behavior evaluation, 11.7%, than in those with assessment, 8.1%.
Comparison of dyads with no follow-up and those with child's assessment of behavior problems at least one time from 7 to 11 years of age
We excluded from analysis children whose mothers admitted to cocaine use in Phase 1 (n=66) but denied use during the detailed interview on enrollment into the follow-up phase, resulting in a total of 962 children included in the analyses. Of the 962 children, 884 (91.9%) had CBCL and/or TRF at 7 years, 905 (94.1%) at 9 years, and 878 (91.3%) at 11 years. The children did not have the TRF-form administered at 13 years but 834 (86.7%) had the CBCL at this age. A total of 730 (75.9%) had measurements for all four years.
3.2. Comparison of teacher and parent ratings
presents the means and standard deviations of the T scores for the outcome measures: externalizing, internalizing, total behavior, and attention problems for TRF and CBCL. Teacher's scores were significantly higher for all scores except for attention problems at 9 years (p=0.803).
Mean parent and teacher ratings (T scores) by age of assessment.
3.3. Latent growth curve modeling
The initial models included levels of prenatal exposures (PCE, tobacco, alcohol, and marijuana), prenatal opiate, maternal age, child's gender, SES, postnatal exposures (tobacco, marijuana, alcohol), BDI, BSI, HOME, family violence, neighborhood scales, and survey of community violence as predictors of intercept and slope of behavior problems observed by the child's teacher and the parent/caretaker. Some variables in the initial modeling did not reach statistical significance at p value <0.10 (postnatal marijuana, BSI, neighborhood scales) and were therefore, dropped from the model. HOME being a direct observation measure, was retained in preference to SES. Reducing the number of predictors improved the parsimony of the latent growth curve models for behavior problems.
shows the parameter estimates from the final latent growth curve models investigating the relationship of prenatal cocaine exposure and other covariates to the intercept and slope of the T scores from the TRF and the CBCL for externalizing, internalizing, total behavior problems, and the syndrome of attention problems. All four models fit well: Externalizing (CFI=0.97, TLI=0.93, RMSEA=0.04), Internalizing (CFI=0.96, TLI=0.91, RMSEA=0.04), Total (CFI=0.97, TLI=0.93, RMSEA=0.04), and Attention (CFI=0.98, TLI=0.96, RMSEA=0.03).
Parameter Estimates from Latent Growth Curve Model of Change in Behavior From Parent and Teacher Ratings
3.4. Relationship between parent and teacher reports
As shown in , the T scores derived from teacher's report were significantly related at baseline (7 years) to ratings by the parent/caretaker for all behavior problems, PCE not withstanding, (parent intercept ↔teacher intercept: externalizing problem, p<0.001; internalizing problem, p=0.004; total behavior problems, p<0.001; attention problems, p<0.001). The findings suggest that children who were rated poorly by their teachers were also rated poorly by their parents/caretakers or vice versa. However, as noted earlier in , the scores from teacher report were significantly higher than T scores from the parent CBCL.
There were significant relationships between the changes in scores over time for externalizing, internalizing, total problems, and attention problems between the CBCL and TRF (parent slope ↔ teacher slope: externalizing, p<0.001; internalizing problem, p=0.005; total problems, p<0.001; attention problems, p<0.001).
3.5. PCE and trajectories of behavior problems
shows the estimated growth curves by PCE. After control for covariates, high PCE predicted the intercept for both the TRF and the CBCL externalizing behavior problems, p<0.034 and p=0.021, respectively. This relationship signifies that externalizing problem scores were higher in children with PCE at 7 years in comparison to non-exposed children, whether observed by the teacher at school or by parent or caretaker at home. PCE, however, did not effect a significant change in externalizing behavior problem scores over time (as indicated by the slope) from observations made by either the teacher or parent/caretaker. The externalizing problem scores remained higher on subsequent years in those exposed to high levels of PCE than those with no PCE.
Estimated linear growth curves for behavior problem T scores from TRF (7, 9, and 11 years) and CBCL (7, 9, 11, and 13 years) by prenatal cocaine exposure.
High PCE did not predict internalizing behavior problems at 7 years from the teacher or parent/caretaker report. PCE, however, marginally predicted total behavior problems at 7 years (intercept) from the TRF (p=0.063 and p=0.063), respectively for high and some PCE) but was not a predictor of total problems from the CBCL. PCE did not predict the slopes of total behavior problems from TRF or CBCL, suggesting that while children with PCE had more behavioral problems initially, they followed similar trajectories over time, remaining higher than those with no PCE at each time point. High PCE only marginally predicted the parent/caretaker intercept for attention problems. Neither was PCE a significant predictor of the change in attention problems over time (i.e., slopes) from either teacher or parent ratings.
3.6. Prenatal opiate and behavior problems
Prenatal opiate was neither a predictor of teacher nor parent intercepts for outcomes externalizing, internalizing, total behavior, and attention problems. However, there was a significant relationship between prenatal opiate exposure and the slopes of the T scores for internalizing and total problems from the parent report (both p values <0.02), and attention problems from both TRF (p=0.012) and CBCL (p=0.015). Therefore, although we found no differences between those with and without prenatal opiate exposure at 7 years of age, the opiate exposed children had significant increase in T scores for internalizing, total, and attention problems over time (slopes) in comparison to children with no opiate exposure.
3.7. Other predictors of behavior problems as reported by parent/caretaker
The other significant predictors of CBCL externalizing behavior intercept included postnatal tobacco exposure, caretaker depression, and the child's exposure to community violence. Also, postnatal tobacco, caretaker depression, and community violence predicted the intercepts of internalizing, total, and attention problems from the CBCL. Postnatal alcohol use also predicted internalizing behavior at 7 years. Male gender was a significant predictor of the intercepts of internalizing and total behavior problems by the CBCL. But the only common significant predictor of the CBCL slope for all the behavior problems, including attention problems, was BDI. The slope indicated a minimal decrease in rate of change in behavior problems' T-scores, as the children became older. Given that BDI was also strongly related to behaviors at 7 years based on parent reports (i.e., parent intercept), the smaller change in problem behaviors over time for parents who are depressed (i.e., smaller parent slopes) may be indicative of a ceiling effect.
3.8. Other predictors of behavior problems from the teacher report
HOME and community violence were significant predictors while young maternal age and male gender were marginal predictors of the intercept of the externalizing behavior problem from the teacher report (p=0.088 and p=0.070, respectively). HOME and male gender also significantly predicted the intercepts of internalizing, total, and attention problems from the TRF. Postnatal alcohol was a marginal predictor of total problems from the TRF. Only postnatal alcohol exposure predicted the slope of attention problems aside from prenatal opiate exposure.