A total of 1388 children were enrolled; 1022 (73.6%) had the CBCL administered at at least 1 visit and had known levels of PCE. Compared with the 366 (26.4%) children excluded from analysis, those included had higher birth weight (P = .048), were more likely to be discharged from the hospital with their mother (P < .001), had mothers of younger age (P = .036) who were more likely to be black (P < .001), and had higher levels of maternal education (P = .002). The proportions of high, some, and no PCE did not differ between subjects included versus not included in the analysis.
shows mean scores of internalizing, externalizing, total, and attention problems for each year according to the 4 exposure categories. The table also includes proportions or means of other prenatal exposures. Of mothers in the high PCE/OD group, 97% used another substance in addition to cocaine. Risk and protective factors were also compared among the exposure groups.
Four Categories of Prenatal Drug Exposure by Specific Drug Exposure, CBCL Scores, and Risk and Protective Factors
Children with high PCE/OD had significantly greater externalizing and total problem scores than those with PCE–/OD– (P < .05) after controlling for risk and protective factors (). Neither some PCE/OD nor PCE–/OD+ predicted the intercept or slope of any of the behavior problems. Male gender was associated with significantly more externalizing, total, and attention scores at 5 years (intercepts); in addition, male children had significantly lower slopes over time for externalizing and internalizing behaviors. Low verbal IQ was linked to higher initial scores on externalizing, total, and attention problems. Continuing caretaker use of tobacco or alcohol predicted externalizing and total behavior problems. Caretaker psychopathology had significant association with all outcomes, and depression predicted internalizing and total problems. Child abuse was a predictor of total and attention problems at 5 years (intercepts) and was also associated with increasing externalizing and total problem scores over time (slopes).
LGC Models of CBCL Scores According to Individual Risk and Protective Factors (Model 1)
High resilience predicted lower scores on all outcomes. Having many friends was associated with lower internalizing and attention problem scores and a decrease in scores at later ages for internalizing (B [SE] = –0.19 [0.090], P = .042) and total problems (B [SE] = –0.81 [0.32], P = .012). Availability of family resources was significantly associated with lower scores on all behavior outcomes except for externalizing problems. Caretaker involvement was significantly associated with decreases on all behavior scores over time.
When the cumulative risk and protective indexes for each subject replaced the individual factors in the LGC model (, model 2), high PCE/OD remained a significant predictor for externalizing problems. Some PCE/OD did not predict behavior outcomes. PCE–/OD+ was associated only with increasing internalizing behavior scores (slope) with increasing age. The risk index was a significant predictor of all behavior problems but did not predict significant changes in scores over time. The protective index predicted lower scores in all problems and a significant decrease in scores over time (slopes) for internalizing, total, and attention problems. For risk index, the coefficient (SE) for the intercept was 2.14 (0.20) (P < .001) for externalizing problems; for the protective index, the coefficient (SE) for the intercept was –0.88 (0.24) (P < .001). Therefore, for externalizing problems, an increase in number of risk factors by 1 would be associated with 2.14-point increase in externalizing scores, whereas an increase of 1 in the number of protective factors would be associated with a 0.88-point reduction in externalizing scores.
LGC Models of CBCL Scores According to Cumulative Risk and Protective Indexes (Models 2 and 3)
With the balance between risk and protective indexes in the final LGC models (, model 3), all drug-exposed groups predicted externalizing behavior, with the highest coefficient in high PCE/OD. Scores declined with age, but slopes did not differ from PCE–/OD–. High PCE/OD also predicted total problems. All drug exposure groups were associated with attention problems (intercepts); slopes were no different from PCE–/OD–.
The balance between risk and protective indexes was significantly associated with all behavior outcomes. A high-risk index in the presence of low protective index added to the already significant effects of drug exposure. Conversely, a high protective index in the presence of a high-risk index mitigated the predicted associated increased behavior scores from drug exposure. Among children with low-risk indexes, those who also had low protective indexes experienced smaller declines (slopes) in behavior problems over time compared with those with high protective indexes for internalizing (B [SE] = 0.29 [0.10], P = .004) and total scores (B [SE] = 1.00 [0.36], P = .005). In addition, although not statistically significant, similar trends were found for externalizing (B [SE] = 0.27 [0.15], P = .062) and attention (B [SE] = 0.12 [0.07], P = .065) problems (, model 3). Furthermore, when compared with children with low risk and high protective index scores, children with high risk and low protective index scores had smaller declines in total problem scores (B [SE] = 1.25 [0.47], P = .008) and attention problem scores (B [SE] = 0.23 [0.09], P = .008).
To illustrate the influence of risk and protective indexes on behavior problems, presents the LGCs of CBCL total problem scores over time according to levels of risk and protective indexes separately in the 4 exposure groups. In the high PCE/OD, the high risk–low protective group had significantly higher total problem scores at 5 years (intercept) than the other groups (P < .05). The groups with low protective scores generally had flatter curves than those with high protective scores, which seemed to decline over time. Although the differences in slopes between these groups were not significant when compared individually, there was a significant difference in slopes when comparing the 2 low protective groups with the 2 high protective groups (t = –2.16, P = .033), supporting this general trend based on level of protective influences. Among youths with some PCE/OD, 3 of the 4 groups differed by initial scores at 5 years (P < .05) and had similar slopes over time. However, although the remaining high risk–high protective group had scores at 5 years similar to the high risk–low protective group, they experienced significantly greater declines in problem behaviors over time (P = .039). Similar patterns of the influence of protective factors were observed among the 2 groups without PCE (). The groups with high protective index scores had significantly different slopes than those with low protective index scores; they declined over time while the low protective groups generally remained flat.
LGC trajectories of CBCL total problems according to prenatal drug exposure and risk and protective factors.