Level of Prenatal Cocaine Exposure and Scores on the Bayley Scales of Infant Development: Modifying Effects of Caregiver, Early Intervention, and Birth Weight
Objectives. The objectives of this study were 1) to assess whether there is an independent association between the level of prenatal cocaine exposure and infants’ developmental test scores after control of potential confounding variables; and 2) if such an association exists, to determine which biological and social variables, individually and in interaction with each other, may modify it.
Methods. In a prospective, longitudinal study of 203 urban term infants, 3 cocaine exposure groups were defined by maternal report and infant meconium assay: unexposed, heavier cocaine exposure (>75th percentile self-reported days of use or meconium benzoylecognine concentration), or lighter cocaine exposure (all others). Examiners, masked to exposure history, tested infants at 6, 12, and 24 months of age with the Bayley Scales of Infant Development.
Results. The final mixed linear regression model included as fixed covariates level of prenatal exposure to cocaine, alcohol, and cigarettes; prenatal marijuana exposure; gestational age and birth weight z score for gestational age; and gender. Age at test, caregiver at time of each test (biological mother, kinship caregiver, unrelated foster caregiver), and any previous child-focused early intervention were included as time-dependent covariates. There were no significant adverse main effects of level of cocaine exposure on Mental Development Index (MDI), Psychomotor Development Index (PDI), or Infant Behavior Record. Child-focused early intervention interacted with level of cocaine exposure such that heavily exposed children who received such intervention showed higher adjusted mean MDI scores than all other groups. Although the sample was born at or near term, there was also a significant interaction of cocaine exposure and gestational age on MDI scores, with those in the heavier exposure group born at slightly lower gestational age having higher mean MDI scores compared with other children born at that gestational age.
There was also a significant interaction on MDI between child’s age and caregiver. At 6 months, the adjusted MDI of children living with a kinship caregiver was 15.5 points lower than that of children living with their biological mother, but this effect was diminished and was no longer significant at 24 months (difference in means: 4.3 points). The adjusted mean MDI of children in unrelated foster care at 6 months was 8.2 points lower than children of biological mothers, whereas it was 7.3 points higher at 24 months.
Early intervention attenuated the age-related decline in PDI scores for all groups. Birth weight <10th percentile was associated with lower PDI scores for children with heavier cocaine exposure and with lower MDI scores for all groups.
Conclusions. Heavier prenatal cocaine exposure is not an independent risk factor for depressed scores on the Bayley Scales of Infant Development up to 24 months of age when term infants are compared with lighter exposed or unexposed infants of the same demographic background. Cocaine-exposed infants with birth weight below the 10th percentile for gestational age and gender and those placed with kinship caregivers are at increased risk for less optimal developmental outcomes. Pediatric clinicians should refer cocaine-exposed children to the child-focused developmental interventions available for all children at developmental risk.
At the beginning of the cocaine epidemic, many predicted that children exposed to cocaine in utero would show lasting developmental impairments. However, peer-reviewed research does not find striking or consistent effects of prenatal cocaine exposure on developmental test scores in the first 2 years of life. Of the 11 studies using masked examiners to evaluate the impact of prenatal cocaine exposure on scores on the Bayley Scales of Infant Development (BSID),1,2 6 found no statistically significant effect,3–8 including 1 study that classified infants according to mothers’ report of level of prenatal exposure.7 Among the 5 studies that did find adverse effects of prenatal cocaine exposure, effects differ by age of assessment and whether exposure to tobacco, alcohol, and marijuana, and other potential confounds were included in the analysis. In a sample of infants whose mothers entered drug treatment during pregnancy, the mean BSID Mental Developmental Index (MDI) and Psychomotor Developmental Index (PDI) scores of infants whose mothers used cocaine/alcohol/marijuana were lower at 6 months than infants whose mothers did not use these drugs, but identical to those of mothers who had used alcohol/marijuana without cocaine. These results suggest that there is no incremental detrimental impact of cocaine use on development quotients beyond that imposed by the concurrent use of alcohol/marijuana.9 No deleterious cocaine effects were detected in this sample on BSID scores at other assessment ages up to 24 months.9 Other investigators found a bivariate association of lower PDI scores at 3 months with prenatal cocaine exposure, but not after statistical control for potential confounds.10 In a sample confined to very low birth weight infants tested at ∼16 months’ corrected age, a negative association with prenatal cocaine exposure was found with MDI and PDI, but in utero exposure to other psychoactive substances was not analytically controlled.11 A study that used the recently revised BSID II2 found no main effects of mothers’ reported level of prenatal cocaine exposure on MDI or PDI at 8 or 18 months.12 In posthoc comparisons, children whose mothers reported higher levels of prenatal cocaine use (2 or more days per week) were found to have significantly lower MDI scores at 18 months than those with no exposure.12 For infants who were unexposed or lightly exposed, higher environmental risk was also associated with a greater decline in MDI scores from 8 to 18 months.12 Singer et al13 found in a large sample a statistically significant decrement in BSID II MDI scores associated with prenatal cocaine exposure, with an increased risk of developmental delay among the exposed. This study contained a substantial number of very low birth weight infants who are excluded from other samples and did not measure whether subjects received developmental intervention.
When adaptive behavior during the testing situation was measured by the Infant Behavior Record (IBR) of the BSID, 2 investigators found cocaine-exposed infants more listless than unexposed infants in bivariate analyses, but not after covariate control.3,11
Contextual factors may partially account for these discrepancies.14 Prenatal cocaine exposure may exert differing impacts on later developmental outcomes depending on the interaction of exposure with other biological and social factors. Discrepant findings are also susceptible to multiple possible methodologic explanations, including differences in study design, varying patterns of sample attrition, misclassification of exposed infants as unexposed because of inaccurate maternal report, and failure to address potential dose effects.15 If exposed infants are misclassified as unexposed, between-group differences may be missed. If lightly and heavily exposed infants are aggregated into a single category, cocaine effects that may occur only at higher or lower levels of prenatal exposure may be obscured.7
Testing meconium for drug metabolites presents a method of assessing prenatal cocaine exposure that enhances accurate identification of newborns exposed to cocaine prenatally, detecting more exposed infants than mothers’ report or peripartum urine assays from mothers or infants.16–19 Although not a precise reflection of the total grams of cocaine used by a mother during pregnancy or the gestational timing of use, the concentration of cocaine metabolites in meconium allows newborns to be rank ordered by their relative level of cocaine exposure. This rank order correlates with newborns’ birth weight, behavior, and risk of subependymal hemorrhage.20–23
The first goal of this analysis was to assess whether, after control of potential confounding variables, there is an independent association between the level of prenatal cocaine exposure (assessed by mothers’ self-report and/or assay of infants’ meconium) and infants’ scores on the BSID at 6, 12, and 24 months of age. The second goal was to determine which biological and social variables, individually and in interaction with each other, may moderate such an association if one is found.
Sample Selection Criteria
The Human Studies Committees of Boston City Hospital (now Boston Medical Center) and the Boston University School of Medicine approved this study. All participants gave informed consent, including the risk of incidental detection and reporting of child maltreatment during study assessments. A writ of confidentiality was obtained from the federal government to protect participants from having research data subpoenaed. The sample was recruited by trained interviewer/recruiters who screened maternity and nursery records 7 days a week on the postpartum floor of Boston City Hospital from October 1990 to March 1993. Unexposed dyads comparable to cocaine-exposed mother-infant dyads in ethnicity (African American/African Caribbean vs other) were preferentially approached for recruitment soon after delivery. All mother-infant dyads met the following criteria based on review of mother and infant medical records and confirmed by interviews, biological markers, and infant physical examinations obtained by study personnel: 1) Infant gestational age greater than or equal to 36 weeks; 2) No requirement for neonatal intensive care; 3) No obvious major congenital malformations; 4) No diagnosis of fetal alcohol syndrome in the neonatal record; 5) No history of human immunodeficiency virus seropositivity noted in the mother’s or infant’s medical record; 6) Mother’s ability to communicate fluently in English; 7) No indication by neonatal or maternal urine toxic screen or history in medical record of mother’s use during pregnancy of illegal opiates, methadone, amphetamines, phencyclidine, barbiturates, or hallucinogens; and 8) Mother aged 18 years or older. These criteria were established to exclude infants with known major risk factors that might confound or obscure the effects, if any, of in utero cocaine exposure. The sample was restricted at recruitment to mothers with English fluency, because many of the neuropsychological measures planned for this cohort at preschool and older ages are not standardized for populations whose first language is not English. Additional details about sample characteristics and recruitment are reported elsewhere.21,23
Method of Exposure Classification
Mothers participating in the study were identified as either heavier, lighter, or nonusers of cocaine by interview and by biological markers obtained by clinicians and study personnel. At intake on the postpartum floor, research assistants using the Addiction Severity Index24 supplemented by study-specific questions interviewed the mothers about pregnancy and lifetime use of cigarettes, alcohol, and illicit drugs. We sought to collect meconium specimens from all enrolled infants to be analyzed by radioimmunoassay for benzoylecognine (a cocaine metabolite), opiates, amphetamines, benzodiazepines, and cannabinoids. The radioimmunoassay used was a modification of the method of Ostrea et al,17,19 published in detail elsewhere.
During the period of study recruitment at Boston City Hospital, urine testing for metabolites of illicit drugs was performed for clinical indications at the discretion of health care personnel, but was not universal. We documented the results (when available in the medical record) of the urine drug enzyme multiplied immunoassay technique assays obtained for clinical purposes during prenatal care or labor and delivery from mother or from the infant after birth. Although cases were targeted for recruitment on the basis of self-report or positive clinical urine assays, comparison mothers were drawn from the maternity population as a whole, most of whom did not have urine assays performed for drug metabolites for clinical purposes. Therefore, after recruitment and informed consent, we collected additional urine samples from all study mothers for analysis for benzoylecognine, opiates, amphetamines, benzodiazepines, and cannabinoids by radioimmunoassy using commercial kits (Abuscreen RIA, Roche Diagnostics Systems, Inc,, Montclair, NJ).
Subject Exposure Classification
All mother-infant dyads had at least 1 biological marker, either urine from mother or infant or meconium, which confirmed their exposure or lack of exposure to cocaine during pregnancy. In this sample, the mean days of self-reported cocaine use during pregnancy was 20.6 days, with a range from 0 to 264. The mean meconium concentration of benzoylecognine/g was 1143 ng with a range from 0 ng to 17 950 ng per g. Before data were analyzed, a composite measure of “heavier” use was a priori defined as the top quartile of meconium concentration for cocaine metabolites (>3314 ng of benzoylecognine/g meconium) and/or top quartile days of self-reported use (>61 days) during the entire pregnancy. All other use was classified as “lighter.”24 This ordinal classification scheme is comparable to that used by other investigators,7,12,25 where use of cocaine more than twice a week during pregnancy is considered “heavier” use.
Pragmatic as well as scientific considerations influenced this definition of exposure level. Because women are more likely to underreport rather than overreport illicit substance use during pregnancy,17,26 we decided a priori that women reporting days of use in the top quartile should be considered heavier users, even if the benzoylecognine levels in meconium were not in the top quartile. Not all infants exposed to cocaine in utero have positive meconium assays.18 Moreover, we were not able to obtain meconium samples from 14% of study infants, whose exposure status was confirmed by maternal or infant urine assay. Therefore, whichever indicator (self-report or meconium assay) demonstrated higher exposure was used to define exposure category.
Outcome and Control Variables
The raw scores of the BSID generate the MDI and a PDI, both standardized to have a mean of 100 and a standard deviation of 16. These standard scores are supplemented by the IBR, a group of descriptive rating scales of infants’ behavior during testing, including interpersonal and affective domains, motivational variables, and interest in specific modes of sensory experience. For the analyses in this study, 3 summary scores (task orientation, extroversion, activity), were generated from the IBR rating according to the method of Matheny.27
In 1993, the BSID was renormed and standardized as the BSID II.2 Because this longitudinal study began before the availability of the new version, we continued to use the original version of the BSID for all assessments to facilitate uniform interpretation.
Ongoing interrater reliability checks on the scoring of the BSID (agreement of pass/fail on test items within the MDI and PDI) averaged above 90% (range: 81%–100%) agreement. At all ages the BSID was performed by examiners masked to the children’s exposure status.
Potential control variables were selected on theoretical grounds at the start of the study and included infant anthropometric characteristics and health parameters repeatedly measured from birth to age 2 years. Within 8 to 72 hours of birth (mean: 48 hours), a study pediatrician, trained to reliability and unaware of the infants’ cocaine exposure, assessed gestational age according the method of Dubowitz et al28 and measured recumbent length on a Holtain Infantometer (Holtain Ltd, Crymych Pembs, United Kingdom) and head circumference with a plastic-coated tape. The pediatrician also reviewed the infant’s medical record for birth weight (measured after delivery by nursery nurses using a Detecto scale [Jericho, NY]) and neonatal medical complications using a list adapted from the work of Hobel.29 Child health data for the first 2 years of life were abstracted in a standard format from caregiver reports and medical records.
At the time of each developmental assessment, trained interviewers unaware of the child’s developmental status administered a follow-up version of the Addiction Severity Index to the caregiver, as well as Saranson’s (1978) Life Experience Survey,30 Norbeck’s (1981) Social Support Questionnaire31 and the Center for Epidemiologic Studies Depression Scale.32 The interviewers also elicited information regarding caregivers’ exposure to violence using a questionnaire developed for the primary care pediatric clinic at Boston City Hospital.33 To characterize from the child’s perspective the social environment over the first 2 years of life, scores on each instrument were averaged, even if different caregivers responded at different ages. To enhance the accuracy of self-reporting of postpartum use of illegal substances,34 respondents’ urine was collected after each interview and assayed by radioimmunoassay for metabolites of cocaine and marijuana. The interviews and urine assays were obtained for research purposes only and were not recorded in the caregiver’s or the child’s medical record. At each assessment, caregivers reported whether the child had received any child-focused developmental intervention since the previous interview. These included a range of intervention services from professionals and paraprofessionals, with delivery models including formal early intervention programs, home health services, parent-child groups, and/or services from individual clinicians such as occupational, speech, and physical therapists.
In addition to completing interviews at the hospital, caregivers of 84% of the unexposed, 83% of the lightly exposed, and 82% of the heavily exposed children (P = .64) permitted at least 1 visit by a research home visitor. These home visitors were masked to the child’s exposure status and developmental data, and performed the Home Observation for Measurement of the Environment (HOME)35 and Nursing Child Assessment Teaching Scale (NCATS).36 After each laboratory or home visit, caregivers received store vouchers valued at $25, and children received an age-appropriate toy.
To assess possible confounding, we performed bivariate analyses to compare the groups at each level of cocaine exposure on the control variables using χ2 tests for categorical variables and 1-factor analysis of variance for measurement variables.
Bivariate analyses were also used to compare the 3 cocaine exposure groups on the BSID MDI and PDI and on the 3 summary scores of the IBR at 6, 12, and 24 months infant age, using 1-factor analysis of variance.
In multivariate analyses, we used mixed linear regression models to analyze repeated observations from each child, which were included in a single analysis for each outcome (MDI, PDI, IBR) by specifying a within-subject correlation structure. This technique maximizes statistical power, allowing the detection of subtle effects in a longitudinal data set with data missing at some points because of subject attrition or noncompliance. These models were also used to allow for time-dependent covariates, such as child’s foster placement or participation in early intervention programs, which may differ at each assessment age. In each mixed model, we first determined a main effects-only model (not shown) that included potential confounding variables. We then examined all possible 2-way interactions between the independent variables in the final model. Interactions that were significant at .05 level were retained in the final model. Adjusted means and effects of level of cocaine exposure or other predictors on BSID scores are shown computed from the estimated parameters from these regression models (Tables 1 and 2). P values noted in the text and not shown in tables were also based on parameter estimates and their standard errors. For interaction effects, adjusted means and effects were based on contrasts involving 2 or more parameter estimates from the model.
In fitting these models, we used different, plausible working correlation structures and compared their computed values for Akaike’s Information Criterion. For each outcome, we selected the working structure that yielded the Akaike’s Information Criterion closest to zero.37 We denote results as statistically significant if 2-tailed P values were less than .05.
The choice of covariates for the final models presented in this study reflect both theoretical and analytic considerations. Prenatal exposure to cigarettes, marijuana, and alcohol were included as covariates in the final model because in human samples, cocaine use rarely occurs without concomitant use of 1 or more of these substances.18 In these models, prenatal cigarette use was measured as the average daily number of cigarettes and the prenatal alcohol use measured in average daily volume. Prenatal marijuana use took the form of a binary variable (“yes”/“no”) based on positive results of urine assay, meconium assay, or self-report. To address potential multicollinearity among these variables characterizing prenatal substance use, we performed formal collinearity diagnostics by the methods of Belsey, Kuh, and Welsch.38 These diagnostics indicated no presence of multicollinearity that would degrade the estimates obtained from our models.
However, caregivers’ postpartum use of alcohol and cigarettes (by self-report), and of marijuana and cocaine (by self-report or urine assay), was so highly correlated with prenatal use that the effects, when prenatal use was excluded from the analysis, were similar to that of prenatal use, and thus could not be included in the same regression model. (Analyses available from author on request.)
Other covariates were retained in the final model if their inclusion altered the unadjusted association between level of cocaine exposure and outcome by >10%. The variables that were tested included infants’ biological characteristics at birth and through the first 2 years of life, and biological mothers’ and other caregivers’ demographic, medical, and psychosocial characteristics, summarized in Tables 3–5. Where relevant, we modeled each of these variables as a time-dependent covariate, eg, HOME score at 5, 11, or 23 months.
We found no significant differences between the 203 children who are the subjects of these analyses and the 49 infants enrolled at birth who did not complete Bayley testing on level of cocaine exposure (P = .91) and other baseline variables measured at intake: birth weight (P = .24); gender (P = .35); gestational age (P = .12); maternal age (P = .79); maternal education (P = .57); ethnicity (P = .11); primiparity (P = .57); and prenatal alcohol (P = .22); cigarette (P = .44); and marijuana exposure (P = .39) via 2-sample t tests and χ2 tests.
Table 3 shows sample characteristics for the 3 prenatal exposure groups. Maternal education did not differ significantly between groups. However, nonusers were significantly younger than users and tended to be primiparas. African American/African Caribbean mothers were disproportionately found in the noncocaine-using or heavier cocaine-using groups, whereas Hispanic/white mothers were more likely to be lighter users. Heavier cocaine use was associated with heavier use of tobacco, alcohol, and marijuana during pregnancy. As summarized in Table 3, the infants in the 3 exposure groups did not differ in gender or in rates of depressed Apgar scores (<7) at 5 minutes. Although the sample was restricted at recruitment to infants born at ≥36 weeks, heavier cocaine use was associated with shorter infant gestation and lower birth weight, length, and head circumference z scores adjusted for gender and gestational age (based on norms calculated from the data file compiled by the National Center for Health Statistics for natality in the United States in 1991).39 In the sample retained for BSID assessment, as in the neonatal sample, heavier cocaine exposure was associated with significantly increased risk of subependymal germinal matrix hemorrhage by neonatal ultrasound and a trend toward less optimal state regulation at 3 weeks of age on the Neonatal Behavioral Assessment Scale.21,23,40 Despite these differences in medical risk at birth, the 3 groups did not differ in mean z score for weight, length, or head circumference at subsequent measurements (data not shown, available from authors on request), or in rates of anemia, elevated lead levels, or recurrent otitis in the first 2 years of life. Children in the “lighter” cocaine exposure category were more likely than the unexposed to have been hospitalized at least once in the first 2 years of life.
Table 5 also shows that in contrast to medical risks, social risks other than homelessness were associated with level of cocaine exposure. Infants with heavier cocaine exposure experienced the greatest number of changes in caregivers and were the most likely to be placed out of biological mother’s custody and with unrelated foster parents.
Table 5 demonstrates that when scores were averaged over the first 2 years of the child’s life, caregivers in the 3 exposure groups did not differ on depressive symptoms, reported negative life events, exposure to violence, or living with a male partner. However, caregivers of cocaine-exposed children used significantly more cocaine, cigarettes, and alcohol during the first 2 years of the child’s life compared with the caregivers of nonexposed children. Caregivers of infants with heavier cocaine exposure were the most likely to use marijuana in the postpartum period. The caregivers of the children with lighter cocaine exposure perceived the lowest levels of social support. As Table 5 shows, children with heavier cocaine exposure were more likely to receive developmental intervention before their first birthday than those with lighter exposure or no exposure (P = .08). The 3 groups did not differ significantly in the receipt of day care in the first 24 months of life. The mothers of the more heavily cocaine-exposed infants were most likely to receive residential drug treatment. Among the 169 children with at least 1 HOME/NCATS36 assessment, there were no significant differences by level of cocaine exposure for average (shown in Table 5), minimum, or maximum HOME or NCATS scores.
Effect of Cocaine on MDI and PDI in Bivariate Analyses
Table 6 summarizes the bivariate relationship between level of cocaine exposure and MDI and PDI scores at each assessment interval. Numbers assessed at each interval vary. No significant differences were observed across cocaine exposure groups at any assessment age for either MDI or PDI.
In Tables 1 and 2, we present adjusted means from the final models for MDI and PDI, including the variables that altered cocaine’s relationship to these outcomes. The final mixed linear regression model included as fixed covariates, level of prenatal exposure to cocaine, alcohol, and cigarettes, prenatal marijuana exposure, birth weight z score, gestational age, and gender. Age at test, caregiver at time of each test (biological mother, kinship caregiver, unrelated foster caregiver), and any previous child-focused early intervention were included in the analysis as time-dependent covariates. Although average psychosocial scores are presented in Table 5, we also tested caregivers’ depressive symptoms and life stress as time-dependent covariates. These time-dependent covariates did not achieve statistical significance, nor demonstrate additional confounding or interaction effects and so were excluded from the final model.
Effect of Cocaine on MDI in Multivariate Analyses
Multivariate analysis showed an apparently paradoxical positive effect of level of cocaine exposure on MDI (Table 1). This effect was explained by a significant interaction between cocaine and early develop-mental intervention (P = .03). Among children with no early intervention, there were no significant differences in adjusted MDI by cocaine exposure. There was no significant effect of early intervention within the groups of unexposed and infants with lighter cocaine exposure, but children with heavier exposure who received early intervention had significantly higher adjusted scores than all other groups.
Multivariate analysis showed additional effects of birth weight, age, early developmental intervention, and caregiver on MDI, regardless of cocaine exposure. Infants with birth weight at the 10th percentile for gender and gestation, the conventional cutoff for “small for dates” had lower adjusted MDI scores (by 2.5 points; P = .02) than those at the 50th percentile. Although the sample was all born at or near term, there was also a significant interaction of cocaine exposure and gestational age on MDI scores (P = .006) with those in the heavier exposure group born at slightly lower gestational age having higher mean MDI scores compared with other children born at that gestational age.
MDI scores declined with age. The extent of the decline varied by type of caregiver. Adjusting for level of prenatal exposure to cocaine and other substances, birth weight z score, and gestational age, there was also a significant interaction (P = .008) between child’s age and caregiver on MDI. For children living with their biological mother or in kinship care, adjusted MDI scores dropped significantly from 6 months to 24 months; for children in unrelated foster care, the decline in scores from 6 to 24 months was not statistically significant.
The caretaker by age interaction reported above can also be interpreted in terms of differences in adjusted mean MDI by caregiver (biological mother, kinship care, and unrelated foster care). Children in kinship care had lower adjusted mean MDI scores than children in the care of their biological mothers at both 6 and 24 months. In contrast, children in unrelated foster care had lower adjusted mean MDI scores at 6 months, but had higher adjusted mean MDI than the children with their biological mothers at 24 months. At 6 months, the adjusted MDI of children living with a kinship caregiver was 15.5 points lower than that of children living with their biological mother (P = .0003), but this effect was diminished and was no longer significant at 24 months (difference in means = 4.3 points; P = .18). The adjusted mean MDI of children in unrelated foster care at 6 months was 8.2 points lower than children of biological mothers (P = .06), whereas it was 7.3 points higher at 24 months (P = .11).
Effect of Cocaine on PDI in Multivariate Analyses
For PDI, multivariate analysis showed no significant main effects of prenatal cocaine exposure. However, a significant interaction between cocaine exposure and birth weight z score (P = .02; Table 2) was found. For example, for infants at the tenth percentile in birth weight, more heavily exposed infants had an adjusted mean PDI that was comparable to those observed for the lighter and unexposed groups of infants. For infants with birth weight at the 50th percentile, however, more heavily cocaine-exposed infants had higher adjusted mean PDI scores than unexposed by 7.1 points and more lightly exposed infants by 6.8 points.
Multivariate analysis showed additional effects of age, early developmental intervention, caregiver, and gestational age on PDI, controlling for prenatal exposure to cocaine and other psychoactive substances. There was a significant interaction between age and early developmental intervention (P = .04). Intervention attenuated the age-related deterioration in PDI scores. For children not receiving intervention, there was a significant 3.8-point drop in adjusted PDI scores from 6 to 24 months (P = .01), whereas for infants receiving intervention there was a 9.6-point increase over the same period (P = .08).
A significant interaction between caregiver and early developmental intervention on adjusted PDI (P = .01) was found. Children living in kinship care and receiving intervention had adjusted PDI scores 14.7 points lower than those not receiving intervention (P = .005). No such significant effect was found for those in biological mothers’ or nonkinship care.
Effect of Cocaine on IBR Scores in Bivariate and Multivariate Analyses
In addition to BSID MDI and PDI, we examined summary scores calculated from the IBR as dependent variables. There was no bivariate relationship between the level of cocaine exposure and the child’s scores on the extroversion, task orientation, or activity level clusters27 of the IBR at 6, 12, or 24 months. In multivariate analyses that included the same variables that were controlled in the multivariate analyses of the MDI and PDI, no significant cocaine effect was identified. (Analyses not shown, but available from the authors on request.)
Effects of Alcohol, Marijuana, and Cigarettes on MDI and PDI in Multivariate Analyses
No significant independent effects of prenatal alcohol, marijuana, or cigarette use on MDI scores were found. In contrast, the adjusted PDI showed a significant paradoxical relationship with prenatal cigarette exposure (P = .007). For example, children whose mothers smoked a pack of cigarettes per day had higher adjusted mean PDI scores by 6.4 points than children of nonsmokers.
This study is one of the first to utilize assays of neonatal meconium for cocaine metabolites in addition to mothers’ self-report to evaluate whether there is a dose-response relationship of level of prenatal cocaine exposure with scores on the BSID from 6 to 24 months. Although adverse dose effects of prenatal cocaine exposure were observed in the neonatal period in this sample,22,24 in bivariate and multivariate analyses, no adverse independent effects of level of cocaine exposure were found on the BSID MDI, PDI, or IBR from 6 to 24 months. However, we observed a number of important interaction effects of level of cocaine exposure with biological and social factors. Three biological factors—older age at test, gestational age at birth, and lower birth weight z score—were associated with less optimal developmental performance. Developmental test scores declined with increasing age at time of testing for both cocaine exposed and unexposed infants, replicating the findings of other longitudinal studies.6,9 Lower birth weight was associated with less optimal MDI scores regardless of cocaine exposure and with lower PDI scores for heavily exposed children. Within this sample of term and near-term infants, gestational age had no independent association with test scores, but paradoxically, lower gestational age seemed to have less adverse effect on MDI scores for children with heavy cocaine exposure than for the other exposure groups.
Two social factors, placement and child-focused developmental intervention, influenced the children’s developmental test scores in interaction with level of prenatal cocaine exposure and with each other. In this sample, there was no difference in infant perinatal characteristics among cocaine-exposed infants discharged to their biological mothers after birth compared with those discharged to kinship or unrelated foster care. (Data not shown, available by request.) However, at later ages, children in kinship care attained lower scores on the BSID than children who remained with their biological mothers or were placed in unrelated foster care, regardless of early intervention. Our clinical experience suggests that children in kinship care receive intervention only after they have developed more severe impairments than those triggering intervention for children in other groups. Others have noted unrecognized and untreated developmental deficits among school-age children in kinship care.41 In Massachusetts, as in other parts of the country, kinship caregivers receive less financial support and less monitoring from social service agencies than unrelated foster parents.42
The developmental interventions received by children in this study were those available from publicly funded early intervention and public health nursing programs in Massachusetts43 to any infant at developmental risk, and were not specifically targeted to infants with prenatal drug exposure. The clinicians that provided the interventions were not associated with the research project, so that the details of the services received by each child are not known. Home visiting and early intervention services in Boston vary in intensity, site of delivery (home or center), and in the areas of expertise (eg, speech, education, occupational, physical therapy, nursing) of the intervention staff, depending on the child’s neighborhood and the referring preferences of the pediatric health provider.41 Caregiver participation is encouraged in all programs. Although heterogeneous, these interventions exerted statistically significant protective effects on the cognitive development of heavily cocaine-exposed infants and on the psychomotor development of infants regardless of their cocaine exposure. These findings are consistent with those of randomized studies showing beneficial impact of such interventions for at-risk infants without known cocaine exposure.44,45 Moreover, positive results of home visiting interventions for cocaine-exposed infants8,46,47 have been reported in uncontrolled, descriptive studies.
Alternate possible explanations are possible for the paradoxical finding that early intervention has the greatest impact on the MDI scores of the infants with the heaviest prenatal cocaine exposure. On the one hand, this finding may only reflect the fact that heavily exposed children received intervention earlier than did unexposed or lightly exposed children (Table 5). We infer that the lower birth weight and less optimal neonatal behavior of infants who were more heavily exposed (Table 4) or clinician awareness of the child’s exposure history may have resulted in earlier and more frequent referrals, although their developmental test scores at 6 months did not suggest that obvious developmental delay was an indication for referral.
On the other hand, intriguing preliminary research suggests that perinatal cocaine exposure enhances animals’ responsiveness to early environmental manipulations such that beneficial early experiences exerted protective effects when cocaine-exposed animals experienced later stressors, effects not detected in unexposed comparisons.48 Whether heavy prenatal cocaine exposure confers on humans similar “supersensitivity” to positive environmental input merits additional study.
Contrary to expectations, there were no statistically significant interactions between level of prenatal cocaine exposure and prenatal exposure to cigarettes, alcohol, or marijuana on BSID scores. There was a paradoxical main effect of cigarettes. Heavier prenatal exposure to cigarettes was associated with higher PDI scores. An analogous positive effect of prenatal cigarette exposure on pegboard tasks has been described in a Canadian cohort.49
These results should be judiciously interpreted in light of the limitations imposed by study design. This study’s site and sample selection criteria limit the generalizability of our findings to similar samples of predominantly African American, urban, term infants who are not acutely ill at birth. Interaction analyses of cocaine effects among subgroups as we have conducted also reduce the effective sample size, and their results may be reliant on the characteristics of the sample. These findings should not be overgeneralized to prematurely born infants or to children who were prenatally exposed to other stimulants or opiates in addition to cocaine. Because the purity and dosage of street drugs vary widely, the actual pharmacologic dose exposure of the lighter and heavier groups in this sample may not be precisely comparable to the dose received by children in other samples who are also classified as “lighter” or “heavier.”7,12,25 To maintain this longitudinal sample, all participants, cocaine exposed or not, received pediatric care and intensive outreach from research staff with offers of referrals for caregivers to address diverse problems ranging from need for emergency food to requests for drug treatment. Therefore, the findings may not apply to children whose families have scant contact with health providers. Because the effect of study participation may minimize between-group differences, differences found are likely to be robust.
Lack of adverse cocaine effects on the BSID, a single apical test, as the sole developmental measure, also warrants cautious interpretation. The first edition of the BSID detects effects of prenatal exposure to potential behavioral teratogens other than cocaine, including alcohol,50 polychlorinated biphenyls,51 and lead,52 and so might be expected to be sensitive to cocaine effects if such effects are of comparable magnitude. However, others have noted effects of heavier cocaine exposure on language precursors25 and on narrow band measures, such as novelty preference or visual habituation, despite no effects on BSID scores.7 We cannot address these domains with our data.
The negative impact of placement with kinship caregivers and the positive impact of early developmental intervention must also be viewed circumspectly, because these factors were not randomly assigned to study subjects. Because of the relatively small number of individuals in each placement/intervention group, we urge particularly cautious interpretation of the significant interaction effects between placement and intervention that we have reported.
Nevertheless, the current study provides some insights to guide pediatric practice. Pediatric clinicians can reassure caregivers and social service agencies that prenatal cocaine exposure is not a potent risk factor for depressed scores on developmental tests up to 24 months when term infants are compared with unexposed infants of the same demographic background. Cocaine-exposed infants with birth weight below the 10th percentile and those placed with kinship caregivers are at increased risk for poorer developmental outcome and warrant particularly close monitoring and support. However, in contrast to birth weight adjusted for gestational age and gender, slightly lower gestational age within a cohort of term infants does not seem to increase risk of adverse outcome after in utero cocaine exposure. This suggests that rate of growth during gestation, rather than absolute duration of gestation within the range of term or near term, is the more important predictor of outcome after in utero cocaine exposure. Thus, infants who are relatively small for dates (10th percentile or below) at term warrant particularly careful scrutiny. Two protective factors that are environmentally modifiable were observed—participation in child-focused developmental intervention, particularly if such participation began in the first year of life, and placement in unrelated foster care. Pediatric clinicians can play a critical role in assuring that high-risk children, including those who have a history of cocaine exposure, are referred to the developmental services available in their community soon after birth. Our data suggest the widely held belief that prenatally cocaine-exposed children need extraordinary interventions different from those available to other children at risk may not be accurate. Pediatricians can also advocate for measures to decrease caregiver burden, increase resources, and enhance supervision for caregivers providing kinship care to the level of that provided for unrelated foster parents.41 Additional research will delineate which risk and protective factors influence developmental outcome beyond infancy among children exposed to cocaine prenatally.
This study was supported by grant DA06532 from the National Institute of Drug Abuse (to Dr Frank) and by grant MO1 RR00533 from the National Institutes of Health/National Center for Research Resources.
We thank Barry Zuckerman, MD, for review of the manuscript and Rebekah Lewis, MPH, for skilled editorial assistance.
- Received March 23, 2001.
- Accepted June 18, 2002.
- Reprint requests to (D.A.F.) FGH #3, Boston Medical Center, 820 Harrison Ave, Boston, MA 02118. E-mail:
- ↵Bayley N. Bayley Scales of Infant Development. San Antonio, TX: Psychological Corporation; 1969
- ↵Bayley N. Bayley Scales of Infant Development. San Antonio, TX: Psychological Corporation; 1993
- ↵Chasnoff IJ, Griffith DR, Friar C, Murray J. Cocaine/polydrug use in pregnancy: two year follow-up. Pediatrics.1992;89 :284– 289
- ↵Mayes LC, Bornstein MH, Chawarska K, Granger RH. Information processing and developmental assessments in three-month-olds exposed prenatally to cocaine. Pediatrics.1995;95 :539– 545
- ↵Frank DA, Augustyn M, Zuckerman B. Neonatal neurobehavioral and neuroanatomic correlates of prenatal cocaine exposure: problems of dose and confounding. In: Harvey JA, Kosofsky BE, eds. Cocaine: Effects on the Developing Brain. New York, NY: New York Academy of Sciences; 1998:40–50
- ↵Ostrea EM Jr, Brady M, Gause S, Raymundo AL, Stevens M. Drug screening of newborns by meconium analysis: a large-scale, prospective, epidemiologic study. Pediatrics.1992;89 :107– 113
- ↵Lester BM, ElSohly M, Wright LL, et al. The maternal lifestyle study: drug use by meconium toxicology and maternal self-report. Pediatrics.2001;107 :309– 317
- ↵Delany-Black V, Covington C, Ostrea EM Jr, et al. Prenatal cocaine and neonatal outcome: evaluation of a dose-response relationship. Pediatrics.1996;98 :735– 640
- ↵Frank DA, McCarten KM, Robson DC, et al. Level of in utero cocaine exposure and neonatal ultrasound findings. Pediatrics.1999;104 :1101– 1105
- ↵Tronick EZ, Frank DA, Cabral H. Mirochnick M, Zuckerman B. Late dose-response effects of prenatal cocaine exposure on newborn neurobehavioral performance. Pediatrics.1996;98 :76– 83
- ↵Singer LT, Arendt R, Minnes S, et al. Developing language skills of cocaine-exposed infants. Pediatrics.2001;07 :1057– 1064
- ↵Saranson IG, Johnson JH, Siegel JM. Assessing the impact of life changes: development of the life experiences survey. J Consult Clin Psychol.1978;46 :942– 946
- ↵Norbeck JS, Lindsey AM, Carrieri VL. The development of the Norbeck social support questionnaire: normative data and validity testing. Nurs Res.1981;32 :2164– 2168
- ↵Caldwell BM, Bradley RH. Administration Manual: Home Observation for the Measurement of the Environment. Rev ed. University of Arkansas; 1984
- ↵Barnard KE. NCAST II: Nursing Child Assessment Teaching Manual. Seattle, WA: NCAST Publications; 1980
- ↵Littell RC, Milliken GA, Stroup WW, Wolfinger RD. SAS System for Mixed Models. Cary, NC: SAS Institute; 1996
- ↵Belsey DA, Kuh E, Welsch RE. Regression Diagnostics. New York, NY: John Wiley and Sons Inc; 1980
- ↵1991 Detail Natality [data tape]. Hyattsville, MD: National Center for Health Statistics; August 1993
- ↵Brazelton BT. Neonatal Behavioral Assessment Scale. Philadelphia, PA: Spastics International Medical Publications, JB Lippincott; 1984
- ↵Gleeson JP. Kinship care as a child welfare service: what do we really know? In: Gleeson PJ, Hairston CF, eds. Kinship Care: Improving Practice Through Research. Washington, DC: Child Welfare League of America, Inc; 1999:3–35
- ↵Massachusetts Department of Public Health. Welcome to Early Intervention in Massachusetts: Your Guide to Early Intervention Programs Throughout the Commonwealth. November 1998
- ↵Olds DI, Henderson CR Jr, Tatelbaum R. Prevention of intellectual impairment in children of women who smoke cigarettes during pregnancy. Pediatrics.1994;93 :228– 233
- ↵Black M, Dubowitz H, Hutcheson J, Berenson-Howard J, Starr RH. Parenting and early development among children of drug-abusing women: effects of a home intervention. Pediatrics.1994;94 :440– 448
- ↵Spear LP, Campbell J, Snyder K, Silveri M, Katovic N. Increased sensitivity to stressors and others environmental experiences after prenatal cocaine exposure. In: Harvey JA, Kosofsky BE, eds. Cocaine: Effects on the Developing Brain. New York, NY: New York Academy of Sciences; 1998:76–88
- Copyright © 2002 by the American Academy of Pediatrics