|Home | About | Journals | Submit | Contact Us | Français|
To determine in a representative sample of full-term urban newborns of English-speaking mothers whether an immediate or late dose-response effect could be demonstrated between prenatal cocaine exposure and newborn neurobehavioral performance, controlling for confounding factors.
The Neonatal Behavioral Assessment Scale (NBAS) was administered by masked examiners to a total sample of 251 clinically healthy, full-term infants at 2 days and/or 17 days. Three in utero cocaine exposure groups were defined: heavily exposed (n = 44, >75th percentile self-reported days of use during pregnancy and/or >75th percentile of meconium benzoylecognine concentration); lightly exposed (n = 79, less than both 75th percentiles); and unexposed (n = 101, no positive biological or self-report marker). At the 3-week examination there were 38 heavily exposed, 73 lightly exposed, and 94 unexposed infants. Controlling for infant birth weight, gestational age, infant age at the time of examination, mothers’ age, perinatal risk, obstetric medication, and alcohol, marijuana, and cigarette use, a regression analysis evaluated the effects of levels of cocaine exposure on NBAS performance.
No neurobehavioral effects of exposure on the newborn NBAS cluster scores or on the qualifier scores were found when confounders were controlled for at 2 to 3 days of age. At 3 weeks, after controlling for covariates, a significant dose effect was observed, with heavily exposed infants showing poorer state regulation and greater excitability.
These findings demonstrate specific dose-related effects of cocaine on 3-week neurobehavioral performance, particularly for the regulation of arousal, which was not observed in the first few days of life.
Early observations suggesting significant neonatal neurobehavioral dysfunction associated with prenatal exposure to cocaine raised clinical and public health concerns.1–3 Subsequent findings were neither as striking nor as consistent as the initial observations.2,4–9 The most consistent behavioral finding is a decrement in the newborn’s ability to habituate in the immediate postpartum period.10,11 However, other studies have not found this effect or have found different effects.1,5,12–15 Several studies have not found any neurobehavioral effects after in utero cocaine exposure.2,16–19 Complicating this issue further, a number of studies1,5,20 have found interactive effects of cocaine with other substances, principally alcohol and marijuana, or with other risk conditions, such as lack of prenatal care.5,10,15 Last, several investigators have found effects later in the newborn period rather than immediately after birth.5,17
These inconsistent findings may be attributable to small sample sizes, which preclude control for confounding variables, to the variability of doses and timing of exposure, and to the timing of assessments.2,3,5,8,9,11,14 Few studies have effectively controlled for confounding variables,9 and in one study in which confounding variables were covaried, the critical variable of birth weight was not controlled for.5 Several investigators have suggested that weak or inconsistent results would be expected, because neurobehavioral effects will be seen only with high levels of exposure.21,22 Studies showing little or no effect may be obscuring effects caused by heavy exposure by evaluating lighter and heavier users (eg, samples identified solely by self-report or by self-report and urine assays) as a single group. It also has been argued that women who are the heaviest users are less likely to participate in studies, leading to ascertainment bias.7
The current study was designed to determine whether a dose effect of prenatal cocaine exposure on neonatal neurobehavioral performance could be demonstrated in a representative sample of urban full-term infants of English-speaking mothers. The study also aimed to see whether effects were found early and/or late in the newborn period. The effects of selected confounding factors were controlled for using statistical techniques.
The sample was recruited on the postpartum floor of Boston City Hospital from October 1990 to March 1993. Cocaine-exposed mother-infant dyads were matched when possible by ethnicity (African-American and African-Caribbean versus other) to unexposed dyads. All mother-infant dyads met the following criteria at delivery: (1) infant gestational age of 36 weeks or older; (2) no requirement for level III (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) the mother’s ability to communicate fluently in English; (7) no indication by toxic screen or history in the medical record of the mother’s use during pregnancy of illegal opiates, methadone, amphetamines, phencyclidine, barbiturates, or hallucinogens; and (8) mother 18 years of age or older. These criteria were established to exclude infants with known major risk factors for developmental impairment that might confound or obscure the effects, if any, of in utero cocaine exposure. In particular, these criteria exclude any preterm infants and infants with significant intrauterine growth retardation (IUGR) whose neurobehavior, compared with the neurobehavior of clinically healthy, full-term infants, might be compromised by these risk factors as well as more vulnerable to the effects of cocaine exposure.
Previous research has demonstrated that a combination of self-report and biological markers—urine or meconium assays—maximizes identification of cocaine use in pregnancy.6 All dyads were classified as exposed or unexposed based on a biological marker of the infant or the mother. After the identification of exposure status, classification of mothers as lighter or heavier users was done with a combination of information from interviews and biological markers obtained by clinicians and study personnel (see below).
The mothers were interviewed about their drug use history at intake on the postpartum floor using an adaptation of the Addiction Severity Index.6 This interview provides information on the chronicity of the mothers’ drug use and the frequency of use of drugs, including cigarettes and alcohol, before and during pregnancy.
As previously described,22 meconium samples were analyzed for the presence of benzoylecgonine (a cocaine metabolite), opiates, amphetamines, benzodiazapines, and cannabinoids by a modification of the methods of Mirochnick et al22 and Ostrea et al.23 Maternal urine samples were analyzed at recruitment for benzoylecognine, opiates, amphetamines, benzodiazepines, and cannabinoids by radioimmunoassay (RIA) using commercial kits (Abuscreen RIA; Roche Diagnostics Systems, Inc, Montclair, NJ).
All mother-infant dyads had at least one biological marker, urine 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 (range, 0 to 264) days. The mean meconium concentration was 1143 (range, 0 to 17950) ng of benzoylecognine/g of meconium. Heavy use was considered a priori to be the top quartile of meconium concentration for cocaine metabolites (>3314 ng of benzoylecognine/g of meconium) and/or the top quartile days of self-reported use (>61 days) during pregnancy. All other use was classified as “lighter.”23
Trained interviewer-recruiters screened maternity and nursery records 7 days a week and then sought informed consent from eligible women who, with their infants, met the selection criteria. Women with clinically documented prenatal cocaine use were recruited as exposed mothers, and those without such use were recruited as provisionally unexposed comparison mothers.
Of the 192 known cocaine-using mothers invited to participate, 123 (64%) agreed to join the study. Of the 646 mothers who were not clinically known cocaine users, 134 (21%) agreed to join the study. Of these 257 participants, 5 were excluded from further analysis, because their unexposed status could not be confirmed by biological assay. One additional woman was dropped because of a meconium test that was deemed unreliable. Of this sample of 251 dyads, 128 were provisionally considered unexposed, and 123 were considered provisionally exposed. Based on the results of the study of biological assays, 15 unexposed dyads were reclassified as exposed, yielding a study sample of 138 exposed and 113 unexposed dyads.
Within the cocaine-exposed group, no statistically significant differences in maternal age, parity, ethnicity, or mode of payment for medical care were observed among the mothers who were recruited, refused, or were not approached. It was not possible to assess levels of use of the exposed mothers who refused to participate. Within the nonexposed group, mothers who were recruited were significantly more likely (P = .03 by X2 test) to be primiparous than those not approached or those who refused to participate (33% vs 20% vs 24%, respectively). These groups did not differ on the other parameters. Thus, aside from the difference in parity, our sample was representative of the general obstetric population from which they were drawn. When infants who had the week 3 Neonatal Behavioral Assessment Scale (NBAS) examination were compared with those who were examined only in the early newborn period, no significant differences in maternal characteristics (including substance use) were found.
A total sample of 251 infants received the NBAS examination in the newborn period and/or at the 3-week visit. In the immediate postpartum period, complete data were available on 224 dyads: 101 of the 224 dyads were nonusers; 79 were light users; and 44 were heavy users. At the week 3 visit, complete data were available on 205 dyads; 94 nonusers, 73 light users, and 38 heavy users. Thus, 7 heavy users, 6 light users, and 6 nonusers were lost to follow-up.
Within. 8 to 72 (mean, 48) hours, a study pediatrician masked to exposure status assessed the infant’s gestational age using the method of Dubowitz et al.24 In addition, each infant was given a physical examination for minor and major congenital anomalies. The pediatrician carried out anthropometric measures of recumbent length on a Holtain Infantometer and head circumference using a plastic-coated tape.6 Reliability was regularly assessed on the anthropometric measures and was maintained at .95 for head and length measurements using an intraclass correlation coefficient. After performing the physical examination, the pediatrician reviewed the infant’s neonatal record and completed the Hobel index25 for neonatal complications. A trained research assistant abstracted the mother’s medical record to complete the prepartum, intrapartum, and postpartum Hobel risk indices for the mothers.26
As the most valid and standardized extant measure of neonatal behavior, the NBAS18,25,27 has been the instrument of choice in the majority of studies evaluating the effects of in utero cocaine exposure on newborn behavior, The NBAS consists of 28 behavioral items scored on a nine-point scale that describe the range of an infant’s behavioral responses to social and nonsocial stimuli as the neonate moves from sleeping to awake and crying states as well as the infant’s central neuro-organizational capacity to process and react to stimuli. Additionally, the NBAS contains 9 supplementary qualifier items. In contrast to the NBAS behavioral item scores, which focus on specific aspects of the infant’s performance, the supplementary items rate the quality of the infant’s performance during the course of the examination (eg, “robustness” varies from the infant appearing very fragile to energetic with no signs of exhaustion during the course of the examination), as well as difficulties the examiner had in examining the infant (eg, “examiner persistence” varies from maximal effort to elicit responses to no special examiner maneuvers needed to elicit responses during the course of the examination). Thus, the qualifier scores are similar to clinical judgments of the infant’s quality of performance.
In this study, the NBAS was administered to the infants during the early neonatal period (mean age, 51 hours) and during the third week of life (mean age, 17 days). Serial NBAS examinations are more accurate and better predictors of behavior28,29 and permit the evaluation of drug effects when the infants are no longer possibly acutely intoxicated by maternal cocaine. The newborn examination was administered in a quiet room off the nurseries. The week 3 examination was administered in a quiet room in the primary care pediatric clinic before the infant received a physical examination. All examiners were masked to infant drug exposure status and trained by a certified NBAS trainer to standard levels of reliability, ie, 90% reliability within one scale point on all items. Reliability was maintained at this level during the course of the study.
The NBAS supplementary qualifier items were analyzed item by item. The NBAS behavioral item raw scores were reduced to cluster scores in two ways. Seven neurobehavioral clusters were calculated using the method of Lester29: habituation, orientation, state regulation, range of state, autonomic stability, motor organization, and reflexes. Two additional clusters (developed by Lester and Tronick30), excitability and depression, were also calculated. The excitability and depression clusters are hypothesized to be related to the direct effects of cocaine on the central nervous system and to indirect effects mediated through IUGR.28 Thirteen NBAS items are used to construct the excitability and depression scales (see “Appendix” and Lester and Tronick30 for details).
Least-squares regression models were carried out on the NBAS data to evaluate the differences among the groups uncontrolled for covariates. These multivariate group analyses are conservative, because they make no assumption about the relations among the three groups. Two-tailed tests (P values) were used. After these analyses, multiple regression analyses were performed on the NBAS measures, controlling for covariates. Some covariates were examined but dropped from the final model because they had no significant relation to outcome. Variables dropped from analyses included infant gender, maternal education, ethnicity, parity, and number of prenatal visits. Covariates entered in the model included age at examination, birth weight, maternal anesthesia (general, local, or none—used only in the newborn analysis), prepartum Hobel risk score (any positive risk item versus none), maternal age, marijuana use, log average daily volume of alcohol, and log average daily number of cigarettes. Birth weight was entered as a z score to adjust for gender and gestational age. Head circumference and length were dropped from the analyses, because birth weight was found to be a better predictor of the dependent measures than these variables.
The birth weight z scores were obtained by taking gender-specific least squares regressions of birth weight on the Dubowitz gestational age, with squared and cubic terms included. Data for these models were obtained in a previous study6 of more than 1200 infants whose mothers had received prenatal care and who were delivered at the hospital. Predicted birth weight values for the current sample were generated using the parameter estimates from the above models. The difference of each infant’s birth weight and the predicted values, referred to as residuals, were computed. Z scores for these residuals were computed using the mean and SD of the residuals from the previous study sample as population values.
Multivariate trend analyses were carried out to evaluate dose-response relations. The three groups (unexposed, lightly exposed, and heavily exposed) were assigned values of 0, 1, and 2, respectively. The values 0, 1, and 2 are not interval values but only ordinal values, representing an increase in exposure. Muitivariate trend analyses were then applied to the slopes of the data for the cluster and qualifier scores over these groups. Dose-response relations would be represented by a significant trend across the means.
Table 1 presents data on maternal characteristics for each exposure group. As would be expected from the matching procedures, there were no significant ethnic differences among the groups. The non-cocaine-using mothers were younger than the light or heavy users (24.8 vs 27.7 vs 26.7 years, respectively; P = .0001). The exposed and unexposed groups did not differ in parity, education, percentage of private insurance, or percentage of cesarean section.
Table 2 presents data on maternal cigarette, alcohol, and marijuana use during pregnancy. Cigarette and alcohol use were determined by self-report on the Addiction Severity Index. Marijuana use was determined by a positive self-report and/or any positive urine or meconium assay. Women who used cocaine prenatally, and heavy users more than lighter users, were significantly more likely to have used cigarettes, alcohol, and marijuana, as reported in other studies.2,6,31–34
Data on infant characteristics in each exposure group are presented in Table 3. The three exposure groups did not differ on Apgar scores at 1 and 5 minutes. No significant differences among the groups on the Hobel neonatal risk index were found. Birth weight and length differed among the three groups (P < .0001). Unexposed infants weighed more and were longer than lightly and heavily exposed infants. Unexposed and lightly exposed infants had significantly larger occipital head circumferences than heavily exposed infants (P < .001} but did not differ from each other. Ponderal indices did not differ significantly among groups. Relatedly, meconium benzoylecgonine concentration was negatively related to birth weight (r = −.27; P = .01) and length (r = −.32; P = .01).22
In Tables 4 and and5,5, analyses of the data on neurobehavioral performance of study infants controlled and uncontrolled for confounders during the newborn period and the third week of life are presented. Confounders included age at examination, birth weight, maternal anesthesia (general, local, or none—used only in the newborn analysis), prepartum Hobel risk score (any positive risk item versus none), maternal age, marijuana use, log average daily volume of alcohol, and log average daily number of cigarettes.
At the newborn examination, uncontrolled for confounders, prenatal cocaine exposure was significantly associated in a dose-related fashion with habituation (P < .03) and with all but three supplementary qualifier items (P = .04; Table 5). Significant effects for habituation (P = .02) and motor maturity (P = .04) were found using multivariate trend analysis. At the 3-week examination, prenatal cocaine exposure had significant univariate associations with NBAS cluster scores for reflexes (P = .04), state regulation (P = .01), autonomic stability (P = .03), excitability (P = .0004), and three of the qualifier scores (P = .02; Table 5). At week 3, trend analyses revealed significant univariate dose-response effects for reflexes (P = .01), state regulation (P = .005), autonomic stability (P = .009), and excitability (P = .001).
After controlling for confounding variables, cocaine exposure was no longer significantly associated with habituation or the NBAS qualifiers at the newborn examination. In contrast, at the 3-week examination, when confounding variables were controlled for, the level of prenatal cocaine exposure continued to be independently associated with state regulation (P = .02) and excitability (P = .007) but not with reflexes (P = .48), autonomic stability (P = .28), or any of the NBAS qualifier scores. Trend analyses continued to be significant for state regulation (P = .01) and excitability (P = .05) after controlling for confounders.
Various covariates were also significantly related to different neurobehavioral scores at the two examinations, but no significant interactions between these covariates and the level of exposure were observed. These covariate effects at the early examination were: (1) habituation cluster—the older the infant at the time of examination, the poorer the performance (P = .004); (2) orientation cluster—infants of mothers who received anesthesia during delivery performed better (P = .04); and (3) reflex cluster—infants with higher birth weight z scores performed better (P = .04).
At the 3-week examination, covariate associations included: (1) reflex cluster—a higher birth weight z score was associated with better performance (P = .05); (2) regulation of state cluster— older infants performed better (P = .006), as did infants with no prenatal risk by Hobel criteria (P = .03); and (3) autonomic stability cluster—infants of mothers who used alcohol performed more poorly (P = .03).
This study is the first to document human dose-response effects of in utero cocaine exposure on the neurobehavioral performance of the infant.7,8,35 In this sample, after controlling for covariates, heavier maternal use of cocaine was associated with poorer regulation of arousal (ie, state regulation and excitability) late in the newborn period but not in the immediate postpartum period. Significant dose effects were found in this sample of infants for habituation and motor performance in the immediate postpartum period and later in the newborn period for reflexes and autonomic stability, as well as on the supplementary qualifier items, but these effects disappeared when potentially confounding variables were controlled for analytically.
The study’s prospective, longitudinal design overcomes, at least in part, many of the methodological problems of earlier research. Examiners were masked to exposure status. Importantly, the sample represents the spectrum of the cocaine-using population of medically indigent, English-speaking women in Boston. The sample was drawn from a general obstetric population of mothers, and the newborns were clinically healthy, full-term, and free of heroin, methadone, barbiturates, phencyclidine, and amphetamines. The sample is not biased against heavier cocaine users, who may not enroll in prenatal care7,36,37 or who do not give birth in hospitals. In Boston virtually all women receive prenatal care and give birth in hospitals. Of course, some of the heaviest using mothers may still have been missed. Missing the heaviest users might have occurred if one were to assume that they are more likely to refuse to participate, even though refusers and compilers did not differ with respect to demographic or clinical characteristics. However, were such a bias to occur, it would work against finding dose-related effects.
Thus, heterogeneity of the amount of use, as indicated by the newly developed composite index based on self-reports and meconium assays,22 was large enough to detect dose-related neurobehavioral effects, even though meconium measurement has been generally thought to be only an all-or-none measurement. Although unmeasured confounders may have affected neurobehavioral performance, the finding of significant dose-response effects after controlling for many covariates suggests a specific effect of cocaine on newborn neurobehavioral performance. Indeed, the dose-response relationships found are likely to underestimate the effects of prenatal exposure because of the exclusion of preterm, IUGR, and other at-risk infants from the sample and the possibility that some of the heaviest-using mothers were still missed.
The emergent independent effects of cocaine at 3 weeks postpartum on state regulation and excitability reflect an impairment in the infant’s ability to modulate arousal. These effects are not likely to be attributable to the direct acute exposure of the infant after delivery. First, cocaine-using women are routinely discouraged from breastfeeding at Boston City Hospital, so that direct mother-to-infant transfer of cocaine is likely to end at delivery, with the exception of passive postnatal exposure. Second, after in utero exposure, cocaine metabolites clear the newborn’s body in 7 days, as assessed by RIA methods.3 Third, if the effect seen were attributable to acute infant intoxication, one would expect to see other neurobehavioral effects. These effects were not observed. Fourth, similar to other studies,5,17 effects are seen late rather than early in the newborn period.
These late-emerging neurobehavioral effects might be expected if in utero cocaine exposure is associated with evolving neuroanatomic damage38 or disruption of the neurotransmitter systems (eg, the mono-aminergic system, which is thought to control the modulation of arousal and attention).37,39 Dysfunction of this or other systems might result in the observed effects and even an infant withdrawal syndrome. Little in these findings supports the hypothesis that the exposed newborn’s neurobehavior is compromised because of IUGR, for two reasons: (1) the effects of cocaine persist after controlling for birth weight adjusted for gender and gestational age4,6; and (2) IUGR was not associated with the depression cluster among the cocaine-exposed infants, as Lester and Tronick4 have hypothesized and found in previous research,9,28
The habituation effect found in this study replicates and extends previous findings.10,11 This effect was no longer significant after controlling for confounders, as was the case in other studies.7,10,11 Similarly, as in other studies,17,40 the significant associations of in utero cocaine exposure with 3-week motor, reflex, and autonomic stability clusters were no longer significant after controlling for covariates. Thus, these effects could not be specifically attributed to cocaine but, rather, are related to other risk factors, such as maternal anesthesia, alcohol, birth weight, and prenatal risk.4,5,17
The effects of a combination of factors are evident in the findings for the NBAS supplementary qualifier items. These clinical judgments demonstrate that exposed newborns at both the early and late examinations were judged to have less regulatory capacity, poorer state regulation, poorer motor tone, and less responsivity than the nonexposed infants. Moreover, the exposed infants were more difficult to examine and less rewarding to the examiner. However, unlike the findings for the specific neurobehavioral items, the clinically recognizable differences for the supplementary qualifier items were not statistically significant after adjusting for covariates. Clinicians correctly judge cocaine-exposed infants to have poorer functional status than unexposed infants, but it is incorrect simply to attribute their dysfunction to maternal cocaine use. Rather, these findings indicate that cocaine exposure is a marker for a number of other adverse lifestyle factors that contribute to the infant’s poorer functioning. Thus, the clinician needs to consider not only cocaine, but these other factors when managing the care of the exposed mother and her infant.
Caution must be exercised in generalizing these results to other populations in which the mix of risk factors and other exposures may be different from those reported here. Moreover, it is important to note that although neurobehavioral effects were found that were consistent with previous reports, several domains of newborn functioning, (eg, orientation), were unaffected. With the exception of the dose-response relationships for excitability and state regulation at 2 to 3 weeks, which indicate specific neonatal neurobehavioral effects of maternal cocaine use, other domains of neurobehavioral functioning of even the heavily exposed infants were indistinguishable from the those of the unexposed infants.14
For the pediatric provider, prenatal cocaine exposure should serve as a flag that the infant is likely to show clinically detectable compromised neurobehavioral status throughout the newborn period. These characteristics make the infants both more sensitive to disruptive care giving and more difficult to care for, leading to a self-reinforcing process that may exacerbate initial neurobehavioral impairments. Thus, in utero cocaine exposure needs to be prevented. However, when prevention fails, intervention with the care giver-child dyad can be beneficial in limiting further neurobehavioral dysfunction caused by the disruption of mutual regulatory processes that determine extrauterine development.9
In summary, this research documents a dose-related association between the level of in utero cocaine exposure and neurobehavioral performance during the late newborn period. In addition, our findings indicate that a combination of other factors, including cocaine, compromises newborn neurobehavioral performance. In utero cocaine exposure thus has a measurable effect on the exposed newborn’s emerging neurobehavioral capacity to regulate arousal. It is not known whether this problem progresses developmentally, but it would be critical to know its course, because regulation of arousal and attention is critical to learning. Longitudinal studies focusing on the development of infants’ capacity to regulate arousal and attention are warranted, especially in the context of social interaction.9
This research was supported by grants from National Institutes of Drug Abuse: “Cocaine Exposure in Utero; Two Year Infant Follow-Up” (R01 DA06532; Deborah Frank, principal investigator) and “Neurornotor Functioning in Cocaine Exposed Infants” (R01 DA6882; Edward Z. Tronick, principal Investigator).
We thank Ruth Rose-Jacobs, Jean Cole, Judith James, Jennifer Scott-Sutherland, Karen Olson, and Linda Burtt for their work on this project. T. Berry Brazelton, M. Katherine Weinberg, Marjorie Beeghly, and Barry M. Lester contributed many ideas to this document.
The 13 NBAS items and their criterion scores that make up the excitability scale are: tone, more than 6; motor maturity, less than 4; cuddliness, less than 3; consolability, less than 4; peak of excitement, less than 7; rapidity of buildup, more than 6; irritability, more than 5; activity, more than 6; tremulousness, more than 5; startles, more than 4; lability of skin color, more than 7; lability of state, more than 3; and self-quieting, less than 3, For the depression scale, the items are: orientation to a ball, less than 4; to a rattle, less than 4; to a face, less than 4; to a voice, less than 4; to a face and voice, less than 4; to alertness, less than 4; tone, less than 4; pull-to-sit, less than 4; defensive reaction, less than 4; rapidity of buildup, less than 4; irritability, less than 3; and activity, less than 4. Infants are assigned one point for each NBAS item on which the infant meets or exceeds the criteria for excitability or depressed behavior. However, rules must also be applied that take account of performances that are dependent on one another, which in turn affect the cluster scores (eg, the failure of an infant to achieve a state of 4 is associated with the infant’s scores on the orientation items, which in turn affects their depression cluster score). Additionally, the score is averaged to allow for missing values. A complete manual with rules of combination is available from the first author, or see Lester and Tronick.30