Growth, Development, and Behavior in Early Childhood Following Prenatal Cocaine Exposure: Title and subTitle BreakA Systematic Review FREE

Recently, the US Supreme Court considered Ferguson et al v City of Charleston, a Fourth Amendment case (unreasonable search and seizure).¹ This case addresses a policy of the Medical University of South Carolina whereby health professionals, in cooperation with the local prosecutor, selectively screened the urine of medically indigent obstetric patients for cocaine metabolites.^{1 - 3} Medical personnel reported positive results to the police, who would then come to the hospital to arrest prenatal and postpartum patients for possession of an illegal drug, delivery of drugs to a minor, or child abuse.^{3 - 4} In the popular press, People magazine reported on C.R.A.C.K. (Children Requiring a Caring Kommunity), a controversial charity that raises money to give mothers with a history of illegal drug use financial incentives to accept long-acting contraception, or, in most cases, sterilization.⁵ This charity and the policies at issue in Ferguson v City of Charleston reflect popular belief that women who use cocaine while pregnant inflict severe, persistent, and unusual impairments on their unborn children, recently described by a newspaper columnist as "blighted by a chemical assault in the womb."⁶

Public expectations of "blighted" children fuel controversial punitive policies directed toward addicted mothers.⁷ Since 1985, more than 200 women in 30 states have faced criminal prosecution for using cocaine and other psychoactive substances during pregnancy.⁷ Scholars and professional organizations have condemned efforts to sterilize or criminally prosecute addicted mothers as ethically and legally flawed, racially discriminatory, and an impediment to providing appropriate medical care to these women and their children.^{3 - 4 ,7 - 9}

Recent reviews^{10 - 15} and articles^{16 - 18} show that most initial predictions of catastrophic effects of prenatal cocaine exposure upon newborns were exaggerated. After controlling for confounders, the most consistent effects of prenatal cocaine exposure are small but statistically significant decrements in 1 or more parameters of fetal growth for gestational age^{12 - 13} and less optimal neonatal state regulation and motor performance.^{10 - 11 ,14} Clinically silent findings on neonatal cranial ultrasounds following prenatal exposure have been found in some studies,^{10 ,16} but not others.¹⁷ Prenatal cocaine exposure without concurrent opiate exposure has not been shown to be an independent risk factor for sudden infant death syndrome.^{15 ,18}

Despite the neonatal data, beliefs about cocaine's teratogenicity impose a stigma on cocaine-exposed infants^{19 - 20} and children at school age.²¹ Teachers fear that "crack kids" will be too developmentally delayed or disruptive to be taught in traditional classrooms.²²

Given the current public concern, health professionals need a critical synthesis of studies of postneonatal outcomes of children exposed to cocaine in utero in 5 domains: (1) physical growth; (2) cognition; (3)language skills; (4)motor skills; and (5) behavior, attention, affect, and neurophysiology.

MEDLINE and Psychological Abstracts were searched for all human studies published in English from 1984 until October 2000 that included the words cocaine, crack/cocaine, crack, pregnancy, prenatal exposure, delayed effects, children, and related disorders. Even if cited in MEDLINE, abstracts or nonreviewed proceedings of scientific meetings²³ were excluded. Seventy-four published articles were identified. ^{24 - 42 43 - 62 63 - 82 83 - 97}

We first applied selection criteria used by others⁹⁸ : all selected studies presented original research published in a refereed English-language journal, used human subjects, and used a control or comparison group. Detailed review was then restricted to studies that also met 3 criteria: (1) samples were prospectively recruited; (2) examiners of the children were masked to their cocaine exposure status; and (3) the cocaine-exposed cohort did not include a substantial proportion of children also exposed in utero to opiates, amphetamines, or phencyclidine, or whose mothers were known to be infected with the human immunodeficiency virus (HIV).

Studies were classified as prospectively recruited if the samples of cocaine-exposed and unexposed mother-infant dyads were identified and enrolled either during pregnancy or immediately after birth. Prospective recruitment obviates recall bias, when caregivers of a child who has experienced an adverse outcome are likely to recall prenatal exposure in greater detail, and selection bias, when caregivers are more likely to enroll children with already suspected developmental impairments. Such biases in retrospective samples can produce an overestimate of the risk of negative developmental outcomes.⁹⁹

In behavioral research, examiners' bias may unconsciously distort measurement of developmental/behavioral outcomes.^{99 - 101} Investigators have shown that evaluators were more likely to code children's videotaped behavior as abnormal if the children were labeled as "crack kids" than if they were not.^{19 - 20}

Lower developmental test scores in infancy and less adaptive behavior at school age have been linked to prenatal opiate exposure.¹⁰² In samples where most cocaine-exposed children are also opiate-exposed, the independent effect of cocaine on outcome cannot be clearly delineated. For the same reason, samples where cocaine exposure was largely confounded with exposure to methamphetamines or phencyclidine were also excluded. Exposure to HIV in utero is correlated with poor developmental outcome not only among infected infants, but also among those who serorevert.¹⁰³ If most cocaine-exposed children in a sample are also offspring of HIV-infected mothers, it cannot be determined whether effects are due to cocaine or HIV exposure.

Two developmental/behavioral pediatricians (D.A.F., M.A.) and a neuropsychologist (W.G.K.) reviewed all articles. After excluding 38 articles according to the above criteria, the same 3 authors abstracted the data from the remaining 36 articles in detail. If a single article covered outcomes in more than 1 domain (eg, cognitive test scores and behavior), each domain was addressed separately. If there was uncertainty, contact was made with the corresponding author of the article to clarify interpretation of data. Disagreements were resolved by consensus.

Of the excluded studies, 20^{24 ,27 ,30 - 31 ,33 - 37 ,41 - 42 ,48 - 49 ,52 - 53 ,55 - 57 ,60 - 61} failed to mask investigators to children's cocaine exposure status. Seven^{24 ,27 - 28 ,36 ,39 - 40 ,53} had no control group. Twenty-six^{24 - 27 ,29 - 30 ,33 - 38 ,40 - 44 ,46 - 49 ,51 ,54 ,58 - 59 ,61} did not use prospective recruitment for some or all of their subjects. Thirteen^{28 ,30 ,32 ,34 ,41 - 42 ,49 - 50 ,54 - 57 ,59} primarily recruited children with in utero exposure to opiates, methamphetamines, or phencyclidine. Two^{32 ,44} reported samples predominantly composed of children of HIV-positive mothers.

The conceptual framework for data extraction was provided by recent theoretical advances in human behavioral teratology^{104 - 105} delineating the implications of various methods of characterizing exposure to possible toxicants and of controlling for potential confounders. Many cocaine-exposed newborns are clinically indistinguishable from their unexposed peers,^{18 ,106} so identification of exposed infants depends on maternal report or measurement of cocaine metabolites in biological matrices. Dose response is a critical issue in the study of all potential teratogens¹⁰⁵ but is difficult to ascertain for cocaine in human studies. Recently, infants' meconium and maternal hair have emerged as useful biological markers for estimating the dose of prenatal cocaine exposure.^{97 ,107 - 111} However, at the time most cohorts available for study in the postneonatal period were recruited, assays of urine from mother or infant for benzoylecognine were the only biological indicators readily available. Urine assays do not reflect cumulative fetal drug exposure. Thus, researchers who address dose response rely on maternal interviews to classify levels of prenatal cocaine exposure, usually classifying 2 or more days a week during pregnancy as "heavier use."^{63 ,66 ,85} For this review, we classified levels of prenatal cocaine exposure as heavier/lighter or as exposed/unexposed.

Even when their mothers do not use opiates, amphetamines, or phencyclidine, most cocaine-exposed infants are also exposed in utero to varying combinations of tobacco, alcohol, and marijuana.¹¹² The heaviest prenatal cocaine users are often the heaviest users of these other substances.¹⁰⁹ If prenatal exposure to tobacco, alcohol, and marijuana is not analytically controlled, their effects on neurodevelopment^{74 ,84 ,113} may be misattributed to cocaine. If these substances are statistically controlled for without regard to the level of use, residual confounding may occur because of overaggregation of light and heavy exposure.^{104 ,114} For this review, we considered whether prenatal tobacco, alcohol, and marijuana exposure are reported or not, are controlled analytically as dichotomous variables (exposed/not exposed), or are statistically controlled in a dose-related manner. However, statistical control in a dose-controlled manner offers the greatest assurance that effects of heavy tobacco, marijuana, or alcohol exposure will not be spuriously attributed to cocaine.

Interpreting cocaine effects is further complicated because the samples studied are, with a few exceptions,^{77 ,90 ,93 ,97} drawn from economically disadvantaged, medically at-risk populations, whose characteristics are associated with high developmental risk without any psychoactive substance exposure. The number of environmental and medical variables, the accuracy of their measurement, and their distribution within the sample may influence the estimation of cocaine effects.¹⁰⁴

The data were derived from 17 independent cohorts from 14 cities. Some cohorts were the subject of multiple articles, either at different ages or with differing analyses of the same data from a single age. Mutually exclusive samples were identified by author and city. For each article, a number of parameters were coded, including number of cocaine unexposed and exposed subjects and the number at varying levels of cocaine exposure if such data were available; how pregnancy exposure to tobacco, alcohol, and marijuana was addressed analytically and whether this exposure was significantly related to outcomes; what other covariates were matched, used as selection criteria, or controlled for statistically; which of these covariates influenced outcomes; and what, if any, statistically significant (P<.05, 2-tailed unless otherwise specified) cocaine effects were identified. Of the included articles, 4 do not report attrition.^{66 ,77 - 78 ,87} In the others, sample retention from birth to the oldest age reported for the cohort ranges from 39%⁷⁰ to 94%.⁶² Of these, 14 articles^{64 - 65 ,67 ,73 - 74 ,81 ,83 ,85 ,89 ,91 - 93 ,96 - 97} from 11 cohorts document the characteristics of those retained compared with those lost to follow-up.

If level of exposure to other substances is not controlled, prenatal cocaine exposure appears to be associated in 2 cohorts with postneonatal decrements in weight or occipitofrontal head circumference,^{64 ,70 ,78 - 79} but not in another⁸⁹ (Table 1). However, in 2 cohorts that did control for dose of prenatal exposure to tobacco and alcohol^{84 ,93} no negative cocaine effect was noted on the children's weight, length, or head circumference. In 1 cohort, full-term unexposed children were longer than exposed or unexposed preterm children and their exposed full-term counterparts.⁷¹

There is little impact of prenatal cocaine exposure on children's scores on nationally normed assessments of cognitive development (Table 2). Findings of cocaine effects depend on contextual factors, such as the child's history of prematurity, age at time of assessment, and the effects of prenatal exposure to other substances. Of the 9 studies evaluating prenatal cocaine effects on developmental test scores in infants, 5 found no effect,^{71 ,77 ,79 ,85 ,89} including 1 that classified infants according to level of prenatal exposure to cocaine, tobacco, and alcohol.⁸⁵ Chasnoff et al⁷⁰ found that the 6-month-old infants whose mothers used cocaine, alcohol, and marijuana attained mean scores lower than infants of controls, but identical to those of infants whose mothers had used alcohol/marijuana without cocaine, suggesting no incremental impact of cocaine use. Mayes et al⁹¹ reported bivariate association of lower psychomotor scores at 3 months with prenatal cocaine exposure, but not after statistical control for potential confounders. Alessandri et al⁶³ found no main effects of level of prenatal cocaine exposure on test scores at 8 or 18 months, but on post hoc comparisons children with the highest level of cocaine exposure in pregnancy (2 or more days a week) obtained significantly lower mental development scores at age 18 months than unexposed infants.

In very low-birth-weight infants, Singer et al⁹⁶ reported a negative association between prenatal cocaine exposure and developmental scores at 16 months corrected age, but in utero exposure to other psychoactive substances was not analytically controlled.

Six reports from 4 cohorts evaluated the association of prenatal cocaine exposure with cognitive test scores in children between the ages of 3 and 6 years.^{64 ,78 ,82 - 83 ,89 ,93} Two articles presented results in a single cohort of 3-year-olds. In one, Azuma and Chasnoff⁶⁴ reported that children whose mothers only used alcohol and marijuana during pregnancy achieved mean IQ scores that were identical to those of children whose mothers had also used cocaine. In a second report of post hoc comparisons from the same cohort, Griffith et al⁷⁸ found that children exposed to cocaine in addition to other substances scored significantly lower than unexposed controls on a verbal reasoning scale of the IQ test. However, these scores were not lower than the scores of children who had been exposed to the other substances but not cocaine and were not statistically controlled for tobacco exposure. Another study found no cocaine effect on IQ.⁸⁹ In the cohort studied by Hurt et al^{82 - 83} there was no impact of prenatal cocaine exposure on children's cognitive test scores at 48 months. In the oldest prospectively recruited cohort studied to date, Richardson et al⁹³ found no effect of prenatal cocaine exposure on any IQ scales at age 6 years, including verbal reasoning, and no association with children's academic skills.

The literature on prenatal exposure to cocaine has not shown consistent effects on cognitive or psychomotor development. However, 7 studies show that environmental factors such as caregiver (biological mothers vs kinship care or foster parents),^{79 ,89} whether or not that caregiver received case management or home visiting services,^{78 ,89} quality of the home environment,^{63 - 64 ,78 ,83} and maternal IQ⁷⁷ were statistically significant correlates of test scores.

Three studies of toddlers^{69 ,81 ,89} showed no association between prenatal cocaine exposure and receptive or expressive language scores on standardized measures (Table 3). Using a naturalistic language sample, Bland-Stewart et al⁶⁹ found that cocaine-exposed children produced different semantic categories than matched unexposed children. However, there were too few subjects to permit confounder control.

Of 6 studies, 3 from 2 cohorts found less optimal motor scores in the first 7 months of life following prenatal cocaine exposure (Table 4).^{75 - 76 ,97} No prospective study has identified a cocaine effect on motor development after age 7 months.^{75 - 76 ,89} Dempsey et al⁷⁴ found mothers' prenatal tobacco use (quantified by urine assays of cotinine rather than by self-report), but not cocaine use (quantified by benzoylecognine levels in meconium), was the major predictor of abnormalities in infant muscle tone at 6 weeks. No other prospective study of motor outcome^{75 - 76 ,79 ,89 ,97} following cocaine exposure used biological markers to measure tobacco exposure. It is not yet clear whether previously reported positive associations between prenatal cocaine exposure and less optimal early motor development may be a misattribution of tobacco effects.

Heterogeneous techniques used to evaluate behavior, attention, affect, and neurophysiology following prenatal cocaine exposure are not readily comparable across studies (Table 5). In the first year of life, visual habituation (an indicator of recognition memory and learning) was negatively associated with higher levels of cocaine exposure in 1 cohort⁸⁵ but not in 3 others.^{63 ,88 ,91} No cocaine effect was found on toddler play⁸⁰ or on observations of behavioral style during an infant motor assessment.⁶⁸ Problem-solving abilities did not differ between cocaine-exposed and unexposed preschoolers.⁶⁷

Differences in affective expression have been correlated with prenatal exposure to cocaine in 4 studies from 3 cohorts of infants younger than age 2 years. Alessandri et al⁶² found that 4- to 8-month-old cocaine-exposed children showed less arousal, interest, joy, or sadness during the learning task. In the same cohort, Bendersky and Lewis⁶⁶ reported no differences in maternal behaviors, but less joy and more negativity among 4-month-old infants with heavy cocaine exposure following a perturbation of the face-to-face interaction between mother and infant. Roumell et al⁹⁴ reported a bivariate association between prenatal cocaine exposure and decreased facial emotion after immunization, uncontrolled for other prenatal exposures. In studies of face-to-face interaction between mothers and infants, Mayes et al⁹² found heavy prenatal cocaine use correlated with less optimal maternal behavior and with decreased readiness for interaction among infants at age 6 months but not 3 months.

Diverse techniques have been used to assess neurophysiology in cocaine-exposed and unexposed infants aged 13 months and younger. Cocaine-exposed infants showed lower basal cortisol levels, but normal cortisol increase in response to the stress of venipuncture and no difference in amount of observed crying.⁸⁶ On electroencephalographic sleep studies at 12 months, cocaine-exposed children did not differ from unexposed children in sleep architecture, but infants whose mothers continued to use cocaine into the third trimester showed subtle reductions in spectral energies.⁹⁵ In 2 reports from a single cohort, assessments of heart and respiratory response to auditory, visual, and social stimulation at age 8 weeks found that cocaine-exposed children showed increased heart rate to social stimulation and a higher baseline respiratory rate, but were not more dysregulated in arousal modulation or observed behavioral state.^{65 ,71} Full-term cocaine-exposed infants showed better arousal modulation than their unexposed counterparts.⁶⁵

Prenatal cocaine exposure, independent of exposure to alcohol, has not been found to be associated with levels of behavioral disturbances detectable by standard scoring of epidemiologic and clinical report measures by parents and teachers.^{64 ,72 - 73 ,77 - 78 ,87 ,93} However, 2 studies in 1 cohort (1 study using a study-specific measure⁷² and the other⁷³ using a new and as-yet unreplicated method of scoring the Teacher Report Form of the Child Behavior Problem Checklist¹¹⁵ ) found less-optimal scores among cocaine-exposed children. Another research group^{90 ,93} found, after covariate control, an association between prenatal cocaine exposure and increased errors of omission, but not commission, on a continuous performance task.

Before summarizing our findings, we must acknowledge the limitations of our approach. Studies that meet our methodologic criteria may still lead to overestimation or underestimation of cocaine's impact. Prospective studies may yield biased results if there is differential attrition.⁹⁹ Less dysfunctional caregivers may be more likely to sustain study participation, creating differential retention of children with more favorable outcomes. Alternatively, caregivers of children with obvious impairments may be more willing to return for repeated assessments, leading to an overestimation of risk for poor outcomes.

Reliance on interviews alone to classify exposure, which was the state of the art when the cohorts reported here were recruited, entails unavoidable imprecision.¹⁴ In the absence of cumulative biological markers some cocaine-exposed children may have been misclassified as unexposed. Conversely, women who do admit cocaine use in interviews tend to be heavier users than those who deny use but whose use is detected by hair assays.¹¹¹ Generalization from atypical cases at the highest levels of exposure will lead to overestimation of the impact of prenatal cocaine exposure in the broader population of users. However, if a sample contains very few infants heavily exposed to cocaine,^{77 ,93} possible effects of heavier use may be statistically "diluted" by overaggregation of various levels of exposure into a single category.¹¹⁴

Four studies with positive^{69 ,75 - 76 ,94} and 1 with negative⁶⁸ findings have small sample sizes and must be interpreted with particular caution since they may overestimate cocaine effects due to the impact of a few outliers or underestimate effects because of insufficient power or sampling variation.

While acknowledging these limitations, we conclude that after control for exposure to tobacco and alcohol, effects of prenatal cocaine on physical growth are not shown.^{64 ,70 - 71 ,79 ,84 ,89 ,93} Researchers have not found a negative association of prenatal cocaine exposure, independent of environmental risk and exposure to other psychoactive substances, with developmental scores from infancy to age 6 years.^{63 - 64 ,70 - 71 ,77 - 79 ,82 - 83 ,85 ,89 ,91 ,93} However, sufficient information is not available to elucidate whether there are specific cocaine effects on developmental scores in the context of prematurity.⁹⁶

Prospective data in the language and motor domains are only available for children up to age 3 years.^{69 ,74 - 76 ,78 - 79 ,81} No effects on standardized language measures have been shown. Less-optimal motor development before age 7 months but not thereafter has been found by some investigators^{75 - 76 ,97} but not others.^{74 ,79 ,89} Recent research suggests that motor findings attributed to cocaine may in fact reflect heavy prenatal tobacco exposure.⁷⁴

Except for the work of 1 investigator,^{72 - 73} prenatal cocaine exposure independent of exposure to alcohol has not yet been found to be associated with levels of behavioral disturbance that are readily detected by standard scoring of epidemiologic and clinical report measures from parents and teachers.^{64 ,72 ,77 - 78 ,87 ,93} However, sophisticated experimental and physiological paradigms of uncertain clinical importance have detected possible effects of prenatal cocaine exposure. Of these, only the finding of decreased emotional expressiveness has been replicated in more than 1 study.^{62 ,66 ,92 ,94}

The differences between our conclusions and those of others show how methodologic rigor influences understanding of prenatal cocaine exposure. For instance, a respected research group recently concluded from a meta-analysis of 6 studies that prenatal cocaine exposure is associated with decreased competence in expressive and receptive language.⁹⁸ However, 5 of these studies^{29 ,37 ,43 ,46 ,51} were retrospective; 2 did not use masked assessors.^{37 ,57} In 2 samples, the majority of cocaine-exposed children were also exposed to opiates and methamphetamines.^{37 ,57} Furthermore, none of these studies analytically controlled for the possible effects of prenatal tobacco exposure, an established correlate of language impairment.¹¹³ Nevertheless, newspaper articles used the conclusions of the meta-analysis to declare that "because of cocaine-related receptive language impairments," "crack babies" would cost taxpayers an additional $42 to $352 million per year in special education services.¹¹⁶

When prenatal cocaine and tobacco exposure are compared dispassionately, it becomes clear how sociopolitical forces shape discrepant interpretations of similar scientific data. The mechanisms of nicotine and cocaine effects on the developing brain are similar, involving vasoconstriction, hypoxia, and perturbations of neurotransmitter networks.¹¹⁷ Prenatal tobacco exposure has been associated with infant mortality,¹¹⁸ moderate impairment of cognitive functioning,¹¹⁹ and a range of behavioral problems (which, unlike those associated with cocaine exposure, are detectable on relatively insensitive epidemiologic measures).¹²⁰ It has been calculated that low birth weight attributable to maternal smoking annually costs $263 million (1995 dollars) in excess direct medical costs for neonatal care alone.¹²¹ Despite increased health care costs imposed by their tobacco use, there are no sterilization campaigns for mothers who use tobacco. No pregnant women have been charged with child abuse for tobacco use in pregnancy. Teachers do not dread having a "tobacco kid" assigned to their class.

We have focused on cocaine as a suspected behavioral teratogen, since exaggerated views of its teratogenicity have provided the rationale for selectively targeting pregnant women who use cocaine for sanctions even more punitive than those imposed on women who use other illicit substances.^{3 ,8 ,122} Our focus omits 2 important considerations beyond the scope of this review. First, even if cocaine were as hazardous to a child's development as some claim, established teratogenicity (eg, that of heavy alcohol use) does not justify policies that violate the usual canons of medical ethics and civil liberties.³ Second, health providers should not ignore that cocaine use in pregnancy is often a marker for a mother-child dyad at risk for poor health and impaired caregiving due to factors ranging from infectious diseases to domestic violence. Addiction to any intoxicant may so impair parents that they abuse or neglect a child.¹²³ However, presumptive punitive sanctions imposed in pregnancy or at birth do not reduce these risks to the child. On the contrary, fear of prosecution may discourage pregnant and parenting women from seeking prenatal care and drug treatment,^{8 ,124} which have been shown to optimize infant outcome.¹²⁵ Stigma and negative expectations generalized from mothers to their children may in themselves impede the children's academic progress.¹⁰¹ Care of families affected by substance abuse should be comprehensive and not irrationally shaped by social prejudices that demonize some drugs and drug users and not others.¹²³

Much is still unknown about the effects of prenatal cocaine exposure. Research on prenatal marijuana and tobacco exposure suggests that, even if no drug effects are found between the ages of 6 months and 6 years, the increasing cognitive demands and social expectations of school or puberty may unmask sequelae of exposure not previously identified.^{126 - 127} Cumulative environmental risk and protective factors may also exacerbate or moderate negative cognitive and behavioral outcomes as children mature.¹²⁸ However, among children up to 6 years of age, there is no convincing evidence that prenatal cocaine exposure is associated with any developmental toxicity different in severity, scope, or kind from the sequelae of many other risk factors. Many findings once thought to be specific effects of in utero cocaine exposure can be explained in whole or in part by other factors, including prenatal exposure to tobacco, marijuana, or alcohol^{64 - 65 ,70 ,74 ,78 ,84 ,86 ,90 ,95} and the quality of the child's environment.^{63 - 64 ,66 ,68 ,71 ,73 ,77 - 79 ,83 - 84 ,86 - 87 ,89 - 90}