European Child & Adolescent Psychiatry

, Volume 16, Issue 4, pp 243–253

Consequences of prenatal toxin exposure for mental health in children and adolescents

A systematic review


    • Department of Child HealthUniversity of Aberdeen Medical School, Royal Aberdeen Children’s Hospital
  • Louise Ross
    • Department of Child HealthUniversity of Aberdeen Medical School, Royal Aberdeen Children’s Hospital

DOI: 10.1007/s00787-006-0596-6

Cite this article as:
Williams, J.H.G. & Ross, L. Eur Child Adolesc Psychiatry (2007) 16: 243. doi:10.1007/s00787-006-0596-6


Drug use during pregnancy is common and the developing foetus may be exposed to a range of environmental toxins that have long-term consequences for neurodevelopment. We conducted a systematic review of the literature to explore the results of longitudinal cohort studies that have examined this question. Out of 2,977 abstracts identified, 7 previous systematic reviews and 95 original articles met further selection criteria. These mostly addressed the neurodevelopmental effects of exposure to lead, polychlorinated biphenyls, mercury, cocaine, alcohol, marijuana, cigarettes and antidepressants. Radiation, opiates, steroids, amphetamines and caffeine have received much less attention. Findings are difficult to interpret because risk factors tend to cluster together and interact. However, some findings are consistent. Lead and PCB’s have a general effect on brain development, whilst marijuana and alcohol appear to have long-term effects specifically on attentional skills. The effects of alcohol increase with maternal age and binge drinking is more important than average intake. The effects of cocaine diminish with age and are largely mediated through psychosocial factors, whilst the relation between smoking and later delinquency is largely mediated by genetically inherited factors. Exposure to toxins during pregnancy may constitute an important but relatively unacknowledged cause of child psychiatric morbidity.


prenataltoxinschildadolescentneuro-developmentmental healthADHDpsychiatric disorderrisk factorintra-uterine


Mental health problems in childhood are increasingly recognised to constitute a serious source of health morbidity with consequences for adult as well as child health [13]. Therefore, measures that might impact upon their prevalence are likely to be of important social and economic benefit. In considering the aetiology of these problems social and genetic factors have received much attention, but environmental factors may also potentially play an important role, especially with respect to stresses occurring during prenatal development. This includes factors such as prenatal infection, birth injury, prematurity or low birth weight, but also some prenatal influences that may have more subtle adverse effects on development. Indeed, the developing foetal brain may be sensitive to relatively minor environmental stressors. One pointer to this is evidence that some childhood behaviours and autistic spectrum disorder may be linked to relatively slight differences in the sex hormone and glucocorticoid environment [48].

Prenatal stressors can result in disorders ranging from severe sensory impairment, cerebral palsy and learning disability that results from infection or prematurity, to (as will be discussed below) mild depression and attentional difficulties in the absence of any clear cut sensory or neurological impairment. Clearly, children with more severe neurodevelopmental impairment are also at increased risk of mental health problems [912], and so in considering risk factors for mental health problems, risk factors for neurological problems cannot be ignored. Similarly, whilst psychiatric disorder may potentially arise as a result of a very broad range of risk factors, it may also be an important question to ask whether risk factors are specific for certain outcomes. Therefore, in considering potentially reversible causes of psychiatric disorder in children, it may be more helpful to classify primarily by risk factor as opposed to disorder, to see whether some psychiatric outcomes are more likely than others. In this article we look specifically at the consequences of prenatal toxin exposure on neurodevelopmental outcomes.

In the National Survey of Drug Abuse conducted in the United States in 2002/2003 [13], 4.3% of pregnant mothers reported using illicit drugs during pregnancy. About 20% of pregnant women smoke [14] and 33% drink alcohol during pregnancy [13]. Furthermore, as discussed further below, drug use is most likely heavily under-disclosed, and actual consumption could be 3–4 times that reported by both pregnant and non-pregnant individuals [1517]. If the drugs consumed during pregnancy do increase the risk of poor subsequent mental health, the frequency of use could mean that they constitute an important risk factor for child neurologic and psychiatric morbidity.

However, whilst the hypothesis that prenatal substance exposure causes adverse outcomes seems very plausible, confirmation of it presents a considerable challenge. Essentially, the hypothesis predicts a relationship between exposure to a risk factor at one time-point and consequent manifestation of problems in the same individual some years later. This relationship must be demonstrated within a naturalistic context where risk factors cluster together rather than occur in isolation. Poverty tends to be associated with a poorer education, different attitudes, poorer living conditions and greater levels of exposure to adverse risk factors both during and after birth. In addition, those factors that drive the exposure to a risk factor may be more important in affecting outcomes than exposure to the risk factor itself. For example, parental attitudes to the foetus may determine whether the mother uses drugs during pregnancy and also influence how she approaches child care. Thirdly, birth may have only a partial effect on exposure to a risk factor. For example maternal smoking may affect the foetus through maternal blood or passive inhalation. Polychlorinated biphenyl compounds may affect the infant via maternal blood or breast milk. Therefore, if the substance is a risk factor for neurodevelopment, it is not easy to be clear if its effects occur before or after birth.

The purpose of this article is to review the evidence that prenatal exposure to toxins leads to compromised neurodevelopment and mental health problems in children. We sought to use a systematic reviewing approach to identify possible risk factors, the methodological problems faced by these studies (and possible solutions employed), the degree of risks these factors posed, and whether there was evidence for specific mental health effects.


A detailed search strategy was devised based on mesh terms and keywords for:
  1. (a)

    Different types of longitudinal studies including cohort, follow-up, longitudinal and epidemiological studies.

  2. (b)

    Risk period including puerperium, prenatal, postnatal, intra-uterine and related keywords.

  3. (c)

    Different words for measures including, measure, predictors, scales inventory.

  4. (d)

    General risk terms including risk, high risk, risk factor.

  5. (e)

    Child infant related terms including child, infant adolescent, pediatric and related words.

  6. (f)

    Specific risk terms for alcohol, tobacco, recreational drugs including cocaine, marijuana, heroin, opiate, methadone, substance abuse, toxin, lead, polychlorinated biphenyls (PCBs), anti-epileptic and anti-depressant, steroid, caffeine.

  7. (g)

    Specific and general mesh words and terms for childhood mental health disorders as specified in the inclusion criteria including but not only conduct disorders, autism and related terms, mental retardation, anxiety disorders, cognitive impairment, behaviour disorders.

A full search strategy devised for MEDLINE and adapted for EMBASE, PSYCHINFO and SSCI is available on request.

Inclusion and exclusion criteria are as follows:


Types of studies

Birth cohort studies, prospective studies, longitudinal studies, twin studies and similar prospective epidemiological studies that examined prenatal, postnatal and/or early childhood risk factors, predictors and measures used to assess the risk factors, and their association with childhood mental health disorders.

Definition/description of types of childhood mental health disorders (using ICD-10 code [18]):
  • F10-19 Mental and behavioural disorders due to psychoactive substance use

  • F32-39 Mood affective disorders excluding mania and bipolar affective disorder

  • F40-48 neurotic stress related and somatoform disorders

  • F50-51 Eating disorders and non-organic sleep disorders

  • F70-79 Mental retardation

  • F80-89 Disorders of psychological development

  • F90-98 Behavioural and emotional disorders with onset usually occurring in childhood and adolescence

However, as children may present with a single comorbidity rather than a single disorder, a combination of disorders may be included too.

Types of participants

The target population was children from 0 up to 18 years of age followed from birth.


  1. (i)

    Any risk factors identified as being associated with the prenatal, perinatal and postnatal period and up to 2 years of age that are linked with disorders specified above

  2. (ii)

    Any measures or scales used to assess the effect of the risk factors on the childhood mental health disorders already specified above.


Exclusion criteria

  1. (i)

    Risk factors not identified as being associated with the prenatal period

  2. (ii)

    The following mental health disorders classified by ICD-10 code

  • F00-09. Organic disorders

  • F20-29. Schizophrenia

  • F30, 31. Manic episode; Bipolar disorder

  • F52-59. Sexual dysfunction; disorders associated with the puerperium; psychological and behavioural factors associated with disorders classified elsewhere; Abuse of non-dependence producing substances; Unspecified behavioural syndromes associated with physiological disturbances and physical factors

  • F60-69. Disorders of adult personality and behaviour


Literature search

An initial search of four data bases, (Medline (1966–2005), Embase (1980–2005), Psychinfo (1967–2005) and Social Science Citation Index (1986–2005) identified 2071 prospective studies that looked at neuro-developmental outcomes for risk factors measured prenatally or within the first year of birth. Of these 270 were identified as being potentially relevant of which 175 of these were factors other than prenatal toxins. These included maternal depression, maternal attitudes and behaviour, maternal physical health problems, infections, intrapartum complications, infant’s physical health problems, infant/toddler behavioural problems and multiple risk factors. A second search for more specific risk factors relating to prenatal toxins yielded 897 abstracts. The second search identified papers dating from after the systematic reviews for cocaine, smoking, antidepressants pcb’s and lead. The search process is outlined in Fig. 1.
Fig. 1

Search process

Seven systematic reviews [1925] were also identified and are summarised in Table 1. For the purposes of this review we elected to incorporate the findings of previous systematic reviews, rather than re-reviewing papers that they had previously covered. After excluding irrelevant or previously reviewed papers, 100 studies were included in this review including seven systematic reviews. In view of the highly varied nature of outcomes, the results did not lend themselves to meta-analysis and the following is a narrative review of the findings. A summary of all the primary studies included is outlined in Supplementary Table 1.
Table 1

Systematic reviews

Authors and year of publication

Risk factor

Study designs included

Outcomes measured/identified


Eskenazi and Castorina 1999 [19]

Prenatal and postnatal environmental tobacco smoke

13 pros cohort 2 case control 1 X section

Neuro-developmental and behavioural problems

Medline Bio-sis Psych-info

Ribas-Fito et al. 2001 [20]


6 Prospective 1 X section


Medline conference reports Refs in inc journals

Goldstein 1999 [21]

Selective serotonin reuptake inhibitors

4 cohort controlled 5 prospective surveys

Neurobehavioural development in one study

Medline, Embase, Derwent drug file, Psychinfo, Sci-search

Peterson et al. 1995 [22], Frank et al. 2001 [23]


33 studies reported on neuro-behavioural outcomes 36 prospective studies

Behaviour assessment psychological abnormalities Cognitive, language and motor skills neurobehaviour and attention

Unspecified medical and psychological databases Medline psychological abstracts

Wisner et al. 1999 [24]


4 prospective studies (1 study looked at CMH)

No differences found in cognitive function, temperament, or general behaviour

Medline healthstar

Pocock et al. 1994 [25]

Environmental lead

5 prospective studies, 21 cross sectional studies

5 prospective studies showed no association of cord blood lead or antenatal blood lead with subsequent IQ

Majority of studies identified by personal knowledge Search of MEDLINE and Toxline revealed no further studies

Risk factors fell broadly into three groups; substance misuse, prescribed drugs and environmental toxins. We found the question of whether environmental toxins are associated with later neurodevelopmental impairment the most straightforward, and so we start at this point.

Environmental toxins


Lead has long been considered as an environmental stressor on brain development and longitudinal studies of its effects were first initiated in the late 70s in Boston [26], and then later in Cincinatti [2728], Port Pirie [29] and Kosovo [30]. These studies consistently confirm that lead exposure is associated with poorer general cognitive outcome, poorer performance on tests of executive function and working memory and increased behaviour problems. Breast feeding appears to offer some neuroprotection. Meta-analyses have set the effect size at 1–2 IQ points [25] or 2–3 IQ points [31] for a doubling of blood lead from 10–20 μg/dl. Whilst this may seem a trivial amount, through the laws of natural distribution, it will be a much larger drop for many individuals in a large population. According to Ris et al. [28], a distinct neurobehavioral phenotype has not emerged from these studies and postnatal levels are stronger predictors than prenatal levels. This would suggest that postnatal exposure is more important than prenatal exposure, but Ris et al. [28] also make the point that blood lead levels stabilise with age, as blood lead levels come into equilibrium with levels sequestered in deep physiological depots such as bone.

Polychlorinated biphenyls (PCBs)

Polychlorinated biphenyls are industrial pollutants that persist in the environment, accumulate in the food chain and have caused much concern because of their widespread toxic effects at low concentrations. Their lipophilic qualities make neurodevelopmental effects a particular concern. Ribas-Fito et al. [20] identified seven follow up studies that evaluated the effect of prenatal and postnatal exposure to PCBs. In two of the studies, children were highly exposed following accidents. They concluded that prenatal exposure to PCB’s had a subtle effect on neurodevelopment.

This Ribas-Fito et al. [20] review included a group of children from Taiwan who were followed up by Lai et al. [32] when the children were 10–13 years of age. Exposure was associated with increased scores on both the Rutter B scale and the Child Behaviour Checklist, as well as a 3 point drop in full scale IQ. Age had a recovery effect on IQ and behaviour as group differences diminished with age. Jacobson and Jacobson [33] followed up a sample exposed to PCBs from Lake Michigan. Consistent with the Lai study, they found effects on executive function tasks requiring sustained attention, attentional flexibility, and working memory. Children also showed evidence of increased impulsivity and poor attention on behavioural questions derived from the infant Bayley scales. Interestingly, even though breast feeding may expose the infant to additional PCBs it appears to have a protective effect. It may be that the growing brain may be more susceptible to the toxins in utero than in breast milk, and breast feeding confers neuroprotective properties which outweigh the effects of additional PCB exposure [20]. Unfortunately no psychiatric assessment has been performed in these samples and the authors do not distinguish the behavioural effects from the motor and cognitive effects.


Two major prospective studies have examined the relationship between mercury exposure and neurodevelopment with apparently conflicting results. Both studies have investigated populations that have diets rich in seafood and are therefore highly exposed to mercury. A Seychelles study [34] has found that increased mercury exposure is associated with better cognitive development and reduced behaviour problems at 5–6 years of age after controlling for multiple confounding factors including the quality of the home environment (see below). They suggest that mercury exposure is actually a proxy measure for quality of nutrition and exposure to long-chain fatty acids and other trace metals such as selenium, which (like breast milk) may confer neuroprotection. In contrast a study in the Faroes [35] has found diminished performance on some cognitive sub-tests at age 7. Again they note confounding factors. The amount of whale meat (main source of mercury) varies geographically across the islands and is eaten less in the capital where there is greater access to day care and education. It is not clear why the two studies reach opposing conclusions. The statistical modelling procedures differ, but it also seems likely that neurodevelopmental risk related to mercury exposure is very much modified by the associated nutritional and cultural environment, as suggested by Davidson et al. [36] and so the immense cultural differences between the Faroes and Seychelles could also offer an explanation.

Prescribed medications

Few drugs have been prescribed to women during pregnancy at levels high enough to be the subject of long-term epidemiological studies. Drugs that have been studied are prenatal steroids, antidepressants and anti-convulsants.


Of the antidepressants, only Serotonergic Re-uptake inhibitors (SRI’s) have been the subject of prospective study. Goldstein and Sundell [21] examined 12 prospective birth cohort studies that examined whether SRI’s had any adverse effect on the foetus if taken during pregnancy. None of the studies found adverse effects of SRI’s on the development of language, IQ or ‘neonatal adaptation’. However, numbers of pregnancies were low (n = 7–686, mean = 125, only 4 had n > 100). Therefore, they would not have been sensitive to an uncommon adverse outcome. Croen et al. [37] have also published some evidence that antidepressant use during pregnancy may increase the prevalence of autistic spectrum disorder after controlling for maternal psychiatric history.


Retrospective studies of sodium valproate (which may have some neuropharmacological effects similar to SRI’s) exposure during pregnancy have identified a range of problems that may develop, that form the foetal valproate syndrome [38] Rasalam et al. [39], using a quasi-prospective design, found that autistic spectrum disorder in the children of mothers exposed to valproate in pregnancy increased by at least 8 times. The studies reviewed by Goldstein and Sundell [21] would most likely not have been of sufficient size to have detected these potentially important effects.

Prenatal steroids

There is some suggestion that endogenous maternal steroids, associated with maternal stress are associated with an increased rate of behaviour problems in children. As this review is concerned with stressors of exogenous origin, such a question is beyond its scope. However, it raises the question as to whether steroids administered prenatally, also lead to long-term adverse behavioural outcomes. According to Owen et al. [6], 7% of mothers receive prenatal administration of synthetic glucocorticoid in an effort to promote lung maturation in foetuses at risk of preterm delivery. Such widespread administration of this medicine may influence levels of childhood psychiatric morbidity. French et al. [40] reviewed a cohort of 541 children who were born preterm (20–32 weeks). Despite no differences in intellectual performance on an IQ test, those infants who had received three or more courses of corticosteroid injection, showed higher levels of aggression on the CBCL and distractibility on the parenting stress index at 3 years, and greater aggression but not hyperactivity at 6 years. These results should be considered very cautiously. Firstly, the later the birth, the more corticosteroid the infants received, so that high corticosteroid babies are not really comparable to low corticosteroid babies, differing in the major factor of gestational age at birth. Also, only a few behavioural measures are reported where many were taken and the group differences of P = 0.04 may have arisen by chance.

Unprescribed drugs

Smoking during pregnancy

The question of whether smoking has effects on the foetus, affect later mental health, is longstanding and epidemiologic studies in this area highlight the problems with prenatal drug exposure research more generally. Eskenazi and Castorina [19] reviewed 17 epidemiologic studies that investigated active smoking during pregnancy. Tobacco smoke during or after pregnancy was associated with slightly poorer academic achievement. The picture with respect to behavioural problems was inconsistent. Although studies found increased symptoms of attention-deficit, hyperactivity and aggression among children of smoking mothers, there are major confounders. Firstly, common inherited factors may predispose to both smoking in adulthood and psychopathology in childhood. This has been termed the latent Cd (conduct) factor by Silberg et al. [41] who investigated a twin study. Maughan et al. [42] also analysed data from twin studies and concluded that the relationship between maternal smoking and later conduct problems is likely to be mediated genetically. In support of this, Eskenazi and Castorina [19] note a Finnish study that found that paternal smoking history was just as strongly related to increased ADHD and aggression as maternal smoking during pregnancy. They also note the relationship between mental health problems such as depression in the mother and maternal smoking. Sexton, 1990, cited in Fried and Watkinson, [43] compared parents who gave up smoking with those who continued to smoke throughout pregnancy. But this is to compare mothers who continued to indulge their habit (presumably against advice) with parents who made a determined effort to change their habits to protect their unborn child. Attitudes to care may therefore also be an important mediating variable to consider, and indeed Wakschlag and Hans [44] found smoking to correlate with the quality of the early care-giving environment and subsequent conduct disorder. Therefore, there are a number of aspects of the maternal mental state, attitudes and personality that predispose towards both continued substance use in pregnancy and increased symptoms of hyperactivity and aggression.

It seems unlikely that any single measure of maternal mental health is likely to successfully control for all these variants as used by Batstra et al. [45] and Weissman et al. [46]. One useful approach may be to compare the differential effects of different drug exposure where dissociations may be informative. The Ottawa Prenatal Prospective Study (OPPS; 47–50) followed up a relatively middle-class low-risk population and compared the effects of marijuana and cigarette exposure prenatally on psychometric test performance between 9 years and 12 years of age [40, 44]. Differences in personality, mental health and attitudes between the populations of women who smoked nothing, cigarettes and marijuana were measured with the Home Environment Questionnaire and the NEO five-factor personality inventory. The Home (Home Observation for Measurement of the Environment) questionnaire [51] is a tool designed for assessing the effects of the home environment in terms of the quantity and quality of stimulation and support available. Women who smoked cigarettes scored lower on conscientiousness, supervision and socio-economic status scores and educational level. After controlling for confounders the IQ of children of heavy smokers was significantly lower than non-smokers but there were no differences in performance on attentional measures. The converse was true for children of marijuana smokers. This group was followed up between 13 years and 16 years and is discussed in more detail below, but here we note that the effects of reduced IQ but not attentional impairment were still evident at that age. In summary, tobacco smoke exposure is associated with poorer academic achievement and heavy but not light exposure is associated with a diminished IQ. The evidence that it predisposes towards psychiatric disorder is poor, as many confounding factors offer alternative more fitting explanations, and psychometric measures do not find effects on attentional skills.


At least since Foetal Alcohol Syndrome was described, with its features of learning disability, hyperactivity and microcephaly, it has been asked what effects lesser exposure to alcohol during pregnancy might have on the developing foetal brain. The first prospective longitudinal study to examine this question commenced in Seattle in 1974. At this time drinking during pregnancy was commonplace, and use of other drugs was low. The study found dose-related effects of alcohol on general intelligence and particularly attentional skills throughout childhood that continued into adolescence. They estimated that an average intake of 50 g of alcohol per day would be associated with a seven point decrement in full scale IQ [52] with deficits in focussed attention, working memory and arithmetic. However, the relationship between exposure and outcome appears to be complex. Goldschmidt et al. [53] found that a threshold rather than a linear model better predicted the effects, as there were virtually no effects in relation to low level drinking. Even so, it may then be peak levels rather than average levels that matter. Jacobson et al. [15] found evidence for sub-groups of more vulnerable children. They found that the risk increased with maternal age and suggested that as the ratio of body fat to water increases, so do peak blood alcohol concentrations. They also found a history of alcohol misuse to be predictive, and that whilst alcohol did not affect full scale IQ, a freedom from distractibility index was very much affected. They found no effects from drinking under seven drinks per week. In their population of black women in the Detroit area, Bailey et al. [54] found that binge drinking but not absolute level of exposure predicted cognitive and behavioural outcomes in 7 year olds. After controlling for co-variates, 41% children who were exposed to ≥5 drinks per occasion at least once per 2-weeks, had an IQ score <70 compared to 24% in children of non-drinkers (according to Bailey et al. [54] low average scores are consistent with African-American children typically scoring 1 SD below the ‘norms’ developed in Caucasian populations) and more likely to exhibit delinquent behaviour [17% vs. 7%). Absolute levels of alcohol exposure were not predictive in their study. This is consistent with concerns expressed by Guerri et al. [55] that the emphasis of public health advice with respect to alcohol consumption during pregnancy should be on consumption level per occasion rather than a weekly average.


Marijuana has been less extensively studied than tobacco and the same problems occur. The OPPS studies [4750], and Fried and Watkinson [43] are informative though, because they carried out very detailed psychological testing on a group of children prenatally exposed to either tobacco or marijuana. They applied a battery of attention and memory tasks including a continuous performance task, the Wisconsin card sorting task, and the WISC arithmetic test. Multiple measures were reduced to five measures of attention and memory by principal component analysis. As mentioned nicotine but not marijuana exposure was associated with ability to encode and retain items in memory, whereas marijuana had greater effects on ability to maintain attention. The children have now been followed up into adolescence and the picture has remained consistent. Furthermore, children who had been exposed to marijuana prenatally were found to exhibit increased depressive symptoms from 10 years to 12 years of age and poor attentional skills were found to mediate the relationship between prenatal marijuana exposure and consequent delinquency [51].


The question of whether cocaine has an effect on the brain of the developing foetus has been the subject of much research. Peterson et al. [22] reviewed 85 studies of which 48 were prospective in design. Low and unrepresentative sample sizes, failure to control for other drugs of misuse and inappropriate statistics (such as multiple t-tests) were common problems, as was unreliable history data. In a sample of over 3,000 mothers, only 11.1% reported illicit drug use during pregnancy. However, meconium analysis found 44.3% of newborns positive for these drugs. The possibility that illicit drug use is so widespread during pregnancy may be an important point to be considered in any birth cohort study. Despite these weaknesses the authors found good evidence that cocaine use is associated with reduced birth weight and head circumference, but not gestational age. Peterson et al. report a range of neurobehavioural effects of prenatal cocaine exposure in infants including tremulousness, irritability, startle, jitteriness, poor feeding and disturbed slow-wave sleep. Since 1995, reports from 15 cohort studies have been published. The most comprehensive of these has been described in a series of reports by Singer et al. [5760]. They measured both maternal IQ and quality of the care and home environment, which have both emerged as important determinants of cognitive development. At 1 year they found a slight decrease in auditory comprehension (86 vs. 89) with no difference in expressive language. Also language scores correlated with meconium cocaine metabolite levels and reported cocaine exposure, whereas alcohol, tobacco and cigarettes did not. Effects on IQ were also evident at 2 years. Cocaine was not associated with decreased IQ at 4 years though they showed some decreased performance on sub-tests and fewer of the cocaine exposed group had an IQ more than 100. Surprisingly they found that the effect of tobacco and alcohol were not significant as confounding factors. A possible reason for this was retrospective reporting of prenatal exposure, which Jacobson et al. have shown can lead to missed effects. Other studies present a largely consistent picture and the most recent [61] does not report long-term adverse effects of prenatal cocaine exposure.


Other risk factors have received relatively little attention. This is presumably either because of a lack of theoretical rationale to justify studies or because of a lack of suitable natural conditions that allow epidemiological studies to take place. For example, we found no articles that addressed the risks of ecstasy, presumably because exposure during pregnancy is too uncommon and difficult to identify. We identified a single cross-sectional case–control study that suggested coumarin exposure may be associated with decreased IQ [62], but no longitudinal study. We identified individual studies that examined radiation, caffeine, amphetamines and opiates. Another case–control study examined offspring of mothers exposed to increased radiation during the Chernobyl disaster and found increased mental retardation which they suggest is secondary to effects on thyroid function [63].

Viggedal et al. [64] followed up 17 infants exposed to benzodiazepines, and suggested a general effect on mental development at least up to 18 months of age. Eriksson et al. [65] reported that 65 children exposed to amphetamines showed increased behavioural problems and poorer academic performance, but confounding factors were not controlled for. Barr and Streissguth [66] did not find caffeine exposure to be associated with any adverse outcome. Curiously, we did not find studies that examined the independent effects of opiate exposure. Several studies did investigate it in the context of polysubstance misuse [6769] and found exposure to be related to poorer cognitive outcome.

Overall conclusions and general points

Despite the complexity of the issues facing epidemiologists who wish to know more about the effects of prenatal toxin exposure, research efforts over the last 30 years have yielded some consistent findings. Alcohol and marijuana taken at levels that are relatively commonly seen in non-addicted individuals have long-standing adverse effects. In contrast cocaine and opiate seem to have little effect in utero and their effects are most likely mediated through the psychosocial functioning of the mother.

It seems that substances may have mechanism-specific effects, rather than just causing psychiatric problems as simply another consequence of a general brain insult. Tobacco use exposes the foetus to a wide range of toxins, which appears to affect IQ but not executive function. Marijuana conversely, affects attentional skills but not IQ, and specifically put older children and adolescents at risk of depression and conduct disorder. However, PCBs and alcohol affect both executive function and IQ suggesting that effects can be more global. In addition the effects of the prenatal toxin exposure vary with age. The effects of PCBs, and possibly cocaine, ameliorate with age, whilst the effects increase with marijuana. Marijuana has been specifically associated with ADHD, whereas sodium valproate has been linked to autistic spectrum disorder. It is possible that the closer study of these effects could inform us with regards to the brain bases for these disorders in addition to their obvious public health implications. Indeed, the emerging picture suggests that specific drugs could be associated with characteristic behavioural phenotypes of the sort currently related to genetic disorders.

A number of factors modulate the adverse effect of the toxin on neurodevelopment. For toxins such as PCB’s and mercury, it appears that a nutritious diet including breast milk, confers neuroprotection. For alcohol, maternal age increases the risk. Parental IQ and the quality of care-giving environment are vital factors for conferring resilience. HOME (Home Observation for Measurement of the Environment) is a widely used tool involving interview and observations carried out in the home environment that appears to be an effective measure of this [51].

A caveat must be that these findings rely heavily on complex models that can make many assumptions. Relationships between variables are not always linear e.g., Goldschmidt et al. [53]—therefore this cannot be assumed when adjusting for nuisance variables. There is also the question as to whether measures accurately reflect the relevant variation of the key variable they intend to measure. For example, the study of alcohol exposure has shown that average dose is less important than the frequency of high doses (the amount of bingeing) and that contemporaneous reporting is more important than retrospective reporting. The variable relevance of lead levels at different ages may reflect varying stability of levels rather than varying vulnerability of the developing brain to exposure. At the level of the outcome variable, few studies have measured psychiatric outcomes in terms of diagnoses, and they have rarely justified their choices of neuropsychological measures in terms of functional effects.

Most studies lack attention to psychiatric diagnosis. We found very few that had used formal diagnosis. A few have used self-report questionnaires, the reliability of which is questionable for many reasons, including literacy problems among parents. The use of attentional measures has been undertaken without reference to the effects of these on day-to-day functioning. Attention to paternity effects is also minimal. There may be understandable methodological reasons for this, but it is obvious that attitudes to substance use and partner choices may bear some common ground.

In conclusion, epidemiological studies of prenatal toxin exposure to date suggest that particular substances can have either general or more specific effects on the developing brain. Substance use during pregnancy is very common and may constitute a more important preventable risk factor for neurodevelopmental disorder in childhood, than has previously been acknowledged. In addition further description of the effects of prenatal exposure will be important in understanding the brain-behaviour relations underlying neurodevelopmental disorder.


We are grateful to Christopher Gillberg, the Scottish ABCD study group and Amudha Poobalan for helpful comments during the design of the search strategy.

Supplementary material

Copyright information

© Steinkopff Verlag 2007