University of Groningen
Gene-environment interactions in disruptive behaviors
Ruisch, Hyun
DOI:
10.33612/diss.136546089
IMPORTANT NOTE: You are advised to consult the publisher's version (publisher's PDF) if you wish to cite from
it. Please check the document version below.
Document Version
Publisher's PDF, also known as Version of record
Publication date:
2020
Link to publication in University of Groningen/UMCG research database
Citation for published version (APA):
Ruisch, H. (2020). Gene-environment interactions in disruptive behaviors. University of Groningen.
https://doi.org/10.33612/diss.136546089
Copyright
Other than for strictly personal use, it is not permitted to download or to forward/distribute the text or part of it without the consent of the
author(s) and/or copyright holder(s), unless the work is under an open content license (like Creative Commons).
Take-down policy
If you believe that this document breaches copyright please contact us providing details, and we will remove access to the work immediately
and investigate your claim.
Downloaded from the University of Groningen/UMCG research database (Pure): http://www.rug.nl/research/portal. For technical reasons the
number of authors shown on this cover page is limited to 10 maximum.
Chapter
2
Maternal substance use during
pregnancy and offspring conduct
problems: a meta-analysis
Published as:
Ruisch IH, Dietrich A, Glennon JC, Buitelaar JK, Hoekstra PJ. Maternal substance use during
pregnancy and offspring conduct problems: a meta-analysis. Neuroscience and
Biobehavioral Reviews. 2018; 84: 325-336.
Acknowledgements, funding & declarations
This work is supported by the European Community's Seventh Framework Programme (FP7/2007-2013) under grant agreement no. 602805 (Aggressotype) and no. 603016 (MATRICS). Jan K. Buitelaar has been in the past years a consultant to / member of advisory board of / and/or speaker for Janssen Cilag BV, Eli Lilly, Medice, Shire, Roche, and Servier. He is not an employee of any of these companies, and not a stock shareholder of any of these companies. He has no other financial or material support, including expert testimony, patents, royalties. All other authors declare that they have no conflicts of interest.
Abstract
We conducted meta-analyses of relationships between highly prevalent substance use
during pregnancy and offspring conduct disorder problems. In total 36 studies were
included. Odds ratios (ORs) were 2.06 (1.67-2.54, 25 studies) for maternal smoking, 2.11
(1.42-3.15, 9 studies) for alcohol use, and 1.29 (0.93-1.81, 3 studies) for cannabis use, while
a single study of caffeine use reported no effects. Our meta-analyses support an association
between smoking and alcohol use during pregnancy, and offspring conduct problems, yet
do not resolve causality issues given potential confounding by genetic factors,
gene-environment interactions, and comorbidity such as with attention deficit hyperactivity
disorders. Future studies should use genetically sensitive designs to investigate the role of
pregnancy substance use in offspring conduct problems and may consider more broadly
defined behavioral problems.
Introduction
Conduct disorder (CD) is a prevalent behavioral disorder characterized by a pattern of
severe antisocial behaviors in which the basic rights of others or major age-appropriate
societal norms or rules are violated, with an estimated prevalence between 4 and 16% in
school age boys and between 1 and 9% in school age girls (1). Symptoms include aggression
to people and animals, destruction of property, theft, and other rule-breaking behaviors (2).
Comorbid externalizing behaviors are frequent, in particular oppositional-defiant disorder
(ODD) and attention-deficit/hyperactivity disorder (ADHD) (1–4). CD is considered a
precursor of antisocial personality disorder in adulthood and about 25% of girls and 40%
of boys with CD are estimated to eventually develop this disorder (4–6). Not surprisingly,
CD causes a considerable societal burden as well as significant negative functional outcomes
for affected individuals (4,7).
CD is a heterogeneous disorder and both genetic and environmental factors are involved in
its etiology (3,4). Early adversities, originating from the prenatal environment, may exert
programming influences on the developing fetus and thereby play a role in the etiology of
many mental disorders including CD (8). A widespread problem is substance use during
pregnancy with estimated prevalence’s of maternal smoking between 20 and 30%, alcohol
use around 15%, and cannabis use between 3 and 10% of pregnant women (9). Cigarette
smoke contains many potentially hazardous agents including carbon monoxide, nicotine,
polycyclic aromatic hydrocarbons, and heavy metals (10,11). Maternal smoking during
pregnancy has been associated with abnormal development of the central nervous system
(12) and behavioral problems in the offspring (13,14). Ethanol, the principal alcohol in
alcoholic beverages, is a neurotoxic agent and intrauterine exposure leads to fetal alcohol
spectrum disorders and related behavioral problems (15,16). Cannabis is one of the most
commonly used recreational drugs (17), probably also during pregnancy (18). Cannabis use
during pregnancy may affect the neurocognitive development in the offspring (19–22).
Furthermore, caffeine is a psychoactive substance used daily by approximately 75 to 93%
of pregnant women (23,24). Human studies regarding caffeine intake during pregnancy and
offspring neuropsychiatric outcomes are mostly inconclusive (25).
Currently, no meta-analytic studies have focused yet on the role of the prenatal environment
in relation to CD. Our goal was therefore to summarize and expand current knowledge on
this topic by conducting meta-analyses of highly prevalent and preventable maternal
substance use during pregnancy including the use of cigarettes, alcohol, cannabis, and
caffeine, and offspring risk of conduct problems.
Methods
Search strategy
We conducted a systematic search in the databases of the U.S. National Library of Medicine
(PubMed/MEDLINE), EMBASE, and PsycINFO, using the following search terms: (“conduct
disorder” OR “conduct problems” OR “conduct symptoms”) AND (pregnancy OR prenatal OR
“intra uterine”) AND (smoking OR cigarettes OR nicotine OR alcohol OR marijuana OR
cannabis OR weed OR caffeine OR coffee). We limited the search to human observational
studies in English, published in peer reviewed journals between January 1
st1990 and
November 1
st2016. Review articles, conference abstracts, posters, whitepapers, and other
grey literature were not included. Table S1 (supplementary material) provides an
overview of the eligibility criteria for the included studies. We did not register a
meta-analysis protocol.
Data collection
Electronic searches and screening of titles and abstracts were conducted by two authors
independently (IHR and AD). Any discrepancies as well as any issues encountered during
literature review were resolved by discussion, if necessary with a third author (PJH). To
determine if a study met our eligibility criteria full text articles were read. Figure 1 provides
an overview of the literature search and review process. From a total of 360 unique citations,
a total of 36 unique studies were included in all meta-analyses.
Data extraction was performed by the first author. The following data was extracted
systematically: study design, sample, demographics, exposure, outcome measures, and
statistical analyses. Study quality was assessed using the Newcastle-Ottowa Scale (NOS) for
non-randomized studies (26). A detailed description of the NOS-criteria, as applied to our
specific sample of studies on pregnancy substance use, is provided in Table S2
(supplementary material). We considered a NOS-score < 6 (out of a maximum of 9) as an
indicator of low quality. Publication bias was graphically assessed by funnel plots when at
least ten studies were included in the particular analysis (27). Missing data was either
imputed, thereby resembling the actual sample as accurately as possible, or authors were
contacted to provide the missing information.
Fig. 1: Study flow diagram.
Data synthesis
Statistical heterogeneity was evaluated by the I
2-statistic and the χ
2-test for homogeneity.
The I
2-statistic estimates the proportion of total variance that can be attributed to
between-studies dispersion, rather than sampling error within individual between-studies. The χ
2-test for
homogeneity estimates the likelihood of true between-studies heterogeneity, yet has low
power when studies are few (27,28). When statistical heterogeneity was considered high
(I
2> 75%) we investigated the source of this heterogeneity by conducting subgroup analyses
based on clinical and/or methodological diversity and study quality within the original pool
of included studies.
Because observational, non-randomized studies cover different study types, samples,
sample sizes, methodology, and measures of CD, we expect more heterogeneity than when
assessing randomized controlled trials and we believe that the included studies do not
estimate one common effect size. Therefore we decided to use a random effects model in our
meta-analyses to calculate an average summary effect size to provide a global impression of
the magnitude of the effect. Note that when statistical heterogeneity (between-studies
variance) decreases, differences between fixed and random effects models decrease,
becoming eventually zero when there is no heterogeneity at all. In cases of very high
heterogeneity the random effects estimate is dominated by the between-studies variance
correction, levelling out study weights, and the summary effect will approach an arithmetic
mean. The between-studies variance (τ
2-parameter) was estimated by the method of
DerSimonian and Laird (28,29). Since available precompiled software would restrict the
analyses to a small selection of studies, thereby introducing non-random selection,
aforementioned models were coded in MATLAB (30) to perform computations and generate
graphics.
349 unique records identified through database searching (PubMed, PsycINFO, EMBASE).
11 unique additional records identified through other sources (such as references of relevant papers).
287 records excluded due to non-relevance.
73 studies assessed for eligibility.
37 studies excluded with reasons: Non-relevance (N=3). No journal article (N=4).
Review article (N=2). Article not available (N=3). No CD or CD-symptoms outcome (N=8). Insufficient data for meta-analysis (N=13). Disordinal statistical interaction effects (N=2).
Overlapping sample (N=2). 360 total unique records screened.
Associations were quantitatively assessed by calculating odds ratio’s (OR) for dichotomous
outcome data and standardized mean differences (SMD) for continuous outcome data. Since
most studies natively provided dichotomous data, SMDs were converted to ORs. We choose
to convert effect sizes to avoid having to exclude a substantial number of studies, which
would potentially cause a substantial degree of selection bias (28).
Meta-analyses
A meta-analysis was performed when an effect size for two or more studies was available.
We combined the maximally adjusted (for the largest number of confounding variables, in
some cases only crude effects were available) effect sizes that were available for each study
(27). For smoking during pregnancy, we considered general, not dose-related effects, from
here on referred to as ‘overall effects’ (none versus any exposure), as well as ‘dose-related
effects’ (light versus heavy exposure). For dose-related effects, ‘light exposure’ was defined
as maternal smoking of ≤ 15 cigarettes per day during pregnancy, whereas ‘heavy exposure’
was defined as maternal smoking of > 15 cigarettes per day. Of note, for dose-related
analyses of smoking, we included one study that was excluded from the non dose-related
meta-analysis due to sample-overlap in that analysis (31). Therefore the total number of
studies regarding non dose-related effects of smoking equals 25 instead of 26.
Further of note is the existence of complex data structures within a number of studies
requiring the combining of subgroups to estimate an overall effect (28). In the few instances
where different gestational trimesters were compared it was decided to use only data from
exposure during the first trimester since the occurrence of fetal organogenesis causes this
to be a vulnerable period (15).
Strength of evidence assessment
To give a global impression of the overall strength of our evidence, we carried out a Grades
of Recommendation, Assessment, Development and Evaluation (GRADE) (32) assessment
for each risk factor. A detailed description of the GRADE-criteria, as applied to our sample
of observational studies, is available in Table S3 (supplementary material). Of note, the
default GRADE-level is ‘low’ for observational studies.
Results
Description of included studies
A total of 36 studies were included in our current analyses. Tables S4-S6 (supplementary
material) present an overview and description of all individual studies.
Samples
Sample sizes varied considerably, from 40 participants to approximately 52,000
participants. Average and total sample sizes were 4,612 and 115,297 (25 studies) for
smoking during pregnancy, 5,625 and 50,626 (9 studies) for alcohol use during pregnancy,
and 421 and 1,263 (3 studies) for cannabis use during pregnancy. The only study regarding
caffeine intake during pregnancy had a sample size of 3,439. Sixteen out of 25 studies with
data on smoking assessed a general population samples, while the other 9 studies assessed
samples preselected based on (psycho)pathological conditions such as ADHD or a familial
predisposition for substance abuse. For alcohol use, 6 out of 9 studies assessed general
population studies, while 3 studies assessed preselected samples. Studies regarding
cannabis use comprised only general population samples.
Demographics
Most studies used samples consisting of both males and females. The distribution of sex
varied roughly between 40 and 100% male, yet was not consistently reported across all
studies. Overall, males appeared to be more prevalent (clearly larger percentage males in at
least six mixed samples, and two male-only samples) in the total pool of analyzed studies.
The age of assessed offspring varied considerably, from approximately 5 to 18 years old.
Globally estimating, fourteen of the studies on smoking during pregnancy assessed offspring
at age 12 or below, five studies assessed conduct problems at an age >12, and the other
studies included an age range across both childhood and adolescence. In the case of alcohol
use during pregnancy four studies assessed offspring at age 12 or below, two studies
assessed at age >12, and the remaining studies included both age ranges.
Exposure to substance use during pregnancy
Data on substance use during pregnancy can be collected during the actual pregnancy
(prospective measurement) or after birth at any time in life (retrospective measurement).
Regarding smoking during pregnancy, five studies measured exposure prospectively, while
nineteen studies measured retrospectively. One study did not specify their method of
exposure measurement. Exposure proportions ranged from 13 to 68%. In the case of alcohol
use during pregnancy, seven studies measured exposure prospectively and three studies
used retrospective assessment methods. Proportions of exposed participants ranged from
13 to about 60%. Regarding cannabis use during pregnancy, all studies measured exposure
prospectively. Exposure proportions ranged from 40 to 50%. Caffeine intake was measured
prospectively and the exposure proportion was 72%.
Confounding variables in multivariable models
Table S7 (supplementary material) provides an overview of control variables and the
number of studies that adjusted for a particular variable. The majority of studies that
provided results adjusted for confounding variables adjusted for offspring age, sex, and
some form of social environmental disadvantage (e.g. low socioeconomic status, family
instability, crowdedness, harsh parenting style) in their multivariable statistical models.
Only few studies adjusted for comorbid ADHD-symptomatology in offspring, and other
maternal substance use during pregnancy.
Outcome measures
About half of the studies (12 studies regarding smoking, 5 studies regarding alcohol) used
diagnostic psychiatric interviews such as the Diagnostic Interview Schedule for Children
(33) or Kiddie Schedule for Affective Disorders and Schizophrenia (34). About 40% of the
studies (10 studies regarding smoking, 4 studies regarding alcohol) used parent-on-child
questionnaires such as the Strengths and Difficulties Questionnaire (35) or Child Behavior
Checklist (CBCL) (36,37), assessing a continuous measure of conduct problems. Regarding
the CBCL, conduct problems were scored using a combination of items from the delinquency
and aggression scales (36,37). In addition to diagnostic interviews and questionnaires,
Gilman et al. created a dimensional scale for conduct problems based on ratings for 15
different behavioral items (38,39). D’Onofrio et al. 2008 used CBCL-derived scales based on
DSM-constructs including seven specific items for conduct problems (40). Agrawal et al.
assessed DSM-IV CD-symptoms in the offspring by self-report and used a cut-off of three
symptoms for CD cases (41).
Quality assessment of individual studies
Results of study quality assessment based on Newcastle-Ottowa Scale (NOS) scores are
presented in Tables S4-S6 (supplementary material) and a detailed breakdown of each
studies’ total score is provided in Table S8 (supplementary material). From 25 studies on
maternal smoking, 11 studies scored a quality rating < 6. For alcohol, five studies scored <
6. No studies on cannabis use scored < 6.
Meta-analysis of maternal smoking during pregnancy and offspring conduct
problems
The overall, non dose-related summary OR for maternal cigarette smoking was 2.06 (95%
confidence interval [CI] 1.67-2.54; I
2=93%, P<0.001; 25 studies; Figure 2). The funnel plot
Fig. 2: Forest plot of studies regarding maternal smoking during pregnancy and offspring conduct problems. The
confidence interval is shown as a horizontal line. Study weight is proportional to the area of the boxes. The width of the summary diamond represents its confidence interval. Maximally adjusted effects. An overview of individual studies is presented in Table S4 (supplementary material).
Smoking during pregnancy (25 studies)
Study: OR (95%-CI): Weight:
Maughan 2004 1.13 (0.94-1.37) 5.38% Agrawal 2010 1.18 (0.75-1.86) 4.46% Brion 2010 1.29 (1.12-1.49) 5.50% Hutchinson 2010 1.42 (1.10-1.84) 5.19% Palmer 2016 1.49 (1.19-1.85) 5.31% Melchior 2015 1.56 (0.99-2.46) 4.45% Larkby 2011 1.56 (0.93-2.63) 4.18% Fergusson 1998 1.58 (1.10-2.27) 4.83% Murray 2010 1.73 (1.46-2.04) 5.44% Tanaka 2016 1.93 (1.15-3.24) 4.19% Chastang 2015 1.94 (1.51-2.49) 5.22% Huijbregts 2008 2.19 (0.68-7.10) 2.03% D’Onofrio 2008 2.25 (1.79-2.82) 5.28% Wakschlag 1997 2.30 (1.09-4.83) 3.30% Freitag 2012 2.32 (0.69-7.72) 1.97% Nigg 2007 2.32 (1.05-5.13) 3.12% Sengupta 2015 2.86 (1.79-4.55) 4.40% Arnold 2005 2.89 (1.60-5.20) 3.90% Gilman 2008 2.97 (2.87-3.07) 5.63% Langley 2007 3.14 (1.54-6.40) 3.41% Biederman 2009 3.30 (1.23-8.85) 2.50% Wakschlag 2002 3.38 (0.99-11.55) 1.91% Gatzke-Kopp 2007 4.59 (1.92-10.96) 2.85% Hill 2000 4.66 (1.38-15.72) 1.94% Wilson 2013 5.02 (2.60-9.69) 3.62% Summary effect 2.06 (1.67-2.54) 100% 0.2 0.5 1 2 5 10 20 50 OR for offspring conduct problems
Summary effect: 𝑍𝑍 = 6.76, 𝑃𝑃 < 0.001
Fig. 3: Funnel plot of studies regarding maternal smoking during pregnancy and offspring conduct problems.
Smoking during pregnancy (25 studies)
Dose-related analyses yielded for light exposure an OR of 1.40 (95%-CI 1.25-1.57; I
2=26%,
P=0.238; six studies; Figure 4A) and for heavy exposure an OR of 1.78 (95%-CI 1.37-2.32;
I
2=73%, P=0.002; six studies; Figure 4B). The 95% CI’s of the dose-effects showed clear
overlap, indicating that effects of light and heavy exposure do not differ significantly.
Fig. 4: Forest plots of studies regarding maternal smoking during pregnancy and offspring conduct problems for light (A) and heavy (B) exposure. The confidence interval is shown as a horizontal line. Study weight is proportional to the
area of the boxes. The width of the summary diamond represents its confidence interval. Maximally adjusted effects. ‘Light exposure’ was defined as maternal smoking of 15 or less cigarettes per day during pregnancy, whereas ‘heavy smoking’ was defined as maternal smoking of 15 or more cigarettes in pregnancy. An overview of individual studies is presented in Table
S4 (supplementary material).
A. Light smoking during pregnancy (6 studies)
Study: OR (95%-CI): Weight:
Maughan 2004 1.07 (0.82-1.40) 13.61% Hutchinson 2010 1.27 (0.97-1.66) 13.72% Fergusson 1998 1.43 (1.02-2.01) 9.58% Maughan 2001 1.48 (1.18-1.86) 17.79% D’Onofrio 2008 1.52 (1.38-1.67) 44.26% Wakschlag 1997 1.60 (0.53-4.83) 1.02% Summary effect 1.40 (1.25-1.57) 100% Summary effect: 𝑍𝑍 = 5.83, 𝑃𝑃 < 0.001 Heterogeneity: 𝑇𝑇2= 0.01, 𝜒𝜒2= 6.77, 𝑑𝑑𝑑𝑑 = 5, 𝑃𝑃 = 0.238, 𝐼𝐼2= 26.16% B. Heavy smoking during pregnancy (6 studies)
Study: OR (95%-CI): Weight:
Maughan 2004 1.27 (1.00-1.60) 20.98% Maughan 2001 1.53 (1.17-2.00) 19.90% Hutchinson 2010 1.60 (1.23-2.09) 20.00% Fergusson 1998 1.92 (1.24-2.97) 14.84% D’Onofrio 2008 2.74 (2.03-3.68) 18.99% Wakschlag 1997 3.30 (1.20-9.08) 5.29% Summary effect 1.78 (1.37-2.32) 100% Summary effect: 𝑍𝑍 = 4.34, 𝑃𝑃 < 0.001 Heterogeneity: 𝑇𝑇2= 0.07, 𝜒𝜒2= 18.69, 𝑑𝑑𝑑𝑑 = 5, 𝑃𝑃 = 0.002, 𝐼𝐼2= 73.25% 0.2 0.5 1 2 5 10 20 50 OR for offspring conduct problems
0.2 0.5 1 2 5 10 20 50 OR for offspring conduct problems
Meta-analysis of maternal alcohol use during pregnancy and offspring conduct
problems
The summary OR for maternal alcohol use was 2.11 (95%-CI 1.42-3.15; I
2=76%, P<0.001; 9
studies; Figure 5).
Fig. 5: Forest plot of studies regarding maternal alcohol use during pregnancy and offspring conduct problems. The
confidence interval is shown as a horizontal line. Study weight is proportional to the area of the boxes. The width of the summary diamond represents its confidence interval. Maximally adjusted effects. An overview of individual studies is presented in Table S5 (supplementary material).
Alcohol use during pregnancy (9 studies)
Study: OR (95%-CI): Weight:
Kelly 2009 0.81 0.54-1.20) 15.52% Niclasen 2014 1.27 (0.96-1.68) 16.79% Disney 2008 1.75 (1.30-2.36) 16.60% Sood 2001 2.04 (1.05-3.95) 12.18% Larkby 2011 2.74 (1.50-5.01) 12.92% Alvik 2013 4.10 (1.50-11.21) 8.48% Hill 2000 4.42 (1.35-14.47) 7.00% Fryer 2007 6.34 (0.74-54.72) 2.89% Ware 2013 7.68 (2.54-23.22) 7.61% Summary effect 2.11 (1.42-3.15) 100% 0.2 0.5 1 2 5 10 20 50 OR for offspring conduct problems
Summary effect: 𝑍𝑍 = 3.67, 𝑃𝑃 < 0.001
Meta-analysis of maternal cannabis use during pregnancy and offspring conduct
problems
The summary OR for cannabis use was 1.29 (95%-CI 0.93-1.81; I
2=0; three studies; Figure
6). For this meta-analysis, there were no studies with a NOS-score < 6.
Fig. 6: Forest plot of studies regarding maternal cannabis use during pregnancy and offspring conduct problems. The
confidence interval is shown as a horizontal line. Study weight is proportional to the area of the boxes. The width of the summary diamond represents its confidence interval. Maximally adjusted effects. An overview of individual studies is presented in Table S6 (supplementary material).
Maternal caffeine use during pregnancy and offspring conduct problems
Due to a lack of studies, we were unable to conduct a meta-analysis and report here the
results of the only identified study on caffeine use during pregnancy. Reported unadjusted
ORs were between 0.62 (95%-CI 0.32-1.17) and 1.12 (95%-CI 0.43-2.94) as well as adjusted
ORs between 0.67 (95%-CI 0.34-1.35) and 1.16 (95%-CI 0.42-3.16) (25).
Analyses investigating heterogeneity
Given high heterogeneity in the meta-analysis of smoking during pregnancy, we stratified
the pool of included studies based on the type of study sample. We distinguished general
population samples, and samples that were preselected based on clinical conditions (such
as ADHD or a familial predisposition for substance abuse). The summary OR for general
population samples was 1.78 (95% CI 1.38-2.29; I
2=96%, P<0.001; 16 studies) and the
summary OR for clinical samples was 3.02 (95% CI 2.35-3.89; I
2=0; 9 studies). The summary
OR’s CI’s did not overlap, indicating a significant difference between general population and
clinical samples.
Since the general population subgroup meta-analyses of smoking during pregnancy still
showed high heterogeneity, we investigated the source of the heterogeneity further by
removing the study of Gilman et al. 2008. This study drew our attention because it appeared
as an outlier (large effect size and one to three orders of magnitude smaller variance than
most studies) and because of its use of a self-constructed outcome measure. The summary
Cannabis use during pregnancy (3 studies)
Study: OR (95%-CI): Weight:
Larkby 2011 1.18 (0.71-1.98) 42.30% Goldschmidt 2000 1.20 (0.73-1.97) 45.60% O’Connel 1991 2.34 (0.89-6.14) 12.10%
Summary effect 1.29 (0.93-1.81) 100%
0.2 0.5 1 2 5 10 20 50 OR for offspring conduct problems
Summary effect: 𝑍𝑍 = 1.51, 𝑃𝑃 = 0.131
OR for the subgroup of general population studies after removal of this study was 1.64 (95%
CI 1.42-1.90; I
2=71%, P<0.001; 14 studies).
Because of substantial heterogeneity in the meta-analysis of alcohol use during pregnancy,
we stratified the studies investigating alcohol use the same way as with smoking during
pregnancy. The summary OR for general population samples was 1.66 (95% CI 1.14-2.40;
I
2=75%, P=0.001; 6 studies) while the summary OR for clinical samples was 5.99 (95% CI
2.81-12.77; I
2=0; 3 studies). The summary OR’s CI’s did not overlap, indicating a significant
difference between general population and clinical samples.
Sensitivity analyses investigating the influence of low quality studies
After excluding studies with low quality (NOS-score < 6) we reported for maternal smoking
during pregnancy a summary OR of 1.94 (95%-CI 1.48-2.54; I
2=96%, P<0.001; 14 studies;
note the substantial heterogeneity), and for alcohol use during pregnancy a summary OR of
1.61 (95%-CI 1.06-2.45; I
2=79%, P<0.001; 5 studies; note the substantial heterogeneity).
GRADE-assessment for each exposure
Taking into account that for observational studies the default GRADE-assessment level is
‘low’, we considered the global strength of evidence as baseline (‘low’) for maternal
smoking, below baseline (‘low to very low’) for maternal alcohol use during pregnancy, and
below baseline (‘low to very low’) for cannabis use during pregnancy. We did not rate the
single study on caffeine, since its quality was already assessed by the NOS (see section
‘Quality assessment of individual studies’). The evidence for smoking and alcohol use
during pregnancy was downgraded because we could only partially explain inconsistency
in results. Results of the GRADE-assessment are shown in Table S9 (supplementary
Discussion
Meta-analyses
We conducted meta-analyses of observational studies concerning substance use during
pregnancy in relation to offspring CD problems. The various analyses were represented by
subgroups of the total pool of included studies of maternal substance use during pregnancy.
Associations were observed between CD problems and both cigarette smoking and alcohol
consumption during pregnancy, while for cannabis use and caffeine intake insufficient
studies were available to draw sensible conclusions yet.
Maternal smoking during pregnancy
We observed an overall, not dose-related, association between maternal smoking during
pregnancy and offspring conduct problems. However, high heterogeneity was noted and
therefore further analysis was required. We found that effects reported in clinical study
samples were clearly higher than those reported in general population studies. Five of the
nine clinical samples involved ADHD-samples. Within these five studies, the two studies
reporting the largest effects controlled for comorbid ADHD, which suggests that comorbid
ADHD does at least not solely drive the higher effect of pregnancy smoking found in clinical
samples. Moreover, in these clinical samples, offspring CD was mostly diagnosed by using a
DSM diagnostic interview, while most general population studies used a screening
questionnaire to assess conduct problems. Expected milder levels of conduct problems in
the general population might explain the lower magnitude of effect when compared to more
severe psychopathology involving a DSM diagnosis of CD.
Furthermore, it should be noted that studies have mostly failed to take into account the
possible role of passive smoking during pregnancy and postnatal smoke exposure (42,43).
Gatzke-Kopp and Beauchaine reported associations of passive smoking during pregnancy
with increased offspring externalizing psychopathology including CD. These associations
were observed in women who did not smoke themselves and thereby indicated an effect of
passive smoking (43). After controlling for multiple confounding variables including
exposure to smoking after birth, Hutchinson et al. reported attenuated but still small effects
of maternal smoking during pregnancy, suggesting that postnatal exposure to smoking
explains only a limited part of the reported association between smoking during pregnancy
and offspring conduct problems (44). Thus, these studies indicate a true effect of exposure
to smoking during pregnancy.
Regarding possible underlying mechanisms, smoking during pregnancy might be related to
abnormal fetal development by inducing hypoxia, nutritional abnormalities, teratogenic
effects, and DNA-damage by exposure to the multitude of toxicants in cigarette smoke (45–
49). More specifically, it has been shown that early nicotine exposure in rodents, equivalent
to gestational exposure in humans, affects development of catecholaminergic and brainstem
autonomic nuclei as well as development of the neocortex, hippocampus and cerebellum.
These observations are supported by clinical data including increased incidence of
disruptive behavior and substance abuse during childhood (14,50). Moreover, another issue
of concern are the adverse effects of tobacco smoking on breastfeeding as many women who
smoke during pregnancy continue to do so after birth. In addition to decreased milk
production and a shorter lactation period, the composition of breast milk as well as the
response of the infant are also affected by maternal smoking, which poses a further health
risk for the infant (51).
Summarizing, we found an association between maternal smoking during pregnancy and
offspring conduct problems, which appears to be stronger in clinical cases of CD. While
hypothesized underlying biological mechanisms underlying effects of smoking are plausible,
the role of various types of confounding variables remains an important issue which needs
to be addressed by future studies.
Alcohol use during pregnancy
Our meta-analyses regarding prenatal ethanol exposure indicated elevated odds on
offspring conduct problems for mothers who consumed alcohol while pregnant.
Furthermore, while the number of studies was relatively low for subgroup analyses, we
found that effects were clearly higher in clinical samples. The studies assessing clinical
samples used diagnostic measures of CD, rather than screening questionnaires. Of further
interest, one study reported that exposure to alcohol during the first but not third trimester
is a risk factor for offspring conduct problems (52). Alcohol use was assessed during
multiple visits in pregnancy in this study and while social desirability might have affected
the reported exposures, it should not have affected trimester-specific reports differently.
Therefore, this finding may indicate more harmful influences of alcohol use early in
pregnancy, reflecting an embryological vulnerable period. Also, early exposure may be more
prevalent because future mothers might be still unaware of their pregnancy, whereas
alcohol consumption after the first gestational trimester decreased in most pregnant women
(53).
Ethanol is a notorious teratogenic substance. It is known to cross both the placenta and fetal
blood-brain barrier and may have detrimental effects on neural development (54–56). In
animal studies structural anomalies are observed in the cerebral cortex in particular (57),
yet also developmental alterations in other structures including the hippocampus and
corpus callosum have been reported (58–60). Human neuro-imaging studies revealed also
abnormalities in these brain regions (61–69) and it is theorized that abnormal hippocampal
and cortical information processing is a major contributing factor for clinical
neuropsychiatric symptoms (70–72). These include a pattern of behavioral disinhibitions
seen in fetal alcohol spectrum disorders, resulting at different ages in symptoms including
irritability, conduct problems, and delinquency (16).
Concluding, we report an effect of alcohol use during pregnancy on offspring conduct
problems. Similar to maternal smoking during pregnancy, this effect appears to be stronger
problems. Upon more close inspection of individual results one study reported a significant
overall association (22) and for another study significant sub effects for different raters and
gestational trimesters were reported, although we did not compute a significant overall
effect for that study (73). However, these results apply to unadjusted data. The only
available confounder-adjusted data did not reveal a significant association and, moreover, it
was suggested by the authors that their unadjusted results may reflect differences in
parental tolerance for problem behavior instead of true offspring behavioral differences
(22). Therefore a more independent rater such as the school teacher was suggested. A
further point of consideration raised by O’Connel et al. is the increase in potency of cannabis
preparations over the last decades (22). A particular increase in ∆
9—tetrahydrocannabinol
(THC), the main psychoactive ingredient in cannabis, has been observed (74). In addition, a
general problem in isolating effects of cannabis is the strong association with use of other
intoxicating substances (75).
Concerning potential biological effects, it is known that cannabinoids cross the placenta (76)
and cannabis use during pregnancy may affect offspring neurocognitive development (19–
22). THC has been shown to disrupt endocannabinoid signaling, resulting in altered fetal
cortical wiring (77) and two large prospective cohort studies showed affected executive
functioning in heavily exposed offspring (78), however, no difference in IQ (79).
Summarizing, we found no overall effect of cannabis use during pregnancy on offspring
conduct problems, however, we suggest that further research is needed considering the
scarcity and somewhat conflicting nature of current evidence.
Caffeine use during pregnancy
The only identified study investigating maternal caffeine intake during pregnancy reported
no association with offspring conduct problems (25). The study measured different doses of
caffeine intake prospectively around the 16
thweek of gestation and controlled for pre-, peri-,
and postnatal confounders. Caffeine (1,3,7-trimethylxanthine) acts primarily as an
adenosine receptor antagonist and stimulates the central nervous system. Neural and
cardiovascular teratogenic effects following prenatal caffeine exposure have been reported
in animal studies (80–84). Moreover, a study by Silva et al., hypothesized that adenosine
antagonism may affect cognitive functioning in mice (85).
To summarize, we identified one study which did not report an effect of caffeine intake
during pregnancy on offspring conduct problems. Although no effects were suggested by
this single study, more studies are probably needed to further investigate potential effects
of caffeine use during pregnancy.
Considerations in current research
In the light of our present results, it is essential to discuss a number of pitfalls emerging from
current research on the role of the prenatal environment in multifactorial behavioral
disorders such as CD. The most important issue is that maternal substance use during
pregnancy does not necessarily exert a causal influence on offspring behavior, but may
merely represent an association, confounded by shared genetic and environmental factors
(39,40).
Genetic confounding
A key issue when investigating the prenatal environment is the possibility that reported
statistical associations are substantially confounded by the maternal genome. More
specifically, this means that the genetic make-up of the mother predisposes for both the
environmental factors (maternal substance use during pregnancy) and the outcome
(offspring conduct problems) (40). Classic approaches to disentangle genetic and
environmental influences use twin study designs, for example twins who grew up in
different environments due to adoption (86). However, to more sophistically test the
hypothesis of genetic confounding for maternal substance use during pregnancy, it would
be necessary to implement a study design in which the prenatal environment was provided
by both genetically related and unrelated mothers. An example of such a design includes the
use of in vitro fertilization (IVF) techniques and surrogacy pregnancies (87,88). A study by
Rice et al. using medical records of children born through IVF (the Cardiff IVF sample)
reported only associations between smoking during pregnancy and offspring antisocial
behavior in genetically related mothers and offspring, thereby providing evidence for
genetic mediation in this association. Cell sizes were, however, unbalanced due to a
relatively low number of mothers who reported smoking during pregnancy (87). Another,
perhaps more feasible approach, may be to correct for polygenic risk scores for CD, although
polygenic risk scores still only capture a limited amount of phenotypic heritability (89,90).
With regard to future studies, the use of genetically sensitive designs is essential, as this will
eventually allow us to gain more insight into the true nature of observed statistical
relationships regarding prenatal adversities and offspring behavioral problems.
Gene x environment interactions
Genetic factors may interact with influences from the environment and subsequently
predispose an individual to a certain condition, also referred to as gene x environment (G x
E) interactions (91,92). Wakschlag et al. reported a sex-specific G x E interaction for a
functional polymorphism in the monoamine oxidase A (MAOA) gene and smoking during
pregnancy for youth antisocial behavior.(93) Another study by O’Brien et al. reported also
sex-specific patterns for a genetic marker in the dopamine transporter (DAT1) gene and
smoking during pregnancy for externalizing behavior (94). Sex-specific G x E interaction
patterns may reflect effects of the genotype on sex-specific prenatal brain development (95).
No gene-environment correlations (rGE) were reported by both studies (93,94). One would
expect rGE if the genetic factor would be responsible for both the environmental factor and
outcome, as discussed in the previous paragraph.
factors such as socioeconomic status, parental psychopathology, and parenting style,
significant confounding likely also comes from unmeasured familial factors (39,40,86),
which are not being taken into account by the propensity scoring. Two studies addressed
this issue by using a sibling-matched design and reported no link between maternal smoking
and offspring behavior when comparing siblings with different prenatal exposure (39,40).
Similar results were reported in more recent studies using sibling-matched designs to
investigate associations between smoking during pregnancy and other offspring
psychopathology such as ADHD or schizophrenia in very large population samples (99,100).
Given that aforementioned studies investigated well-powered general population samples
this could imply that our current results may be explained by familial confounding, resulting
from both shared social environmental factors and genetic factors (as discussed previously).
Further pregnancy factors
In addition to substance use, other maternal factors during pregnancy such as medication
use, somatic health issues and anxiety could also contribute to offspring risk for CD related
problems (101–103), and might also be related to unhealthy maternal behaviors such as
smoking (104). It would therefore be important for future studies to concurrently consider
a broader range of factors during pregnancy to adjust for potential confounding effects. Only
a few studies adjusted for some form of medication use or life events stressors during
pregnancy, which could imply a slight overestimation of present results.
Furthermore, most studies in our meta-analyses did not control for potential effects of
maternal parity or birth order of the child. While literature findings show that risk for
emotional disorders and attention problems appears to be higher for the first child, no such
effect has been observed for conduct disorders (105,106). Regarding substance use during
pregnancy, both maternal smoking and alcohol consumption during pregnancy have been
associated with grand multiparity (more than five pregnancies), which itself is also related
to a number of pregnancy complications (107). Such highly multiparous pregnancies
constitute, however, only a minority of all births in developed countries (108). Therefore,
we conclude that potential confounding effects of maternal parity in the relationship
between substance use during pregnancy and offspring conduct problems are likely to be
minor.
Comorbidity among behavioral disorders
As already covered in the discussion about smoking during pregnancy, high rates of
comorbidity among behavioral disorders may constitute another general, important issue
and reflect a more fundamental problem of psychiatric classification in general. Descriptions
of mental illnesses are merely based on clinical symptomatology and therefore do not
necessarily infer etiologically separate entities. This may imply a common underlying
etiology among a broader spectrum of disruptive behavioral disorders, as suggested by a
case-control study on ADHD, that reported substantial comorbidity among ADHD, ODD, and
CD (42). Two other studies implied a common underlying etiology for CD and ODD
(109,110). Furthermore, assessing comorbid disruptive behavior can, in addition to
revealing more generic effects, also aid in investigating whether or not observed
associations with CD may be better explained by mediation effects, for example due to ADHD
(111–114). Goldschmidt et al. reported that inattention symptoms mediated the association
between maternal cannabis use during pregnancy and offspring delinquency (73). However,
three other studies that investigated maternal smoking during pregnancy did not report
such effects (31,115,116). Unfortunately, in general, most studies did not assess comorbid
disruptive behavior in the offspring. Therefore, to summarize, present meta-analytic results
are not yet conclusive with regard to the specificity of effects of maternal substance use on
CD only versus their role within the broader domain of disruptive behavioral disorders.
Methodological considerations of this study
Given limited time and resources we had to restrict our search parameters to CD specific
outcome measures, and thereby excluded potentially relevant and related constructs such
as delinquency, aggression, affective dysregulation and the broader domain of externalizing
problems. Also of note is that the prevalence of CD is higher in males than females, which
may lead to an overrepresentation of boys in clinical samples (1,4). While in the present
selection of studies most samples covered both sexes, males appeared to dominate in
general in our meta-analytical sample. Sex-specific effects were not analyzed due to
insufficient available data. Further considerations include the onset-age of conduct
problems. It has been suggested that an early onset-age is predictive for the persistence of
conduct problems (1), and recent research also suggested different correlates for a different
age of onset (117). We were unable to analyze the role of the age of onset given the lack of
data. Moreover, while CD has been suggested to be strongly related to instrumental
aggression and callous-unemotional traits (118), subphenotypes of aggression were also not
available and should be included in future studies (1–3).
Most studies assessed maternal substance use using retrospective methods, which are
subject to recall bias, thereby potentially causing additional inaccuracy and/or error of
exposure measurement. Although recall delays were reported up to multiple years, previous
studies have indicated satisfactory long-term recall of maternal smoking (119). Another
potential issue with self-reports of pregnancy substance use is social desirability bias, which
may even occur in computerized surveys (120), potentially resulting in underestimates of
exposure. A possible solution could be the inclusion of biomarkers in exposure
measurement, such as maternal or fetal cotinine levels (121). Furthermore, the outcome
measure was based on both diagnostic psychiatric interviews and parental or self-report
questionnaires such as the child behavior checklist (CBCL) (36,37) and SDQ (35). Although
the SDQ has shown high sensitivity and specificity for detecting CD (122), it contains only a
limited number of items for conduct problems, and it probably captures the broader domain
of CD and ODD. Further of note is the inclusion of samples, whose subjects were preselected
clinical samples, and within general population samples one particular study ((39))
explained about 25% of heterogeneity. This study of Gilman et al. is a good quality study
concerning a very large (50,000+) total sample size, which makes it of considerable value.
However, the outcome measure used to assess CD symptoms was self-constructed and also
screened for more broadly defined disruptive behavior. We suspect that the use of this
outcome measure might be a possible explanation for the heterogeneity caused by this
study. Furthermore, remaining, unexplained heterogeneity is still considerable (about
70%), and we suspect that unexplored factors such as differences in control variables,
sex-differences, or exposure ascertainment methods at least partially explain the remaining
inconsistency. Because we were not able to further investigate these factors we downgraded
our strength of evidence assessment. However, as all effects of the individual studies were
in the same direction (with by far most effects also being significant), and summary
estimates were not affected by low quality studies, we consider our results of sufficient
robustness and meaningful as an average effect estimate.
Furthermore, we analyzed effect sizes based on multivariable statistical models with the
largest number of confounding variables (27). As mentioned before, it was, however, not
possible to adjust for individual differences in number or types of confounding variables. Of
further consideration is that some potentially important results may have been excluded
because the respective studies did not provide sufficient data for our meta-analyses. In
addition, except for cigarette smoking during pregnancy, a more general issue was the low
availability of studies, resulting in limited power and sample representability for
meta-analytic data synthesis, and highlighting the need for more studies (28). Finally, for a
number of studies it was necessary to combine multiple subgroups to obtain an overall
estimate. It is important to recall that the subgroups represent in fact adjusted data for a
particular moderating variable, such as sex or age. By combining these effects we computed
an approximation of a general effect, which potentially could have slightly affected the
accuracy of our results (28).
Conclusion
This study provides evidence for a link between both maternal cigarette smoking and
alcohol use during pregnancy, and offspring conduct disorder problems. We did not
encounter clear effects of cannabis use and caffeine intake during pregnancy. However, it
must be emphasized that the paucity of studies does not yet allow for drawing firm
conclusions. Future research needs to address important unresolved issues concerning
confounding by shared genetic and socio-environmental factors, the role of possible G x E
interactions, and comorbidity among behavioral disorders to advance our understanding of
true effects of maternal substance use during pregnancy on offspring conduct problems.
References
1. Olsson M. DSM diagnosis of conduct disorder (CD)--a review. Nord J Psychiatry. 2009;63(Cd):102–12.
2. American Psychiatric Association. Diagnostic and statistical manual of mental disorders. 5th ed. Washington, DC: American Psychiatric Association; 2013.
3. Blair RJR, Leibenluft E, Pine DS. Conduct Disorder and Callous–Unemotional Traits in Youth. N Engl J Med. 2014;371:2207–16. 4. Martin A, Volkmar FR, editors. Lewis’s child and adolescent psychiatry: a comprehensive textbook. 4th ed. Wolters Kluwer: Lippincott
Williams & Wilkins; 2007.
5. Black D. The Natural History of Antisocial Personality Disorder. Can J Psychiatr. 2015;60(7):309–14.
6. Robins L. Epidemiology of antisocial personality disorder. In: Michels R, CavenarJO, editors. Psychiatry. Volume 3. Philadelphia (PA): Lippincott; 1987. p. 1–14.
7. Olsson M, Hansson K, Cederblad M. A long-term follow-up of conduct disorder adolescents into adulthood. Nord J Psychiatry. 2006;60(6):469–79.
8. Kim DR, Bale TL, Epperson CN. Prenatal Programming of Mental Illness: Current Understanding of Relationship and Mechanisms. Curr Psychiatry Rep. 2015;17(2):5.
9. Lamy S, Thibaut F. Psychoactive substance use during pregnancy: a review [article in French]. Encephale. 2010;36(1):33–8. 10. National Center for Chronic Disease Prevention and Health Promotion; Office on Smoking and Health (U.S.). How Tobacco Smoke
Causes Disease: The Biology and Behavioral Basis for Smoking-Attributable Disease: A Report of the Surgeon General. Atlanta (GA): Centers for Disease Control and Prevention (US); 2010. Available from: http://www.cdc.gov/
11. Rodgman A, Perfetti TA, editors. The Chemical Components of Tobacco and Tobacco Smoke. 1st ed. CRC Press, Taylor & Francis Group; 2009.
12. Honein MA, Paulozzi LJ, Watkins ML. Maternal smoking and birth defects: validity of birth certificate data for effect estimation. Public Health Rep. 2001;116(4):327–55.
13. Haustein K. Cigarette smoking, nicotine and pregnancy. Int J Clin Pharmacol Ther. 1999;37(9):417–27.
14. Wakschlag LS, Pickett KE, Cook Jr E, Benowitz NL, Leventhal BL. Maternal smoking during pregnancy and severe antisocial behavior in offspring: a review. Am J Public Health. 2002 Jun;92(6):966–74.
15. Kumar P, Clark M, editors. Clinical Medicine. 6th ed. Elsevier Saunders; 2005.
16. Dörrie N, Föcker M, Freunscht I, Hebebrand J. Fetal alcohol spectrum disorders. Eur Child Adolesc Psychiatry. 2014;23(10):863–75. 17. United Nations Office on Drugs and Crime. 2006 World Drug Report, Volume 1: Analysis. United Nations Publications; 2006. 18. National Institue on Drug Abuse (U.S.). National pregnancy and health survey. Rockville, MD; 1996. Available from:
https://www.drugabuse.gov/
19. Day N, Richardson G, Goldschmidt L, Robles N, Taylor P, Stoffer D, et al. Effect of prenatal marijuana exposure on the cognitive development of offspring at age three. Neurotoxicol Teratol. 1994;16(2):169–75.
20. Fried P, Watkinson B. 36- and 48-month neurobehavioral follow-up of children prenatally exposed to marijuana, cigarettes, and alcohol. J Dev Behav Pediatr. 1990;11(2):49–58.
21. Fried P, Watkinson B, Gray R. A follow-up study of attentional behavior in 6-year-old children exposed prenatally to marihuana, cigarettes, and alcohol. Neurotoxicol Teratol. 1992;14(5):299–311.
22. O’Connell CM, Fried PA. Prenatal exposure to cannabis: A preliminary report of postnatal consequences in school age children. Neurotoxicol Teratol. 1991 Nov;13(6):631–9.
23. Frary C, Johnson R, Wang M. Food sources and intakes of caffeine in the diets of persons in the United States. J Am Diet Assoc. 2005;105(1):110–3.
24. Kaiser L, Allen L. Position of the American Dietetic Association: nutrition and lifestyle for a healthy pregnancy outcome. J Am Diet Assoc. 2008;108(3):553–61.
25. Loomans EM, Hofland L, Van der Stelt O, Van der Wal MF, Koot HM, Van den Bergh BR, et al. Caffeine intake during pregnancy and risk of problem behavior in 5- to 6-year-old children. Pediatrics. 2012 Jan;130(2):e305–13.
26. Wells G, Shea B, O’Connell D, Peterson J, Welch V, Losos M, et al. The Newcastle-Ottawa Scale (NOS) for assessing the quality of nonrandomised studies in meta-analyses. The Ottawa Hospital - Research Institute. Available from: http://www.ohri.ca/programs/clinical_epidemiology/oxford.asp
27. Higgins JPT, Green S, editors. Cochrane Handbook for Systematic Reviews of Interventions (online version). 5.1.0. John Wiley & Sons, Ltd; 2011.
28. Borenstein M, Hedges L V., Higgins JPT, Rothstein HR, editors. Introduction to Meta-Analysis. 1st ed. John Wiley & Sons, Ltd; 2009. 29. DerSimonian R, Laird N. Meta-analysis in clinical trials. Control Clin Trials. 1986;7(3):177–88.
30. The Mathworks Inc. MATLAB and Statistics Toolbox Release 2014a. Natick, Massachusetts, United States: The Mathworks, Inc.; 2014. 31. Maughan B, Taylor C, Taylor A, Butler N, Bynner J. Pregnancy smoking and childhood conduct problems: A causal association? J Child
Psychol Psychiatry. 2001;42(8):1021–8.
32. Schunemann H, Brozek J, Oxman A. GRADE handbook for grading quality of evidence and strength of recommendation. Version 3. The Grade Working Group; 2009. Available from: http://www.gradeworkinggroup.org/index.html
33. Shaffer D, Fisher P, Dulcan MK, Davies M, Piacentini J, Schwab-Stone ME, et al. The NIMH Diagnostic Interview Schedule for Children Version 2.3 (DISC-2.3): description, acceptability, prevalence rates, and performance in the MECA Study. J Am Acad Child Adolesc Psychiatry. 1996;35(7):865–77.
34. Orvaschel H. Schedule for Affective Disorder and Schizophrenia for School-Age Children Epidemiologic Version. 5th editio. Lauderdale: Nova Southeastern University, Center for Psychological Studies; 1994.
35. Goodman R. Psychometric properties of the Strengths and Difficulties Questionnaire. J Am Acad Child Adolesc Psychiatry. 2001;40:1337–45.
36. Achenbach TM. Manual for the Child Behavior Checklist/4-18 and 1991 Profile. Burlington, VT: University of Vermont Department of Psychiatry; 1991.
44. Hutchinson J, Pickett KE, Green J, Wakschlag LS. Smoking in pregnancy and disruptive behaviour in 3-year-old boys and girls: An analysis of the UK millennium cohort study. J Epidemiol Community Health. 2010;64(1):82–8.
45. Chernoff N. Teratogenic effects of cadmium in rats. Teratology. 1973;8(1):29–32.
46. Mochizuki M, Maruo T, Masuko K, Ohtsu T. Effects of smoking on fetoplacental-maternal system during pregnancy. Am J Obstet Gynecol. 1984;149(4):413–20.
47. Lammer EJ, Shaw GM, Iovannisci DM, Van Waes J, Finell RH. Maternal smoking and the risk of orofacial clefts: susceptibility with NAT1 and NAT2 polymorphisms. Epidemiology. 2004;15(2):150–6.
48. Munger RG, Sauberlich HE, Corcoran C, Nepomuceno B, Baack-Hirsch S, Solon FS. Maternal vitamin B-6 and folate status and risk of oral cleft birth defects in the Philippines. Birth Defects Res Part A, Clin Mol Teratol. 2004;70(7):464–71.
49. Ziaei S, Nouri K, Kazemnejad A. Effects of carbon monoxide air pollution in pregnancy on neonatal nucleated red blood cells. Paediatr Perinat Epidemiol. 2005;19(1):27–30.
50. Dwyer J, McQuown S, Leslie F. The dynamic effects of nicotine on the developing brain. Pharmacol Ther. 2009;122(2):125–39. 51. Napierala M, Mazela J, Merritt TA, Florek E. Tobacco smoking and breastfeeding : Effect on the lactation process , breast milk
composition and infant development . A critical review. Environ Res. 2016;151:321–38.
52. Larkby CA, Goldschmidt L, Hanusa BH, Day NL. Prenatal alcohol exposure is associated with conduct disorder in adolescence: Findings from a birth cohort. J Am Acad Child Adolesc Psychiatry. 2011;50(3):262–71.
53. Hill SY, Lowers L, Locke-Wellman J, Shen S. Maternal smoking and drinking during pregnancy and the risk for child and adolescent psychiatric disorders. J Stud Alcohol. 2000;61(5):661–8.
54. Taléns-Visconti R, Sanchez-Vera I, Kostic J, Perez-Arago MA, Erceg S, Stojkovic M, et al. Neural differentiation from human embryonic stem cells as a tool to study early brain development and the neuroteratogenic effects of ethanol. Stem cells Dev. 2010;20(2):327– 39.
55. Guizzetti M, Zhang X, Goeke C, Gavin DP. Glia and neurodevelopment: focus on fetal alcohol spectrum disorders. Front Pediatr. 2014;11(2):123.
56. Rubert G, Miñana R, Pascual M, Guerri C. Ethanol exposure during embryogenesis decreases the radial glial progenitorpool and affects the generation of neurons and astrocytes. J Neurosci Res. 2006;84(3):483–96.
57. Abbott CW, Kozanian OO, Kanaan J, Wendel KM, Huffman KJ. The Impact of Prenatal Ethanol Exposure on Neuroanatomical and Behavioral Development in Mice. Alcohol Clin Exp Res. 2016;40(1):122–33.
58. Livy D, Elberger A. Alcohol exposure during the first two trimesters- equivalent alters the development of corpus callosum projection neurons in the rat. Alcohol. 2008;42:285–93.
59. Livy D, Miller E, Maier S, West J. Fetal alcohol exposure and temporal vulnerability: effects of binge-like alcohol exposure on the developing rat hippocampus. Neurotoxicol Teratol. 2003;25:447–58.
60. Maier S, West J. Regional differences in cell loss associated with bing-like alcohol exposure during the first two trimesters equivalent in the rat. Alcohol. 2001;23:49–57.
61. Bookstein F, Sampson P, Connor P, Streissguth A. Midline corpus callosum is a neuroanatomical focus of fetal alcohol damage. Anat Rec. 2002;269(3):162–74.
62. Bookstein F, Connor P, Huggins J, Barr H, Pimentel K, Streissguth A. Many infants prenatally exposed to high levels of alcohol show one particular anomaly of the corpus callosum. Alcohol Clin Exp Res. 2007;31(5):868–79.
63. Mattson S, Riley E, Sowell E, Jernigan T, Sobel D, Jones K. A decrease in the size of the basal ganglia in children with fetal alcohol syndrome. Alcohol Clin Exp Res. 1996;20(6):1088–93.
64. Archibald S, Fennema-Notestine C, Gamst A, Riley E, Mattson S, Jernigan T. Brain dysmorphology in individuals with severe prenatal alcohol exposure. Dev Med Child Neurol. 2001;43(3):148–54.
65. Riikonen R, Salonen I, Partanen K, Verho S. Brain perfusion SPECT and MRI in foetal alcohol syndrome. Dev Med Child Neurol. 1999;41:652–9.
66. Sowell E, Thompson P, Mattson S, Tessner K, Jernigan T, Riley E, et al. Regional brain shape abnormalities persist into adolescence after heavy prenatal alcohol exposure. Cereb cortex. 2002;12(8):865–65.
67. Sowell E, Thompson P, Peterson B, Mattson S, Welcome S, Henkenius A, et al. Mapping cortical gray matter asymmetry patterns in adolescents with heavy prenatal alcohol exposure. Neuroimage. 2002;17(4):1807–19.
68. Zhou D, Lebel C, Lepage C, Rsmussen C, Evans A, Wyper K, et al. Developmental cortical thinning in fetal alcohol spectrum disorders. Neuroimage. 2011;58(16–25).
69. Yang S, Fombonne E, Kramer MS. Duration of gestation, size at birth and later childhood behaviour. Paediatr Perinat Epidemiol. 2011 Jul;25(4):377–87.
70. El Shawa H, Abbott C, Huffman K. Prenatal ethanol exposure disrupts intraneocortical circuitry, cortical gene expression, and behavior in a mouse model of FASD. J Neurosci Off J Soc Neurosci. 2013;33(48):18893–905.
71. Granato A. Altered organization of cortical interneurons in rats exposed to ethanol during neonatal life. Brain Res. 2006;1069(1):23– 30.
72. Streissguth A, Barr H, Sampson P, Bookstein F. Prenatal alcohol and offspring development: the first fourteen years. Drug Alcohol Depend. 1994;36(2):89–99.
73. Goldschmidt L, Day NL, Richardson GA. Effects of prenatal marijuana exposure on child behavior problems at age 10. Neurotoxicol Teratol. 2000;22(3):325–36.
74. Trimbos Instituut. THC, CBD en gezondheidseffecten van wiet en hasj: recente inzichten. april 2012. Niesink R, Van Laar M, editors. Trimbos Instituut; 2012. Available from: http://www.trimbos.nl
75. Hill M, Reed K. Pregnancy, Breast-feeding, and Marijuana: A Review Article. Obstet Gynecol Surv. 2013;68(10):710–8.
76. Tennes K, Avitable N, Blackard C, Boyles C, Hassoun B, Holmes L, et al. Marijuana: prenatal and postnatal exposure in the human. NIDA Res Monogr. 1985;59:48–60.
77. Tortoriello G, Morris C V, Alpar A, Fuzik J, Shirran SL, Keimpema E, et al. Miswiring the brain: Δ9-tetrahydrocannabinol disrupts cortical development by inducing an SCG10/stathmin-2 degradation pathway. EMBO J. 2014;33(7):668–85.
78. Fried P, Watkinson B. Differential effects on facets of attention in adolescents prenatally exposed to cigarettes and marihuana. Neurotoxicol Teratol. 2001;23(5):421–30.
79. Fried P, Watkinson B, Gray R. Differential effects on cognitive functioning in 9- to 12-year olds prenatally exposed to cigarettes and marihuana. Neurotoxicol Teratol. 1998;20(3):293–306.
80. Bruyere H, Michaud B, Gilbert E, Folts J. The effects of cardioteratogenic doses of caffeine on cardiac function in the 3-day chick embryo. J Appl Toxicol. 1987;7(3):197–203.
81. Ma Z, Qin Y, Wang G, Li X, He R, Chuai M, et al. Exploring the caffeine-induced teratogenicity on neurodevelopment using early chick embryo. PLoS One. 2012;7(3):e34278.
82. Ma Z, Wang G, Cheng X, Chuai M, Kurihara H, Lee K, et al. Excess caffeine exposure impairs eye development during chick embryogenesis. J Cell Mol Med. 2014;18(6):1134–43.
83. Nikoui V, Ostadhadi S, Kazare N, Nabavi S, Giorgi M, Bakhtiarian A. The role of clomipramine in potentiating the teratogenic effects of caffeine in pregnant rats: a histopathological study. ScientificWorldJournal. 2013;2013:382434.
