Antidepressant Exposure During Pregnancy and Congenital Malformations: Is There an Association? A Systematic Review and Meta-Analysis of the Best Evidence

Article (PDF Available)inThe Journal of Clinical Psychiatry 74(4):e293-308 · April 2013with63 Reads
DOI: 10.4088/JCP.12r07966 · Source: PubMed
Abstract
Depression is often not optimally treated during pregnancy, partially because of conflicting data regarding antidepressant medication risk. This meta-analysis was conducted to determine whether antenatal antidepressant exposure is associated with congenital malformations and to assess the effect of known methodological limitations. EMBASE, CINAHL, PsycINFO, and MEDLINE were searched from their start dates to June 2010. Keywords of various combinations were used, including, but not limited to depressive/mood disorder, pregnancy, antidepressant drug/agent, congenital malformation, and cardiac malformation. English language studies reporting congenital malformations associated with antidepressants were included. Of 3,074 abstracts reviewed, 735 studies were retrieved and 27 studies were included. Two reviewers working independently assessed article quality. Data on use of any antidepressant, including fluoxetine and paroxetine specifically, were extracted. Outcomes included congenital malformations, major congenital malformations, cardiovascular defects, septal heart defects (ventral septal defects and atrial septal defects), and ventral septal defects only. Nineteen studies were above quality threshold and make up the primary meta-analyses. Pooled relative risks (RRs) were derived by using random-effects methods. Antidepressant exposure was not associated with congenital malformations (RR = 0.93; 95% CI, 0.85-1.02; P = .113) or major malformations (RR = 1.07; 95% CI, 0.99-1.17; P = .095). However, increased risk for cardiovascular malformations (RR = 1.36; 95% CI, 1.08-1.71; P = .008) and septal heart defects (RR = 1.40; 95% CI, 1.10-1.77; P = .005) were found; the RR for ventral septal defects was similar to septal defects, although not significant (RR = 1.54; 95% CI, 0.71-3.33; P = .274). Pooled effects were significant for paroxetine and cardiovascular malformations (RR = 1.43; 95% CI, 1.08-1.88; P = .012). These results are contrasted with those addressing methodological limitations but are typically consistent. Overall, antidepressants do not appear to be associated with an increased risk of congenital malformations, but statistical significance was found for cardiovascular malformations. Results were robust in several sensitivity analyses. Given that the RRs are marginal, they may be the result of uncontrolled confounders. Although the RRs were statistically significant, none reached clinically significant levels.
© 2013 COPYRIGHT PHYSICIANS POSTGRADUATE PRESS, INC. NOT FOR DISTRIBUTION, DISPLAY, OR COMMERCIAL PURPOSES.
e293J Clin Psychiatry 74:4, April 2013
Focus on Womens Mental Health
Meta-Analysis
Antidepressant Exposure During Pregnancy and
Congenital Malformations: Is There an Association?
A Systematic Review and Meta-Analysis of the Best Evidence
Sophie Grigoriadis, MD, PhD, FRCPC; Emily H. VonderPorten, MPH;
Lana Mamisashvili, MSW; Michael Roerecke, PhD; Jürgen Rehm, PhD;
Cindy-Lee Dennis, PhD; Gideon Koren, MD, FRCPC, FACMT; Meir Steiner, MD, PhD, FRCPC;
Patricia Mousmanis, MD, CCFP, FCFP; Amy Cheung, MD, MSc, FRCPC; and Lori E. Ross, PhD
ABSTRACT
Objective: Depression is often not optimally treated during pregnancy,
partially because of conflicting data regarding antidepressant medication
risk. This meta-analysis was conducted to determine whether antenatal
antidepressant exposure is associated with congenital malformations
and to assess the effect of known methodological limitations.
Data Sources: EMBASE, CINAHL, PsycINFO, and MEDLINE were
searched from their start dates to June 2010. Keywords of various
combinations were used, including, but not limited to depressive/mood
disorder, pregnancy, antidepressant drug/agent, congenital malformation,
and cardiac malformation.
Study Selection: English language studies reporting congenital
malformations associated with antidepressants were included. Of 3,074
abstracts reviewed, 735 studies were retrieved and 27 studies were
included.
Data Extraction: Two reviewers working independently assessed article
quality. Data on use of any antidepressant, including fluoxetine and
paroxetine specifically, were extracted. Outcomes included congenital
malformations, major congenital malformations, cardiovascular defects,
septal heart defects (ventral septal defects and atrial septal defects), and
ventral septal defects only.
Results: Nineteen studies were above quality threshold and make up
the primary meta-analyses. Pooled relative risks (RRs) were derived
by using random-effects methods. Antidepressant exposure was not
associated with congenital malformations (RR = 0.93; 95% CI, 0.85–1.02;
P = .113) or major malformations (RR = 1.07; 95% CI, 0.99–1.17; P = .095).
However, increased risk for cardiovascular malformations (RR = 1.36; 95%
CI, 1.08–1.71; P = .008) and septal heart defects (RR = 1.40; 95% CI, 1.10–
1.77; P = .005) were found; the RR for ventral septal defects was similar
to septal defects, although not significant (RR = 1.54; 95% CI, 0.71–3.33;
P = .274). Pooled effects were significant for paroxetine and cardiovascular
malformations (RR = 1.43; 95% CI, 1.08–1.88; P = .012). These results are
contrasted with those addressing methodological limitations but are
typically consistent.
Conclusions: Overall, antidepressants do not appear to be associated
with an increased risk of congenital malformations, but statistical
significance was found for cardiovascular malformations. Results were
robust in several sensitivity analyses. Given that the RRs are marginal,
they may be the result of uncontrolled confounders. Although the RRs
were statistically significant, none reached clinically significant levels.
J Clin Psychiatry 2013;74(4):e293–e308
© Copyright 2013 Physicians Postgraduate Press, Inc.
Submitted: June 21, 2012; accepted November 16, 2012 (doi:10.4088/JCP.12r07966).
Corresponding author: Sophie Grigoriadis, MD, PhD, FRCPC, Womens Mood
and Anxiety Clinic: Reproductive Transitions, Department of Psychiatry, FG 29,
Sunnybrook Health Sciences Centre, 2075 Bayview Ave, Toronto, ON M4N 3M5,
Canada (Sophie.Grigoriadis@sunnybrook.ca).
T
he consequences of a major depressive episode
during pregnancy include effects on the mother,
her infant, and family.
1–3
Although treatment is essen-
tial, the use of antidepressant medication during
pregnancy has been found to be lower than other times
of the life cycle.
4
One of the contributing factors for
poor uptake of pharmacologic interventions appears
to be concerns regarding the safety of antidepressant
exposure for the fetus.
5
For example, preliminary
reports of neonatal cardiovascular malformations fol-
lowing paroxetine exposure in early pregnancy
6
led
to both the US Food and Drug Administration
6a
and
Health Canada
6b
issuing advisories in 2005 warning
about potential risks associated with the use of anti-
depressants during pregnancy. However, the warnings
suggested a direct conflict between safety of the fetus
and a mother’s need for antidepressant medication, per-
haps without regard to the documented negative effects
of untreated depression on pregnancy outcomes.
3,7
The results of prior meta-analyses have been con-
flicting. Several have found no evidence of increased
risk of major congenital malformations above the base-
line
8–10
rate, which has been widely cited as 1%–3% for
any pregnancy in North America
11–14
and less than 4%
for minor congenital malformations.
15
The most recent
meta-analyses did report an increased risk for congeni-
tal malformations
16
and cardiac malformations
13,16
with
paroxetine exposure specifically. Unfortunately, many
individual studies have serious methodological limita-
tions
17
that were not taken into account in the previous
meta-analyses. For example, raw data unadjusted for
confounders were used or studies were based on conve-
nience samples. Furthermore, it is important to note the
distinction between statistical significance and clinical
significance when interpreting scientific evidence for
clinical practice implications. The aim of this study
was to synthesize the available data on congenital mal-
formations, including cardiac, in infants of mothers who
took antidepressant medications during pregnancy. In
order to address the methodological limitations of the
available research, we excluded studies below a quality
threshold, used adjusted data where possible, exam-
ined for the effect of exposure contamination with any
© 2013 COPYRIGHT PHYSICIANS POSTGRADUATE PRESS, INC. NOT FOR DISTRIBUTION, DISPLAY, OR COMMERCIAL PURPOSES.
e294 J Clin Psychiatry 74:4, April 2013
Grigoriadis et al
antidepressant use in the control group, and examined the
influence of whether or not the samples were population
based or obtained by convenience.
DATA SOURCES AND STUDY SELECTION
Following the Meta-analysis of Observational Studies in
Epidemiology (MOOSE) guidelines,
18
2 professional librar-
ians with expertise in the areas of psychopharmacology and
psychiatry independently conducted literature searches.
A variety of keyword combinations were utilized, such as
depressive/mood disorder; pregnancy/pregnancy trimesters;
tricyclic antidepressant drugs; antidepressant drug/agent; selec-
tive serotonin reuptake inhibitors, SSRIs; monoamine oxidase
inhibitors; prenatal or antenatal, infant/outcomes; congenital
malformation, cardiac malformation (see supplementary
material for full list of keywords). Strategies were based on
the subject headings specific to the individual databases
searched, combined with appropriate keywords and keyword
phrases, and truncated if necessary. Concepts were combined
with Boolean operators (AND, OR) to either broaden “like
concepts (OR) or narrow them (AND) to ensure more than
1 concept was included in the results. The explode feature
(available for all of the databases searched, with the excep-
tion of MEDLINE In-Process and Scopus) was also used
to broaden like concepts. The databases used included
MEDLINE (Ovid), MEDLINE In-Process (Ovid) to access
current literature (keyword searching only), PsycINFO
(American Psychological Association; Ovid), CINAHL
(Nursing and Allied Health), EMBASE (Excerpta Medica,
Elsevier; Ovid), and Scopus (Elsevier) to access current
literature (keyword searching only). The databases were
searched from their start date to June 30, 2010. Reference
lists of reviews and meta-analyses were also searched for
further articles, but none were found.
Inclusion and Exclusion Criteria
Cohort and case-control studies published in English
were eligible for inclusion if they (1) reported original data
and at least 1 malformation of interest, (2) reported on any
pharmacologic antidepressant agent exposure (ie, selective
serotonin reuptake inhibitors, tricyclic antidepressants,
and monoamine oxidase inhibitors), (3) had a nonexposed
comparison group of pregnant women for the antidepres-
sant examined, and (4) provided sufficient data to calculate
an effect size. Abstracts and conference proceedings were
excluded, and unpublished data were not searched because
the volume of potentially eligible studies would have made
doing so infeasible. Outcomes of interest were identified by
the research team as well as the advisory committee of key
stakeholders for this program of research, which included
representatives from psychiatry, family medicine, obstetrics,
neonatology, public health, patient advocacy, and policy. The
following outcomes were included in this meta-analysis:
any congenital malformation, major congenital malforma-
tion (structural defects present at birth that have surgical,
medical, or cosmetic significance or a significant effect on
function or social acceptability
15,19
), cardiovascular malfor-
mation, septal cardiac defect (atrial septal defects or ventral
septal defects), and ventral septal defects only as defined by
the authors of the original publication. When more than 1
study had been published on the same cohort, we selected
the study with the largest number of cases of malformations
within each malformation outcome.
DATA EXTRACTION
The study was part of a larger research program developing
a reference guide for physicians to assist in treatment discus-
sions regarding antidepressant use during pregnancy with
their patients. The data extraction and quality assessment
processes for this program of research have been previously
published.
7
Two independent research assistants completed
the screening for all articles by their title and abstract, and
eligible articles were retrieved. Data extraction forms were
completed for all eligible studies and were based on the
Strengthening the Reporting of Observational Studies in
Epidemiology (STROBE)
20
checklist. Data extracted included
source, study design, participants (sample, control, demo-
graphics, and clinical characteristics), inclusion/exclusion
criteria, antidepressants examined, dosage, duration of expo-
sure, primary and secondary outcomes, outcome assessment
methods, and loss to follow-up. Authors of original publi-
cations were sent requests for raw data if not provided in the
article. Eight authors were contacted and 3 replied. Of these,
1 was unable to share data because of confidentiality issues,
1 was not able to meet our timeline, and the last 1 did not
respond to our request for further clarification.
Quality Assessment
The Systematic Assessment of Quality in Observational
Research (SAQOR)
7
tool used for this investigation was
developed by our team and based on previously published
quality assessment tools (Downs and Black
21
and the New-
castle-Ottawa Scale
22
) and adapted to assess study quality in
this area of research. Nineteen aspects of each study under
the following categories were evaluated by outcome: (1)
sample, (2) control group, (3) quality of exposure/outcome
measure, (4) follow-up, and (5) distorting influences. Spe-
cifically, the distorting influences category included whether
analyses controlled for confounders such as depression,
other psychotropic medications, smoking, alcohol, or illicit
drug use. On the basis of the aforementioned criteria, we
assigned a final quality rating of high, moderate, low, or very
Clinical Points
Antidepressants do not appear to be associated with an
increased risk of congenital malformations overall.
Antidepressants may be associated with cardiovascular
malformations.
Clinical significance does not always follow from statistical
significance, and clinicians must consider the risks of
depressive illness prior to making any treatment decisions.
© 2013 COPYRIGHT PHYSICIANS POSTGRADUATE PRESS, INC. NOT FOR DISTRIBUTION, DISPLAY, OR COMMERCIAL PURPOSES.
e295J Clin Psychiatry 74:4, April 2013
Antidepressants and Congenital Malformations
low using a modification of the Grading of Recommenda-
tions Assessment, Development and Evaluation (GRADE)
system.
23
Studies were then categorized as above quality
threshold(high, moderate, or low quality) or “below quality
threshold” (very low quality). Results of the data extraction
and quality assessment were compared between raters for
each study, with any differences resolved through consensus
by the principal investigators.
Statistical Analyses
Adjusted risk estimates were preferred when these were
available, and odds ratios (ORs) and relative risks (RRs)
were considered as equivalent measures of risk. We calcu-
lated ORs from the raw number of cases and controls when
no such risk estimate was reported. For each outcome, we
calculated 1 estimate of the OR for malformations for each
article and pooled these risks across studies using random-
effects models.
24
We added 0.5 to cells with 0 cases when
calculating the OR.
25
All analyses were conducted on the
log scale. Visual inspection of funnel plots depicting the
risk estimates (on the log scale) against their standard
error and Egger regression-based test
26
were used to assess
publication bias. Cochrane Q and I
2
were used to quantify
between-study heterogeneity.
27,28
I
2
may be interpreted as
the proportion of the total variance due to between-study
heterogeneity. For studies reporting multiple exposure or
control groups, these groups were combined in order to cal-
culate a single risk estimate for each study where possible, if
they were independent. For each outcome, our main analysis
consisted of the pooled risk for the studies determined to be
above quality threshold; further, we excluded studies that
included some antidepressants in their control group (for
example, when only selective serotonin reuptake inhibitor
[SSRI] use was used as the exposure group but other anti-
depressants were not assessed and controls could have been,
and in some cases were reported to have been, exposed to
any antidepressants in their pregnancy) and also reran the
analysis with articles that matched or adjusted their data
in any way. Because of the generally low study quality, we
also estimated the pooled risk excluding all convenience
samples. We selected 2 specific antidepressants (fluoxetine
and paroxetine) for a separate analysis of their effect on
malformations where sufficient data were available, as more
information was available for them compared to others. All
analyses were conducted with Stata statistical software, ver-
sion 10.1 (StataCorp LP)
29
and similar to our other work in
Ross et al.
30
RESULTS
Of the 3,074 abstracts reviewed, 2,339 were excluded
based on title and abstract. In total, 735 articles were
retrieved and assessed for eligibility and 31 articles met
the inclusion criteria (Figure 1).
31
Of these, 4 articles were
excluded because they were duplicate analyses of studies
already included in our quantitative analysis, leaving 27
studies
32–58
for a quantitative analysis (Table 1). Nineteen
of the 27 studies were above the quality threshold, whereas
8 were below. Most studies reported data on more than 1
outcome (14 above quality threshold reported on any con-
genital malformation, 11 on any major malformation, 13
on any cardiovascular malformation, 9 on any septal car-
diac defect [atrial septal defects or ventral septal defects],
and 5 on ventral septal defects specifically). Eleven studies
used a convenience sample, while 16 studies used either a
Figure 1. Identification of Independent Studies for Inclusion in Meta-Analysis (adapted from
PRISMA 2009 flow diagram
30
)
a
Other outcomes of interest in our research program.
3,073 Records identified through
database searching
1 Additional record identified
through other sources
2,339 Records excluded
3,074 Records screened
27 Studies included in
quantitative synthesis
(meta-analysis)
735 Full-text articles
assessed for eligibility
IdentificationScreeningEligibilityIncluded
708 Full-text articles excluded, with reasons
75 Nonpharmacologic interventions
a
178 Risk factors for antenatal depression
a
242 Impact of maternal depression
a
120 Postpartum treatment
a
47 Case reports/series
23 Other outcomes of interest
a
19 Insufficient data
4 Utilized same cohorts as other studies
© 2013 COPYRIGHT PHYSICIANS POSTGRADUATE PRESS, INC. NOT FOR DISTRIBUTION, DISPLAY, OR COMMERCIAL PURPOSES.
e296 J Clin Psychiatry 74:4, April 2013
Grigoriadis et al
Table 1. Study Characteristics of 27 Studies in the Congenital Malformations Meta-Analyses
Article
Quality
Threshold Country Sample Size Confounders
a
Drug Results Outcome Definition
Pastuszak et al,
32
1993
b,c,d,e,f,g
B Canada,
United States
Fluoxetine, n = 128 and 74
TCA, n = 74
NTC, n = 128
Matched for age and date of
consultation (collected data
on alcohol and smoking, and
reported groups had “similar
distribution”)
Fluoxetine, TCA Major malformations: 2% in fluoxetine
vs 1.8% in NTC (P = .38) and 3.4% in
fluoxetine, 0% in TCA, vs 3% in NTC
(P = .8)
h
Cardiac malformations: fluoxetine, 1 VSD
event; TCA, 0 events; NTC, 1 VSD event
h
Chambers
et al,
33
1996
b,c,g
B United States Fluoxetine, n = 164
NTC, n = 226
(Collected data on alcohol,
smoking, and other drugs)
Fluoxetine Major malformations: 5.5% in fluoxetine vs
4.0% in NTC (P = .63)
Major malformations: structural
defect with cosmetic or
functional importance
seen in less than 4% of
the general population
Kulin et al,
34
1998
b,c,d,g,i
A Canada,
United States
SSRI, n = 267
NTC, n = 267
Groups “matched”
(No difference between groups
for alcohol or age, parity,
or previous spontaneous
abortion)
Sertraline, paroxetine,
fluvoxamine, fluoxetine
Major malformations: 4.1% in SSRI
vs 3.8% in NTC (RR = 1.06; 95% CI, 0.43 to
2.62; P = .91)
Cardiac malformations: SSRI , 2 cardiac
events; NTC, 4 cardiac events
Major malformations: structural
or functional anomalies with
significant medical or social
consequences
Einarson et al,
35
2001
b,c,d,g
B Canada,
United States,
Italy,
Brazil
Venlafaxine, n = 150
SSRI, n = 150
NTC, n = 150
(Reported no significant
difference for age, alcohol and
smoking)
Venlafaxine, fluoxetine,
paroxetine, sertraline,
fluvoxamine
Major malformations: 1.6% in venlafaxine,
2.4% in SSRI vs 0.7% in NTC (P = .89)
Venlafaxine vs NTC (OR = 2.21; 95% CI, 0.20
to 24.69; P = .93)
Venlafaxine vs SSRI (OR = 0.66; 95% CI, 0.11
to 3.99; P = .99)
h
Cardiac malformations: venlafaxine, 0 events;
SSRI, 1 event; NTC, 1 event
h
Major malformations: any
anomaly with adverse
effect on the function or
social acceptability of the
individual
Simon et al,
36
2002
b,c,d,i
A United States SSRI, n = 185;
control, n = 185
TCA, n = 209;
control, n = 209
Matched for maternal age,
year of delivery, length of
enrollment in health plan,
lifetime antidepressant
prescriptions filled and
refilled, lifetime history
of psychiatric treatment
(compared smoking, alcohol,
other drug use)
Fluoxetine, fluvoxamine,
sertraline, paroxetine,
amitriptyline,
imipramine, doxepin,
nortriptyline,
protriptyline,
desipramine
Major malformations: SSRI vs control
(OR = 1.36; 95% CI, 0.56 to 3.30), TCA vs
control (OR = 0.82; 95% CI, 0.35 to 1.95)
h
Minor malformations: SSRI vs control
(OR = 1.14; 95% CI, 0.56 to 2.31), TCA vs
control (OR = 0.76; 95% CI, 0.37 to 1.58)
h
Cardiac malformations: 0.5% in TCA vs 1.0%
in unexposed (OR = 0.50; 95% CI, 0.05 to
5.53), 2.2% in SSRI vs 0.0% in unexposed
h
Einarson et al,
37
2003
b,c,d,e,f,g
B Canada,
United States,
Italy
Trazodone/
nefazodone, n = 147
Nonteratogenic
antidepressants, n= 147
NTC, n = 147
Matched for timing of call
to Teratogen Information
Service (compared age,
smoking, alcohol use)
Trazodone, nefazodone,
other nonteratogenic
antidepressants
Major malformations: 1.6% in nefazodone/
trazodone, 2.4% in other nonteratogenic
antidepressants vs 3.0% in NTC (P = .75)
h
Cardiac malformations: trazodone/
nefazodone, 0 events; other antidepressants,
1 VSD event; NTC, 1 VSD event
h
Major malformations: any
anomaly with adverse
effect on the function or
social acceptability of the
individual
Chun-Fai-Chan
et al,
38
2005
b,c,g
B Canada,
United States,
United
Kingdom
Bupropion, n = 91
Other antidepressant,
n = 89
NTC, n = 89
Matched for age, alcohol use,
smoking
Bupropion, other
antidepressants
Major malformations: 0 in bupropion,
1 in other antidepressants vs 2 in NTC
(P = .51)
h
continued
© 2013 COPYRIGHT PHYSICIANS POSTGRADUATE PRESS, INC. NOT FOR DISTRIBUTION, DISPLAY, OR COMMERCIAL PURPOSES.
e297J Clin Psychiatry 74:4, April 2013
Antidepressants and Congenital Malformations
Table 1 (continued). Study Characteristics of 27 Studies in the Congenital Malformations Meta-Analyses
Article
Quality
Threshold Country Sample Size Confounders
a
Drug Results Outcome Definition
Malm et al,
39
2005
b,c,i
A Finland SSRI, n = 1,398
Control, n = 1,782
Matched for year of pregnancy
ending, age, parity, geographic
area, social status
Adjusted for maternal age,
smoking, low social status,
nulliparity, purchase of other
reimbursed drugs
Citalopram, fluoxetine,
paroxetine, sertraline,
fluvoxamine
Major malformations: SSRI vs control
(AOR = 1.0; 95% CI, 0.6 to 1.7; P = .4)
Major congenital anomalies:
significant congenital
structural anomaly,
congenital hypothyroidism,
or chromosomal defect
Minor anomalies were not
included
Sivojelezova
et al,
40
2005
b,c,d,e,g,i
A Canada Citalopram, n = 108
Other SSRIs, n = 115
NTC, n = 118
Matched for maternal age,
gestational age at time of
recruitment (collected data on
alcohol, smoking, and other
drugs)
Citalopram, other SSRIs Major malformations: 0.9% in citalopram,
2.6% in other SSRIs vs 0.8% in NTC
(P = .52)
h
Cardiac malformations: citalopram, 1 ASD
event; other SSRIs, 2 cardiac events
(1 ASD); NTC, 0 events
h
Major malformations:
structural and/or functional
anomalies requiring surgical
correction or altering social
acceptability of the child
Djulus et al,
41
2006
b,c,g
B Canada,
United States,
Israel,
Australia,
Italy, United
Kingdom
Mirtazapine, n = 104
Other antidepressant,
n = 104
NTC, n = 104
Matched for maternal age at
time of conception, tobacco
use, alcohol use, gestational
age at the first contact,
chronic conditions
Mirtazapine, other
antidepressants
Major malformations: 1.9% in mirtazapine,
1.0% in other antidepressants vs 1.9% in
NTC; mirtazapine vs other antidepressants
(P = .50); mirtazapine vs NTC (P = .69)
h
Major malformations: structural
abnormality that was
lethal, required medical or
surgical treatment, or was of
importance cosmetically and
would affect quality of life.
Excluded genetic disorders
and chromosomal defects
Levinson-Castiel
et al,
42
2006
b,c,d,e,f,i
A Israel SSRI, n = 60
Unexposed, n = 60
Matched for sex, gestational
age, mode of delivery, birth
weight (excluded other
medications, recreational
drugs, alcohol, conditions or
congenital anomalies affecting
the central nervous system,
preterm infants)
Any SSRI Major malformations: 5% in SSRI vs 1.7% in
unexposed (P = .60)
Cardiac malformations: SSRI, 2 VSD events;
unexposed, 0 events
Wen et al,
43
2006
b,c
A Canada SSRI, n = 972
Unexposed, n = 3,878
Matched for year of birth, type
of institute at birth, first 3
digits of mother’s postal code
Adjusted for maternal age,
parity, drug dependence,
multigestation, receipt of
provincial social assistance
Citalopram, fluoxetine,
fluvoxamine, paroxetine,
sertraline
Major malformations: SSRI vs unexposed
(AOR = 0.98; 95% CI, 0.59 to 1.64)
Minor malformations: SSRI vs unexposed
(AOR = 1.02; 95% CI, 0.69 to 1.51)
h
ICD-9 codes used to identify
malformations (major and
minor)
Davis et al,
44
2007
b,d,i
A United States SSRI, n = 805
Unexposed, n = 49,031
TCA, n = 167
Unexposed, n = 49,669
Stratified by health system,
maternal age, birth season
SSRIs, TCAs One or more malformation risk: SSRIs
(RR = 0.97; 95% CI, 0.81 to 1.16); TCAs
(RR = 0.86; 95% CI, 0.57 to 1.30)
h
Cardiac malformations: SSRI, 17 cardiac
events; TCA, 2 cardiac events
h
Used ICD-9 codes and reviewed
charts for 3 categories of
congenital anomalies
Källén and
Olausson,
45
2007
b,i
A Sweden SSRI, n = 6,555
Unexposed, n = 873,876
Adjusted for year of birth,
maternal age, parity, smoking,
previous miscarriages
Paroxetine, fluoxetine,
citalopram, sertraline,
fluvoxamine,
escitalopram
Congenital malformations: SSRI (AOR = 0.89;
95% CI, 0.79 to 1.07)
Utilized definitions by the
Swedish health registers
(some used ICD codes),
chromosomal anomalies
excluded
continued
© 2013 COPYRIGHT PHYSICIANS POSTGRADUATE PRESS, INC. NOT FOR DISTRIBUTION, DISPLAY, OR COMMERCIAL PURPOSES.
e298 J Clin Psychiatry 74:4, April 2013
Grigoriadis et al
Table 1 (continued). Study Characteristics of 27 Studies in the Congenital Malformations Meta-Analyses
Article
Quality
Threshold Country Sample Size Confounders
a
Drug Results Outcome Definition
Louik et al,
46
2007
d,e,i
A Canada,
United States
Infants with birth defects,
n = 9,849
Infants without birth
defects, n = 5,860
Adjusted for maternal age, race/
ethnic group, education, year
of last menstrual period,
study center, smoking,
alcohol, family history of
birth defect, prepregnancy
BMI, parity, seizures, diabetes
mellitus, hypertension,
infertility, use of folic acid
SSRI, non-SSRI
antidepressants
Cardiac malformations:
SSRIs (AOR = 1.2; 95% CI, 0.9 to 1.6);
non-SSRIs (AOR = 0.8; 95% CI, 0.5–1.5)
h
Septal heart defects:
SSRIs (AOR = 1.2; 95% CI, 0.8 to 1.8),
non-SSRIs (AOR = 1.1; 95% CI, 0.6 to 2.4)
h
ICD-9-CM codes
Boucher et al,
47
2008
b,d,e,f
B Canada Any antidepressant, n = 73
Unexposed, n = 73
Matched for same hospital,
gestational age, and date at
delivery
(No group difference for
smoking, alcohol, or age)
Citalopram, paroxetine,
sertraline, fluoxetine,
fluvoxamine, venlafaxine,
amitriptyline, trazodone,
mirtazapine
Congenital malformations: 3% in any
antidepressant vs 5% in unexposed (P = .68)
Cardiac malformations: exposed, 1 VSD
event; unexposed, 0 events
Diav-Citrin
et al,
48
2008
b,c,d,g
B Israel, Italy,
Germany
Paroxetine, n = 463
Fluoxetine, n = 346
NTC, n = 1,467
Adjusted cardiac analysis
for gestational age at call,
maternal age, smoking,
previous miscarriages, origin,
concomitant psychiatric
medications, multifetal
gestation, SSRI dose
Paroxetine, fluoxetine Major malformations: 5.2% in paroxetine,
4.7% in fluoxetine vs 2.5% in NTC
(P < .05)
h
Cardiac malformations: 2.0% in paroxetine,
2.8% in fluoxetine vs 0.6% in NTC (P < .05)
(paroxetine exposed: AOR = 2.66; 95% CI,
0.80–8.90) (fluoxetine exposed: AOR = 4.47;
95% CI, 1.31–15.27)
h
Major anomalies: structural
abnormalities with serious
surgical, medical, or cosmetic
consequences. Significant
neurodevelopmental or
functional abnormalities
were classified as major
anomalies when they
necessitated special
education or interventions
Einarson et al,
49
2008
d,g,i
A Canada,
United States,
Italy,
Switzerland,
Australia,
Germany,
Israel,
Finland
Paroxetine, n = 1,174
NTC, n = 1,174
(Reported “similar
characteristics,” including
smoking and alcohol use)
Paroxetine Cardiac malformations: 0.7% in paroxetine
vs 0.7% in NTC (OR = 1.1; 95% CI, 0.36 to
2.78)
Oberlander
et al,
50
2008
b,c,d,e,f
A Canada SSRI, n = 2,625
Unexposed, n = 107,320
Adjusted for age, prenatal care
visits, diagnosis of depression
in year before pregnancy,
number of psychiatric
visits and physician visits
in year before pregnancy,
diseases and complications
of pregnancy, depression in
first trimester, prescription fill
after pregnancy known
Citalopram, fluoxetine,
fluvoxamine, paroxetine,
sertraline, venlafaxine
Major malformations: 2.9% in SSRI vs 3.1%
in unexposed (regression adjusted risk
difference = 0.61; 95% CI, 1.44 to 0.21)
Cardiac malformations: SSRI, 17 events vs
control, 512 events (regression adjusted risk
difference = 0.21, 95% CI, 0.14 to 0.55)
Septal heart defects: SSRI, 12 events (6 VSD
events); control, 294 events (219 VSD
events) (VSD: regression adjusted risk
difference = 0.10; 95% CI, 0.12 to 0.33)
Major congenital anomalies
identified by ICD-9 codes
(740.0–759.9)
Excluded specific minor
anomalies
continued
© 2013 COPYRIGHT PHYSICIANS POSTGRADUATE PRESS, INC. NOT FOR DISTRIBUTION, DISPLAY, OR COMMERCIAL PURPOSES.
e299J Clin Psychiatry 74:4, April 2013
Antidepressants and Congenital Malformations
Table 1 (continued). Study Characteristics of 27 Studies in the Congenital Malformations Meta-Analyses
Article
Quality
Threshold Country Sample Size Confounders
a
Drug Results Outcome Definition
Ramos et al,
51
2008
b,c
A Canada Any antidepressant,
n = 1,101
Unexposed, n = 1,228
Adjusted for maternal age,
receiving welfare, urban
dweller, living alone,
psychiatric disorders and
comorbidities unrelated to
psychiatric disorders before
and during pregnancy,
diabetes and hypertension
before and during pregnancy,
baby’s gender, year of
pregnancy, prenatal visits
SSRI, TCA, bupropion,
mirtazapine,
moclobemide,
nefazodone, trazodone,
venlafaxine
Major malformations: any antidepressant
(AOR = 1.10; 95% CI, 0.75 to 1.62)
ICD-9 codes
Einarson et al,
52
2009
b,c,g,i
A Canada
Any antidepressant,
n = 928
NTC, n = 928
Matched for maternal age,
smoking, alcohol use
Bupropion, citalopram,
escitalopram,
fluvoxamine,
nefazodone, paroxetine,
mirtazapine, fluoxetine,
trazodone, venlafaxine,
sertraline
Major malformations: 2.5% in any
antidepressant vs 2.6% in NTC (risk
ratio = 0.96; 95% CI, 0.55 to 1.67)
Merlob et al,
53
2009
d,e,f
A Israel SSRI, n = 235
Unexposed, n = 67,636
(Collected data on smoking and
alcohol)
Paroxetine, fluoxetine,
citalopram, escitalopram,
sertraline, fluvoxamine,
venlafaxine
Cardiac malformations: 3.4% in SSRI vs 1.6%
in unexposed (RR = 2.17; 95% CI, 1.07 to
4.39; P = .023)
Septal heart defects: exposed, 6 VSD events;
unexposed, 656 VSD events
Screened infants with heart
murmurs for cardiac
malformations (excluded
cases of congenital
syndromes)
Functional murmurs, isolated
persistent foramen ovale,
peripheral pulmonic stenosis,
and patent ductus arteriosus
excluded
Pedersen et al,
54
2009
c,i
A Denmark SSRI, n = 1,370
Unexposed, n = 493,113
Adjusted for age, year, income,
marital status, smoking
Fluoxetine, citalopram,
paroxetine, sertraline,
fluvoxamine
Major malformations: SSRI exposure
(AOR = 1.21; 95% CI, 0.91 to 1.62; NS)
ICD-10 codes
Malformations detected at
birth for live-born infants or
within year 1 after birth were
considered
Wichman et al,
55
2009
d,e,f
A United States Exposed, n = 808
Unexposed, n = 24,406
Citalopram, escitalopram,
paroxetine, fluoxetine,
sertraline, venlafaxine
Cardiac: 0.4% in exposed vs 0.8% in
unexposed (P = .23)
VSD: 0.0% in exposed vs 0.1% in unexposed
(P > .99)
Congenital heart
disease—abnormality in
cardiocirculatory function
or structure, present at
birth. VSD—hole within the
septum between the heart
ventricles
Heart disease diagnosed
immediately following birth
and prior to discharge home
included
continued
© 2013 COPYRIGHT PHYSICIANS POSTGRADUATE PRESS, INC. NOT FOR DISTRIBUTION, DISPLAY, OR COMMERCIAL PURPOSES.
e300 J Clin Psychiatry 74:4, April 2013
Grigoriadis et al
Table 1 (continued). Study Characteristics of 27 Studies in the Congenital Malformations Meta-Analyses
Article
Quality
Threshold Country Sample Size Confounders
a
Drug Results Outcome Definition
Bakker et al,
56
2010
d,e,f
A Netherlands Exposed, n = 16
Unexposed, n = 1,266
Adjusted for year of birth
(compared age, smoking,
alcohol, number of other
characteristics)
Paroxetine Cardiac malformations: 10 events in exposed
cases (AOR = 1.5; 95% CI, 0.5 to 4.0;
P = .476)
Septal heart defects: exposed, 4 (AOR = 1.6;
95% CI, 0.4 to 5.6; P = .493)
VSD: exposed, 1 (AOR = 0.5; 95% CI, 0.1 to
4.2; P = .528)
ICD-9 and ICD-10
Kornum et al,
57
2010
b,d,e,i
A Denmark SSRI, n = 2,062
Non-SSRI antidepressant,
n = 358
Unexposed, n = 213,712
Adjusted for smoking, maternal
age, birth order, birth year
Paroxetine, fluoxetine,
sertraline, citalopram,
escitalopram, non-SSRI
antidepressant
Congenital malformations: 5.1% in SSRI, 3.5%
in unexposed (AOR = 1.3; 95% CI, 1.1 to
1.6); non-SSRI exposure (AOR = 0.5; 95%
CI, 0.2 to 1.1)
h
Cardiac malformations: 1.3% in SSRI vs 0.7%
in unexposed (AOR = 1.7; 95% CI, 1.1 to
2.5)
h
Septal heart defects: SSRI, 18 events
(AOR = 1.4; 95% CI, 0.8 to 2.3); non-SSRI,
0 events
h
ICD-8 and ICD-10 codes
Malformations had to be
registered within the first
year of life and a singleton
birth, excluded chromosomal
anomalies
Reis and
Källén,
58
2010
c,d,e,i
A Sweden SSRI, n = 10,170
TCA, n = 1,662
SNRI, n = 1,351
MAOI, n = 37
Unexposed, n = 1,236,053
Adjusted for year of birth,
maternal age, parity, smoking,
BMI
SSRIs, TCAs, SNRIs,
MAOIs
Severe malformations: TCAs (AOR = 1.36;
95% CI, 1.07 to 1.72), SSRIs (AOR = 1.08;
95% CI, 0.97 to 1.21), SNRIs (AOR = 1.00;
95% CI, 0.73 to 1.37), antidepressants
(AOR = 1.10; 95% CI, 1.00–1.22)
Cardiac malformations: TCAs (AOR = 1.63;
95% CI, 1.12 to 2.36), SSRIs (AOR = 0.99;
95% CI, 0.82 to 1.20), SNRIs (AOR = 1.33;
95% CI, 0.84 to 2.09), antidepressants
(AOR = 1.11; 95% CI, 0.94–1.30)
Septal heart defects: TCAs (AOR = 1.84;
95% CI, 1.13 to 2.97), SSRIs (AOR = 1.00;
95% CI, 0.77 to 1.29), SNRIs (AOR = 1.26;
95% CI, 0.63 to 2.26), antidepressants
(AOR = 1.11; 95% CI, 0.89–1.38)
ICD-9 and ICD-10 codes
Excluded preauricular
appendix, patent ductus
(preterm infant), tongue
tie, single umbilical artery,
hip subluxation, and nevus,
undescended testicle.
Considered these conditions
common, recorded variably
and of lower significance
clinically
a
Refers to matching, exclusions, or adjusted data for outcome of interest. Individual studies may have collected or compared groups on other characteristics or adjusted data on other outcomes.
b
Included in our any congenital malformations meta-analysis.
c
Included in our major malformations meta-analysis.
d
Included in our cardiovascular malformations meta-analysis.
e
Included in our septal heart defects meta-analysis.
f
Included in our ventral septal defects meta-analysis.
g
Convenience sample.
h
We combined the drug groups in our analyses.
i
Included in no antidepressant exposure in control group analysis.
Abbreviations: A = above quality threshold; AOR = adjusted odds ratio; ASD = atrial septal defects; B = below quality threshold; ICD-8 = International Classification of Diseases, Eighth Revision;
ICD-9-CM = International Classification of Diseases, Ninth Revision, Clinical Modification; ICD-10 = International Classification of Diseases, Tenth Revision; MAOI = monoamine oxidase inhibitor;
NS = nonsignificant; NTC = nonteratogen control; OR = odds ratio; RR = relative risk; SNRI = serotonin-norepinephrine reuptake inhibitor; SSRI = selective serotonin reuptake inhibitor; TCA = tricyclic
antidepressant; VSD = ventral septal defects.
© 2013 COPYRIGHT PHYSICIANS POSTGRADUATE PRESS, INC. NOT FOR DISTRIBUTION, DISPLAY, OR COMMERCIAL PURPOSES.
e301J Clin Psychiatry 74:4, April 2013
Antidepressants and Congenital Malformations
population- or hospital-based sample. Of all studies avail-
able, 13 clearly stated that no antidepressants were used in
the control group and 21 reported adjusted data or applied
matching. Information on fluoxetine was available from 0 to
7 studies depending on the outcome, and 0–8 studies were
available for paroxetine.
Congenital Malformations
Overall, we pooled results from 12 studies* that were
above quality threshold; the RR for congenital malforma-
tions was 0.93 (95% CI, 0.85–1.02; P = .113; Figure 2). Results
were similar when studies were restricted to those studies
that clearly excluded antidepressants from the control group.
When studies with any adjustment were analyzed, all the
studies regardless of quality, or 10 studies that did not use
convenience samples (regardless of study quality), yielded
similar results, with RRs ranging from 0.92 to 0.95 (Table
2). The RRs from the meta-analyses with the individual
antidepressants (fluoxetine and paroxetine) were similar in
magnitude (ranging from 1.02 to 1.15) and not statistically
significant (Supplementary eTable 1).
Major Congenital Malformations
When only major malformations were analyzed, the
pooled risk was small and not significant for the studies
that were above quality threshold (RR = 1.07; 95% CI, 0.99–
1.17; P = .095; Supplementary eFigure 1). Overall, the RRs for
the other analyses were small, between 1.07 and 1.10 (Table
2), but statistically significant depending on the subanalysis
conducted. Results were not significant for any of the analy-
ses for paroxetine use; however, the risk associated with the
*Pedersen et al
54
and Reis and Källén
58
were excluded from this analysis,
as these studies were population based and from the same country as
Kornum et al
57
and Källén and Olausson,
45
respectively.
use of fluoxetine was similar to the overall analyses, with
slightly higher RRs between 1.20 and 1.29 (Supplementary
eTable 1).
Cardiovascular Malformations
With regard to cardiovascular malformations (13 studies
above quality threshold) as the disease outcome, the pooled
RR was small (RR = 1.36; 95% CI, 1.08–1.71; P = .008; Figure
3) but statistically significant in all analyses, ranging from
1.26 to 1.39 (see Table 2). One study
58
in particular had a
strong influence on the results because of its sample size,
detailed exposure measurement, and moderate study qual-
ity and was included in most subanalyses. Few studies were
available for analyses stratified by medication. The risk asso-
ciated with paroxetine was slightly higher (RR = 1.43; 95% CI,
1.08–1.88; P = .012; RRs between 1.43–1.47) and consistently
statistically significant. The risk associated with fluoxetine
was slightly lower and not statistically significant (RRs from
1.17 to 1.33; Supplementary eTable 1).
Septal Heart Defects (atrial septal defects or ventral
septal defects)
Results involving the 9 studies above quality threshold
indicated a significant association between exposure to
antidepressants during pregnancy and an increased risk for
any septal heart defect (atrial septal defects or ventral septal
defects) (RR = 1.40; 95% CI, 1.10–1.77; P = .005; Supple-
mentary eFigure 2). In subanalyses, the RRs were similar
to the findings for any cardiovascular malformation (RRs
between 1.17 and 1.40; Table 2) and statistically significant.
Two studies
46,58
had a particularly strong influence on the
pooled estimate. There was no evidence suggesting that
paroxetine or fluoxetine were associated with septal defects
(Supplementary eTable 1); however, only 3 and 2 studies,
respectively, were available for such an analysis.
Figure 2. Exposure to Any Antidepressant and the Risk of Congenital Malformations:
Meta-Analysis Results for Studies Above the Quality Threshold
Study
Relative Risk
(95% CI)
Weight,
%
a
Kulin et al,
34
1998 1.06 (0.43–2.62) 1.02
Simon et al,
36
2002 0.78 (0.45–1.36) 2.69
Malm et al,
39
2005 1.00 (0.59–1.68) 3.07
Sivojelezova et al,
40
2005 2.05 (0.15–27.74) 0.12
Levinson-Castiel et al,
42
2006 3.00 (0.11–79.14) 0.08
Wen et al,
43
2006 1.01 (0.74–1.37) 8.77
Davis et al,
44
2007 0.95 (0.81–1.12) 31.69
Källén and Olausson,
45
2007 0.89 (0.76–1.04) 36.16
Oberlander et al,
50
2008 0.81 (0.58–1.14) 7.12
Ramos et al,
51
2008 1.10 (0.75–1.62) 5.61
Einarson et al,
52
2009 0.96 (0.55–1.67) 2.70
Kornum et al,
57
2010 0.89 (0.35–2.24) 0.98
Overall (I
2
= 0.0%, P = .984) 0.93 (0.85–1.02) 100.00
a
Weights are from random effects analysis.
1.00.5 1.5 2.0
© 2013 COPYRIGHT PHYSICIANS POSTGRADUATE PRESS, INC. NOT FOR DISTRIBUTION, DISPLAY, OR COMMERCIAL PURPOSES.
e302 J Clin Psychiatry 74:4, April 2013
Grigoriadis et al
Ventral Septal Defects
Results for ventral septal defects were similar to any car-
diac malformations (RR = 1.54; 95% CI, 0.71–3.33; P = .274;
Supplementary eFigure 3), but they were not statistically
significant; the few studies that investigated this outcome
were generally small and reported very few or sometimes
no case in either the exposure or the comparison group
(Table 2). There were not enough studies to conduct a meta-
analysis stratified by paroxetine or fluoxetine.
Publication Bias and Influential Studies
When the analyses were repeated with all identified stud-
ies regardless of study quality, we found very similar results.
We did not find evidence for the presence of publication
bias in any of our analyses (see Table 2). When we recalcu-
lated the pooled risk, excluding studies one by one, 1 study
58
was found to have had a particular influence on the derived
pooled risk. This finding was most apparent for cardiac
outcomes, for which exclusion of this study resulted in a
slightly higher risk. For all other analyses, the estimate was
well within the pooled CIs.
DISCUSSION
We conducted a systematic review and meta-analysis of
the literature, examining associations between the use of any
antidepressant and rates of overall congenital malformations,
overall major congenital malformations, and, more specifi-
cally, cardiac malformations (as defined by the authors of the
individual articles; see Table 1). Furthermore, on the basis
of the available data for cardiac outcomes, we were then
able to examine possible associations with antidepressant
exposure and septal defects, as well as ventral septal defects
alone. In addition to any antidepressant, we were also able
to conduct analyses on 2 individual antidepressant medica-
tions (paroxetine and fluoxetine). These 2 medications were
chosen on the basis of available evidence for meta-analyses
on the outcomes of interest; the evidence reflects the focus
of previously published literature. We found no evidence that
Table 2. Exposure to Any Antidepressant and Risk of Malformation: Meta-Analyses Results
Analysis
No. of
Studies
No. of
Cases
Total
Sample Size
Relative
Risk 95% CI P Value
P Value for
Heterogeneity I
2
(%)
P Value for
Publication Bias
Congenital malformations
Studies above quality threshold 12 52,572 1,223,210 0.93 0.85–1.02 .113 .984 0.0
Studies above quality threshold,
no antidepressants in controls
9 48,675 1,102,717 0.92 0.83–1.02 .115 .988 0.0
Studies above quality threshold
with adjusted data
12 52,572 1,223,210 0.93 0.85–1.02 .113 .984 0.0
All studies 20 52,687 1,226,756 0.95 0.87–1.04 .308 .684 0.0 .439
All studies excluding convenience samples 10 52,506 1,220,536 0.92 0.84–1.01 .099 .953 0.0
Major malformations
Studies above quality threshold 11 56,334 1,940,124 1.07 0.99–1.17 .095 .857 0.0
Studies above quality threshold,
no antidepressants in controls
8 52,605 1,817,081 1.10 1.01–1.21 .032 .986 0.0
Studies above quality threshold
with adjusted data
11 56,334 1,940,124 1.07 0.99–1.17 .095 .857 0.0
All studies 18 56,443 1,943,538 1.09 1.01–1.18 .033 .666 0.0 .984
All studies excluding convenience samples 8 56,262 1,936,142 1.08 0.99–1.17 .091 .650 0.0
Cardiovascular malformations
Studies above quality threshold 13 20,444 1,547,012 1.36 1.08–1.71 .008 .134 31.1
Studies above quality threshold,
no antidepressants in controls
9 17,945 1,338,913 1.33 1.02–1.75 .037 .147 33.9
Studies above quality threshold
with adjusted data
10 19,128 1,450,406 1.35 1.07–1.70 .011 .181 28.6
All studies 18 20,473 1,550,271 1.26 1.07–1.47 .005 .313 11.8 .238
All studies excluding convenience samples 11 20,419 1,542,707 1.39 1.09–1.79 .009 .089 39.1
Septal heart defects
Studies above quality threshold 9 10,195 1,703,561 1.40 1.10–1.77 .005 .075 44.0 .144
Studies above quality threshold,
no antidepressants in controls
5 8,954 1,494,368 1.17 1.03–1.33 .013 .527 0.0
Studies above quality threshold
with adjusted data
7 9,509 1,608,759 1.35 1.08–1.68 .009 .105 42.9
All studies 12 10,200 1,704,652 1.37 1.11–1.69 .003 .168 28.2 .173
All studies excluding convenience samples 9 10,194 1,703,049 1.40 1.10–1.77 .006 .073 44.3
Ventral septal defects
Studies above quality threshold 5 1,096 207,467 1.54 0.71–3.33 .274 .232 28.4 .557
Studies above quality threshold,
no antidepressants in controls
1
a
2 120 7.51 0.46–121.51 .155
Studies above quality threshold
with adjusted data
3 410 114,382 1.39 0.48–4.07 .542 .275 22.6
All studies 8 1,101 208,316 1.67 0.94–2.97 .081 .473 0.0 .565
All studies excluding convenience samples 6 1,097 207,613 1.65 0.81–3.36 .168 .294 18.4
a
Insufficient number of studies for meta-analyses.
© 2013 COPYRIGHT PHYSICIANS POSTGRADUATE PRESS, INC. NOT FOR DISTRIBUTION, DISPLAY, OR COMMERCIAL PURPOSES.
e303J Clin Psychiatry 74:4, April 2013
Antidepressants and Congenital Malformations
antidepressants were associated with congenital malforma-
tions or major malformations in our primary analyses. Results
were generally confirmed in several subanalyses, although
there was an indication that antidepressants in general and,
more specifically, fluoxetine exposure was associated with
major congenital malformations. It is important to consider
that these classifications of congenital and major malforma-
tions include many different malformations with different
timelines of development, which can potentially confound
relationships with a specific malformation, if one exists.
54
It
is also important to note that we were able to analyze only
2 specific antidepressants, and so we cannot draw conclu-
sions regarding the safety of other individual antidepressant
drugs that we did not examine separately. Any antidepressant
exposure was found to be associated with cardiac malforma-
tions in general. More specifically, when we looked at septal
heart defects, we also found any antidepressant exposure to
be associated. However, it is important to consider that sta-
tistical significance is not equivalent to clinical significance.
There were only a few studies with which to further examine
categorizations of cardiac malformations, and these studies
had a small number of cases. We were able to analyze only
ventral septal defects as a subcategory of septal defects, and
these analyses were not significant, although the RRs were
similar. Further research of high methodological quality is
needed to confirm the results presented here.
There may be different associations between specific anti-
depressants and malformations. For example, antidepressants
as a whole were associated with risk for any cardiac malforma-
tion, as was paroxetine, yet fluoxetine was not. Regardless of
whether a class or individual effect exists, it is important to
note that, overall, these results are largely reassuring, as even
the nonsignificant meta-analyses did not return an RR higher
than 1.67, with the highest significant one being 1.47, which
is below the 2-fold increase benchmark that has been cited
for clinical significance in this field.
52
The findings also are
relieving, as depression during pregnancy is common, with
rates as high as 20% being reported,
59
and 1.8%2.1% of preg-
nancies having reported exposure to an SSRI,
60,61
among the
most frequently used of the antidepressants in pregnancy.
62
As far as we are aware, our study is the first meta-analysis
to report a small increase in major malformations with flu-
oxetine, the oldest of the SSRIs in use. Note, however, that
this was not a consistent finding, as our main analysis, which
was with the higher quality studies, was not significant,
although it was reduced to 4 studies. Several meta-analyses
previously did report an increased risk for cardiovascular
malformations with paroxetine (ie, Bar-Oz et al,
13
Wurst et
al
16
) but, unlike Wurst et al,
16
we did not find an increased
risk for congenital malformations with this drug. Ours is the
first meta-analysis, to our knowledge, to specifically examine
the type of malformation implicated, and our results suggest
that septal heart defects are associated with antidepressant
exposure. However, although paroxetine was associated with
cardiovascular malformations in general, it was not associated
with septal heart defects in our analysis. Isolated ventricular
malformations are a common type of congenital cardiovas-
cular malformation, with an incidence rate of approximately
2%–5%.
63
These are typically small muscular defects, though
they can be large or complex, and are often asymptomatic
and/or close spontaneously early on and thus are frequently
not detected.
63
Atrial septal defects, on the other hand, are
less common
63
and are approximately 7% of congenital car-
diac lesions.
64
As there are more severe and less severe types,
some will require surgical intervention in infancy to avoid
negative outcomes,
65
while others close spontaneously.
63
In
contrast to the majority of ventricular septal defects that do
close by the first year, it is often the atrial septal defects that
are not diagnosed early and present in adulthood.
63
Previous
research suggested that the inclusion of common reversible
Figure 3. Exposure to Any Antidepressant and the Risk of Cardiovascular Malformations:
Meta-Analysis Results for Studies Above the Quality Threshold
Study
Relative Risk
(95% CI)
Weight,
%
a
Kulin et al,
34
1998 0.50 (0.06–4.43) 1.06
Simon et al,
36
2002 2.52 (0.31–20.29) 1.16
Sivojelezova et al,
40
2005 4.52 (0.41–50.11) 0.88
Levinson-Castiel et al,
42
2006 7.51 (0.46–121.51) 0.66
Davis et al,
44
2007 0.93 (0.57–1.52) 12.86
Louik et al,
46
2007 2.16 (1.16–4.02) 9.54
Einarson et al,
49
2008 1.10 (0.40–3.06) 4.32
Oberlander et al,
50
2008 1.44 (0.83–2.49) 11.28
Merlob et al,
53
2009 2.17 (1.07–4.40) 7.90
Wichman et al,
55
2009 0.44 (0.12–1.68) 2.67
Bakker et al,
56
2010 1.50 (0.53–4.24) 4.20
Kornum et al,
57
2010 1.70 (1.13–2.56) 15.78
Reis and Källén,
58
2010 1.11 (0.94–1.31) 27.69
Overall (I
2
= 31.1%, P = .134) 1.36 (1.08–1.71) 100.00
a
Weights are from random effects analysis.
1.00.5 2.0 3.0
© 2013 COPYRIGHT PHYSICIANS POSTGRADUATE PRESS, INC. NOT FOR DISTRIBUTION, DISPLAY, OR COMMERCIAL PURPOSES.
e304 J Clin Psychiatry 74:4, April 2013
Grigoriadis et al
heart defects such as the ventral ones may inflate any existing
association found with antidepressant exposure.
13,49,52,63
We
attempted to differentiate between atrial septal defects and
ventral septal defects, but there were not enough studies.
Although our ventral septal defects analysis was not signifi-
cant, the RRs were in a similar range to the overall septal
defects analysis, which was significant. Note that the CIs
were wide for the ventral septal defects-only analyses, and,
thus, perhaps the results are biased. Sources of uncontrolled
bias leading to the marginal increased risk for septal defects
may include detection bias.
13
Women with depression may
be more likely to have an ultrasound and echocardiogram in
pregnancy and after birth. They may also have an increased
use of health care services due to the depression itself or due
to the publicity surrounding the risks of congenital malfor-
mations with SSRIs. Furthermore, many more women with
depression may request diagnostic testing of their infants
compared to healthy mothers.
13
All of the aforementioned
potential sources of bias could increase the chance of false
detection of an increased risk for septal defects and SSRI
exposure. More research is certainly needed that focuses on
these types of malformations, with the goal of determining
whether an association with the type of cardiovascular effect
exists, if it is dependent on a particular drug, or if indeed it
is a class effect.
Strengths and Limitations
Perinatal researchers and clinicians have poor evidence-
based information on effective treatments for perinatal
women, as conducting the evidence-based gold standard,
the randomized controlled trial, especially with pharmaco-
logical interventions, has ethical barriers among others.
66
Our reliance on observational studies and registries
67,68
may
propagate biases, as they have methodological limitations;
thus, we keenly await new studies for replication and to con-
firm previous results. One of the strengths of our analysis
was our attention to study quality. We limited our analyses to
studies determined to be above a certain threshold of qual-
ity, according to the SAQOR, excluding studies below that
threshold. Further, we ran subanalyses to examine whether
limiting the data to those studies that (1) confirmed no anti-
depressant use in the control group, (2) had any adjustments
made to the data, or (3) did not use convenience sampling
among all the studies influenced any of the study results.
We found few differences associated with study quality, sug-
gesting that the findings presented in our analysis are quite
robust. One exception was our analysis for major congenital
malformations. While the pooled risk measure was not sta-
tistically significant in our main analysis using only studies
above our quality threshold, the pooled estimates became
statistically significant in 2 subanalyses. Regardless of the
statistical significance, however, it is important to note that
the RRs increased from 1.07 to only 1.10. As such, a very
small and only marginally significant effect was present, even
in these secondary analyses. Thus, largely, differences were
small when examining all available data versus studies above
our quality threshold. This is noteworthy given the known
methodological flaws in these data as well as our concerns
with our inability to make definitive conclusions based on
previous research.
A further strength of our work rests on the fact that we
used adjusted or matched data where possible in our meta-
analyses, as well as running subanalyses for studies with any
adjustments. Research in the area has been criticized for not
taking into account non-iatrogenic confoundersin preg-
nancy.
17
Indeed, many studies either did not adjust for any
confounders, including depression, or applied matching in
very small samples, which makes it difficult to assess whether
such matching was successful. However, the most influential
studies were also the large ones, and thus heavily influenced
the results; these studies did adjust for several potential con-
founders.
45,46,50,57,58
Thus, most of our pooled risk estimates
were heavily influenced by studies that did use adjusted data,
including our subanalyses. Although this is relieving to a cer-
tain extent, these studies used data from population-based
registries; they provide better evidence than that provided by
data derived from convenience samples, but they are not ran-
domized controlled trials, which provide the best evidence.
The absence of evidence from randomized controlled trials
highlights the importance for more research. There seems
to be a clear need for more studies of high methodological
quality that clearly separate antidepressant exposure across
intervention and control groups and accurately adjust for
the potential confounding effect of other risk factors for
malformations.
Our study was limited by the quality of the articles