|
|
Background Information regarding the safety of selective serotonin-reuptake inhibitors (SSRIs) in human pregnancy is sparse. Concern has been raised about the risk of congenital heart defects associated with the use of SSRIs in pregnancy.
Methods We obtained data on 9622 case infants with major birth defects and 4092 control infants born from 1997 through 2002 from the National Birth Defects Prevention Study. Case infants were ascertained through birth-defects surveillance systems in eight U.S. states; controls were selected randomly from the same geographic areas. Mothers completed a standardized telephone interview regarding exposure to potential risk factors, including medications, before and during pregnancy. Exposure to SSRIs was defined as treatment with any SSRI from 1 month before to 3 months after conception. Birth defects were assigned to 26 categories and subcategories.
Results There were no significant associations between maternal use of SSRIs overall during early pregnancy and congenital heart defects or most other categories or subcategories of birth defects. Maternal SSRI use was associated with anencephaly (214 infants, 9 exposed; adjusted odds ratio, 2.4; 95% confidence interval [CI], 1.1 to 5.1), craniosynostosis (432 infants, 24 exposed; adjusted odds ratio, 2.5; 95% CI, 1.5 to 4.0), and omphalocele (181 infants, 11 exposed; adjusted odds ratio, 2.8; 95% CI, 1.3 to 5.7).
Conclusions Maternal use of SSRIs during early pregnancy was not associated with significantly increased risks of congenital heart defects or of most other categories of birth defects. Associations were observed between SSRI use and three types of birth defects, but the absolute risks were small, and these observations require confirmation by other studies.
We used data from the National Birth Defects Prevention Study (NBDPS) to evaluate the relationship between maternal SSRI use in early pregnancy and the occurrence of selected birth defects.
Methods
Study Subjects and Data Collection
The NBDPS is an ongoing, multisite, case–control study of environmental and genetic risk factors for more than 30 selected categories of major birth defects. Case infants had received a diagnosis of at least one selected birth defect and were ascertained by population-based birth-defects surveillance systems at eight study sites (Arkansas, California, Georgia, Iowa, Massachusetts, New Jersey, New York, and Texas).9 We used data from infants who were born on or after October 1, 1997, and who had an estimated date of delivery on or before December 31, 2002. Case infants were born alive (all participating sites) or died at 20 weeks or more of gestation (Arkansas, California, Georgia, Iowa, Massachusetts, and Texas). Pregnancies with reliably ascertained defects that were electively terminated were also included in Arkansas, California, Georgia, Iowa, and Texas. Infants with recognized or strongly suspected chromosomal abnormalities or single-gene conditions were excluded from the study.
The controls were live-born infants with no major birth defects who were randomly selected from hospital or state birth-certificate records from the same geographic areas. Only one case or control infant was included from each multifetal pregnancy.
Birth defects for which at least 200 case mothers were interviewed were selected for the analysis. This number was determined as that needed to obtain 80% power with an odds ratio of 2.5, given a 2.1% exposure rate to SSRIs among control mothers. Information on the infants in each defect category was reviewed by clinical geneticists who were unaware of the infants' exposure status, who confirmed case eligibility and classified the cases as isolated (no additional major unrelated defect) or multiple (more than one major unrelated birth defect).10 To reduce pathogenetic heterogeneity, cases with complex sequences (e.g., omphalocele–exstrophy–imperforate anus–spinal defects phenotype)11 were excluded, and isolated defects were analyzed separately. A total of 709 infants who had more than one eligible birth defect were included in multiple analyses. Each case of a cardiac birth defect was reviewed by a team of experts in pediatric cardiology and the epidemiology of heart defects and was assigned to a single cardiac diagnostic category.
Demographic information and information on exposure to SSRIs and other potential risk factors during pregnancy were collected by standardized telephone interviews with mothers of case and control infants conducted in English or Spanish from 6 weeks to 2 years after the estimated date of delivery.9 Infants for whom complete maternal interviews were unavailable were excluded. The participation rates among case and control mothers were 71.1% and 68.6%, respectively. The mothers were asked whether they had taken any of three specific SSRIs, identified by brand name — Prozac (fluoxetine), Zoloft (sertraline), and Paxil (paroxetine) — during and before pregnancy, and when each medication was taken. They were also asked about their use of other medications during this period, which enabled us to analyze associations with other SSRIs. Exposure was defined as reported use of any SSRI from 1 month before to 3 months after conception (the date of conception was calculated as 266 days before the estimated date of delivery). Women were considered unexposed if they did not take an SSRI at any time during pregnancy or during the 3 months before conception. Women who took non-SSRI antidepressants were included in the unexposed group.
Statistical Analysis
Our hypothesis for this exploratory study was that SSRI exposure early in pregnancy is associated with one or more of the selected categories of birth defects. Crude analyses were performed by Pearson's chi-square test, and odds ratios and (if the expected number in a cell was less than five) Fisher's exact confidence limits were calculated with SPSS software, version 10.0, and with SABER, with the use of the same control group for every comparison.
The following potential confounders were evaluated: maternal race or ethnic group (non-Hispanic white vs. other), maternal age (under 35 years of age vs. 35 or older), maternal education (12 years or less vs. more than 12 years), prepregnancy obesity (body-mass index [the weight in kilograms divided by the square of the height in meters] <30 vs. 
30), maternal smoking or alcohol use from 1 month before to 3 months after conception, maternal folic acid use from 1 month before to 1 month after conception, annual family income (less than $20,000 vs. $20,000 or more), singleton versus multiple pregnancy, parity (no previous live births vs. one or more previous live births), and presence or absence of prepregnancy hypertension. Potential confounders were first evaluated for associations with SSRI exposure and with each of the specific defects, and they were excluded from the logistic regression if their removal resulted in a change in risk estimate of less than 10%. All confounders retained in the model for any of the defects were included in the final models for all defects. Infants of mothers with prepregnancy type 1 or 2 diabetes were excluded from adjusted analyses because of the strong association of diabetes with birth defects.
For the defects with positive associations, we performed post hoc analyses of SSRIs stratified according to factors that were associated both with maternal SSRI use and with the outcomes. These factors were maternal race or ethnic group, age, education, presence or absence of obesity, presence or absence of smoking, presence or absence of hypertension, and singleton versus multiple pregnancy. After identification of effect modification by obesity for craniosynostosis, as indicated by the Breslow–Day test, we further evaluated the effect of obesity on all defects by performing analyses with a four-level variable (obesity alone, SSRI exposure alone, both SSRI exposure and obesity, neither SSRI exposure nor obesity) by asymptotic and Fisher's exact logistic regression with the use of LogXact software, version 4.0.12
We also performed crude and adjusted analyses for isolated defects only and assessed the associations between four specific SSRIs (fluoxetine, sertraline, paroxetine, and citalopram [Celexa]) and 4 combined birth-defect categories and between three specific SSRIs (fluoxetine, sertraline, and paroxetine) and 16 categories of birth defects when there were at least three exposed infants. Additional analyses were performed assessing the association between the use of any SSRI during two other time intervals of pregnancy and 16 categories of birth defects. No adjustment was made for the multiple comparisons performed.
Results
Eighteen categories of birth defects and eight subcategories of cardiac birth defects met the criterion of being represented by at least 200 interviews with case mothers. Interviews were incomplete for 163 case or control mothers, and 5 mothers reported depression but did not report use of antidepressants. These subjects were excluded from the analysis, leaving 9622 case and 4092 control infants.
Of the 13,714 case and control mothers included, 408 (3.0%) reported using an SSRI at some point before or during pregnancy. The reported frequency of SSRI use in the period from 1 month before to 3 months after conception was 2.3%; 230 mothers of case infants (2.4%) and 86 mothers of control infants (2.1%) reported using an SSRI. The SSRI most commonly used by control mothers was sertraline (0.8%), followed by fluoxetine (0.7%), paroxetine (0.5%), and citalopram (0.2%).
Table 1 gives the characteristics of the case and control mothers and their reported use of SSRIs. Control mothers were significantly more likely than case mothers to be younger than 35 years of age, to have more than 12 years of education, and to have a higher family income. Case mothers were more likely than control mothers to smoke, have diabetes, have hypertension, and be obese.
|
|
Restricting the analyses to infants with isolated defects resulted in wider confidence intervals and a loss of statistical significance, although there were only slight reductions in the odds ratios (see Supplementary Table 2). Restricting the exposure period to the first 2 months of pregnancy did not change the estimates appreciably (data not shown). Since the critical period for the development of craniosynostosis may extend beyond the first trimester, we examined SSRI use during the second and third trimesters and found a lower odds ratio (17 exposed; adjusted odds ratio, 1.9; 95% CI, 1.0 to 3.5). Sixteen of these women were also exposed to an SSRI during the first trimester.
In additional post hoc analyses, we assessed the associations between the use of four specific SSRIs and the risks of all 18 evaluated birth defects combined and of specific groups of birth defects (Table 3). None of the individual SSRIs were associated with significantly increased risks of all 18 birth defects combined, the 4 cardiac birth defects combined, or the 14 noncardiac birth defects combined. The use of paroxetine or citalopram significantly increased the risk of the pooled group of anencephaly, craniosynostosis, and omphalocele (Table 3).
|
Discussion
Using data from a population-based case–control study, we found no significant associations between the use of SSRIs in early pregnancy and the risks of the majority of birth defects assessed in this study, including congenital heart defects. However, we observed associations between SSRI use and the occurrence of anencephaly, craniosynostosis, and omphalocele, defects that had not been previously associated with maternal SSRI use in pregnancy.
Studies in animals have demonstrated delayed ossification in offspring after maternal treatment with sertraline.13,14 A specific role of serotonin in cardiac and craniofacial morphogenesis in the rodent embryo has also been established.15,16,17 Several recent reports have suggested possible associations between maternal use of paroxetine during pregnancy and birth defects in humans. A report from the Swedish Medical Birth Register showed an increased risk of heart defects overall, and of ventricular and atrial septal defects specifically, among infants whose mothers took paroxetine (but not other SSRIs).18 That report did not find the associations we observed between SSRI use and craniosynostosis or abdominal wall defects (omphalocele and gastroschisis were not examined separately).
An unpublished analysis of administrative records from a managed care organization showed an increased risk of major birth defects overall, and of heart defects specifically, among infants whose mothers received prescriptions for paroxetine as compared with those whose mothers received prescriptions for other antidepressants during pregnancy.5 Similar results were shown in another unpublished cohort study among women treated with paroxetine during the first trimester as compared with women who did not take SSRIs.19 A cohort study based on administrative data from Quebec, Canada, showed that the risks of major birth defects and of heart defects were greater among infants whose mothers took high doses of paroxetine during early pregnancy than among those whose mothers took non-SSRI antidepressants, but this association was not seen among infants whose mothers took lower doses of paroxetine.7
In a Finnish cohort study, the overall risk of major malformations was not significantly increased among infants born to women treated with fluoxetine during the first trimester, but an increase in the risk of heart defects by a factor of three was noted.20 A Danish cohort study showed an increased risk of congenital malformations in general among women given prescriptions for any SSRI.6 Other epidemiologic studies have not shown a significant association between major birth defects and maternal SSRI use during pregnancy.21,22,23,24,25 However, all previous studies have had limitations, such as insufficient power, issues with ascertainment or classification of birth defects, inability to address potential confounding, or poor information on exposure.
Some, but not all, of our findings are consistent with those of another large case–control study in this issue of the Journal.26 Like our report, that study showed no significant associations between SSRI use overall and congenital heart defects. It did show significant associations between paroxetine use and right ventricular outflow tract obstruction defects (six infants; adjusted odds ratio, 3.3; 95% CI, 1.3 to 8.8) and neural-tube defects (four infants; adjusted odds ratio, 3.3; 95% CI, 1.1 to 10.4; anencephaly and spina bifida were not examined separately). However, that report found no significant association between craniosynostosis and SSRI use, and for omphalocele it found a significant association only with sertraline.
Our study is population-based and includes a large birth sample, allowing the evaluation of the relationship between SSRI use and specific types of birth defects. The study also uses careful case definitions and review and excludes infants with chromosomal abnormalities and single-gene disorders. Although the large size of our study overall allowed for consideration of several potential confounders and effect modifiers, the small number of exposed infants for each individual defect remains a limitation.
Because of the large number of comparisons evaluated in our analysis, it is likely that some of the observed associations reflect chance variation. We performed a total of 265 tests, with 54 positive results at the 0.05 significance level; we would have expected 14 positive results to occur by chance. Not all positive tests are reported, and it is not possible to identify which, if any, of the observed associations are due to chance alone. Analyses of other data sets are warranted to replicate our findings.
The effect modification by prepregnancy obesity that we observed post hoc for the association between SSRI use and the occurrence of some birth defects has not, to our knowledge, been reported previously, and confirmation is needed. However, maternal obesity itself has been associated with increased risks of neural-tube defects,27,28 congenital heart defects,27,29 and other defects.27,30 The effect modification we observed may reflect differences in the pharmacokinetics of these lipophilic drugs in women with various percentages of body fat.31
An important limitation of this study is our inability to separate the effect of maternal SSRI use from that of the underlying depression. Only 39 case or control mothers reported having had depression when asked a general question about illnesses during pregnancy. All but five of these women reported taking an antidepressant medication before or during pregnancy. Data on dosage were unavailable, so that analysis of dose–response relationships could not be performed. When they were questioned about the use of medications, the mothers were prompted by brand names for the three most commonly used SSRIs, leading to potential underreporting of other SSRIs. Exposures were determined by maternal report, which introduces a potential for recall bias. Selection bias may have occurred, because participation was less than 100%; however, it is unlikely that participation varied according to SSRI exposure status, and therefore any selection bias was probably in favor of the null hypothesis.
Our study did not show an increased risk of most birth defects, and SSRI exposure was present in only a small number of cases of certain defects. The absolute risks associated with SSRIs appear small in comparison with the baseline risks of birth defects that exist in every pregnancy. Maternal stress and depression during pregnancy have been associated with adverse reproductive outcomes,32,33 and discontinuation of antidepressant treatment in pregnant women with serious depressive illness may have adverse effects on the mother and her baby.34 Thorough assessment of the potential risks and benefits of SSRI use is necessary to allow women of reproductive age to make informed decisions about such therapy.
Dr. Friedman reports receiving honoraria for consulting from I3 Research. No other potential conflict of interest relevant to this article was reported.
The findings and conclusions in this article are those of the authors and do not necessarily represent the views of the Centers for Disease Control and Prevention.
We thank all the families participating in the National Birth Defects Prevention Study (NBDPS) and all NBDPS investigators. We also thank Drs. Margaret A. Honein and Owen Devine for their valuable input. Coding of drug information in the NBDPS used the Slone Drug Dictionary, under license from the Slone Epidemiology Center at Boston University, Boston.
Source Information
From the Department of Medical Genetics, University of British Columbia, Vancouver, BC, Canada (S.A., J.M.F.); and the National Center on Birth Defects and Developmental Disabilities, Centers for Disease Control and Prevention, Atlanta (J.R., S.A.R., R.S.O.).
Address reprint requests to Dr. Reefhuis at the National Center on Birth Defects and Developmental Disabilities, Centers for Disease Control and Prevention, 1600 Clifton Rd., MS E-86, Atlanta, GA 30333, or at JReefhuis{at}cdc.gov.
References
| |||||||||||||||||||||||||||||||||||||||||||||||||||
This article has been cited by other articles:
HOME | SUBSCRIBE | SEARCH | CURRENT ISSUE | PAST ISSUES | COLLECTIONS | PRIVACY | HELP | beta.nejm.org Comments and questions? Please contact us. The New England Journal of Medicine is owned, published, and copyrighted © 2008 Massachusetts Medical Society. All rights reserved. |
Previous
