Selective serotonin reuptake inhibitor (SSRI) use during pregnancy and risk of preterm birth: a systematic review and meta-analysis

Abstract

Background

Depression is a prevalent condition in pregnancy affecting about 10% of women. Maternal depression has been associated with an increase in preterm births (PTB), low birthweight and fetal growth restriction, and postnatal complications. Available treatments for depressive disorders are psychotherapeutic interventions and antidepressant medications including selective serotonin inhibitors (SSRIs). SSRI use during pregnancy has been associated with several fetal and neonatal complications; so far, however, the risk of PTB in women using SSRIs during pregnancy is still a subject of debate.

Objective

To evaluate the risk of preterm birth (PTB) in cases of exposure to SSRIs during pregnancy.

Search strategy

Electronic databases (MEDLINE, Scopus, ClinicalTrials.gov, the PROSPERO International Prospective Register of Systematic Reviews, EMBASE and the Cochrane Central Register of Controlled Trials) were searched from their inception until May 2015 with the use of a combination of the following text words ‘depression’, ‘pregnancy’, ‘exposure’, ‘antidepressant’, ‘SSRI’, ‘selective serotonin reuptake inhibitor’, ‘preterm birth’, ‘small for gestational age’ and ‘prematurity’.

Selection criteria

We included studies evaluating the effect of SSRIs exposure in utero and pregnancy outcomes. All cohort and case-control studies were eligible to be included if they reported the incidence of PTB after any exposure to SSRIs and had a comparison group of unexposed pregnant women. Studies without a control group were excluded.

Data collection and analysis

The primary outcome was the incidence of PTB <37 weeks. Subgroup analysis of studies in which controls were defined as women with depression but without SSRI exposure during pregnancy were planned.

Main results

Eight studies (1 237 669 women) were included: 93 982 in the exposure group and 1 143 687 in the control group. After adjusting for confounders, the incidence of PTB was significantly higher in the group of women treated with SSRIs compared with controls (i.e. both women with depression but without SSRI exposure and women without depression) (adjusted OR (aOR) 1.24, 95% CI 1.09–1.41). In the subgroup analysis of studies in which controls were defined as women with depression but without SSRI exposure during pregnancy, an increased risk of PTB (6.8 versus 5.8%; OR 1.17, 95% CI 1.10–1.25) in the SSRI group was found compared with controls (i.e. depressed women treated with psychotherapy alone).

Conclusions

Women who received SSRIs during pregnancy had a significantly higher risk of developing PTB compared with controls. This higher risk remained significant even when comparing depressed women on SSRI with women not on SSRI.

Tweetable abstract

Selective serotonin reuptake inhibitors may be associated with preterm birth.

Introduction

Depression is a prevalent condition in pregnancy affecting about 10% of women.¹ Untreated antenatal depression is associated with poor self-care during pregnancy, risk of postpartum depression as well as risk of impaired maternal–infant bonding when it persists into the postpartum period.² Maternal depression has also been associated with an increase in preterm births (PTB), low birthweight, fetal growth restriction, and postnatal complications.² Several explanations for the concept that maternal depression may contribute to PTB have been postulated: higher circulating levels of inflammatory markers such as C-reactive protein and the pro-inflammatory cytokines interleukin (IL)-1b, IL-6 and tumour necrosis factor (TNF)-α; higher level of placental CRH due to the increase of maternal pituitary-adrenal stress hormones (e.g. ACTH, cortisol); higher risk of bacterial vaginosis in women with maternal psychosocial stress.³

Available treatments for depressive disorders are psychotherapeutic interventions and antidepressant medications including selective serotonin inhibitors (SSRIs). SSRI use during pregnancy has been associated with several fetal and neonatal complications such as pulmonary hypertension,^4,5 cardiac malformations and spontaneous abortion.^6,7 However, so far the risk of PTB in women using SSRIs during pregnancy is still a subject of debate.

The aim of this meta-analysis is to evaluate the risk of PTB in cases of in utero exposure to SSRIs.

Methods

Eligibility criteria

This review was performed according to a protocol designed a priori and recommended for systematic reviews and meta-analyses.⁸ Two authors (A.E., G.S.) identified studies by searching independently the electronic databases MEDLINE, Scopus, ClinicalTrials.gov, the PROSPERO International Prospective Register of Systematic Reviews, EMBASE and the Cochrane Central Register of Controlled Trials with the use of a combination of the following text words: ‘depression’, ‘pregnancy’, ‘exposure’, ‘antidepressant’, ‘SSRI’, ‘selective serotonin reuptake inhibitor’, ‘preterm birth’, ‘small for gestational age’ and ‘prematurity’ from inception of each databases until May 2015. No restrictions as to language or geographic location were applied.

Study selection

We included studies evaluating the effect of SSRIs exposure in utero and pregnancy outcomes. All cohort and case-control studies were eligible to be included if they reported the incidence of PTB after any exposure to SSRIs and had a comparison group of unexposed pregnant women. We included both studies in which controls were defined as ‘all women’, i.e. without depression and without SSRI exposure, as well as studies in which controls were defined as women with depression but no SSRI exposure during pregnancy. Studies were excluded if they lacked the outcome of interest (i.e. incidence of PTB). Studies without a control group were also excluded.

Risk of bias

The risk of bias of the included studies was assessed via the Methodological Index for Non-Randomized Studies (MINORS).⁹ Seven domains related to risk of bias were assessed in each study: (1) aim (i.e. clearly stated aim), (2) rate (i.e. inclusion of consecutive patients and response rate), (3) data (i.e. prospective collection of data), (4) bias (i.e. unbiased assessment of study end points), (5) time (i.e. followup time appropriate), (6) loss (i.e. loss to follow-up), (7) size (i.e. calculation of the study size).⁹ Review authors’ judgments were categorised as ‘low risk’, ‘high risk’ or ‘unclear risk of bias’.

Data abstraction

Two review authors (A.E., G.S.) independently assessed inclusion criteria, risk of bias and data extraction. Discrepancies were resolved by discussion with a third reviewer (V.B.). Data from each eligible study were extracted without modification of original data onto custom-made data collection forms. Data not presented in the original publications were requested from the principal investigators.

Primary and secondary outcomes were defined before data extraction. The primary outcome was the incidence of PTB (i.e. PTB <37 weeks). Secondary outcome included birthweight and the incidence of respiratory distress syndrome (RDS). We planned to assess the primary outcome in sensitivity analyses according to type of study, type of SSRIs used and gestational age at antidepressant exposure. We also planned to assess the incidence of PTB in an a priori subgroup analysis of studies in which controls were defined as women with depression but without SSRI exposure during pregnancy.

Data analysis

The data analysis was completed independently by two authors (A.E., G.S.) using REVIEW MANAGER 5.3 (Copenhagen: The Nordic Cochrane Centre, Cochrane Collaboration, 2014),⁸ and STATA command metandi (Stata Corp. College Station, TX, USA: 2013). The completed analyses were then compared and any difference was resolved with review of the entire data and independent analysis. Statistical heterogeneity between studies was assessed using the Higgins I² statistic.⁸ In case of statistically significant heterogeneity (I² ≥ 50%) the random effect model of DerSimonian and Laird was used to obtain the pooled risk estimate, otherwise a fixed effect model was planned.⁸ The summary measures were reported as odds ratio (OR) or as mean difference (MD) with 95% confidence interval (CI). A P-value <0.05 was considered statistically significant. Potential publication biases were assessed graphically using the funnel plot, and statistically using Begg’s and Egger’s tests.

For studies which reported both unadjusted and adjusted risk for confounders statistically proven, we performed an aggregate data meta-analysis using generic inverse variance method in order to obtain the adjusted OR for the primary outcome and for the secondary outcomes in the main analysis^8,10.

The meta-analysis was reported following the Preferred Reporting Item for Systematic Reviews and Meta-analyses (PRISMA) statement.¹¹ Before data extraction, the review was registered with the PROSPERO International Prospective Register of Systematic Reviews (registration No.: CRD42015027379) following the PRISMA guidelines for protocols (PRISMA-P).¹²

Results

Study selection and study characteristics

The flow of study identification is shown in Figure S1. Table 1 shows the characteristics of the included studies. Eight studies including 1 237 669 women, were analysed.^13–20 Of the 1 237 669 women included, 93 982 used SSRIs as antidepressant during pregnancy, and 1 143 687 formed the non-exposed group (control group). Three studies came from USA,^17,18,20 two from Canada,^15,16 and the other three from Northern Europe.^13,14,19 Five studies used women with no depression or SSRI exposure as controls.^{13–15,17,18} In the other three studies, controls were defined as women with depression but without SSRI exposure during pregnancy;^16,19,20 these 97 303 controls were prescribed no drugs and were treated with psychotherapy alone.^16,19,20 The quality of the studies included in our meta-analysis was assessed by the MINORS tool for assessing the risk of bias (Figure S2). Most of the included studies had low risk of bias in ‘aim’, ‘rate’, and ‘size’. Three of them were prospective cohort studies,^15,18,19 two retrospective,^14,20 while the other three were large, high-quality population-based cohort studies.^13,16,17 Regarding the type the controls, all studies were judged as low risk of bias regarding ‘contemporary groups’ and ‘baseline equivalence group’.⁹ Two studies included only women who received fluoxetine as SSRI,^18,20 two only paroxetine,^15,19 and the others included women who received other SSRIs. Five studies included women who received SSRIs only in the first trimester,^{13,14,16,19,20} one only in the third trimester.¹⁵ Chambers et al.¹⁸ stratified data for gestational age, reporting data both for the first and the third trimester. Hayes et al.¹⁷ did not report gestational age at antidepressant treatment. The majority had PTB as primary outcome. In all of the included studies, there was a high risk for bias due to confounding by indication.

Table 1.

Descriptive data of included studies

	Study location	Type of study	Number of included women	SSRI used	GA at exposure	Confounders adjusted	Control group	Definition of depression	Primary outcome
Chambers 199618	USA	Prospective cohort	228 vs 2S4	Fluoxetine	First and third trimester	None	No depression, no SSRIs	Prescription of SSRI during the first or third trimester	PTB
Costei 200215	Canada	Prospective cohort	SB vs B4	Paroxetine	Third trimester	None	No depression, no SSRIs	Prescription of SSRI during the third trimester	PTB and RDS
Simon 200220	USA	Retrospective cohort	209 vs 209	Fluoxetine	First trimester	None	Women with depression but no SSRIs exposure	Not reported	PTB
Kallen 200413	Sweden	Population-based cohort	28 634 vs B3B 022	Various type*	First trimester	Maternal age, smoking, parity	No depression, no SSRIs	Prescription of SSRI during the first trimester	GA at delivery
Oberlander 200616	Canada	Population-based cohort	1 B 68B vs 92 192	Various type*	First trimester	Pre-pregnancy counselling, maternal age, smoking	Women with depression but no SSRIs exposure	One 3- or 4-digit ICD-9 code	PTB
Lund 200919	Denmark	Prospective cohort	329 vs 4902	Paroxetine	First trimester	None	Women with depression but no SSRIs exposure	Self-administered questionnaire	PTB
Kallen 201214	Sweden	Retrospective cohort	1B 04B vs 31B 97B	Various type*	First trimester	None	No depression, no SSRIs	ICD-10 code	PTB
Hayes 201217	USA	Population-based cohort	33 797 vs 19B 079	Various type*	N/R	Maternal age, race, smoking, education, parity	No depression, no SSRIs	ICD-9 code of 296.2, 296.3, 300.4, or 311	GA at delivery, PTB

Data are presented as number of exposed versus number of non-exposed. GA, gestational age; ICD, International Classification of Diseases; N/R, not reported; PTB, preterm birth; RDS, respiratory distress syndrome; SSRI, selective serotonin reuptake Inhibitor.

^*Varlous type of SSRIs, including citalopram, paroxetine, fluoxetine or sertraline.

Risk of publication bias was assessed by visual inspection of funnel plot, and the symmetric plot suggested no publication bias (Figure S3). Publication bias, assessed using Begg’s and Egger’s tests, showed no significant bias (P = 0.17 and P = 0.14, respectively).

Synthesis of results

The meta-analysis showed that the incidence of PTB was significantly higher in the group of women treated with SSRI than in controls (11.6 versus 5.2%; OR 1.45, 95% CI 1.24–1.68, Figure 1), even after adjusting for statistically proven confounders, including maternal age, smoking, parity, prepregnancy counselling, race and education (aOR 1.24, 95% CI 1.09–1.41, Figure 2). Neonates from women who received SSRIs during pregnancy had a significantly higher risk of RDS (3.7 versus 1.4%; OR 1.33, 95% CI 1.14–1.56, Figure 3; aOR 1.22, 95% CI 1.19–1.58) and significantly lower birthweight (MD −117.12 g, 95% CI −125.99 to −108.24, Figure S4).

Figure 1.

Unadjusted estimates Forest plot for the risk of preterm birth in women treated with selective serotonin reuptake inhibitor. CI, confidence interval; M-H, Mantel-Haenszel test; PTB, preterm birth; SSRI, selective serotonin reuptake inhibitor.

Figure 2.

Adjusted estimates forest plot for the risk of preterm birth in women treated with selective serotonin reuptake inhibitor. CI, confidence interval; IV, independent variable; PTB, preterm birth; SE, standard error; SSRI, selective serotonin reuptake inhibitor.

Figure 3.

Forest plot for the risk of respiratory distress syndrome in women treated with selective serotonin reuptake inhibitor. CI, confidence interval; M-H, Mantel-Haenszel test; RDS, respiratory distress syndrome; SSRI, selective serotonin reuptake inhibitor.

We found an increased risk of PTB in sensitivity analysis of prospective cohort studies (OR 1.83, 95% CI 1.30–2.59);^15,18,19 of retrospective cohort studies (OR 1.51, 95% CI 1.31–1.75);^14,20 of population-based cohort studies (OR 1.14, 95% CI 1.11–1.17);^13,16,17 of women who received SSRIs in the first trimester (OR 1.67, 95% CI 1.25–2.23);^{13,14,16,18–20} of women who received SSRIs in the third trimester (OR 1.86, 95% CI 1.13–3.61);^15,18 of studies in which fluoxetine was used (OR 1.91, 95% CI 1.07–3.41);^18,20 and in studies in which paroxetine was used (OR 2.07, 95% CI 1.42–3.02).^15,19 Women who received paroxetine had a similar risk of PTB to those who received fluoxetine (OR 1.42, 95% CI 0.88–2.31). Women who received SSRIs in the third trimester had a significantly higher risk of PTB compared with those who received SSRIs only in the first trimester (OR 4.17, 95% CI 2.75–6.30).

In subgroup analysis of studies in which controls were defined as women with depression but without SSRI exposure during pregnancy,^16,19,20 we found an increased risk of PTB in the exposed group (6.8 versus 5.8%; OR 1.17, 95% CI 1.10–1.25; I² = 0%) compared with controls (i.e. depressed women treated with psychotherapy alone).

Discussion

Main findings

This meta-analysis showed that women who received SSRIs during pregnancy had a significantly higher risk of developing PTB. This remained significant even when comparing depressed women on SSRI versus depressed women not on SSRI. This is important, as depression itself is associated with preterm delivery.³ Neonates from women who received SSRIs during pregnancy had a significantly higher risk of RDS and significantly lower birthweight compared with controls. The risk of PTB seems to be higher if the SSRIs were given in the third trimester compared with an earlier exposure.

Comparison with existing literature

To date, most meta-analyses have found that exposure to SSRI during pregnancy may be associated with several pregnancy complications including neonatal pulmonary hypertension and cardiac defects.^4–7 These meta-analyses did not analyse the correlation between SSRI and PTB. To our knowledge, this is the first systematic review evaluating this possible association. Other meta-analyses have reported an association between antidepressants and PTB.^21,22 Huybrechts et al. pooled data from 14 studies and found an increased risk of preterm delivery in women taking antidepressants, including tricyclic and noradrenergic antidepressant, during the second and third trimester of pregnancy.²²

Strengths and limitations

The most important strength of our work rests on the attention to potential confounding factors. Generic inverse variance method was used to obtain the aOR for studies which adjusted for statistically proven confounders. No prior meta-analysis on this issue is as large, up-to-date or comprehensive. The number of the included women is very high. Most of the included studies had low risk of bias. Subgroup analyses and sensitivity analyses were performed to reduce the clinical heterogeneity within the studies. Primary outcome was assessed in subgroup analysis of studies in which controls were defined as women with depression but without SSRI exposure during pregnancy, which was the most clinical meaningful analysis.

Although meta-analytical techniques pool all available data, limitations include those of the original articles. All the included studies were cohort studies; no randomised controlled trials were included in this systematic review and this is a major shortcoming of this study. None of the included studies stratified data for PTB aetiology, so data regarding this outcome referred to both spontaneous and indicated. This limitation places the biologic plausibility of a drug–preterm delivery association in question, as the two types of PTB have distinct aetiologies and pathogeneses. There were different control groups and this point raises the question of selection of the overall control group. Only three studies adjusted for confounders and these may not have been sufficient to control for a study of risk factors for PTB.^13,16,17 Even within the subgroup analysis, there remained confounding by severity of depression that had the potential to bias results significantly. Confounding factors that influence birth outcomes were variably controlled; these potential confounders, such as poor prenatal care and drug, nicotine and alcohol use, occur at a higher rate in depressed than non-depressed women.^23,24 Because SSRI use occurs in the context of maternal depression, we were not able to study the effects of SSRI exposure independent of exposure to depression alone. This bias cannot be reliably eliminated with a multivariable analysis. To avoid this limitation we performed subgroup analysis of studies in which controls were defined as women with depression but without SSRI exposure during pregnancy. This subgroup analysis concurs with the main analysis. However, only three studies (113 526 women) were included; moreover, restricting unexposed controls to depressed patients without SSRI exposure does not remove the possibility of confounding by indication, as treatment groups were not randomised. For example, it is possible that women in the SSRI treatment group had more severe disease compared with those who were not given medication, simply because their physicians selected them for SSRI treatment. Based on the characteristics of the included studies and the summary statistics for heterogeneity there was a large amount of both statistical and clinical heterogeneity. The studies vary markedly by overall study design, analysis (e.g. some not controlling for confounders), drug exposure timing, drug type(s), and study population, especially with regard to controls. This may impact the validity of the pooled results. The statistical heterogeneity within the studies for the primary outcome was high (I² = 86%). For this reason, random effects models were used in most of the analyses performed. Another major issue is the small OR (1.45 for the primary outcome). The OR seems to be smaller in the largest and best-designed studies (Figure 1).

Interpretation

There are many methods to treat depression during pregnancy.²⁴ Many patients with mild-to-moderate depression can be treated by psychosocial approaches such as individual and group psychotherapy instead of medication. Interpersonal and cognitive behavioural psychotherapy have been shown to be effective for depression in pregnant women,²⁵ and are recommended by The American College of Obstetricians and Gynecologists (ACOG) and The American Psychiatric Association (APA).²⁴ Regarding antidepressant treatment during pregnancy, SSRIs are the most frequently prescribed drugs;²⁶ physicians rarely use monoamine oxidase inhibitors or tricyclic antidepressants.^26,27 However, there are no randomised controlled trials of antidepressant drug efficacy in depressed pregnant women.

The biological plausibility of our findings is not completely clear. With respect to prematurity, some studies have suggested that maternal stress may increase the risk of preterm delivery.²⁸ Women who received SSRIs during pregnancy in whom the rate of PTB was increased, may have had more severe depression or anxiety and therefore been at higher risk for PTB (i.e. bias due to illness severity). In this case, if the antidepressant treatment lessened the effect of maternal depression, then in the absence of treatment, the outcome of these pregnancies may have been even worse. Alternatively, SSRIs could have affected the outcome of PTB separately from the effect of depression. Regarding birthweight, the findings of decreased birthweight are consistent with the results of a study in which pregnant rats treated with fluoxetine delivered smaller pups.²⁹ Sometimes, the use of SSRIs during pregnancy is inevitable. Untreated antenatal depression has been associated with suicidal tendencies, including death. Studies have shown that stopping SSRI treatment in pregnant women with a previous history of depression leads to relapse in the majority of women. Untreated depression during pregnancy has also been shown to be a strong risk factor for subsequent postpartum depression. Hence, in situations where it may be necessary to use SSRIs in pregnancy (prior history of severe antepartum or postpartum depression, current history of suicidal tendencies in the setting of depression in pregnancy, or in depressed patients dependent on high doses of antidepressants in pregnancy), the decision to use SSRIs during pregnancy must be weighed against the risks of untreated depression and this risk/benefit ratio, including the risk of preterm birth. This must be carefully discussed with the patient, and should ideally be done in collaboration with the patient’s psychiatrist.

Conclusion

This meta-analysis showed that women who received SSRIs during pregnancy had a significantly higher risk of developing PTB compared with controls. This higher risk remained significant (17% increase in PTB) even when comparing depressed women on SSRI versus depressed women not on SSRI, which is the most clinical meaningful analysis.

In summary, these data warrant caution in the use of SSRIs during pregnancy. Women should be informed about possible risks of antidepressant medication during pregnancy. However, as other confounding cannot be excluded, these data warrant further research on possible effects of SSRI on PTB.

Supplementary Material

Figure S1_Flow diagram of studies identified in the systematic review

Figure S1. Flow diagram of studies identified in the systematic review.

Figure S3_Funnel plot for assessing publication bias. OR, odds ratio; SE, standard error

Figure S3. Funnel plot for assessing publication bias. OR, odds ratio; SE, standard error.

Figure S2_Assessment of risk of bias

Figure S2. Assessment of risk of bias.

Figure S4_Forest plot for birthweight in women treated with selective serotonin reuptake inhibitor

Figure S4. Forest plot for birthweight in women treated with selective serotonin reuptake inhibitor.

Video S1_Author Insights.

Video S1. Author Insights.