The first-trimester interview took place before 16 completed weeks of pregnancy; the second-trimester interview took place at 28 (±2) weeks, and the third-trimester interview took place at 8 (±4) weeks after delivery. The table below the graph provides the corresponding number of interviews that were missing or provided insufficient data to make a diagnosis for that interval.
eAppendix. Participating Institutions
eTable 1. Demographic Characteristics, Pregnancy History, Substance Use by Panic Disorder and Generalized Anxiety Disorder
eTable 2. Edinburgh Postnatal Depression Scale (EPDS) Scores by Maternal and Birth Outcomes and Benzodiazepine and/or Serotonin Use
eTable 3. Edinburgh Postnatal Depression Scale (EPDS) Scores by Neonatal Outcomes and Benzodiazepine and/or Serotonin Use
eTable 4. Edinburgh Postnatal Depression Scale (EPDS) Scores by Maternal Outcomes and Panic and General Anxiety Disorders
eTable 5. Edinburgh Postnatal Depression Scale (EPDS) Scores by Birth Outcomes and Panic and General Anxiety Disorders
eFigure. Participant Screening and Study Flow
Customize your JAMA Network experience by selecting one or more topics from the list below.
Yonkers KA, Gilstad-Hayden K, Forray A, Lipkind HS. Association of Panic Disorder, Generalized Anxiety Disorder, and Benzodiazepine Treatment During Pregnancy With Risk of Adverse Birth Outcomes. JAMA Psychiatry. 2017;74(11):1145–1152. doi:10.1001/jamapsychiatry.2017.2733
What is the possible association of anxiety disorders with maternal and fetal outcomes?
In this cohort study of 2654 pregnant women followed up through pregnancy and the postpartum period, neither panic disorder nor generalized anxiety disorder was associated with maternal or neonatal outcomes of interest. With maternal benzodiazepine treatment, rates of ventilatory support increased by 61 of 1000 neonates and duration of gestation was shortened by 3.6 days; with maternal serotonin reuptake inhibitor use, gestation was shortened by 1.8 days, 152 of 1000 additional newborns required minor respiratory interventions, and 53 of 1000 additional women experienced hypertensive diseases of pregnancy.
Neither panic disorder nor generalized anxiety disorder increased adverse birth outcomes in this analysis; although antidepressant and benzodiazepine treatment increased some adverse outcomes, the rate for most events was modest.
Registry data show that maternal panic disorder, or anxiety disorders in general, increase the risk for adverse pregnancy outcomes. However, diagnoses from registries may be imprecise and may not consider potential confounding factors, such as treatment with medication and maternal substance use.
To determine whether panic disorder or generalized anxiety disorder (GAD) in pregnancy, or medications used to treat these conditions, are associated with adverse maternal or neonatal pregnancy outcomes.
Design, Setting, and Participants
This cohort study conducted between July 1, 2005, and July 14, 2009, recruited women at 137 obstetric practices in Connecticut and Massachusetts before 17 weeks of pregnancy and reassessed them at 28 (±4) weeks of pregnancy and 8 (±4) weeks postpartum. Psychiatric diagnoses were determined by answers to the World Mental Health Composite International Diagnostic Interview. Assessments also gathered information on treatment with medications and confounding factors, such as substance use, previous adverse birth outcomes, and demographic factors.
Panic disorder, GAD, or use of benzodiazepines or serotonin reuptake inhibitors.
Main Outcomes and Measures
Among mothers: preterm birth, cesarean delivery, and hypertensive diseases of pregnancy. Among neonates: low birth weight, use of minor respiratory interventions, and use of ventilatory support.
Of the 2654 women in the final analysis (mean [SD] age, 31.0 [5.7] years), most were non-Hispanic white (1957 [73.7%]), 98 had panic disorder, 252 had GAD, 67 were treated with a benzodiazepine, and 293 were treated with a serotonin reuptake inhibitor during pregnancy. In adjusted models, neither panic disorder nor GAD was associated with maternal or neonatal complications of interest. Most medication exposures occurred early in pregnancy. Maternal benzodiazepine use was associated with cesarean delivery (odds ratio [OR], 2.45; 95% CI, 1.36-4.40), low birth weight (OR, 3.41; 95% CI, 1.61-7.26), and use of ventilatory support for the newborn (OR, 2.85; 95% CI, 1.2-6.9). Maternal serotonin reuptake inhibitor use was associated with hypertensive diseases of pregnancy (OR, 2.82; 95% CI, 1.58-5.04), preterm birth (OR, 1.56; 95% CI, 1.02-2.38), and use of minor respiratory interventions (OR, 1.81; 95% CI, 1.39-2.37). With maternal benzodiazepine treatment, rates of ventilatory support increased by 61 of 1000 neonates and duration of gestation was shortened by 3.6 days; with maternal serotonin reuptake inhibitor use, gestation was shortened by 1.8 days, 152 of 1000 additional newborns required minor respiratory interventions, and 53 of 1000 additional women experienced hypertensive diseases of pregnancy.
Conclusions and Relevance
Panic disorder and GAD do not contribute to adverse pregnancy complications. Women may require treatment with medications during pregnancy, which can shorten the duration of gestation slightly. Maternal treatment with a serotonin reuptake inhibitor is also associated with hypertensive disease of pregnancy and cesarean delivery.
During pregnancy, up to 5% of women have panic disorder and as many as 10% have generalized anxiety disorder (GAD).1 Although the impact of a maternal major depressive episode (MDE) has been explored extensively with regard to adverse birth outcomes,2 attention devoted to anxiety disorders has been sparse. Several population-based registry studies show associations between panic disorder and adverse birth outcomes, such as delivery of a small for gestational age3 or preterm infant.3,4 A small study found that pregnant women with panic disorder had smaller babies and shorter gestations than women without either any psychiatric illness or GAD specifically.5 The presence of any anxiety disorder, as determined by medical record review, has also been linked to risk of preterm birth and delivery by cesarean section.6
The investigations reported above have not considered possible confounding effects from maternal medication use and other factors. Serotonin reuptake inhibitors (SRIs), which are used to treat panic disorder and GAD in pregnancy, are associated with some adverse birth outcomes. A previous study showed that neonates exposed to an SRI in pregnancy are more likely to be born preterm than are unexposed neonates,7 a finding in line with most research,8,9 including work that used rigorous methods to control for confounding.10 Use of an SRI in pregnancy has also been associated with delivery of a low-birth-weight baby in some11,12 but not all13,14 studies. Mild and time-limited respiratory problems have also been found in neonates exposed to SRIs in utero.15 Maternal complications linked with SRI use include gestational hypertension,16-18 preeclampsia,12,18,19 and cesarean delivery.20,21 Benzodiazepines, also used to treat anxiety disorders, can confound associations between birth outcomes and SRI use22 or anxiety disorders. They have been implicated in the risk of preterm birth,22-25 delivery of a low-birth-weight infant,25 and neonatal respiratory distress.22,25
In this report, we extend prior work7,26 and specifically probe the association between exposures of panic disorder, GAD that endured 1 month rather than 6 months, SRI and benzodiazepine treatment, and a broader array of outcomes that include several maternal and neonatal pregnancy complications, including (1) hypertensive diseases of pregnancy (gestational hypertension or preeclampsia), (2) cesarean delivery, (3) preterm birth, (4) delivery of a low-birth-weight infant, (5) minor neonatal respiratory interventions, and (6) neonatal ventilatory support. We included interactions between an MDE and anxiety disorders given a prior finding that an MDE moderates the effect of posttraumatic stress disorder (PTSD) on the risk of preterm birth.26 We hypothesized that panic disorder but not GAD would increase the risk of preterm or low-birth-weight delivery. We also hypothesized that benzodiazepine use would increase the risk of preterm birth, cesarean delivery, and the use of neonatal ventilatory support, and that SRI use would be associated with cesarean delivery and hypertensive diseases of pregnancy. A previous study showed an association between SRI use and preterm birth.7
Details of the study have been reported.7 Briefly, this was a cohort study conducted between July 1, 2005, and July 14, 2009, at the Yale University School of Medicine, New Haven, Connecticut, and 137 collaborating hospitals and private practices throughout Connecticut and southern Massachusetts. All delivery hospitals provided human participants approval; a complete list is given in the eAppendix in the Supplement. In addition, participants provided verbal consent for screening and written consent for interviews.
Women were eligible if they were at least 18 years of age and had not yet reached 17 weeks of pregnancy. They were ineligible if they had a known multifetal pregnancy, were receiving insulin treatment for diabetes, did not speak English or Spanish, did not have access to a telephone, or had plans to relocate or voluntarily terminate their pregnancy.
The aim of the parent study was to assess birth outcomes among women with an MDE or a trauma history or who were undergoing antidepressant treatment. Women were given a letter at their usual obstetrical visit and invited to participate in a study of pregnancy and health. Possible participants provided a signature and telephone number, and study staff called possible participants and screened them for an MDE or trauma or use of antidepressant agents. All women who screened positive for these exposures during a telephone interview were offered participation. We also included a randomly selected nonexposed comparison group.
After screening, participants were interviewed at home prior to 17 weeks of pregnancy and reinterviewed by telephone at 28 (±2) weeks’ gestation and again 8 (±4) weeks after delivery. Women with a psychiatric condition who were not receiving care were offered treatment referrals or, if there was a question of safety, we emergently conducted a psychiatric assessment.
Interviewers received a minimum of 4 days of training in the administration of the interview and completed at least 6 practice interviews and 4 supervised interviews. Interviews were taped and a subset was reviewed for quality control.
At each interview, we administered the World Mental Health Composite International Diagnostic Interview, version 2.1, to identify likely diagnoses of an MDE, GAD, and panic disorder. This valid, reliable, and fully structured lay interview27 has been administered to more than 150 000 people from 28 countries, including pregnant women.28,29 It is reliable when administered via the telephone.30 We required a minimum of 1 month, rather than 6 months, of symptoms to meet the criteria for GAD in pregnancy since the interval of follow-up was short. Previous work showed that this duration is associated with substantial functional impairment.31 We excluded a diagnosis of panic disorder if women had a medical condition that could account for panic attacks (eg, asthma attacks). We assessed PTSD with a module from the Mini-International Neuropsychiatric Interview.32 All women were administered the Edinburgh Postnatal Depression Scale (EPDS),33 which includes an anxiety subscale.34
Pregnancy dating used an algorithm that prioritized ultrasonography results prior to 18 weeks and, if unavailable, used the last menstrual period. A birth was designated as preterm if it occurred before 36 completed weeks of gestation (36 weeks and 6 days). Birth weight and delivery mode were abstracted from the medical record. Low-birth-weight deliveries were designated as those less than 2500 g. Hypertensive diseases of pregnancy were operationalized as incident hypertension after 20 weeks of pregnancy (and after exposure to illness or medication for those exposed) or preeclampsia as designated in the medical record. Minor neonatal respiratory interventions included delee suction or tracheal suction and oxygen administration; ventilatory support included use of a ventilator or continuous positive airway pressure.
At the home interview, we obtained information about height, participant-reported race/ethnicity, prepregnancy weight, and preceding pregnancy outcomes as well current medical complications, such as asthma, hyperemesis, hyperthyroidism, and preexisting hypertension and use of any substances or prescribed medications. This information was obtained at each follow-up interview. After delivery, data on maternal illnesses in pregnancy were also abstracted from medical records.
Possible confounding variables were operationalized as follows: heavy alcohol use was defined as 4 or more drinks per week or 3 or more drinks in 1 sitting any time in pregnancy; illicit substance use as any use in pregnancy; previous pregnancy outcomes were coded as a 2-level categorical variable (previous preterm birth vs term births only or no previous live births); age was categorized (<25, 25-34, and ≥35 years); race/ethnicity included Hispanic, black, white, and other; educational level was categorized (<12, 12, 13-15, and ≥16 years); and body mass index was calculated as weight in kilograms divided by height in meters squared.
We adopted several conventions for missing data. If a woman met criteria for a psychiatric diagnosis at any point in pregnancy, she was classified as positive for that disorder. If a woman had partial data (eg, missed 1 or 2 interviews) and was negative for a disorder, we considered her negative for that disorder during pregnancy. We also conducted a sensitivity analysis using complete cases. Rates of missing data by visit are in the Figure.
To estimate associations between exposures and outcomes, we used logistic regression and built separate unadjusted and adjusted models for each outcome. All adjusted models included main exposures (panic disorder, GAD, benzodiazepine use, and SRI use) and possible confounders (age, race/ethnicity, and educational level, as well as smoking, heavy drinking, illicit drug use, MDE, and PTSD during pregnancy, plus the interaction of MDE with panic disorder, GAD, or PTSD). Body mass index was included as an explanatory variable in models for all outcomes except neonatal ventilatory support and minor neonatal respiratory interventions. When modeling preterm birth and low birth weight, we adjusted for previous preterm birth. For the cesarean delivery outcome, previous cesarean delivery was included as an explanatory variable. For all models, the interaction between illness (panic disorder, GAD, and MDE) and SRI or benzodiazepine use, as well as between SRI and benzodiazepine use, was tested to determine moderating effects of illness or medication. Adjusted risk differences were calculated using the modified ordinary least squares method by Cheung.35 We used SAS, version 9.4 (SAS Institute Inc) for analyses. P < .05 (2-sided) was considered significant.
We screened 9525 women and invited 3517 to participate (women with an MDE in the past 5 years, with PTSD, or undergoing SRI treatment, as well as randomly selected nonexposed participants). In this secondary analysis, exposure groups were recoded to include panic disorder, GAD, SRI use, or benzodiazepine treatment. Of the women invited to participate, 2793 underwent at least a home interview. This analysis includes 2654 singleton live births.7 Data on birth outcomes were available for 2634 of the 2654 singleton births (99.2%). A figure for participant flow is provided in a previous publication26 and the the eFigure in the Supplement.
The Figure shows the proportion of women in the cohort who met criteria for panic disorder (98 [3.7%]) or GAD (252 [9.5%]) during pregnancy. Rates were highest in the first trimester (61 [2.3%] for panic disorder and 173 [6.5%] for GAD). At any point in pregnancy, 67 women underwent benzodiazepine treatment, including 56 in the first trimester, 15 in the second trimester, and 11 in the third trimester. A total of 293 women used SRIs at some point during pregnancy. More women used SRIs during the first (n = 257) compared with the second (n = 157) and third (n = 147) trimesters.
Participant characteristics by exposure group are in Table 1. Most participants were 25 to 34 years of age (1509 [56.9%]), white (1957 [73.7%]), and non-Hispanic (2271 [85.6%]). Women undergoing treatment with medications were likely to have multiple psychiatric conditions and to use illegal substances (eTable 1 in the Supplement). Illness severity according to the EPDS33 and maternal and neonatal outcomes are in eTables 2-5 in the Supplement.
Unadjusted and adjusted models are presented in Table 2 and Table 3. No interactions between benzodiazepine or SRI use were significant. Adjusted models showed no significant associations between panic disorder or GAD and maternal or neonatal complications of interest. The co-occurrence of either panic disorder or GAD with an MDE did not increase the risk of these outcomes.
Exposure to a benzodiazepine resulted in the following associations in adjusted models independent of panic disorder, GAD, and SRI use: risk of cesarean delivery (odds ratio [OR], 2.45; 95% CI, 1.36-4.40) (Table 2), delivery of a low-birth-weight infant (OR, 3.41; 95% CI, 1.61-7.26), and neonatal ventilatory support (OR, 2.85; 95% CI, 1.17-6.94) (Table 3). Estimated probabilities translate to 200 additional cesarean deliveries per 1000 births (95% CI, 151-249), 117 additional low-birth-weight infants per 1000 births (95% CI, 96-138), and 61 additional newborns needing ventilatory support per 1000 births (95% CI, 46-76).
We reran these analyses using complete cases and found similar significant associations. However, the association between panic disorder and neonatal ventilatory support became significant, and the association between benzodiazepine use and this outcome lost significance.
To better understand the correlation between benzodiazepine use and fetal growth, we conducted a post hoc logistic regression analysis adjusted for all covariates and used delivery of small for gestational age infant as the outcome (this outcome considers size for birth age rather than low birth weight at any age). The association was not significant (OR, 0.85; 95% CI, 0.35-2.10). Similarly, because benzodiazepine use in pregnancy was marginally associated with preterm birth (OR, 1.98; 95% CI, 0.97-4.04), we used a multiple linear regression model to test the difference in mean duration of gestation in days between those using and not using benzodiazepines adjusted for the same potential confounders included in the preterm birth model. Exposure to benzodiazepines shortened gestational duration by 3.6 days (β = –3.6, SE = 1.7; P = .03).
Use of an SRI was associated with preterm birth (OR, 1.56; 95% CI, 1.02-2.38) (Table 2) as previously reported,7 as well as the need for minor neonatal respiratory interventions (OR, 1.81; 95% CI, 1.39-2.37) (Table 3). In addition, SRI exposure was associated with hypertensive diseases of pregnancy (OR, 2.82; 95% CI, 1.58-5.04) (Table 2). Gestation was 1.8 days shorter in women who received SRIs during pregnancy compared with those who did not. Estimated probabilities were 43 additional preterm births per 1000 births (95% CI, 35-51), 152 additional newborns needing minor respiratory interventions per 1000 births (95% CI, 47-158), and 53 additional cases of maternal hypertension per 1000 births (95% CI, 46-60). Sensitivity analyses that added the anxiety subscale of the EDPS were consistent with the models presented above.
In this carefully characterized cohort of women who were longitudinally followed up through pregnancy, neither maternal panic disorder nor GAD was associated with the maternal or neonatal pregnancy complications of interest. Maternal use of a benzodiazepine was associated with delivery of a low-birth-weight infant, cesarean delivery, and neonatal ventilatory support. Use of an SRI was associated with preterm birth, minor neonatal respiratory interventions, and hypertensive diseases of pregnancy.
To our knowledge, few reports focus specifically on anxiety disorders and birth outcomes.3-5 Our results differ from those that showed an association between panic disorder and preterm birth3,4; reliance on registry data for referenced studies, which can be inaccurate compared with systematic diagnostic interviews, may be a factor. Indeed, panic attacks can occur in patients with several psychiatric conditions, including PTSD. A previous study showed a robust association between preterm birth and comorbid PTSD and an MDE,26 which is reflected in the present analyses. However, with only 98 participants who met the criteria for panic disorder, we may have had limited power to find an association. The limits of power are exacerbated by the need to control for other possible confounders. The postadjustment changes in ORs, particularly for maternal outcomes among those with panic disorder, were substantial. This finding reflects the difficulty of determining the offending exposure when comorbidity among the various psychiatric disorders is frequent. We diagnosed 252 women as having GAD, but we shortened the required symptomatic period from 6 months to 1 month, and women may have been mildly afflicted in this length of time. Prior work has found that case status is not substantially different for GAD that endures 1 vs 6 months.36
The literature on benzodiazepine use in pregnancy is dominated by reports on the risk of fetal malformations. However, some work25,37 has found that neonates exposed to benzodiazepines in utero are more likely to have respiratory difficulties, particularly if exposure is late in gestation. This finding aligns with our finding of increased need for ventilatory support among neonates exposed to benzodiazepines, although it was uncommon.
Several reports show an association between preterm birth and benzodiazepine treatment in pregnancy22,23,38 although there are inconsistencies.39 Calderon-Margalit et al22 reported that the risk of preterm birth is greater if women were treated later rather than earlier in pregnancy. Women in our cohort had predominantly early exposures to benzodiazepines, and the duration of gestation was shortened by only 3.6 days.
Confounding by indication may explain the linkage in other studies between late benzodiazepine use and the risk of preterm birth. Women who had pregnancy complications may have been prescribed benzodiazepines for anxiety related to obstetrical risks. Alternatively, benzodiazepine treatment proximal to delivery may be more problematic than early treatment in terms of the risk of preterm delivery. γ-Amino butyric acid agonists can trigger apoptosis,40 which could theoretically weaken fetal membranes, but the finding of apoptosis was in the central nervous system and it is not clear whether this finding extends to fetal membranes.
Other studies have found that maternal benzodiazepine use is linked with delivery of a low-birth-weight infant,22,25 but these investigations showed that the infants were appropriately sized for gestational age.
The associations we found for SRI use in pregnancy (risk of preterm birth, minor neonatal respiratory interventions, and hypertension diseases of pregnancy) have also been reported by others.8,17,18,41-43 We found that the absolute difference in the duration of gestation among women who were or were not treated with an SRI was only 1.8 days.
Several maternal and neonatal outcomes can be influenced by illness severity and unmeasured confounding factors. An earlier report investigated the role of illness severity (early onset of depression, hospitalization, and suicidal ideation) on the risk of preterm birth among women treated with SRIs in pregnancy.7 The association was attenuated only slightly. As shown in eTable 2 in the Supplement, the mean EPDS score was not higher for women treated with SRIs in pregnancy than for those who were not; a sensitivity analysis that included the EPDS anxiety subscale did not change the significance of our findings, suggesting that illness severity did not confound associations with this exposure.
Our study is one of a few that evaluated the possible association of common anxiety disorders with maternal and neonatal risk of hypertensive diseases of pregnancy, cesarean delivery, preterm birth, low birth weight, neonatal ventiliatory support, and minor neonatal respiratory intervention. Our data collection was longitudinal and prospective and relied on structured interviews to assign likely psychiatric diagnoses. It is the only study to our knowledge that evaluated the effects of benzodiazepines and SRIs independently from the anxiety disorder. An additional strength was inclusion of information about other potential confounders, such as licit and illicit substance use. Limitations include administration of structured interviews by trained lay interviewers and reliance on self-report for substance use; medical record data provide some accuracy compared with self-report, but important information may not be included in the record. In addition, there were too few women taking medication to determine dosages that might be problematic. It is reassuring that there were no adverse maternal and neonatal outcomes directly associated with illness. The use of medication shortened gestation by only a small amount, and we saw no association between medication use and fetal growth. Respiratory effects in neonates appeared to be more severe in neonates exposed to benzodiazepines than those exposed to SRIs.
In general, women are reluctant to take medication in pregnancy, and medication exposures are typically the result of necessity. Often, mothers have the clearest sense of their risk of relapse if medication is discontinued, and this relapse bears on her well-being and that of her offspring. In addition, treatment may be necessary for the mother to carry the pregnancy to term.44 Clinicians and patients need to consider both maternal illness and treatment needs in addition to risks associated with offspring when making treatment decisions in pregnancy.
Corresponding Author: Kimberly Ann Yonkers, MD, Department of Psychiatry, Yale School of Medicine, 40 Temple St, Ste 6B, New Haven, CT 06510 (email@example.com).
Accepted for Publication: July 15, 2017.
Published Online: September 13, 2017. doi:10.1001/jamapsychiatry.2017.2733
Author Contributions: Dr Yonkers and Ms Gilstad-Hayden had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: Yonkers.
Acquisition, analysis, or interpretation of data: All authors.
Drafting of the manuscript: Yonkers, Gilstad-Hayden, Lipkind.
Critical revision of the manuscript for important intellectual content: Yonkers, Forray, Lipkind.
Statistical analysis: Gilstad-Hayden.
Obtained funding: Yonkers.
Administrative, technical, or material support: Yonkers.
Study supervision: Yonkers, Lipkind.
Conflict of Interest Disclosures: Dr Yonkers reported receiving consulting fees from Juniper and Marinus Pharmaceuticals and receiving royalties for a chapter written for Up To Date. No other disclosures were reported.
Funding/Support: This study was supported by grant 5 R01HD045735 from the Eunice Kennedy Shriver National Institute of Child Health and Human Development entitled, “Effects of Perinatal Depression on PTD and LBW” (Dr Yonkers).
Role of the Funder/Sponsor: The funding source had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and the decision to submit the manuscript for publication.