Association Between Incident Exposure to Benzodiazepines in Early Pregnancy and Risk of Spontaneous Abortion

Key Points

Question  Which type of benzodiazepine places women in early pregnancy at increased risk of spontaneous abortion?

Findings  In this nested case-control study of the 442 066 pregnancies included in the Quebec Pregnancy Cohort, an association between benzodiazepine exposure during early pregnancy and risk of spontaneous abortion was observed in all 3 independent models that quantified benzodiazepine use by drug class, duration of action, and specific benzodiazepine agents.

Meaning  The findings suggest that any benzodiazepine use during early pregnancy is associated with spontaneous abortion and that health care clinicians should carefully evaluate the risk-benefit ratio of benzodiazepine use for the treatment of mood and anxiety disorders or insomnia during early pregnancy.

Importance  Benzodiazepine use in early pregnancy is associated with spontaneous abortion (SA). However, to date, the association between specific benzodiazepine agent exposure and the risk of SA has not been examined.

Objective  To quantify the risk of SA associated with gestational benzodiazepine incident use by drug class, duration of action, and specific benzodiazepine agent.

Design, Setting, and Participants  This nested case-control study within the Quebec Pregnancy Cohort, Montreal, Quebec, Canada, includes all pregnancies covered by the Quebec Prescription Drug Insurance Plan from January 1, 1998, through December 31, 2015. Each case was randomly matched with up to 5 controls. Statistical analysis was performed from January 1, 1998, through December 31, 2015.

Exposures  Benzodiazepine exposure was defined as 1 or more filled prescriptions between the first day of the last menstrual period and the index date (the calendar date of the SA diagnosis). Benzodiazepine exposure was categorized by overall use, long- or short-acting benzodiazepine, and specific benzodiazepine agents.

Main Outcomes and Measures  Spontaneous abortion defined as a pregnancy loss between the beginning of the sixth week of gestation and the 19th completed week of gestation. Conditional logistic regression models were used to calculate odds ratios (OR) and 95% CIs.

Results  Of the 442 066 pregnancies included in the Quebec Pregnancy Cohort, 27 149 (6.1%) ended with SA, with a mean (SD) maternal age of 24.2 (6.5) years. Among pregnancies ending with SA, 375 (1.4%) were among women exposed to benzodiazepines in early pregnancy compared with 788 (0.6%) of the 134 305 matched control pregnancies (crude OR, 2.39; 95% CI, 2.10-2.73). Adjusting for potential confounders, including maternal mood and anxiety disorders before pregnancy, and compared with nonuse, benzodiazepine exposure in early pregnancy was associated with an increased risk of SA (adjusted OR, 1.85; 95% CI, 1.61-2.12). The risk was similar among pregnancies exposed to short-acting (284 exposed cases; adjusted OR, 1.81; 95% CI, 1.55-2.12) and long-acting (98 exposed cases; adjusted OR, 1.73; 95% CI, 1.31-2.28) benzodiazepines during early pregnancy. All benzodiazepine agents were independently associated with an increased risk of SA (range of adjusted ORs, 1.13-3.43).

Conclusions and Relevance  An increased risk of SA was observed among early pregnancies with incident exposure to short- and long-acting benzodiazepines and all specific benzodiazepine agents during early pregnancy. Insomnia, anxiety, and mood disorders are prevalent during pregnancy; clinicians should carefully evaluate the risk-benefit ratio of prescribing benzodiazepines in early pregnancy since alternative nonpharmacologic treatments exist.

Benzodiazepines are psychoactive medications that are frequently prescribed during pregnancy for the treatment of anxiety, insomnia, and mood disorders.^1-4 Benzodiazepines cross the placental barrier and accumulate in the fetal circulation at levels that are 1 to 3 times higher than the maternal serum levels.⁵ Although there have been inconsistencies among studies, benzodiazepines were classified as category D drugs based on findings suggesting an increased risk of cleft lip or palate, urogenital, and neurologic congenital malformations.⁶

Benzodiazepines are anxiolytic medications that act as agonists at the GABA receptors, with inhibitory postsynaptic signaling.⁷ Benzodiazepines also bind to peripheral tissues through peripheral benzodiazepine receptors⁸ and may play a role in steroidogenesis and cell proliferation.⁹ Benzodiazepines freely and readily cross human placenta^10,11 and may accumulate substantially in the embryo and fetal tissues,¹¹ causing adverse effects.¹² Animal studies^13-15 show that benzodiazapines can interfere with fetal development, including neurodevelopment and development of the immune system.

A population-based study from the United Kingdom found that women exposed to benzodiazepines during pregnancy had a 60% higher risk of spontaneous abortion (SA) compared with women with unmedicated depression or anxiety during the first trimester.¹⁶ In addition, women who continued benzodiazepine use during the first trimester had a 50% higher risk of SA compared with those who discontinued medication use.¹⁶ Higher rates of SA among benzodiazepine users have also been reported in a prospective study from Israel.¹⁷ In these studies,^16,17 the effect of benzodiazepines was considered as a class. Previous studies^12,18 concerning benzodiazepine exposures in early pregnancy and the risk of major congenital malformation have shown considerable differences in the safety profiles among different benzodiazepines. To our knowledge, the risk of SA associated with specific prenatal benzodiazepine agents or by duration of action has not been examined. Spontaneous abortion is the most common complication in early pregnancy and occurs in 12% to 15% of clinically recognized pregnancies.¹⁹ Data concerning the safety of specific benzodiazepine agents are of major clinical relevance. Thus, we conducted a nested case-control study to quantify the risk of SA associated with prenatal benzodiazepine use by class, duration of action, and specific benzodiazepine agents.

We conducted a nested case-control study within the Quebec Pregnancy Cohort (QPC). The QPC is described in detail in Bérard and Sheehy.²⁰ In brief, the QPC is an ongoing population-based cohort with prospective data collection on all pregnancies of women covered by the Quebec Public Prescription Drug Insurance Plan from January 1, 1998, to December 31, 2015. Individual-level information was obtained from provincewide databases and linked using the unique health care personal identifiers. The Ethics Review Board of Centre Hospitalier Universitaire Sainte-Justine, Montreal, Quebec, Canada, approved the study, and the Commission d’accès à l’information authorized the linkage between databases. Since the data were encrypted, no consent was needed from the study participants. The Commission d’accès à l’information gives authorization to use such data for research.

The QPC was constructed by identifying all pregnancies in the Régie de l’assurance maladie du Québec (RAMQ) and the Quebec hospitalization archives (MedEcho) databases. The first day of the last menstrual period (LMP) was defined using data on gestational age, which was validated against ultrasound measures in patients’ medical records.²¹ Prospective follow-up was available from at least 1 year before LMP, during pregnancy, and until December 31, 2015.

The data sources for this study included the medical service database RAMQ (diagnoses, medical procedures, socioeconomic status of women, and prescribers), the Quebec Public Prescription Drug Insurance Plan database (drug name, start date, dose, and duration), the MedEcho database (in-hospital diagnoses and procedures, including gestational age for planned abortions, SAs, and deliveries), and the Quebec Statistics database for patient sociodemographics and birth weight. Validity studies were performed and have shown the accuracy and the high quality of the data sources.^21,22

We included pregnancies of women aged 15 to 45 years who were continuously insured by the Quebec Public Prescription Drug Insurance Plan for at least 12 months before their LMP and during their pregnancy. We excluded pregnancies among women exposed to known teratogens during the first trimester (defined as 0-14 completed weeks of gestation) (eTable 1 in the Supplement).³ We also excluded pregnancies among women with (1) epilepsy without pharmacologic treatment (International Classification of Diseases, Ninth Revision [ICD-9] diagnosis code 345 or International Statistical Classification of Diseases and Related Health Problems, Tenth Edition [ICD-10] diagnosis codes G40-G41) because the pregnancies (99.4%) among women exposed to antiepileptics, which are considered to be teratogens, have already been excluded; (2) a history of epilepsy because some benzodiazepines are also prescribed for the treatment of this epilepsy and we aimed to only include women with anxiety, insomnia, and mood disorders; (3) previous use of benzodiazepines; and (4) planned or induced abortions. We were interested in the pharmacologic effect of the benzodiazepine in association with pregnancy loss and not the decision to end the pregnancy.

We defined the SA case as any pregnancy loss between the beginning of the sixth week of gestation and the 19th completed week of pregnancy using ICD-9 diagnosis code 634 or ICD-10 diagnosis code O03 in the RAMQ or MedEcho databases. This approach is consistent with the definitions provided by the World Health Organization²³ and the United Nations.²⁴

We excluded pregnancies of women who had SA occur at fewer than 6 weeks’ gestation; these women were potentially misclassified because of unrecognized early pregnancy losses. We did not consider pregnancy losses after 20 weeks’ gestation because they were categorized as stillbirth deliveries.²⁵

We defined the index date as the calendar date of the SA diagnosis. Using incidence density sampling, we randomly selected 5 controls for each case at the index date and matched them with the case pregnancy by gestational age and calendar year.

A pregnancy was considered to have incident benzodiazepine exposure if the mother had filled at least 1 prescription for any type of benzodiazepine from the LMP until the index date. Women filling prescriptions for benzodiazepines before becoming pregnant with a prescription duration overlapping the LMP were also considered to be exposed during pregnancy. Benzodiazepine exposure was also dichotomously categorized based on duration of action, namely short-acting (half life ≤24 hours) or long-acting (half life >24 hours) benzodiazepines.^26-28 Specific benzodiazepine agents included in this study were short acting (alprazolam, bromazepam, lorazepam, oxazepam, temazepam, and triazolam) and long acting (chlordiazepoxide, clonazepam, diazepam, flurazepam hydrochloride, and nitrazepam).^27,28 To evaluate the dose-response effect, we expressed all benzodiazepine doses as diazepam equivalents.^29,30 We calculated the mean daily dose for each pregnancy exposed using the cumulative diazepam-equivalent dose divided by the cumulative duration of the filled prescription. We categorized mean daily dose as follows: 5 mg or less, 6 to 20 mg, and more than 20 mg. Pregnancies of women not exposed to benzodiazepines between the LMP and the index date were used as a reference category. Data on prescription fills during pregnancy have been validated and compared with maternal reports.²²

Potential confounders were selected a priori based on known risk factors or predictors of SA or/and benzodiazepine exposure and included (1) maternal sociodemographic variables measured by the LMP, (2) maternal chronic conditions measured by diagnostic codes or prescribed medications (eTable 2 in the Supplement) in the year before the LMP and during pregnancy, (3) health care resources utilization in the year before the LMP and during pregnancy, and (4) pregnancy-associated variables.

To control for potential confounding by the main indications (mood and anxiety disorders or insomnia), we adjusted for the presence of physician-based diagnoses for these conditions (ICD-9 diagnosis codes 296, 309, 311, 300.0, and 300.4 and ICD-10 diagnosis codes F30-F43 for mood and anxiety disorders and ICD-9 diagnosis codes 307.4, 327.0, 327.3, and 780.5 and ICD-10 diagnosis codes G47 and F51 for insomnia) in the year before pregnancy until the index date. We also adjusted for concomitant exposure to antidepressant and antipsychotic medications between the LMP and the index date. Since information about lifestyles were incomplete in Quebec's administrative databases, we used diagnoses of tobacco, alcohol, and other drug dependencies to adjust for smoking status, alcohol consumption, and illicit drug use (eTable 3 in the Supplement). We then adjusted for exposure to folic acid supplementation in the 6 months before the LMP and in early pregnancy to take into account the benefits of this supplementation on the risk of SA.

We used descriptive statistics to summarize characteristics of the study population and the pattern of benzodiazepine prescription filled during early pregnancy. The association between benzodiazepine use in early pregnancy and the risk of SA was quantified using conditional logistic regression models with 95% CIs. The conditional logistic regression models allow for a more comparable measure of exposure between case and control pregnancies by providing the ability to adjust for gestational age at the time of the SA.³¹ Four independent models were used for (1) overall benzodiazepine exposure between the LMP and index date, (2) benzodiazepine exposure by duration of action (short and long acting), (3) specific benzodiazepine agents, and (4) mean daily diazepam-equivalent dose (eTable7 in the Supplement). We assessed the dose-response trend using the Cochran-Armitage trend test. The multivariate models were adjusted for the potential confounders listed above. In addition, several sensitivity analyses were performed to evaluate the robustness of estimates using the overall incident benzodiazepine exposure in early pregnancy model (eMethods and eTables 8-14 in the Supplement). We also used the E-value to measure the robustness of the association between incident benzodiazepine exposure and SA for unmeasured or unadjusted confounding using the new measure proposed by VanderWeele and Ding.³² All analyses were conducted using SAS software, release 9.1 (SAS Institute Inc). Statistical analysis was performed from January 1, 1998, through December 31, 2015.

The QPC included 442 066 pregnancies between January 1, 1998, and December 31, 2015. The prevalence of SA among the 262 070 eligible pregnancies was 7.0%, and 27 149 pregnancies with SA met the inclusion criteria and were defined as cases. The mean (SD) age of the case pregnancies was 24.2 (6.5) years. The mean (SD) gestational age at index date was 16.7 (3.1) weeks. Cases were matched to 134 305 control pregnancies; 5 controls were found for 26 789 (98.7%) of the cases (Figure 1).

Compared with the matched control pregnancies, women experiencing pregnancies with SA were more likely to be older, welfare recipients, have hypertension or asthma, and have diagnoses of drug, tobacco, and alcohol dependence in the year before or during early pregnancy (up to index date). Women of case pregnancies were more likely than controls to have hospitalizations or emergency department visits, general practitioner or specialist visits, diagnoses of mood and anxiety disorders in the 12 months before the LMP, concomitant antidepressant or antipsychotic exposures, and more pregnancies in the year before the LMP (Table 1). Women of case pregnancies were less likely to be exposed to folic acid before pregnancy compared with controls.

Among the 1163 pregnancies exposed to benzodiazepines in early pregnancy, 1128 (97.0%) women received only 1 specific benzodiazepine agent and 897 (77.1%) had only 1 prescription filled (eTable 3 and eTable 4 in the Supplement). The most frequently prescribed benzodiazepines were lorazepam (768 of 1715; 44.8%) and clonazepam (402 of 1715; 23.4%). The mean duration of benzodiazepine use was 16.5 days (median, 15.0 days), and the mean number of pills provided was 24.3 (median, 20.0 pills) (eTable 5 in the Supplement).

Overall, 375 of the 27 149 cases of SA (1.4%) had at least 1 filled prescription for benzodiazepines during early pregnancy compared with 788 of the 134 305 matched controls (0.6%) (crude odds ratio [OR], 2.39; 95% CI, 2.10-2.73). After adjustment for all potential confounding variables, benzodiazepine exposure in early pregnancy was independently (of mood and anxiety disorders) associated with an increased risk of SA (375 exposed cases; adjusted OR [aOR], 1.85; 95% CI, 1.61-2.12) (Table 2).

The use of short-acting (284 exposed cases; aOR, 1.81; 95% CI, 1.55-2.12) and long-acting (98 exposed cases; aOR, 1.73; 95% CI, 1.31-2.28) benzodiazepines was associated with an increased risk of SA (eTable 6 in the Supplement).

When we categorized pregnancies according to specific benzodiazepine agents, we observed an increased risk of SA associated with each agent (Table 3). The increased risk of SA ranged from an aOR of 1.13 for women exposed to flurazepam hydrochloride to an aOR of 3.43 for women exposed to diazepam. The association between benzodiazepines and SA was strengthened with increasing diazepam-equivalent daily dose (<5 mg: 202 exposed cases; aOR, 1.73; [95% CI, 1.44-2.08]; 6-20 mg: 163 exposed cases; aOR, 1.96 [95% CI, 1.59-2.43]; and >20 mg: 10 exposed cases; aOR, 2.55 [95% CI, 1.08-6.01]; P < .01) (eTable 7 in the Supplement).

The E-value obtained for the association between incident benzodiazepine exposure and SA was 3.1 with a lower limit of 2.6, suggesting that unmeasured confounding was unlikely to explain the findings. Figure 2 summarizes sensitivity analyses estimates with corresponding CIs (eMethods and eTables 8-14 in the Supplement). All CIs included the main association estimate (aOR, 1.85).

To our knowledge, this was the first study to investigate the risk of SA associated with specific benzodiazepine agent exposures during early pregnancy as well as the first study considering the association of duration of action with SA during early pregnancy. Early pregnancy incident exposure to any benzodiazepine agent was associated with an increased risk of SA; the risk of SA increased with increasing daily dose of benzodiazepines, which may suggest a dose-response effect. Overall, 7.0% of pregnancies ended with a clinically recognized SA, which was comparable with the 8% to 20% of pregnancies reported in the literature.³³

Our findings on class effect of benzodiazepines were consistent with a population-based UK study,¹⁶ in which women exposed to benzodiazepines during pregnancy had a 60% higher risk for SA compared with women with unmedicated depression or anxiety during the first trimester of pregnancy. However, in the UK data, a prescription for benzodiazepines during early pregnancy was likely to indicate more severe anxiety or depression than an unmedicated diagnostic record of depression or anxiety because the diagnoses were recorded by general practitioners in The Health Improvement Network, United Kingdom.³⁴ Ban et al¹⁶ have acknowledged that they did not quantify the severity of the anxiety or depression, and the findings may have been attributable to confounding by illness severity.¹⁶ In our study, we adjusted for diagnoses of mood and anxiety disorders and insomnia as well as for several documented proxies of these diseases, such as concomitant exposure to antidepressants or antipsychotics, visits to a psychiatrist, comorbidities, and hospitalizations. Furthermore, crude and adjusted analyses showed that our statistical adjustment for potential confounders corrected, at least partially, for this form of bias. The large E-value obtained for the strength of potential residual confounding reinforced the validity of the observed association. In addition, in the study by Ban et al,¹⁶ exposure status was defined as a prescription for benzodiazepine compared with a prescription fill for a benzodiazepine in our study. Given that some prescriptions do not get filled, the study by Ban et al¹⁶ could be subjected to exposure misclassification. In our study, data on filled prescriptions were used to define exposure, and within the Quebec Drug Plan, pregnant women made a copayment for their medication, which increased the likelihood of taking at least 1 dose. Medication exposure in the QPC was validated by Zhao et al,²² with overall positive and negative predictive values above 80%.

In our study, lorazepam, clonazepam, oxazepam, and alprazolam were the most frequently filled benzodiazepines, which was similar to what has been reported in other studies.^17,35,36 Slight variations in exposure prevalence were likely attributable to differences in health care systems, access to health care, and benzodiazepines available in the specific countries. To our knowledge, this study was the first to examine benzodiazepine exposures according to specific agents and mode of action.

Using the population-based QPC, we were able to obtain accurate information on filled prescriptions rather than rely on maternal recall for a large number of pregnancies with detailed information regarding outcome and potential confounders.²² Validated gestational age²¹ allowed us to calculate the exact timing of the beginning of pregnancy and exact date of benzodiazepine prescription fills. We only included incident users of benzodiazepines during pregnancy because they were more likely to take at least 1 pill during early pregnancy and thus be considered exposed, decreasing exposure misclassification bias. We also used prospectively and routinely collected physician-based diagnoses and records of procedures related to pregnancy outcomes, which limited the potential for detection and misclassification biases on outcome status. The association between benzodiazepines and SA was also adjusted for the most important potential confounders (ie, concomitant exposure to antidepressants and/or antipsychotic medications).

Limitations include missing information about potentially important confounders such as smoking and alcohol intake. Tobacco and alcohol dependence and prescribed folic acid were used as proxies for these variables. Adjusting only for proxies of these variables most likely resulted in the adjustment of severe instances of abuse, which could result in residual confounding. The data on assisted reproduction were not available in the cohort. In any case, women treated with assisted reproduction were routinely monitored, with pregnancy more likely to be detected before the sixth week of gestation. We only included SA between 6 and 20 weeks of gestation in our study, as was the case in similar studies,^23,24 to avoid outcome misclassification. Furthermore, we performed several sensitivity analyses, which showed the robustness of our findings.

Although the main study cohort included a large sample size, subgroup analyses on specific benzodiazepine agents were performed on a smaller sample, which resulted in lower statistical power. All benzodiazepine agents were associated with increased risk of SA, but some associations were not statistically significant. Because we only considered pregnant women insured by the prescription drug insurance program, generalizability of results to those insured by private drug insurance could be affected. However, validation studies have shown that pregnant women receiving medication insurance from Quebec’s public system have characteristics and comorbidities similar to those of women who are covered by private medication insurance.³⁷ Although socioeconomic status might differ between the 2 groups, it does not affect internal validity given that all have similar socioeconomic status.³⁷ Large, well-conducted cohort studies are needed to confirm our results.

An association between any benzodiazepine exposure during early pregnancy and the risk of SA was observed in all 4 independent models that quantified benzodiazepine exposure as a class, by duration of action, for specific benzodiazepine agents, and by cumulative diazepam-equivalent dose. The findings suggest that health care clinicians should carefully evaluate the risk-benefit ratio of benzodiazepines for the treatment of insomnia and mood or anxiety disorders in early pregnancy. Alternative nonpharmacologic treatments exist and are recommended, but if benzodiazepines are needed, they should be prescribed for short durations.^38,39

Accepted for Publication: March 11, 2019.

Corresponding Author: Anick Bérard, PhD, Research Center, Centre Hospitalier Universitaire Sainte-Justine, 3175 Chemin de la Côte-Sainte-Catherine, Montreal, QC H3T 1C5, Canada (anick.berard@umontreal.ca).

Published Online: May 15, 2019. doi:10.1001/jamapsychiatry.2019.0963

Author Contributions: Ms Sheehy had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Concept and design: All authors.

Acquisition, analysis, or interpretation of data: All authors.

Drafting of the manuscript: All authors.

Critical revision of the manuscript for important intellectual content: All authors.

Statistical analysis: Sheehy, Bérard.

Obtained funding: Bérard.

Administrative, technical, or material support: Sheehy, Bérard.

Supervision: Sheehy, Bérard.

Conflict of Interest Disclosures: Dr Bérard reported being a consultant for plaintiffs in litigations involving antidepressants and birth defects. No other disclosures were reported.

Funding/Support: This study was funded by the Canadian Institutes of Health Research, CAnadian Network for Advanced Interdisciplinary Methods (CAN-AIM) grant, Fonds de la recherche du Québec–Santé (FRQS), and the Réseau québécois de recherche sur les medicaments. All grants were given to Dr Bérard and her research team. Dr Bérard is the holder of a FRQS research chair on medications and pregnancy.

Role of the Funder/Sponsor: The funding sources 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 decision to submit the manuscript for publication.