First-trimester benzodiazepine prescriptions not linked to more birth defects in 4.2 million US pregnancies

Benzodiazepines, the family of sedatives that includes lorazepam, alprazolam and diazepam, are prescribed for anxiety and insomnia, and both are common in pregnancy. For decades, studies have disagreed about whether taking them in the first months raises the chance of a birth defect. A study of more than 4.2 million US pregnancies now reports that it found no such increase. Major birth defects were recorded in about 4 to 5 of every 100 babies, whether or not a benzodiazepine had been dispensed in the first trimester.

The study, published on October 12 in Annals of Internal Medicine, answers a narrow question. It counted defects apparent around birth in babies born alive. It did not look at miscarriage or at how children developed later. And its results for benzodiazepines taken together with other psychoactive medicines, meaning those that act on the brain, were mixed.

Why the evidence on benzodiazepines in pregnancy conflicted

The paper’s starting point is one sentence: existing data on malformations associated with benzodiazepine use in pregnancy are conflicting. A congenital malformation is a structural problem in the body, such as a heart defect, that is present from birth.

Use in pregnancy is not rare. A 2020 review found that maternal use of benzodiazepines during pregnancy is common and has increased over the last decades. Its authors pooled the available surveys, drawing on 32 studies reporting on 28 countries, and the worldwide figure came to 1.9% of pregnancies.

The reasons for prescribing are plain enough. US researchers have noted that benzodiazepine medications can be used to treat anxiety, a condition affecting 15% of women of childbearing age in the United States. Leaving illness untreated has costs as well. A team at Massachusetts General Hospital wrote that untreated maternal mental illness can cause harm to women and their children.

The worry has a biological basis. Korean researchers summed it up in 2022: benzodiazepines readily cross the human placenta, and regulators advise against them in pregnancy because of their potential to harm the fetus. They added that the evidence was uncertain, because pregnant women are usually excluded from clinical trials.

That leaves observational studies, and those disagreed from the start. A 1998 analysis in the BMJ, a British medical journal, sifted the literature to that date. Of over 1400 studies reviewed, 74 were retrieved and 23 included. The answer depended on the kind of study. Pooled data from cohort studies showed no association between fetal exposure to benzodiazepines and the risk of major malformations or oral cleft. Case-control studies pointed to roughly three times the odds of a major malformation.

The two designs work differently. A cohort study means following exposed and unexposed pregnancies forward and counting what happens. A case-control study means starting with affected and unaffected babies and asking mothers afterwards what they took. The second design is more open to distortion, since the mother of a child with a birth defect may recall her medicines more carefully.

How the benzodiazepines in pregnancy study worked

The new work lists Kelly Fung as first author and Krista Huybrechts as senior author. The paper calls it a population-based U.S. cohort study. It drew on two sources of insurance records: Medicaid, the public program for people on low incomes, for 2000 to 2018, and the MarketScan database of commercial insurance claims for 2003 to 2020.

To be counted, a pregnancy had to end in a live birth with continuous insurance coverage around it. The paper describes its participants as pregnant persons enrolled from 90 days before pregnancy to 30 days after delivery, and their babies had to be enrolled for 90 days after birth. Enrolled here means covered by the insurance plan.

Exposure was defined as at least 1 fill of a benzodiazepine in the first trimester. The first trimester means roughly the first three months of pregnancy, when the organs form. A fill is a prescription dispensed by a pharmacy.

The outcomes were any major congenital malformation and organ-specific malformations previously associated with benzodiazepine exposure.

Women who are prescribed a benzodiazepine differ from those who are not, in health and in other medicines taken. To deal with that, the team used propensity score fine stratification weights to control for confounders, meaning other factors that could explain an apparent link. A propensity score is an estimate of how likely each woman was to receive the drug, given everything else in her record. It lets researchers compare like with like.

The team also decided in advance what size of result would matter. A risk in exposed pregnancies 1.3 times that in unexposed ones, or higher, was set beforehand as the level that would count as clinically important.

What the study found on benzodiazepines and birth defects

Exposure was uncommon. In Medicaid, 51,339 of 2,478,434 pregnancies, or 2.1%, had a benzodiazepine dispensed in the first trimester. In the commercial data it was 27,779 of 1,750,721, or 1.6%. Together that makes 79,118 exposed pregnancies out of 4,229,155.

The main result is easiest to read as plain risks. After adjustment, a major malformation was recorded in 4.0% of exposed pregnancies and 4.1% of unexposed pregnancies in Medicaid. In the commercial data the figures were 4.9% and 4.8%.

Pooled across both sources, the risk in exposed pregnancies was 1% lower than in unexposed ones, with a range from 4% lower to 3% higher. That range sits far below the 1.3 mark the team had set.

The finer analyses agreed. Any associations with defects of particular organs were not consistently observed when the team repeated the work under different assumptions, nor was a dose-response relationship observed. A link that comes and goes between such repeats is treated as unreliable. A dose-response relationship is a pattern in which more of a drug goes with more of an effect, and its absence counts against a causal link. The paper adds that analyses of different benzodiazepines yielded similar results.

One part was less tidy. In some analyses of women who took a benzodiazepine along with other psychoactive medicines, the risk of certain organ-specific malformations was raised, but a consistent pattern did not emerge. The authors say those results were inconsistent and warrant further study.

Their overall conclusion is worded with care. No substantial increase in the risk for malformations was consistently identified across data sources.

How this fits with earlier benzodiazepine studies

The closest comparison is the Korean study from 2022, which used a national insurance database and a similar method. Among a total of 3,094,227 pregnancies, 40,846 (1.3%) were exposed to benzodiazepines during the first trimester. It reported a small increase: after adjustment, the risk of any malformation was 9% higher in exposed pregnancies, with a range from 5% to 13%, and the risk of heart defects 15% higher. The increase was larger at higher daily doses.

Even so, the Korean authors played down the size of what they saw. In their words, the absolute risks and population attributable fractions were modest. A population attributable fraction is the share of all cases in a population that could be put down to one exposure.

The US study did not reproduce the Korean result. Its range, topping out at 3% higher, leaves little room for a 9% increase, and it saw no rise with dose. Why two large studies differ is not settled by either. They cover different countries, different health systems and different prescribing habits.

Pooled analyses have not shown a clear increase either. A 2019 review combined eight cohort studies and concluded that prenatal benzodiazepine use was not associated with an increased risk of congenital malformations. Its estimate was 13% higher odds, with a range from 1% lower to 30% higher, which does not rule out a small effect.

A US registry that enrolls pregnant women with psychiatric illness reported much the same in 2022, in far smaller numbers. There were 5 (3.21%) major malformations in the exposure group and 32 (3.46%) in the comparison group.

A US case-control study, the National Birth Defects Prevention Study, raised flags for a few rare defects in 2019. Its authors stressed that exposure to benzodiazepines during pregnancy was rare in their data, with 93 exposed mothers among 11,614, and that the estimates were therefore loose.

The research group behind the new paper has used the same two databases before. Last December it reported on the sleep medicines known as Z-drugs, such as zolpidem, noting that sleep disturbances are common in pregnancy and often treated with nonbenzodiazepine sedative hypnotics (Z-drugs). That study also found no overall increase in malformations.

Benzodiazepines taken with other medicines in pregnancy

The unsettled corner of the new paper, combined use, has a history. The 2019 review found a signal there. There may be an increased risk of congenital malformations when benzodiazepines are used in conjunction with antidepressants, its authors wrote, on the basis of three studies.

A larger study from Taiwan, published in 2024, tested that directly in more than 2.6 million pregnancies. Its authors began by saying that the association between such concomitant use during pregnancy and the risk of congenital malformations remains inadequately explored. Concomitant means taken over the same period. Malformations were recorded in 3.81 of every 100 pregnancies exposed to both a benzodiazepine and an antidepressant and 2.87 of every 100 unexposed. After weighting for differences between the groups, the results did not suggest an increased risk for overall malformations. Digestive system malformations were the one exception.

So the new US study and the Taiwanese one have both looked at combinations, and neither found a clear, repeatable pattern. Neither ruled one out.

What the benzodiazepines study cannot show about pregnancy

Defects seen at birth only. The paper lists as a limitation analyses restricted to malformations prevalent at birth. Problems that come to light later in infancy may not be counted.

Live births only. The study included babies born alive. A pregnancy that ended in miscarriage or stillbirth was not in the data, so a defect severe enough to end a pregnancy would be missed. The Korean authors listed the same limit in their own work: restriction to live births.

A prescription filled is not a pill taken. Exposure meant a pharmacy fill. Some women will have stopped on learning they were pregnant. The paper names potential nondifferential exposure and outcome misclassification as a limitation. Misclassification means sorting some people into the wrong group. When it affects both groups alike, it tends to make a real difference look smaller.

One outcome. The study concerned structural birth defects. It did not examine miscarriage, early birth, withdrawal symptoms in newborns or later development.

Rare defects. Even in millions of pregnancies, individual rare defects occur in small numbers, and estimates for them are less certain than the overall figure.

Observational design. Weighting can balance the factors that insurance records capture. It cannot balance those they do not, such as how severe a woman’s anxiety was.

The study was funded by the National Institute of Child Health and Human Development, part of the US National Institutes of Health.

None of this turns a group average into a statement about one pregnancy. Do not stop or change a prescribed medicine without talking to your doctor.

Whether a benzodiazepine is suitable in a particular person who is pregnant or planning a pregnancy is assessed by their doctor.

On the question it set out to answer, a very large US dataset points the same way in both of its halves: about as many birth defects with a first-trimester benzodiazepine prescription as without one.