Migraine antibody drugs in early pregnancy linked to higher miscarriage rate in US claims study

A newer group of migraine prevention drugs works by blocking a signaling molecule called CGRP. Because migraine is most common in women of childbearing age, some pregnancies begin while these drugs are in use, and a study of US insurance records has now compared those pregnancies with others. Miscarriage was more common when a pregnancy was exposed to the antibody form of the drugs in the first 20 weeks, and no more common with propranolol, an older preventive.

The study, published on October 12 in JAMA Neurology, is observational and cannot show that the drugs caused the losses. Miscarriage is common in any group of pregnancies, and people prescribed these drugs may have more severe migraine than those who are not. Insurance claims also record only that a prescription was filled, which is a step short of knowing it was used. The authors ask for further evaluation and stop short of a verdict.

Why migraine drugs in pregnancy are an open question

Migraine affects many people of reproductive age, and the new paper’s authors point out that treatment options during pregnancy are limited because of safety worries. An earlier study estimated that 17% to 24% of females of reproductive age have migraine.

CGRP, short for calcitonin gene-related peptide, is a small protein released by nerves that plays a part in migraine pain, and drugs that block it have been available to prevent attacks since 2018. The ones studied here are monoclonal antibodies, laboratory-made proteins built to latch on to one target in the body. There are four: erenumab, fremanezumab, galcanezumab and eptinezumab.

The worry in pregnancy comes from what else CGRP does, and the new paper notes that CGRP is important in placental development. A laboratory study in 2007 found the molecule and its receptor at the human implantation site during early pregnancy, the place where the embryo attaches to the womb, and its authors concluded that CGRP there could modify the growth of new blood vessels in the placenta. A later paper credited the molecule with helping blood reach the placenta, saying it increases utero-placental blood flow and lowers resistance in the blood vessels of the womb.

A drug that blocks CGRP might therefore matter to an early pregnancy, and these drugs linger: their half-life, the time the body takes to clear half of a dose, is long. The team running a pregnancy safety study of galcanezumab pointed to that long half-life and to how common migraine is in women of childbearing age, and concluded that galcanezumab exposure may occur during pregnancy.

Until now there was little to go on. A review published in September found that the evidence on pregnancy came only from pharmacovigilance, the monitoring of a medicine’s safety after it reaches the market, and from case reports of unintended exposures.

How the study tracked migraine drugs and miscarriage

Leah Flatman is the first author of the new work, and Anick Bérard the senior author. Rather than recruiting anyone, the team looked back through existing records, comparing what happened in pregnancies that were and were not exposed to the drugs. Their records came from the AM-PREGNANT cohort, which is derived from the Merative MarketScan Commercial Database, a store of claims submitted to commercial health insurers in the United States, and covered 2018 to 2021.

The team started with 7,991,200 pregnancies and narrowed them down to 7579 singleton pregnancies, each with a single baby, in 7119 individuals who met the predefined eligibility criteria. Those included were aged 15 to 45, had continuous insurance, and had migraine before pregnancy. The paper defined that as a migraine diagnosis, or a prescription for a triptan or an ergot (medicines taken to stop an attack once it starts), within 6 months before the last menstrual period. The average age was 31.6 years.

Two kinds of preventive were examined: the four antibodies, and propranolol, a beta-blocker, which is a type of heart and blood pressure medicine long used to prevent migraine. Including propranolol gave the researchers a second comparison, with people on a different preventive whose migraine also called for one.

A pregnancy counted as exposed once there was 1 or more filled prescription or injection claim for a drug during follow-up. Exposure was treated as time-varying, so a pregnancy counted as unexposed until the first such claim and as exposed from then on. The outcome was miscarriage at 20 weeks of pregnancy or earlier, which the paper calls spontaneous abortion, and the statistical models adjusted for comorbidities, meaning other illnesses, and for other medicines taken at the same time.

What the study found about miscarriage

Across all 7,579 pregnancies, 1862 spontaneous abortions occurred, or about one in four.

There were 100 miscarriages during time exposed to the antibodies, which the paper reports as 28.6%, and 37 during time exposed to propranolol without the antibodies, reported as 12.8%. Those two percentages are not a like-for-like measure of anyone’s chances. Each is calculated on time spent exposed, so neither is a simple share of a counted group of pregnancies, and neither is adjusted for differences between the groups or for when in pregnancy exposure began. The adjusted analysis is the paper’s main comparison.

After adjustment, the rate of miscarriage during antibody-exposed time was 42% higher than during unexposed time, with a range from 12% to 80%. The estimate held up in bias-adjusted analyses, which simulate how errors in the records might have shifted the answer, coming out 33% higher, with a range from 23% to 45%. That second range comes from the simulations and measures a different kind of uncertainty from the first, so its narrower span does not mean a firmer result.

Propranolol looked different. Its adjusted estimate was 2% higher than no exposure, with a range from 29% lower to 47% higher, which is consistent with no effect but too wide to rule out a modest rise. The bias-adjusted figure showed no difference at all, and the paper concludes that no association was observed with propranolol.

The authors do not claim to have shown harm. They call for further evaluation of the antibodies’ safety and say their results emphasize counseling for individuals of childbearing potential, meaning discussion of the drugs and pregnancy with patients who could become pregnant.

What earlier monitoring of CGRP migraine drugs in pregnancy showed

Earlier evidence had not shown a problem, though it was a weaker kind of evidence. One research team has twice searched the World Health Organization’s global database of safety reports, which collects accounts of suspected side effects from doctors and patients. In 2023 the team described safety data on the use of these antibodies in pregnancy as limited. Up to the end of 2021, 286 safety reports were found, and sixty-three of them mentioned a miscarriage. Miscarriage was reported no more often with these drugs than across the database as a whole, or than with triptans.

An update in 2024 added newer CGRP-blocking pills known as gepants. It found 467 reports, mostly originating from the United States of America, and again no signal compared with triptans. The team was frank about the limits of such counts: disproportionality analysis, which checks whether a problem is reported more often with one drug than with others, is a hypothesis-generating method and cannot settle questions of drug safety on its own. Because nobody knows how many pregnancies were exposed in total, it cannot produce a rate.

The September review drew on sixteen studies and reached the same place. The reports and case studies had not identified specific patterns of harm to mother, fetus or newborn, and its authors called for pregnancy registries.

Gathering better data has been slow. One of two ongoing pregnancy safety studies of galcanezumab, an insurance-claims study designed to look at birth defects, reported in 2024 that it had found 207 galcanezumab-exposed pregnancies in women with migraine over more than four years. It needs 430 with linked infant records, and its final report is planned for December 2027. The new study is a different kind of evidence, because it has a comparison group and can estimate a rate.

Migraine itself, other migraine drugs and miscarriage

Any study of this question has to reckon with how common miscarriage is. A Lancet review defines it, generally, as the loss of a pregnancy before viability. The review put the pooled risk at 15.3% of all recognized pregnancies, and its list of risk factors includes very young or older female age, previous miscarriages and smoking.

Migraine may carry a small risk of its own. A UK study published in January matched 193,208 pregnancies of women with migraine one-to-one with pregnancies of women without it, and miscarriage was recorded in 10.8% of the migraine group and 10% of the other. The same study looked at medicines. Triptans, the antidepressant amitriptyline and anti-inflammatory painkillers were each tied to more miscarriages among women with migraine, while beta-blockers were not associated with a higher risk, a result that matches the new paper.

A North American study of 7,890 women who were planning a pregnancy adds a caution. Among them, a history of migraine before conception made no appreciable difference to miscarriage risk. The largest increase was among those taking prescription preventives, at 43% higher, but with a range from 28% lower to 184% higher, which is too wide to lean on. The authors concluded that patterns of medication use suggesting greater underlying migraine severity went with a higher risk of miscarriage, which suggests the medicine may be a marker of how severe the migraine is.

Miscarriage rates in these studies run from 10% to about 25%, a spread that may reflect how early a pregnancy is recognized and recorded in each data source.

What the migraine drug study cannot show about pregnancy

Cause. This was an observational study of insurance records. It shows that miscarriage was more frequent during exposed time. It cannot show that the antibodies were the reason.

Who is given these drugs. People prescribed an antibody for migraine may differ from those who are not, above all in how severe their migraine is. The propranolol comparison helps, since those patients also needed prevention, but the two groups may still differ.

What a claim records. A filled prescription shows that a drug was dispensed. Whether and when it was used is not recorded. Because the antibodies clear slowly, a dose given shortly before pregnancy could still be present in the body during a pregnancy counted as unexposed.

Which losses are counted. Claims capture a miscarriage only when it leads to medical care. Very early losses may be missed in both groups.

Small numbers. The antibody result rests on 100 miscarriages. Its range, from 12% to 80% higher, is wide.

One result for four drugs. The estimate is for the four antibodies taken together. The gepant pills were not part of the study.

Who was studied. These were commercially insured people in the United States between 2018 and 2021.

The study is about a pattern across thousands of pregnancies and says nothing about any single one. Do not stop or change a prescribed medicine without talking to your doctor.

Whether a migraine preventive is continued, changed or paused in a particular person who is pregnant or planning a pregnancy is assessed by their doctor.

For drugs that had only voluntary reports behind them in pregnancy, the study supplies a measured comparison, and its authors treat it as a reason to look harder.