GLP-1 drugs in early pregnancy raise preterm birth risk - but only for diabetes patients

Women who inadvertently took GLP-1 receptor agonists around the time of conception face a higher risk of preterm birth - but only if they were taking the drugs to treat diabetes. When the same medications were used for weight management, no increased risk was detected.

That's the central finding of a Danish nationwide cohort study published in Human Reproduction Open, which analyzed data from 756,636 singleton pregnancies among 480,231 women between October 2009 and December 2023. The results suggest it may be diabetes itself, rather than the drugs, that drives the elevated risk of premature delivery.

The scale of inadvertent exposure

GLP-1 receptor agonists - a class that includes semaglutide (sold as Ozempic and Wegovy) and liraglutide (sold as Saxenda) - were originally developed for type 2 diabetes. Their dramatic effectiveness for weight loss has turned them into one of the fastest-growing medication classes worldwide. In Denmark, roughly 70% of people prescribed weight loss medications are women, many of childbearing age.

Current clinical guidance recommends stopping GLP-1 receptor agonists eight weeks before planning a pregnancy. But this recommendation rests on early-phase animal studies, not real-world human evidence. And with millions of women now taking these drugs, inadvertent exposure during early pregnancy - before a woman knows she's pregnant - is inevitable.

The Danish study defined inadvertent exposure as redeeming a prescription for liraglutide or semaglutide within eight weeks before or after the date of last menstruation - a 16-week window that captures the period when a pregnancy might not yet be detected and when early embryonic organ development is underway. A total of 529 pregnancies met this exposure definition.

Separating the drug from the disease

The study's most important contribution is distinguishing between women who took GLP-1 drugs for diabetes and those who took them for weight management. Previous research has lumped these groups together, which made it impossible to tell whether adverse outcomes were caused by the medication or by the underlying metabolic condition.

After adjusting for maternal age, body mass index, smoking status, geographic region, education level, pre-existing diabetes, and temporal trends, the researchers found that the increased preterm birth risk was confined entirely to women taking GLP-1 drugs for diabetes treatment.

The numbers were specific. Compared to unexposed women, those who took liraglutide for diabetes had a 70% higher risk of preterm birth (delivery before 37 weeks). For semaglutide for diabetes, the risk was 84% higher. In absolute terms, this translated to roughly a 9% preterm birth rate for liraglutide-exposed diabetic women and approximately 11% for those exposed to semaglutide - compared to the background rate among unexposed women.

Among women taking the same drugs for weight management alone, no statistically significant increase in preterm birth was found. The drugs were the same. The doses overlapped. The difference was the reason for the prescription - and, by extension, the metabolic health of the patient.

Diabetes as the underlying driver

Kathrine Hviid, the study's first author and a PhD student at Copenhagen University Hospital Hvidovre, described this as an extremely important finding for future research and clinical counseling. The observation that elevated preterm birth risk was confined to women treated for diabetes, and not to those using GLP-1 receptor agonists for weight management, suggests that the underlying condition of diabetes, rather than the medication, may be driving this association.

This interpretation aligns with what's already known about diabetes and pregnancy. Pre-existing diabetes is an established risk factor for preterm birth, pre-eclampsia, macrosomia (large-for-gestational-age babies), and placental complications. The metabolic disruptions of diabetes - chronic hyperglycemia, insulin resistance, vascular damage, and systemic inflammation - create a hostile environment for pregnancy independent of any medication.

The study doesn't prove that GLP-1 drugs are safe during pregnancy. What it shows is that, in this large dataset, the signal for harm appears only in a subgroup already at elevated risk due to their disease. Separating drug effects from disease effects is one of the hardest challenges in observational pharmacoepidemiology, and this study represents a meaningful step in that direction.

What the study could and couldn't measure

The strength of the Danish study lies in its size and data quality. Denmark's nationwide health registries link prescriptions, diagnoses, and birth outcomes for the entire population, creating datasets that are difficult to match elsewhere. The 756,636-pregnancy sample provides statistical power that smaller studies lack.

But several limitations constrain interpretation. The study is observational, meaning it shows association, not causation. It's possible that unmeasured differences between diabetic and non-diabetic women - factors not captured by the adjustment variables - account for some of the risk difference.

The researchers also couldn't confirm that women actually took the medications after filling their prescriptions. In Denmark, GLP-1 drugs carry a significant out-of-pocket cost - approximately 180 euros for a 1mg injection of Wegovy or Ozempic, with a state subsidy available for diabetes treatment. The cost is high enough that researchers assume compliance is very high among those who fill prescriptions, but assumption is not confirmation.

The 529 exposed pregnancies, while drawn from a massive overall sample, represent a relatively small group for subgroup analyses. Breaking that number down further by drug type, indication, and outcome leaves some cells with limited statistical power. The absolute risk estimates should be interpreted with appropriate caution.

The study examined only periconceptional exposure - the period around conception and very early pregnancy. It doesn't address what happens when GLP-1 drugs are taken later in pregnancy, or whether stopping them well before conception (as currently recommended) eliminates any residual risk.

Clinical counseling in a data-sparse field

Henriette Svarre Nielsen, professor in the Department of Gynaecology and Obstetrics at Copenhagen University Hospital Hvidovre and the study's senior author, emphasized that it was too soon to change the existing recommendation to stop GLP-1 drugs before planned pregnancy, regardless of indication. But she noted that these findings could guide clinical counseling for women who discover they've been inadvertently exposed.

For a woman taking semaglutide for weight loss who finds out she's pregnant, the data may provide some reassurance - or at least a basis for a more nuanced conversation with her obstetrician. For a woman taking the same drug for diabetes, the discussion would appropriately emphasize the importance of tight glycemic control and monitoring for preterm labor, recognizing that her underlying condition already confers elevated risk.

In an accompanying commentary in the same journal, Yeyi Zhu and Monique Hedderson of Kaiser Permanente Northern California wrote that the findings support more balanced, individualized discussions with patients who may have experienced inadvertent periconceptional exposure. They also stressed that for patients with diabetes, the study reinforces the importance of prioritizing metabolic health and glycemic management before and during pregnancy.

The bigger prescribing picture

GLP-1 receptor agonists are the fastest-growing prescription drug class in many countries. Tens of millions of people now take them, and the demographics skew heavily toward women of reproductive age. The probability of inadvertent periconceptional exposure will only increase as prescriptions grow.

Nielsen made this point explicitly: future studies must account for the reason a woman has been prescribed these medications, because the risks differ between diabetes and weight management indications. Failing to make that distinction - as most prior studies have - risks either falsely reassuring diabetic patients or unnecessarily alarming those using the drugs for weight loss.

Randomized controlled trials of GLP-1 drugs during pregnancy are ethically complicated and unlikely in the near term. For now, large observational studies like this one represent the best available evidence. The Danish data won't be the last word, but they shift the conversation in an important direction: away from asking whether GLP-1 drugs are dangerous in pregnancy, and toward asking for whom, under what conditions, and why.