GLP-1s and Pregnancy: The Data Behind the Label's One-Line Warning

A one-line warning, and a much longer real story

Every GLP-1 label carries some version of the same instruction: stop before pregnancy, or as soon as you know. It reads like a formality, tucked into section 8.1 alongside a dozen other prescribing details. It is not a formality. Behind that line sits a genuine animal-safety signal, an almost complete absence of prospective human trial data, and — increasingly — a body of real-world evidence from tens of thousands of pregnancies that happened anyway, because these drugs are now common enough that "I found out I was pregnant while on Ozempic" has become one of the most searched questions in this entire category.

This piece is for two different readers, and it is worth saying so upfront. One is planning ahead and wants to know the runway. The other already has a positive test and a medication in her system, and wants to know what the data actually says before she panics. Both deserve a straight answer, not a hedge.

Why the label says what it says

The caution starts in animal studies, not human ones. GLP-1 receptor agonists in pregnant animals have been linked to reduced fetal growth and to skeletal, visceral, and embryonic abnormalities². The Wegovy label translates that into plain instruction: weight loss offers no benefit to someone who is pregnant and may cause fetal harm, so the drug should be discontinued once pregnancy is recognized¹. What the label does not have is a body of human trial data to weigh against that animal signal, because no pharmaceutical company runs a pregnancy-safety trial on a weight-loss drug. When the animal data raises a flag and the human trial evidence is simply absent, obstetric medicine defaults to caution — that is the entire logic of the warning, not squeamishness dressed up as science.

Worth noting: the two major molecules do not even agree with each other on timing. Semaglutide's label calls for stopping at least two months before a planned pregnancy, citing the drug's long half-life. Tirzepatide's label sets no such advance window at all — it simply says to discontinue once pregnancy is recognized. A reader assuming one rule covers both drugs is working from the wrong label.

What happens to the women who didn't stop in time

This is where the story gets more interesting than the warning label suggests, and it is the part most coverage of this topic skips. Because inadvertent early exposure has become common, two research teams independently pooled the real-world outcomes in 2026.

The larger of the two combined six studies and 43,577 women exposed around the time of conception, and found no significant difference in the risk of major congenital malformations between exposed and unexposed pregnancies³. A second, separate meta-analysis pooled seven cohort studies covering more than 40,000 exposed pregnancies and reached the same broad conclusion — no statistically significant increase in malformations, stillbirth, miscarriage, small-for-gestational-age birth, or preterm delivery⁴. Women in the cohort were exposed at a range of gestational points, not uniformly in the first trimester, which is itself part of why both research teams are careful to call the finding reassuring rather than conclusive.

That distinction matters enough to spell out. Both papers explicitly warn against reading their results as proof of safety, and both call for prospective registry data rather than the retrospective, observational pooling that produced these numbers³⁴. A broader 2026 systematic review of GLP-1 safety across preconception, pregnancy, and lactation lands in the same place — the human evidence so far is reassuring in aggregate, and still too thin to call the question settled⁵.

The one signal worth watching

Not every finding in this literature washed out to null. One of the two meta-analyses flagged a possible association with urinary-tract malformations, though the authors attribute it to residual confounding, since it rests entirely on unadjusted estimates rather than analyses that controlled for maternal weight, diabetes, or other confounders⁴. It is the kind of result that responsible researchers report rather than bury, and it is exactly the kind of signal a future adjusted study could either confirm or make disappear. Nobody currently knows which.

What actually happens next, if it happens to you

The practical sequence is short and every source agrees on it. Stop the medication as soon as pregnancy is confirmed — do not take another dose while you sort out next steps¹. Call your OB or prescriber and tell them exactly which drug, at what dose, and when your last dose was; this is a routine conversation for them by now, not a novel one. From there, standard early-pregnancy care takes over — prenatal vitamins with folate, the normal first prenatal visit, whatever monitoring your OB recommends — and the goal shifts from weight loss to healthy gestational weight gain.

If pregnancy is still ahead of you rather than already here, the better version of this story is the planned one: contraception while you're still on the drug, a deliberate stop on your specific molecule's timeline, and then trying. Our companion piece on trying to conceive on a GLP-1 walks through that sequence and the washout-timing paradox in full, and if PCOS is part of your fertility picture, GLP-1s and PCOS explains why ovulation can return faster than expected once metabolic health improves.

The honest read

Two claims sit on this page, and they carry different weight. "Do not use a GLP-1 in pregnancy" is settled — built on reproductive toxicity across species, an explicit label instruction, and the plain fact that intentional weight loss offers a pregnant patient nothing¹². "An early, accidental exposure likely did not harm your baby" is newer and still improving — grounded in tens of thousands of real pregnancies, genuinely reassuring, and still observational data that both research teams themselves decline to call proof³⁴. The gap between those two sentences is not a flaw in the reporting. It is the actual shape of what is currently known.