Start with what this is not. These drugs are not approved in pregnancy, and none of the evidence below is a reason to take one while pregnant or trying to conceive. Every study here describes pregnancies that were already exposed, usually before anyone knew about them. The size of that group is set by the contraception question in what tirzepatide does to oral contraception.
What the pooled cohorts found
A systematic review and meta-analysis covered seven cohort studies and more than 40,000 exposed pregnancies [1]. Exposure at any point was not associated with congenital malformations overall, at an odds ratio of 1.11 with an interval from 0.82 to 1.51.
First-trimester exposure and major malformations came in at 1.39, with an interval from 0.73 to 2.65. Stillbirth, spontaneous abortion, small for gestational age and preterm birth all showed no significant increase.
One outcome cleared significance. Urinary malformations came in at 1.24 with an interval from 1.05 to 1.47 — and that estimate rests exclusively on unadjusted data. The authors read it as likely residual confounding rather than a real effect, which is a stronger caveat than most papers attach to their own positive finding.
The question of continuing versus stopping
A target trial emulation asked something narrower and more practical: among pregnancies already exposed, what happens if the prescription continues into the first trimester [2]. It used US insurance claims from 2011 to 2024 and covered 3,572 pregnancies, 41.1% of them in women with type 2 diabetes.
| Outcome | Continuing vs not | 95% CI |
|---|---|---|
| Nonlive birth | 29.7% vs 27.1% (aRR 1.09) | 0.98 to 1.23 |
| Major congenital malformation | 1.21 | 0.83 to 1.82 |
| Small for gestational age | 1.29 | 0.82 to 2.06 |
| Large for gestational age | 1.08 | 0.84 to 1.40 |
Why confounding is especially hard here
People taking these drugs into a pregnancy have higher baseline BMI, more diabetes and more cardiometabolic disease than the general obstetric population. Every one of those independently affects pregnancy outcomes.
The target trial emulation names its own version of the problem: potential residual confounding by prior glycemic control. Separating the drug from the reason it was prescribed is the same difficulty this site set out in what happens when the control group never started treatment.
What a reader can actually take from this
If a pregnancy was exposed before it was known about, the pooled evidence so far does not show an increase in the outcomes that question usually concerns. That is worth knowing and it is genuinely the most common situation.
It is a different statement from the drug being safe in pregnancy, which nobody has established and no randomized trial is likely to test. The same "not shown" reading applies in the Alzheimer's evidence. The contraception interaction that puts people in this position is covered in what tirzepatide does to oral contraception.