Metformin in pregnancy was not linked to developmental delay in 177,409 children
A population-linked Australian cohort followed children to school entry after antenatal metformin. The adjusted relative risk of developmental vulnerability was 0.97 — no different from unexposed children.
| Group | Value (%) |
|---|---|
| Metformin-exposed | 18.2 |
| Not exposed | 13.9 |
Metformin is prescribed in pregnancy more often than it used to be — for gestational diabetes, for type 2 diabetes, and increasingly to blunt excess weight gain in high-metabolic-risk pregnancies. It crosses the placenta, and in animal models it influences embryonic cortical development, which has left an open question no trial was designed to answer: does exposure in the womb change how a child develops [s1]? A cohort study published in JAMA Network Open on September 30 looked for that signal in a whole population of schoolchildren and did not find one [s1].
What the study did
The researchers used population-level pregnancy and birth records from Victoria, Australia, for births between January 1, 2009, and December 31, 2020, and linked them to the Australian Early Development Census (AEDC) — a standardised national assessment teachers complete every three years during a child's first year of full-time school, at age 4 to 6 years [s1].
A child was counted as "developmentally vulnerable" if they scored below the 10th percentile in at least two of the AEDC's five domains: physical health and well-being, social competence, emotional maturity, school-based language and cognitive skills, and communication skills and general knowledge [s1]. The analysis used inverse probability of treatment weighting combined with regression adjustment to account for the many ways mothers prescribed metformin differ from those who are not [s1].
From 871,627 singleton births, 177,409 children had linked developmental outcomes [s1]. Of those, 1,095 (0.6%) had been exposed to metformin before birth [s1]. That is a small exposed group, and it matters for what the study can and cannot rule out.
The result
Developmental vulnerability was observed in 199 of the metformin-exposed children (18.2%), against 24,379 (13.9%) of those not exposed [s1]. On its face that looks like a gap. But the exposed mothers differ systematically — more diabetes, more obesity, more of the socioeconomic factors that independently track with developmental scores — and once those were accounted for, the gap closed.
After adjustment, metformin exposure was not associated with developmental vulnerability: adjusted relative risk 0.97 (95% CI 0.74-1.29) [s1]. Nor was there an association with any of the five individual domains taken separately [s1]. The authors conclude the finding "may provide reassurance to clinicians and women considering metformin use during pregnancy" [s1].
How this fits the wider evidence
This is a reassuring null from an observational dataset, and it sits alongside recent randomised evidence on a different question. A patient-level meta-analysis in NEJM Evidence in August pooled trials of metformin in high-metabolic-risk pregnancies, examining its effects on pregnancy outcomes rather than long-term child development [s2]. The two together sketch a drug whose short-term obstetric effects are being mapped by trials while its long-term offspring effects are, necessarily, being tracked in cohorts like this one.
The limits, stated plainly
A confidence interval running from 0.74 to 1.29 is the heart of the matter [s1]. It is centred almost exactly on no effect, which is genuinely reassuring, but its upper bound is still compatible with a 29% relative increase in developmental vulnerability [s1]. With only 1,095 exposed children [s1], the study is better at excluding a large effect than a small one.
It is also observational. Inverse probability weighting adjusts for the confounders that were measured; it cannot adjust for those that were not, such as glycaemic control during pregnancy or the underlying severity of the condition metformin was treating. The indication for the drug is tangled up with the outcome in ways no weighting fully unpicks.
And the AEDC is a screen for developmental vulnerability at school entry, not a clinical diagnosis of any specific condition [s1]. It captures something real and population-relevant, but it is a blunt instrument for detecting subtle, domain-specific effects that might emerge later.
What to watch
Whether longer follow-up — into the school years, where learning and attention outcomes can be measured more finely — reproduces this null; whether cohorts large enough to study metformin's separate indications (gestational diabetes versus obesity versus type 2 diabetes) find the same reassurance across all of them; and whether the randomised trials now following their offspring report developmental endpoints directly.
This article describes one observational cohort study and is informational only. It is not medical advice, and nothing here should be used to start, stop or change any medication in pregnancy.
Sources
- [s1] Metformin Exposure in Pregnancy and Childhood Developmental Outcomes, JAMA Network Open, 2026;9:e2637054, published online 2026-09-30. PMID 42814460.
- [s2] Metformin in High Metabolic Risk Pregnancies — A Patient-Level Meta-Analysis, NEJM Evidence, 2026;5, published 2026-08-25.
Sources
- Metformin Exposure in Pregnancy and Childhood Developmental Outcomes — JAMA Network Open, 2026;9:e2637054 , September 30, 2026
- Metformin in High Metabolic Risk Pregnancies — A Patient-Level Meta-Analysis — NEJM Evidence, 2026;5 , August 25, 2026
More on
Metformin did not prevent gestational diabetes in a pooled analysis of 2,297 pregnancies
An individual-participant meta-analysis found metformin given in pregnancy did not lower the odds of gestational diabetes, though it modestly lengthened gestation and reduced preterm birth.
Baby-led weaning vs spoon-feeding: what the trials actually found
The main randomised trial found babies who fed themselves were no slimmer, no more iron-deficient and no more likely to choke than spoon-fed babies — provided parents were coached to offer safe, iron-rich foods.
Higher pre-pregnancy weight tracks with offspring heart disease decades on
In 2.5 million Swedish births followed to adulthood, mothers' early-pregnancy obesity was linked to a stepwise rise in offspring cardiovascular disease, though sibling analyses temper the size.
What actually reduces childhood obesity: the evidence, and its limits
Intensive family-based behavioural programmes produce small, real cuts in children's BMI; the newer weight-loss drugs produce much larger ones, but the weight returns when they stop.