This was an individual participant data meta-analysis — the strongest form of pooling, because it reanalyses each woman's data rather than each trial's summary. Ten double-blind placebo-controlled trials of metformin in pregnancies without diabetes were identified, seven supplied individual data, and 2,297 pregnancies were analysed after harmonisation: 1,159 on metformin, 1,138 on placebo.
Metformin did not reduce gestational diabetes. By WHO 1999 criteria the adjusted odds ratio was 1.00 (95% CI 0.71 to 1.41), and by NICE 2015 criteria it was identical. Only using IADPSG thresholds did an adjusted analysis suggest benefit (adjusted OR 0.71, 95% CI 0.52 to 0.98) — a result that depends on which diagnostic criteria you apply, which is a warning sign rather than a finding. Fasting glucose was lower by 0.06 mmol/L, a difference with no clinical meaning.
What metformin did do was extend pregnancy: gestation was 0.30 weeks longer (95% CI 0.06 to 0.54) and preterm birth was less common (adjusted OR 0.64, 95% CI 0.47 to 0.89). Neonatal head circumference was marginally larger. Gastrointestinal side effects were more frequent, as expected.
The practical reading is that metformin should not be offered to a woman with obesity or polycystic ovary syndrome on the promise that it will prevent gestational diabetes, because across seven trials it did not. The preterm birth signal is interesting and deserves its own trial, but it is a secondary outcome from a study designed to answer a different question, and it should not become the new reason to prescribe.
- Stop counselling metformin as gestational diabetes prevention in high-risk pregnancy — the pooled adjusted odds ratio is exactly 1.00
- Where metformin is already in use for polycystic ovary syndrome, this is not a reason to stop it, but do not extend it for glycaemic prevention
- Continue what does work: early risk assessment, weight and activity support before conception, and timely oral glucose tolerance testing
- Warn about gastrointestinal side effects, which were more common on metformin and drive discontinuation
- Treat the preterm birth finding as hypothesis-generating, not as a new indication
The statistics, in plain English
An adjusted odds ratio of 1.00 with a confidence interval of 0.71 to 1.41 is a clean null: metformin is as likely to have been slightly harmful as slightly helpful, and the trial data cannot support a claim in either direction. The one positive result — adjusted OR 0.71 using IADPSG criteria — has an upper limit of 0.98, barely clearing 1.0, and appears in only one of three diagnostic definitions tested. When a result survives under one definition and vanishes under two others, the safest interpretation is that it reflects the definition rather than the drug.
Read the rest in the app
You have read your two free briefings this month. The app carries all 27 specialties, every morning, free — and this finding is waiting in it.
Scan to keep reading on your phone. No account needed to start.
Tomorrow morning, before your first patient
One edition a day for diabetes & endocrinology, written by the desk, every claim tied to its paper. Six minutes.
Get the app — free