This was a patient-level meta-analysis of double-blind, placebo-controlled trials of metformin in pregnancies without diabetes. Ten trials met the criteria, covering 2,695 women, and seven supplied individual participant data on 2,485 of them, which is 92.2 per cent of the data available. After harmonising the datasets, 2,297 pregnancies were analysed, 1,159 randomly assigned to metformin and 1,138 to placebo. Models were adjusted for maternal age, body mass index, gestational age when metformin was started, and baseline blood glucose.
Metformin did not reduce gestational diabetes. On World Health Organization 1999 criteria the odds ratio was 1.04 (95% CI 0.80 to 1.36), and 1.00 (0.71 to 1.41) after adjustment. On National Institute for Health and Care Excellence 2015 criteria it was 0.98 (0.76 to 1.27), and again 1.00 (0.71 to 1.41) adjusted. Only on adjusted analysis using International Association of Diabetes and Pregnancy Study Groups thresholds did anything appear: adjusted odds ratio 0.71 (0.52 to 0.98), while the unadjusted figure of 0.83 (0.65 to 1.06) did not reach significance. Fasting glucose was lower by 0.06 mmol/L (-0.10 to -0.01), which is real but too small to matter clinically, and there was no difference in the two-hour value after a glucose load.
Where metformin did do something was the neonate. Gestation was 0.30 weeks longer (0.06 to 0.54), preterm birth was less likely (adjusted odds ratio 0.64, 0.47 to 0.89), and head circumference was 2.43 percentiles larger (0.13 to 4.72). Gastrointestinal side effects were more common. The practical conclusion is narrow and firm: do not start metformin in a pregnancy without diabetes in order to prevent gestational diabetes. In India, where prevalence is high and onset is younger and at lower body mass index, this removes a tempting shortcut and leaves screening and lifestyle support as the tools that actually work.
- Do not prescribe metformin in early pregnancy with the aim of preventing gestational diabetes.
- Continue screening on your usual pathway - this changes what you prescribe, not when you test.
- Where metformin is already running for another indication, such as polycystic ovary syndrome or pre-existing type 2 diabetes, this analysis does not apply.
- Warn about gastrointestinal side effects, which were reported more often on metformin.
- Record which diagnostic criteria your unit uses - the result changed depending on whether WHO, NICE or IADPSG thresholds were applied.
The statistics, in plain English
An odds ratio of 1.00 means no difference at all, and the confidence intervals here sit squarely across 1.0 on both of the commonly used sets of criteria, so this is a genuinely negative result rather than a study too small to tell. The one positive signal, an adjusted odds ratio of 0.71 using IADPSG thresholds, is weaker than it looks for two reasons: the unadjusted version of the same comparison (0.83, crossing 1.0) was not significant, and it is one of several definitions tested, so the chance of one crossing the line by luck is high. The fasting glucose difference of 0.06 mmol/L is a good example of statistically significant but clinically irrelevant - it is smaller than the variation between two readings on the same meter.
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