Lifestyle changes in pregnancy cut gestational diabetes — by how much depends on the test

Two papers published two days apart in January examine gestational diabetes from opposite ends: whether it can be prevented, and whether preventing it saves babies' lives. Together they show a field where the diagnostic threshold does more work than most people realise.

The prevention question

An individual participant data and network meta-analysis published in The BMJ on 6 January pooled randomised trials of lifestyle interventions in pregnancy — physical activity based, diet based, or mixed — searching major databases from January 1990 to April 2025 [s1].

It included 104 randomised trials covering 35,993 women, with individual participant data available for 68% of participants (24,391 women across 54 studies) [s1]. That level of individual data is what allows the analysis to test whether effects differ by maternal characteristics rather than only estimating an average.

The headline effect depends on which definition of gestational diabetes is used, and the paper reports all of them.

Under any criteria, lifestyle interventions reduced gestational diabetes by 10% in the individual-data trials — an odds ratio of 0.90 (95% CI 0.80 to 1.02), with an absolute risk reduction of 1.3% (95% CI −0.3% to 2.6%) [s1]. Both intervals cross the null. Combining individual data with aggregate data from trials that did not supply it, the reduction was 20% (odds ratio 0.80, 95% CI 0.73 to 0.88; absolute risk reduction 2.6%, 95% CI 1.6% to 3.6%) [s1].

Under International Association of Diabetes in Pregnancy Study Group criteria, the reductions were 14% in the individual-data trials (odds ratio 0.86, 95% CI 0.75 to 0.97; absolute risk reduction 2.7%, 95% CI 0.6% to 5.0%) and 18% when combining data sources (0.82, 0.72 to 0.93; absolute risk reduction 3.5%, 1.3% to 5.7%) [s1].

Under UK NICE criteria, there was no reduction at all: odds ratio 0.98 (95% CI 0.84 to 1.13) [s1].

That divergence is not a statistical artefact to be argued away. It means that the same intervention, in the same women, either works or does not depending on where a country draws its diagnostic line — and it means the answer to "does this work" cannot be given without naming the criteria.

What varied and what did not

The analysis tested whether effects differed by maternal body mass index, age, parity, ethnicity or education, and by intervention characteristics [s1].

Effects did not vary by maternal characteristics, with one exception: education [s1]. Women at all educational levels benefited, but the benefit was smaller in those with low education — the low versus middle interaction odds ratio was 0.68 (95% CI 0.51 to 0.90) and low versus high 0.71 (0.54 to 0.93) [s1].

Among intervention features, two mattered. Group format was more effective (odds ratio 0.81, 95% CI 0.68 to 0.97; absolute risk reduction 2.5%, 0.4% to 4.3%), as were newly trained facilitators (0.82, 0.69 to 0.96; absolute risk reduction 2.4%, 0.5% to 4.2%) [s1]. Physical activity based interventions ranked highest in the network meta-analysis, with a mean rank of 1.1 (95% CI 1 to 2) [s1].

The finding on newly trained facilitators is the kind of result that could reflect enthusiasm in trial settings rather than something transferable to routine care, and the paper offers it as an implementation consideration rather than a mechanism [s1].

The authors' conclusion is that lifestyle interventions in pregnancy are likely to prevent gestational diabetes, with effects varying by diagnostic criteria, and that implementation strategies should address inequalities by maternal education [s1].

The stillbirth question, and its surprise

A systematic review and meta-analysis published in eClinicalMedicine on 8 January asked what gestational diabetes does to stillbirth risk, noting that around 2 million pregnancies a year end in stillbirth, the majority in low- and middle-income countries [s2].

Searching seven databases from inception to 9 May 2025 with no restriction on language, location or time, it identified 101 studies presenting unadjusted results (92,915,856 women) and 19 reporting adjusted results (63,629,536 women) [s2].

The pooled adjusted result did not show evidence of an association between a gestational diabetes diagnosis and stillbirth worldwide — odds ratio 0.81 (95% CI 0.68 to 0.97), with substantial heterogeneity (I² = 87.7%) across 19 studies [s2].

Stratified by country income, the picture separates. In high-income countries, a gestational diabetes diagnosis was associated with reduced odds of stillbirth (odds ratio 0.73, 95% CI 0.65 to 0.82, I² = 31.9%, 13 studies) [s2]. In upper and lower middle income countries, no association was observed (1.17, 0.71 to 1.93, I² = 59.7%, 6 studies) [s2]. There were no adjusted estimates from low-income countries at all [s2]. Screening method made no difference [s2].

The interpretation offered is that in high-income countries, increased screening, timely diagnosis and effective management — including induction of labour and increased antenatal care — may be responsible for the reduced risk [s2]. In other words, the diagnosis is not protective; the care that follows it may be.

Reading them together

Preventing gestational diabetes reduces a diagnosis whose consequences, in well-resourced systems, are already largely managed. That is not an argument against prevention — gestational diabetes carries burdens beyond stillbirth, for both parent and child, and the diagnosis itself involves intervention that women may reasonably prefer to avoid. But it does complicate the framing in which prevention is sold primarily as a way to prevent the worst outcomes.

The place where the case is strongest is the place with the least evidence. The stillbirth review found no adjusted estimates from low-income countries [s2], and calls for research to identify optimum screening and management strategies in low- and middle-income settings [s2]. The prevention analysis found the smallest benefit among women with low education [s1]. Both gaps point the same way.

What to watch

Whether diagnostic criteria converge internationally, since the same interventions currently produce different measured effects under different definitions, and whether trials of lifestyle prevention are run in the settings where the stillbirth evidence is missing.

Sources

1. [s1] Effects of lifestyle interventions in pregnancy on gestational diabetes: individual participant data and network meta-analysis. BMJ, 6 January 2026. https://doi.org/10.1136/bmj-2025-084159
2. [s2] Gestational diabetes and stillbirth: a systematic review and meta-analysis. eClinicalMedicine, 8 January 2026. https://doi.org/10.1016/j.eclinm.2025.103751