REF-1712 · Li et al., 2026 · British Journal of Sports Medicine, · Paper · PubMed

Supervised exercise in pregnancy lowered gestational diabetes by about a quarter: in a meta-analysis of 19 randomised trials (6,213 women; moderate certainty) supervised exercise against usual care cut gestational diabetes (RR 0.74, 95% CI 0.57-0.96), gestational hypertension (RR 0.55, 0.40-0.77), excessive weight gain (RR 0.75, 0.61-0.92) and macrosomia (RR 0.61, 0.46-0.81); subgroups pointed to larger effects with an early start or more than 20 weeks, adherence of 80% or more, combined aerobic and resistance training and at least 150 min a week, and the weight-gain benefit peaked near 500-550 MET-min a week

Main caveat: Not in the abstract: the risk-of-bias tool and its results, certainty per outcome, the exercise protocols and intensities, results for pre-eclampsia, caesarean…
fairly strong evidenceMeasured on non-athletes and recreational peopleRead from the abstract only

Li and colleagues asked whether supervised exercise in pregnancy prevents gestational diabetes and other adverse outcomes, and whether more exercise does more.

They searched four databases to November 2025 for randomised trials comparing supervised exercise with usual care that reported gestational diabetes, and extracted gestational hypertension, pre-eclampsia, excessive gestational weight gain, macrosomia, caesarean delivery, preterm birth and low birth weight where available.

Nineteen trials with 6,213 women were included, with moderate certainty of evidence overall. Supervised exercise lowered the risk of gestational diabetes (RR 0.74), gestational hypertension (RR 0.55), excessive gestational weight gain (RR 0.75) and macrosomia (RR 0.61).

Subgroups suggested larger effects with exercise started early in pregnancy or lasting more than 20 weeks, adherence of 80% or more, combined aerobic and resistance training and at least 150 min a week; for gestational diabetes, the subgroup differences by start time (p = 0.009) and duration (p = 0.010) were significant.

The dose-response analysis put the peak benefit for excessive weight gain at about 500-550 MET-min a week, with a plateau at higher doses. The authors conclude that supervised exercise lowers adverse pregnancy outcomes, most when it begins early, is kept up, stays at a moderate dose and mixes aerobic with resistance work.

Participants

19 RCTs, 6,213 pregnant women, 48.6% in the intervention groups.
Age, BMI, parity, risk status and prior activity are not in the abstract.

Protocol

Search: four databases (not named in the abstract) to November 2025.
Inclusion: RCTs in pregnant women comparing supervised exercise with usual care and reporting GDM.
Outcomes: GDM (primary); gestational hypertension, pre-eclampsia, excessive gestational weight gain, macrosomia, caesarean delivery, preterm birth and low birth weight (secondary, where available).
Analysis: random-effects meta-analysis (risk ratios), subgroup analyses (start time, duration, adherence, exercise type, weekly duration), meta-regression and dose-response analysis in MET-min a week; certainty rated (overall moderate).
Not in the abstract: the risk-of-bias tool and results, the exercise protocols, results for pre-eclampsia, caesarean delivery, preterm birth and low birth weight, registration and funding.

Key Results

Outcome (supervised exercise vs usual care)RR (95% CI)I²
Gestational diabetes (primary)0.74 (0.57 to 0.96)47%
Gestational hypertension0.55 (0.40 to 0.77)1%
Excessive gestational weight gain0.75 (0.61 to 0.92)70%
Macrosomia0.61 (0.46 to 0.81)27%
Subgroups (the four outcomes above): greater effects with intervention started in early pregnancy or lasting more than 20 weeks, adherence ≥ 80%, combined aerobic and resistance training, and ≥ 150 min a week; for GDM the subgroup differences were significant for start time (p = 0.009) and duration (p = 0.010).
Dose-response: peak effect on excessive weight gain at about 500-550 MET-min a week, plateauing at higher doses.
Not reported in the abstract: results for pre-eclampsia, caesarean delivery, preterm birth and low birth weight.

Practical Application

Supervised exercise through pregnancy lowers the risk of gestational diabetes and high blood pressure. Across 19 trials and 6,213 women, it cut gestational diabetes by about a quarter (RR 0.74), gestational hypertension by about 45% (RR 0.55), excessive weight gain (RR 0.75) and macrosomia (RR 0.61), at moderate certainty; starting early, keeping it up, mixing aerobic and strength work and doing at least 150 min a week went with larger effects.
Consistent with [SYNTH-95]'s default: keep training, aerobic and strength, once the obstetric team agrees (REF-567), and at least 150 min of moderate activity a week as a public-health floor (REF-568). This review adds trial evidence for that floor and for combined training, in women who were not described as athletes; it changes no line of the guide and sets no ceiling for a trained woman.
Inference, not tested here: a woman who already trains and is cleared to continue may already be above 150 min a week and above the 500-550 MET-min point where the weight-gain benefit levelled off; for her the features this review links to larger effects are an early start and keeping training going through pregnancy. Subgroup and dose-response findings compare trials, so they show which programmes went with larger effects, not that the dose caused them.
Not treatment advice: whether and how much to exercise in a given pregnancy, and any question of diabetes, blood pressure or the baby's size, is decided by the obstetric provider; warning signs stop the session and go to them (SYNTH-95).

Limitations

Not in the abstract: the risk-of-bias tool and its results, certainty per outcome, the exercise protocols and intensities, results for pre-eclampsia, caesarean delivery, preterm birth and low birth weight, and funding.
Heterogeneity: I² 47% for GDM and 70% for excessive weight gain.
Dose and timing: subgroup, meta-regression and dose-response findings across trials are associations between programme features and effects, not tested comparisons within trials.
Comparator: usual care, so part of the effect may come from supervision and contact rather than exercise alone (our reading).
Written from the abstract and the PubMed record; full text not read.
How this entry was checked and changed
Why this level

a meta-analysis of randomised trials whose abstract reports its certainty assessment and the result (moderate), so the abstract-only cap does not apply; held at A- because the risk-of-bias tool and its findings, the per-outcome certainty and the outcomes reported as not significant are not in the abstract, heterogeneity reaches I² 70% for excessive weight gain, and the timing, adherence and dose findings come from subgroup and dose-response analyses across trials