REF-1783 · Zhang et al., 2026 · Birth · Paper · PubMed

Supervised moderate exercise three times a week through an uncomplicated pregnancy went with fewer caesarean sections (12.6% against 28.9%) and no more miscarriages than in the control group, in a randomised trial of 256 women analysed per protocol (120 exercise, 136 control; only women with at least 75% attendance kept in the exercise group); no confidence interval in the abstract, and a rate in one trial, not a promise for one woman; from the abstract

Main caveat: Per-protocol only: exercisers below 75% adherence were excluded from the analysis.
moderate evidenceMeasured on non-athletes to trained athletesRead from the abstract only

Zhang, Barakat and colleagues randomised women with uncomplicated pregnancies to a supervised group exercise programme or to a control group, and examined delivery outcomes with particular attention to caesarean section and miscarriage. Of 400 women assessed for eligibility, 256 were analysed: 120 in the exercise group and 136 controls.

The exercise group trained three times a week for 60 min, at moderate intensity, in a multicomponent programme delivered both face-to-face and online, from 8-10 weeks of gestation until 38-40 weeks. Only exercisers who attended at least 75% of sessions were kept in the per-protocol analysis.

Among women who completed the pregnancy, the caesarean rate was 12.6% with exercise against 28.9% in controls (χ² = 10.42, p = 0.005). Exercise was not associated with more miscarriages (p = 1.000), and instrumental vaginal delivery and the other maternal and neonatal outcomes did not differ significantly.

The authors conclude that the programme reduced caesarean rates without raising miscarriage risk and recommend it in routine prenatal care.

Participants

400 assessed, 256 analysed: exercise 120, control 136.
Uncomplicated pregnancies; the eligibility criteria, age, parity, body mass, previous caesarean, pre-pregnancy activity and the number randomised are not in the abstract.

Protocol

Exercise group: supervised, moderate-intensity, multicomponent, group-based, hybrid (face-to-face and online); 3 sessions a week of 60 min; from 8-10 to 38-40 weeks of gestation.
Control group: what controls received is not in the abstract.
Analysis: per protocol; exercisers needed at least 75% adherence to be included; the delivery outcomes are reported among women who completed the pregnancy.
Not in the abstract: how intensity was set and monitored, the exercise components, randomisation and allocation concealment, blinding, how the caesarean decision was made, the number of miscarriages per group, and the number lost or excluded per arm.

Key Results

OutcomeExerciseControlTest
Caesarean section (women who completed pregnancy)12.6%28.9%χ² = 10.42, p = 0.005
Absolute difference (the base's arithmetic, not in the abstract)16.3 percentage points fewer with exerciseno confidence interval in the abstract
Miscarriagenot givennot givenp = 1.000 (no association)
Instrumental vaginal deliverynot givennot givennot significant
Other maternal and neonatal outcomesnot givennot givennot significant (outcomes not named)
The abstract gives percentages without counts or denominators for each group.

Practical Application

In uncomplicated pregnancies, supervised moderate exercise three times a week from early pregnancy to term went with fewer caesarean sections (12.6% against 28.9%) and no rise in miscarriage, in one randomised trial analysed per protocol — a group rate, not a promise that any one woman will avoid a caesarean. Whether a caesarean is needed depends on many things this trial did not control for each woman.
Consistent with [SYNTH-95]: the guide keeps a cleared woman with an uncomplicated pregnancy training; in this trial regular moderate exercise was not associated with more miscarriages (p = 1.000; counts not in the abstract), and its caesarean result points the same way as the meta-analysis of 88 trials in the base (RR 0.86; REF-1710), though the effect here is larger than the pooled one.
The size of the effect is uncertain: only the women who kept up 75% of sessions were analysed in the exercise group, no interval is given, and the per-protocol design can enlarge a difference between groups (an inference from the design; the abstract gives no intention-to-treat result).
Inference, not tested here: it says nothing about vigorous or high-volume training, which an athlete may be doing; for volume see REF-1782, for vigorous work into the third trimester REF-1396.
Not treatment advice. Clearance to exercise in pregnancy, and the mode of delivery, are decided by the obstetric provider (SYNTH-95; REF-567).

Limitations

Per-protocol only: exercisers below 75% adherence were excluded from the analysis; the abstract gives no intention-to-treat result, no number randomised and no dropouts per arm, so the comparison may favour the women most able to keep training.
No precision: no confidence interval, counts or denominators for the caesarean rates; miscarriage counts are not given, so "no increase" rests on a p value only. The reported χ² of 10.42 with p = 0.005 fits a test with two degrees of freedom (with one it would give about p = 0.001), so the test may compare the whole spread of delivery modes rather than caesarean alone (the base's arithmetic; the abstract does not say).
Not in the abstract: randomisation, allocation concealment and blinding (participants cannot be blinded to exercise), what controls did, parity and previous caesarean, how the decision for a caesarean was made, the other outcomes tested, and funding and conflicts of interest.
Transfer: moderate supervised group exercise in uncomplicated pregnancies of women whose training status is not reported; not vigorous, high-volume or sport-specific training, and not athletes.
How this entry was checked and changed
Why this level

a parallel randomised trial with more than 100 analysed per arm, risk of bias not read (abstract only); a trial reaches A or A- only when read in full; the analysis is per protocol with a 75% adherence cut, and the number randomised and the dropouts per arm are not in the abstract