Design: Systematic review and meta-analysis of 9 randomised trials
Population: 1,823 healthy women at 36 weeks or more in the first stage of labour
Primary outcome: Use of regional analgesia
Effect: RR 0.91 (95% PI 0.82 to 1.01); instrumental birth RR 0.78 (95% PI 0.61 to 0.98)

A systematic review and meta-analysis in the Japan Journal of Nursing Science (1 October 2026) pooled nine randomised trials, run between 1994 and 2018 in eight countries, of warm-water bathing in the first stage of labour for 1,823 healthy women at 36 weeks or more with uncomplicated singleton cephalic pregnancies.

Bathing did not clearly reduce the use of regional analgesia (risk ratio 0.91, 95% prediction interval 0.82 to 1.01). It made no detectable difference to perineal trauma, caesarean birth, abnormal fetal heart rate patterns or neonatal infection. Instrumental vaginal birth was less common with bathing (RR 0.78, 95% PI 0.61 to 0.98). Satisfaction leaned towards bathing, but only two small trials measured it.

The authors graded the certainty low to very low, mainly because nothing about a bath can be blinded and several trials were small and old. What the review supports is offering a bath as a comfort measure that appears safe, not promising it will avoid an epidural.

Offer warm-water immersion in the first stage to low-risk women who want it.
Do not present bathing as a way to avoid an epidural; the trials did not show that.
Set and record water temperature, and keep intermittent fetal monitoring going in the bath.
Have a tub cleaning protocol before introducing bathing on a labour ward.

Why it matters

Many units ration baths for want of evidence; this finds no safety signal and a possible fall in instrumental births.

The statistics, in plain English

A prediction interval estimates the range in which the effect in a new setting would probably fall, which is wider and more honest than a confidence interval. For regional analgesia it runs to 1.01, so no effect remains plausible. For instrumental birth it stays below 1, but low-certainty evidence means the true effect could be quite different.