Results 2: primary analysis

Descriptives by parity

A total of 60,402 birth records were eligible for inclusion in the main analysis, of which 39,627 (65.6%) were waterbirths and 20,775 (34.4%) could have remained in the water but left the pool to birth on land. These births were from all low-risk women who had no current or underlying risk factor and had no complications arising in labour. Of the 27,386 nulliparous women, 15,176 (55.4%) had a waterbirth compared with 24,451 (74.1%) of the 33,016 women who had had a previous birth.

Table 13 describes the maternal and infant characteristics of waterbirths and for those who left the pool, for all births and broken down by parity. The two groups were comparable on average maternal age (years), ethnic background, BMI at booking, gestational age at birth and infant birthweight. Differences between the two groups were observed in some maternal characteristics, such as parity, where birthing in water was more prevalent in women who had previously given birth (61.7%) compared to women who had left the pool to birth (41.2%) and also in the prevalence of clinical concern identified shortly prior to birth (0.7% waterbirth vs 9.0% left pool). For this reason, we additionally adjusted for clinical concern in the regression models. Similar patterns in maternal and infant characteristics were observed for waterbirths and for those who left the pool, within the nulliparous and multiparous women with the exception, again, of clinical concern.

Primary outcomes

Maternal outcome: obstetric anal sphincter injury

Main analysis

Of the 60,402 women who were included in this analysis, 60,338 (99.9%) contributed to the analysis with perineal trauma recorded. Fewer nulliparous women in the waterbirth group had a record of OASI than in those who left the pool [730 of 15,176 women (4.8%) vs. 641 of 12,210 women (5.3%); UOR 0.91, one-sided 95% CI, −∞ to 0.99] (Table 14). After adjustment for covariates, the aOR was 0.97 with a one-sided 95% CI, −∞ to 1.08. Similarly, fewer multiparous women in the waterbirth group had a record OASI use than in those who left the pool [269 of 24,451 women (1.1%) vs. 144 of 8,565 women (1.7%); UOR 0.65, one-sided 95% CI −∞ to 0.77; aOR 0.64, −∞ to 0.78]. When pooled for all women, the rate of OASI in women giving birth in water was again lower than in those who left the pool [999 of 39,627 patients (2.5%) vs. 785 of 20,775 patients (3.8%); aOR 0.89, one-sided 95% CI, −∞ to 0.98].

Since the upper limit of the CIs for the difference in the proportion of OASI between the groups were less than the prespecified non-inferiority margin for both nulliparous and multiparous women of 1% (OR < 1.23) and 0.6% (OR < 1.38), respectively, non-inferiority can be concluded (Figure 7). This suggests that birthing in water is no worse than birthing out of water according to OASI status.

FIGURE 7 Forest plot of maternal and infant primary outcomes. a, The infant primary outcome included: any NNU admission requiring respiratory support, intravenous antibiotic administration within 48 hours of birth (with or without culture proven infection) or intrapartum stillbirth and all deaths prior to NNU/postnatal ward discharge. Note: dashed lines represent the non-inferiority margin for each comparison, and shaded areas represent the equivalence zone. (Figure not reproduced here; see the original article.)

Sensitivity analyses

A number of planned sensitivity analyses were performed to assess the robustness of the results to factors that may introduce bias [i.e. imputing missing data, impact of giving birth in water in the absence of any midwife concerns (group 1a vs. 2a) and using propensity score methods]. The main findings were consistent in these prespecified sensitivity analyses (see Table 14).

Instrumental variable analysis

The proportion of women using water for labour or birth at each unit was chosen as an IV due to (1) being associated with exposure (significantly different for births in water compared to those out of water) and (2) not being associated with the rate of OASI. These two conditions only held for nulliparous and the pooled cohort of women; for multiparous women, there was no association between the proportion of women using water for labour and exposure. For nulliparous women, the aOR was 0.96 (one-sided 90% CI −∞ to 1.17).

Secondary analyses

The lower rate of OASI in water for multiparous women and all women was sufficient that, when tested in superiority analysis as prespecified in the SAP, it met the thresholds of statistical significance (aOR 0.64, two-sided 95% CI 0.51 to 0.80, p-value < 0.001, 0.89, 0.80 to 0.997, 0.045, respectively). For nulliparous women, the lower rate of OASI did not meet the thresholds of statistical significance (aOR 0.97, two-sided 95% CI, 0.86 to 1.11, 0.684).

Subgroup analyses

A planned and powered subgroup of the primary maternal outcome was conducted to compare rates of OASI separately for nulliparous and multiparous women. There was evidence of a differential effect for nulliparous and multiparous women’s rates of OASI (interaction p-value = 0.001).

Another planned subgroup was that of the relationship between the proportion of women using a pool during labour at individual sites and the incidence of the adverse maternal primary outcome. There was no association between the rate of site pool use and the rate of OASI and neither did it significantly differ by birthing in water or not (interaction p-value = 0.205).

Infant outcome: adverse infant outcome or treatment

Main analysis

The primary infant outcome of ‘adverse infant outcome or treatment’ was captured in prospective maternal and NNRD data. Of the 16,223 infants born after site opening, 14,946 (92%) contributed to the analysis. Data for four sites were excluded as they did not routinely collect postnatal data and could not therefore contribute to the outcome. Fewer infants in the waterbirth group had an adverse infant outcome than in those who left the pool [263 of 9868 infants (2.7%) vs. 224 of 5078 infants, respectively (4.4%); aOR 0.65; one-sided 95% CI, −∞ to 0.79] (Table 15). The upper limit of the CI for the difference in infant outcome between the groups was less than the prespecified non-inferiority margin of 1.0% (OR < 1.21), suggesting that non-inferiority can be concluded and birthing in water is no worse than birthing out of water (see Figure 7).

Individual components of the infant primary outcome

The individual components of the infant primary outcome are shown in Table 15. Of the infants who experienced an adverse infant outcome in waterbirth (n = 263) and in those who left the pool (n = 224), 3 (1.1%) in the waterbirth group had died, with no deaths in those who left the pool. Fewer infants in the waterbirth group had a NNU admission, requiring respiratory support than in those who left the pool (0.9% vs. 2.0%, aOR 0.44; one-sided 95% CI −∞ to 0.60). All infants were administered with intravenous antibiotics within 48 hours of birth.

Sensitivity analyses

Sensitivity analyses were performed to assess the robustness of the results to factors that may introduce bias. The main findings were consistent in these prespecified sensitivity analyses (Table 15). IVA could not be performed since the assumption of the IV (proportion of women who used a pool by site), being associated with exposure, was not met.

Secondary analyses

The lower rate of adverse infant outcomes in water was sufficient that, when tested in superiority analysis as prespecified in the SAP, it met thresholds of statistical significance (aOR 0.65, two-sided 95% CI 0.52 to 0.82, p-value < 0.001).

An important secondary analysis of the infant primary composite outcome used NNU admissions requiring respiratory support, intrapartum stillbirth and early neonatal death captured over both periods of data collection in both data sources, thus increasing the sample size and power of the analysis. Administration of intravenous antibiotic within 48 hours of birth among babies not admitted to a NNU was not captured over the retrospective data period. The finding was consistent in this prespecified secondary analysis (aOR 0.69, one-sided 95% CI −∞ to 0.78, p-value < 0.001).

Subgroup analyses

A planned (but unpowered) subgroup analysis of the primary infant composite outcome was conducted by parity to compare rates separately for infants born to nulliparous and multiparous women. There was no evidence of a differential effect (of birthing in water compared to out of water) for nulliparous and multiparous women (interaction p-value = 0.932).

Subgroup analyses

A planned and powered subgroup of the primary maternal outcome was conducted to compare rates of OASI separately for nulliparous and multiparous women. There was evidence of a differential effect (of birthing in water compared to out of water) for nulliparous and multiparous women (interaction p-value = 0.001). There was no association between the rate of site pool use and the rate of infant primary outcome and neither did it significantly differ by birthing in water or not (interaction p-value = 0.126).

Secondary outcomes

Maternal intrapartum outcomes

Fewer women in the waterbirth group experienced adverse maternal intrapartum outcomes than those leaving the pool to give birth, suggesting that birthing in water is no worse than birthing out of water according to these outcomes (Table 16).

Management of the third stage of labour

Following waterbirth, active management of the third stage of labour was used less frequently than following births out of water (72.8% vs. 89.0%), and physiological management was used more frequently (27.1% vs. 10.9%); both were non-inferior to birthing out of water (aOR 0.31, one-sided 95% CI −∞ to 0.33; 0.69, −∞ to 0.78, respectively). An important subgroup were women who left the pool prior to delivery of the placenta compared with women who remained in the pool.

Of the 10,760 waterbirths recorded after site opening (where the fields on intention/actual delivery of placenta were added), 910 (8.5%) of midwives intended to deliver the placenta in water, whereas 7422 (69%) intended to deliver out of water. Overall, the placenta was delivered in water following 926 (8.6%) of all waterbirths, including on 58.4% of the occasions when it was intended to deliver the placenta in water. Among women who gave birth in water after site opening for whom data were collected, 37.5% (4031/10,760) received oxytocin for third-stage management before leaving the pool.

Rates and management of postpartum haemorrhage

The PPH (defined as the total blood loss at and after delivery of ≥ 500 ml and also ≥ 1 l) was more prevalent in those who left the water than those staying in water to give birth (13.1% vs. 16.0%; 2.9% vs. 3.8%, respectively). Major obstetric haemorrhage (≥ 1500 ml) was recorded in 1.1% of waterbirths compared to 1.3% among births out of water. For PPH ≥ 1 l (the only tested PPH outcome), the aOR was 0.90 (one-sided 95% CI −∞ to 0.98), indicating that birth in water was non-inferior to birth out of water. For women experiencing a PPH ≥ 500 ml, IV therapy was given in 30.4% of all waterbirths compared to 38.5% of those who left the water to birth [IV fluids: 21% vs. 27.1%; Oxytocin infusion: 19.0% vs. 22.3%; plasma expanders: 4.6% vs. 5.0%, blood transfusions: 1.4% vs. 1.3%, respectively].

Need and reason for obstetric involvement in woman’s care following birth

Evidence of obstetric involvement in woman’s care following birth included a record of obstetric treatment for haemorrhage, manual removal of placenta (MORP) or obstetric repair of a third-/fourth-degree perineal tear or other complex trauma. Obstetric involvement was recorded in fewer mothers in the waterbirth group than those who left the pool (6.5% vs. 9.6%; aOR 0.78, one-sided 95% CI −∞ to 0.82).

Incidence and management of perineal trauma

Perineal trauma was more prevalent in births out of water compared to waterbirths (32% vs. 38%, respectively; aOR 0.84, one-sided 95% CI −∞ to 0.88), specifically for second-, third- and fourth-degree trauma and episiotomies, whereas first-degree traumas were higher in waterbirths (18% vs. 14%). Ordinal regression showed that births out of water have more severe perineal trauma (aOR, 0.72, one-sided 95% CI −∞ to 0.74). The prevalence of genital tears was comparable between the two groups, with the majority (24%) experiencing labial trauma/lacerations. A record of perineal repair was observed in 32% of waterbirths and 44% of births out of water. Perineal trauma was also broken down by parity (Table 17).

Maternal position at birth

The most common birthing position in both in and out of water was semi-recumbent/sitting (40% and 49%, respectively). For waterbirth, this was followed by kneeling or leaning forward (24% and then all fours 21%), both of which were more prevalent than those out of water (7% and 12%, respectively). Most of the other positions were more prevalent in birth out of water (e.g. lateral left and right, lithotomy, use of a birthing chair and McRoberts); squatting was more prevalent in waterbirth (5%), accounting for the partial births.

Incidence of shoulder dystocia and required management

When a shoulder dystocia occurred during a waterbirth, if the baby’s head had been born into water, they remained in the waterbirth group for analysis (group 1a or 1b, depending on whether prior midwife concerns had been present) even if the birth was completed out of water. Shoulder dystocia was recorded in fewer births in water than among births out of water [221 of 39,627 births (0.6%) vs. 651 of 20,775 births (3.1%); aOR 0.16; one-sided 95% CI, −∞ to 0.18]. As non-inferiority was shown, a superiority analysis was performed; this showed lower rates of shoulder dystocia among births in water (aOR 0.16; −∞ to 0.19, p-value < 0.001) compared to births out of water. Information describing the management of shoulder dystocia was not routinely collected for around 20% of events, either due to non-completion, or as these data were not included in the MIS at some sites. For the remainder, there were differences in management between groups (35% of waterbirths with shoulder dystocia used the McRoberts position compared to 69% of shoulder dystocia cases occurring out of water; 10% used suprapubic pressure for waterbirths compared to 29% out of water). In the waterbirth group, 20% of cases of shoulder dystocia were managed by a change of position or exiting the pool only, whereas a change in position was the only management used in 2.6% of case of shoulder dystocia occurring out of water. The head-to-body birth interval was comparable by group (waterbirth 3.5 minutes vs. birth out of water 3.3 minutes).

For the infant primary outcome, a higher rate of adverse infant outcome or treatment was observed following shoulder dystocia than in those without shoulder dystocia (11.2% vs. 3.0%, respectively) (Table 18). Cases of shoulder dystocia had a higher prevalence of NNU admissions and required respiratory support when compared to births without shoulder dystocia (12.4% vs. 4.3%, respectively; 4.2% vs. 1.0%, respectively). The prevalence of therapeutic hypothermia and antibiotics administered within 48 hours of birth were also higher (14.8% vs. 2.5%, respectively; 7.6% vs. 2.0%, respectively). No neonatal deaths resulted from shoulder dystocia and brachial plexus injuries were rare (see Table 18).

Management of the third stage of labour following waterbirth

Of the 10,760 waterbirths recorded after site opening (where the fields on intentional/actual delivery of placenta were added), 910 (8.5%) of midwives intended to deliver the placenta in water, whereas 7422 (69.0%) intended to deliver out of water (Table 19). Overall, the placenta was delivered in water following 926 (8.6%) of all waterbirths, including on 59.5% of the occasions when it was intended to deliver the placenta in water. The prevalence of PPH between waterbirths where the placenta was delivered in water compared to those who left the pool during the third stage is shown in Table 20. The rate of PPH was lower for women who had their placenta delivered in water compared to those leaving the pool during the third stage, but when tested (total blood loss ≥ 1000 ml), there was no evidence of a significant difference (aOR 0.70, two-sided 95% CI 0.42 to 1.18, p-value = 0.181).

Maternal postnatal outcomes

Table 21 reports the maternal postnatal outcomes.

Duration of postnatal stay (hours)

Women who had waterbirths experienced shorter durations of postnatal stay (hours) than for those who left the pool (11.5 vs. 15.5 hours), indicating that waterbirth was non-inferior to birthing out of water (IRR 0.83, one-sided 95% CI −∞ to 0.85) (see Table 20).

Breastfeeding initiation and continuation (at community discharge of care)

The rate of not initiating breastfeeding was 16.2% in waterbirths and was 20.2% in those leaving the pool; and it had increased at community discharge of care to 27.9% and 32.2%, respectively (Table 21). Waterbirth was non-inferior to birthing out of water at both time points (initiation and community discharge of care) (aOR 0.79, one-sided 95% CI −∞ to 0.83; aOR 0.83, one-sided 95% CI −∞ to 0.88, respectively).

Higher-level care

Less than 0.5% of mothers in groups 1 and 2 were transferred to higher level of care (including transfer to critical care, high-dependency unit) (aOR 0.92, one-sided 95% CI −∞ to 1.21).

Maternal re-admission to hospital

Maternal re-admission to hospital or a hospital assessment was recorded in around 1% in women (1.1% waterbirth vs. 1.4% left pool to give birth) (aOR 0.92, one-sided 95% CI −∞ to 1.07). Reasons for admission includes breastfeeding problems, secondary PPH, abdominal wound infection, perineal problems, postdural headaches, urinary incontinence or retention, or for evacuation of retained products of conception.

Infant outcomes

Table 22 reports the infant secondary outcomes.

Snapped umbilical cord prior to clamping

There was a higher rate of snapped umbilical cords prior to clamping in the infants born in water compared to those born out of water (1.0% vs. 0.3%, respectively), with the null hypothesis accepting that waterbirth is inferior to leaving the pool (aOR 3.89, one-sided 95% CI −∞ to 6.88). Table 23 describes key outcomes for infants with snapped cord (n = 122) compared to infants with no record of snapped cord (n = 16,101). All outcomes were more prevalent for infants with snapped umbilical cords (including NNU admissions with respiratory support: 13% vs. 5%); antibiotics were administered within 48 hours of birth, 5.7% vs. 2.3%, but there were no deaths among babies with a snapped cord prior to clamping.

Timing of cord clamping

Cord clamping of < 60 seconds after birth was observed in a low proportion of each group (waterbirth: 5.4% vs. birth out of water: 8.1%), demonstrating that waterbirth was non-inferior to birthing out of water after adjusting for confounders (aOR 0.73, one-sided 95% CI −∞ to 0.85).

Apgar scores

Apgar scores at 1 and 5 minutes were dichotomised into 7–10 (normal) vs. 0–7 (low). Apgar score at 10 minutes was not available from the data. A high proportion of both groups had normal Apgar scores at each time point, with 3.6% vs. 5.1% with low scores at 1 minute and 0.5% vs. 0.9 at 5 minutes (waterbirth and out of water, respectively). After adjustment for confounders, there was evidence to show that waterbirth was non-inferior to birthing out of water at 1 and 5 minutes (aOR 0.72, one-sided 95% CI −∞ to 0.78; 0.62, −∞ to 0.77, respectively).

Resuscitation at birth

Evidence of resuscitation was seen in 4.1% of infants born in water and 6.3% born out of the pool, demonstrating that waterbirth was non-inferior to birthing out of water (aOR 0.61, −∞ to 0.65). The most common method of resuscitation was stimulation/suction bag and mask ventilation (basic resuscitation), with the remaining using intubation, drug administration or cardiac massage.

Intrapartum stillbirth or deaths prior to neonatal unit/postnatal ward discharge and cause

Intrapartum or neonatal deaths were reported for 13 infants: 7 following waterbirth (0.18 per 1000 births) and 6 (0.29 per 1000 births) among babies born out of water, with evidence of non-inferiority (aOR 0.22, one-sided 95% CI −∞ to 0.80). There were no records that suggested the babies who died and did so from a congenital anomaly.

Skin-to-skin contact at birth

Skin-to-skin contact between the infant and mother or father was recorded in the majority of births; 5.6% of all waterbirths had no record of skin-to-skin contact compared to 7.6% in those leaving the pool. This suggests that waterbirths are non-inferior to birthing out of water (aOR 0.74, one-sided 95% CI −∞ to 0.80).

First breastfeed within first hour

There were 17.6% of waterbirth babies who were not breastfed within the first hour compared to 27.7% babies born out of water; waterbirths are non-inferior to birthing out of water (aOR 0.63, one-sided 95% CI −∞ to 0.70).

Neonatal unit admissions and respiratory support

The NNU admissions were observed in 3.7% of waterbirths compared to 5.8% of births out of water, with the average length of stay for NNU admissions in the waterbirth group being 48 hours compared to 50.3 hours in births out of water. Respiratory support was provided in 0.8% of waterbirths and in 1.5% of births out of water. For all outcomes, waterbirth was found to be non-inferior to birthing out of water [aOR 0.66, one-sided 95% CI −∞ to 0.71; IRR 0.88, one-sided 95% CI −∞ to 0.94; aOR 0.58, one-sided 95% CI −∞ to 0.68, respectively].

Therapeutic hypothermia, brachial plexus injury, treatment for jaundice, successful/attempted lumbar puncture

Therapeutic hypothermia was recorded in 0.07% of waterbirths and in 0.26% of births out of water (aOR 0.33, one-sided 95% CI −∞ to 0.53). No brachial plexus injuries were recorded in the data for waterbirth and four were recorded in births out of water. Treatment for jaundice was observed in 1.1% of waterbirths and in 1.7% of births out of water (aOR 0.75, one-sided 95% CI −∞ to 0.86). Successful or attempted lumbar punctures were recorded in 0.5% of waterbirths and in 0.8% of births out of water (aOR 1.00, one-sided 95% CI −∞ to 1.62).

Administration of intravenous antibiotics, including timing and duration

From the NNRD records, antibiotics were administered within 48 hours of birth for 1.8% (n = 629) of all waterbirths and 2.9% (n = 535) for infants born out of water, with similar rates recorded in the MIS data from site opening (n = 181, 1.8% and n = 149, 2.9%, respectively). Being born into water was found to be non-inferior to being born out of water (aOR 0.69, one-sided 95% CI −∞ to 0.77; aOR 0.74, one-sided 95% CI −∞ to 0.94, respectively). The average duration of antibiotic administration for infants admitted to NNU administered within 48 hours of birth was 3 days (IQR: 2–5) and 4 days (3–6), respectively.

Highest C-reactive protein and blood culture positive with a recognised pathogen

Among the 1164 (waterbirth: n = 629; out of water: n = 535) babies who had been recorded on the NNRD as those who received IV antibiotics within 48 hours of birth, blood culture results were extracted from the NNRD. Blood cultures were sent for 84 (13.4%) and 61 (11.4%) of waterbirths and births out of water, respectively, of which 4 (4.8%) and 2 (3.3%), respectively, had growth recorded (either clear or unclear pathogenicity).

After site opening, data relating to CRP and blood culture results were also captured in the MIS.

Overall, 2.0% (n = 329/16,223) of babies born after site openings received IV antibiotics within 48 hours of birth; the mean highest CRP result was similar among babies born into water and out of water (8.6 mg/dl vs. 8 mg/dl, respectively). Of the 374 blood cultures recorded as being sent after site opening, four babies had a blood culture performed and had a growth confirmed infection (one baby born in water, and three babies born out of water).

Re-admission to hospital within 7 days of birth

As babies requiring re-admission to hospital are usually admitted to a ward outside of maternity, infant re-admissions to hospital were not reported in the MIS or NNRD data received.