Delayed versus early pushing in women with epidural analgesia: a systematic review and meta-analysis



Epidural analgesia is highly effective in relieving the pain of labour and childbirth, but it also interferes with the normal mechanism of labour.1,2 Trials of delayed pushing have occurred in response to concerns about the association between epidural analgesia and unwanted and potentially harmful outcomes, particularly instrumental delivery.1 Instrumental deliveries, especially forceps, have been associated with increased risk of urinary and faecal incontinence, sexual problems and organ prolapse.3–8 The mechanism for the association between epidural analgesia and increased instrumental deliveries is likely to be multifactorial but may include a weakened desire to push due to diminution of the bearing down reflex, reduced uterine activity and altered clinical practice.9–12 Contemporary obstetric practice has women begin pushing as soon as the cervix is fully dilated. Delaying the onset of pushing has been proposed as an alternative that may allow spontaneous descent and rotation of the fetal head, thereby reducing the instrumental delivery rate. On the other hand, delayed pushing prolongs the second stage of labour and this too has been implicated in pelvic floor trauma and subsequent maternal morbidity.6,13 Further, second stage is considered to be a time of particular risk to the fetus.14 A systematic review and meta-analysis of three small studies published in 1992 found both a tendency towards decreased perineal trauma and a decrease in rotational forceps with a policy of delayed pushing, however, there were insufficient data on infant outcomes and no data on pelvic floor morbidity.15 Since then several more trials of delayed pushing have been published and we undertook to update the systematic review and meta-analysis with the aim of assessing the effectiveness of delayed pushing among women with epidural analgesia in reducing instrumental deliveries and on other measures of maternal and infant morbidity. We aimed to compare the potential benefits and harms of a policy of delayed pushing among women with uncomplicated pregnancies and effective epidural analgesia established in the first stage of labour.


We identified relevant studies by searching MEDLINE, EMBASE and CINAHL databases up to October 2003. We also searched the Cochrane Central Register of Controlled Trials. The search terms included: obstetric analgesia, epidural, extradural, intrathecal, labour (or labour), push(ing) or bear(ing). We supplemented our search by cross checking the reference lists of published papers. No attempt was made to identify unpublished studies.

Randomised controlled trials of delayed pushing in labouring women who had effective epidural analgesia in the first stage was the pre-specified eligibility criteria. Quasi-randomised designs, such as alternate allocation and use of record numbers, were excluded. Two reviewers independently assessed each study for inclusion in the review. Each study was also assessed for allocation concealment, loss to follow up and intention-to-treat analysis. Blinding was not possible. Discrepancies were resolved by consensus.

All outcomes were pre-specified. The primary outcome was any instrumental delivery. Secondary maternal outcomes included rotational or mid-pelvic instrumental deliveries, second stage caesarean section, length of second stage (time from full dilatation to delivery), duration of pushing (time from commencement of pushing until delivery), episiotomy, perineal laceration (second, third or fourth degree tears), postpartum haemorrhage (estimated blood loss >500 mL), maternal fever, satisfaction with labour care and longer term outcomes including urinary or faecal incontinence and sexual problems. Infant outcomes included Apgar scores, umbilical arterial pH, need for positive pressure ventilation, admission to neonatal intensive care (NICU), birth trauma and perinatal death.

We also pre-specified subgroup analyses where clinical heterogeneity might be expected including by: parity (first birth and second or subsequent births), type of epidural analgesia (intermittent bolus, continuous infusion or combined spinal epidural), type of analgesic agent (local anaesthetic alone or local anaesthetic plus opioid) and high study quality (adequate allocation concealment, <20% loss to follow up and intention-to-treat analysis). The review of eligible trials resulted in the identification of several other potential sources of clinical heterogeneity. Consequently, we undertook several post hoc analyses: immediate pushing in the control group versus‘early pushing’ (within 1 hour of full dilatation); duration of pushing limited to 1 hour and then instrumental delivery versus no stated limits; and comparative use of forceps and vacuum extraction (comparatively high vs comparatively low forceps use as assessed by the vacuum to forceps ratio in each study).9

Data were independently extracted from each paper by two reviewers on to a standard data extraction form. We performed statistical analyses using STATA (STATA statistical software version 7.0, STATA, College Station, USA). Where data were missing (incomplete follow up on all women), we used the results reported in the studies as the numerator and denominator. For each dichotomous outcome of interest within individual studies, relative risks (RR) and 95% confidence intervals (CIs) were calculated according to the intention to treat. For continuous variables, the weighted mean differences and 95% CIs were calculated. The assumption of homogeneity of treatment effect between studies was tested using Cochran's Q test statistic and the I2 test.16 Overall estimates of effect were calculated with a fixed effect model (Mantel–Haenszel), but if the assumption of homogeneity was rejected (P < 0.1), a random effects model (DerSimonian) was used.17


The search strategy yielded 131 citations. Of these 117 were excluded (52 duplicate citations and 65 did not meet the eligibility criteria), leaving 14 potentially eligible studies. Five of these were subsequently excluded (four before–after studies18–21 and one quasi-randomised trial22), leaving nine eligible trials with a total of 2953 participants.12,23–30 One of these trials reported the results for nulliparous and multiparous women separately so these have been analysed separately in the meta-analysis.27 All studies included only women at term with spontaneous or induced labour and effective epidural analgesia giving birth to a single, cephalic-presenting fetus. Additional inclusion criteria are outlined in Table 1. Most studies specified that women with major obstetric complications (such as hypertension, diabetes, fetal complications or indications for a short second stage of labour) were excluded. Seven of the studies were limited to nulliparous women (Table 1). Eight of the studies compared delayed pushing with immediate pushing from full dilatation while one30 used early pushing (within 1 hour of full dilatation) as the control group. The duration of the delay until pushing in the experimental group varied between the studies, as did the baseline instrumental delivery rate in the control group and the overall vacuum to forceps ratio (Table 1).

Table 1.  Characteristics of the included randomised controlled trials of delayed versus early pushing among women with epidural analgesia.
Study period and locationEpiduralStudy sizeInclusion criteriaDelay in pushingOVD rateVacuum/forceps*Second stage management
  1. OVD = operative vaginal delivery (instrumental deliveries) in the control group; PCEA = patient controlled epidural analgesia; FHR = fetal heart rate; FBS = fetal blood sampling.

  2. *Vacuum to forceps ratio.

Buxton et al.23 1986 UK Single centreIntermittent bolus41Parity < 4 17–35 yearsUp to 3 hours or until head visible at introitus37%All forcepsOxytocin as required in first and second stage
Fitzpatrick et al.24 1998–1999 Ireland Single centreContinuous infusion; 0.1% bupivacaine and 2 μm/mL fentanyl178Nulliparae1 hour39%1:1.1Active management of labour Maximum for active pushing 1 hour then decision regarding need for instrumental or caesarean section delivery. No rotational forceps
Fraser et al.25 1994–1996 Canada MulticentreContinuous infusion; 0.125% bupivacaine and 2 μg/mL fentanyl1862Nulliparae≥2 hours40%1:0.9Stratified by oxytocin use in first stage + standardized indications for oxytocin use in second stage
Goodfellow and Studd26 UK Single centreIntermittent bolus of 0.25% bupivacaine37Nulliparae > 157cm Receiving oxytocinHead visible at introitus or 1 hour75%All forcepsCo-intervention of increased oxytocin only in the delayed pushing group After 1 hour of pushing forceps were applied unless spontaneous delivery was imminent
Hansen et al.27 USA Single centreContinuous infusion; bupivacaine252All paritiesHead visible at introitus or 1 hour (multiparas), 2 hours (nulliparas)22%Nullipara 1:5.5 Multipara 1:3.4Coached Valsalva directed pushing
Mayberry et al.12 1996 USA MulticentreContinuous infusion; bupivacaine and fentanyl153Nulliparae1 hour, earlier if involuntary urge to push29%Not availableBreath-holding pushing no longer than 6–8 seconds; adequate contraction pattern, change bed positions every 20–30 minutes
McQueen and Myrlea28 UK Single centreNot reported99Not reportedUntil head visible at introitus or rotation and descent ceased56%All forcepsAll women internally monitored
Plunkett et al.29 1999 UK Single centrePCEA with 0.0625% bupivacaine and 2 μg/mL fentanyl202NulliparaeUntil strong urge, or 90 minutes if no strong urge19%Not availableActive management of labour Epidural continued throughout second stage and was only decreased at discretion of anaesthetist
Vause et al.30 1993–1996 UK Single centreNot reported135NulliparaeUp to 3 hours, earlier if head visible at introitus43%1:2.9Unrestricted use of oxytocin, continuous FHR monitoring in all women, FBS available

All studies were unblinded randomised controlled trials that used intention-to-treat analyses. Only one study reported adequate allocation concealment with central randomisation,25 four used opaque envelopes12,24,29,30 and allocation concealment was not reported for the remaining four studies.23,26–28 In four studies, randomisation occurred at full dilatation,23–26 in four randomisation occurred during the first stage of labour12,27,29,30 and the timing was not reported for one study.28 Three studies enrolled women before full dilatation27,29,30 and one of these reported 19% exclusions for medical indications or caesarean section in the first stage of labour.27 One small study excluded infants weighing >4 kg but the number excluded was not reported.26

The only outcome measured in all nine studies was instrumental deliveries. There was a non-significant reduction in instrumental deliveries in six of the nine studies. Meta-analysis showed an overall small reduction which failed to reach statistical significance (Fig. 1). However, meta-analysis of the results from studies which measured the rates of rotational or mid-pelvic instrumental deliveries showed a 31% reduction which was statistically significant (Fig. 1). There was also a non-significant reduction in the rates of second stage caesarean section (Fig. 1). Because the trial by Fraser et al. contributed ≥68% of the weight of these three outcomes, we undertook a sensitivity analysis excluding the Fraser trial. The consequent pooled results were consistent with the overall findings: instrumental delivery (RR 0.94, 95% CI 0.80 to 1.11); rotational or mid-pelvic instrumental deliveries (RR 0.59, 95% CI 0.36 to 0.98); and second stage caesarean section (RR 0.49, 95% CI 0.23 to 1.04). Overall, the reduction in operative deliveries was offset by a significant increase in spontaneous vaginal births (RR 1.22, 95% CI 1.05 to 1.42).

Figure 1.

Delivery outcomes for delayed versus early pushing among women with epidural analgesia.

Seven studies reported either the mean or median duration of the second stage of labour and the duration of pushing (Table 2). The duration of second stage was significantly longer among women allocated to the delayed pushing group in eight studies individually,23–30 but the impact on duration of pushing varied (Table 2). Delayed pushing was associated with a significant decrease in the duration of pushing in three studies,25,27,30 a significant increase in nulliparas in the Hansen study27 and a non-significant decrease in three studies23,24,29(Table 2). Only studies reporting means could be pooled and both these outcomes showed significant heterogeneity (P < 0.01). Using a random effects model, there was a statistically significant increase in the duration of second stage of labour (58 minutes) but no difference in the duration of pushing associated with delayed pushing (Table 3).

Table 2.  Duration of second stage and of pushing in minutes.
StudyOutcome measureResultsP
Delayed pushingEarly pushing
  1. NS = (reported as) not significant.

Duration of second stage
Buxton et al.23Mean [standard deviation]209 [81]118 [50]<0.001
Fitzpatrick et al.24Median (interquartile range)120 (57–225)60 (0–148)<0.001
Fraser et al.25Median (10th–90th centile)187 (86–314)123 (49–248)<0.001
Hansen et al. (nullips)27Mean [standard deviation]171 [57]76 [41]<0.001
Hansen et al. (multips)27Mean [standard deviation]63 [32]24 [23]<0.001
Mayberry et al.12Mean [standard deviation]120 [65]106 [73]0.225
Plunkett et al.29Median (interquartile range)99 (48–160)69 (42–135)<0.05
Vause et al.30Median (interquartile range)214 (149–252)119 (89–155)<0.002
Duration of pushing
Buxton et al.23Mean [standard deviation]79 [44]81 [48]>0.05
Fitzpatrick et al.24Median (interquartile range)56 (8–130)60 (0–148)0.37
Fraser et al.25Median (10th–90th centile)68 (17–175)110 (37–228)0.0001
Goodfellow and Studd26Median (range)42 (1–87)66 (45–75)NS
Hansen et al. (nullips)27Mean [standard deviation]58 [44]76 [41]0.021
Hansen et al. (multips)27Mean [standard deviation]13 [14]24 [23]<0.001
Plunkett et al.29Median (interquartile range)57 (34–126)62 (33–112)>0.05
Vause et al.30Median (interquartile range)52 (31–90)73 (48–115)0.026
Table 3.  Summary of other maternal and infant outcome measures.
OutcomeStudies*No. of subjectsStatistical measure (model type)Effect size (95% CI)
  • RR = relative risk; WMD = weighted mean difference.

  • *

    Bracketed studies recorded the outcome but had no events.

  • Specifies whether a fixed or random effects model was used.

  • 1-minute Apgar <723,25 or <526.

  • §

    5-minute Apgar <723,29 or <825.

Maternal and delivery outcomes
Duration of second stage (minutes)12,23,27446WMD (Random)58.2 (21.51 to 94.84)
Duration of pushing (minutes)23,27293WMD (Random)1.11 (−20.19 to 22.40)
Episiotomy23–25,302209RR (Fixed)0.97 (0.88 to 1.06)
Perineal laceration12,24,25,29,302530RR (Fixed)0.90 (0.70 to 1.17)
Postpartum haemorrhage25,29,302199RR (Fixed)1.04 (0.86 to 1.26)
Intrapartum maternal fever25,292064RR (Random)1.36 (0.68 to 2.73)
Postpartum maternal fever251862RR (Fixed)0.91 (0.33 to 1.96)
Dyspareunia at 3 months postpartum24162RR (Fixed)1.15 (0.63 to 2.10)
Faecal incontinence at 3 months postpartum24178RR (Fixed)1.47 (0.94 to 2.29)
Maternal satisfaction with labour care24178RR (Fixed)0.97 (0.82 to 1.13)
Infant outcomes
Low Apgar at 1 minute23,25,261939RR (Fixed)0.96 (0.74 to 1.24)
Mean Apgar at 1 minute27249WMD (Fixed)0.07 (−0.20 to 0.33)
Low Apgar at 5 minutes§(23),25,292104RR (Fixed)0.82 (0.50 to 1.36)
Mean Apgar at 5 minutes27402WMD (Fixed)−0.01 (−0.09 to 0.08)
PPV for resuscitation25,26,302032RR (Fixed)1.12 (0.80 to 1.57)
Admission to NICU24,25,29,302375RR (Fixed)1.00 (0.70 to 1.42)
Umbilical artery pH23,27,29411WMD (Random)0.03 (−0.01 to 0.06)
Infant trauma251862RR (Fixed)0.90 (0.66 to 1.22)
Perinatal death(30),251997RR (Fixed)4.95 (0.24 to 102.90)

Only two studies reported intrapartum fever.25,29 Plunkett et al.29 found no significant difference (RR 0.93, 95% CI 0.54 to 1.61) but Fraser et al.25 found that women in the delayed pushing group were significantly more likely to have intrapartum fever (RR 1.88, 95% CI 1.31 to 2.71), and these findings were significantly heterogeneous. The pooled RR (using a random effects model) was for a non-significant increase in risk (Table 3).

There were no other statistically significant maternal morbidity results, with similar rates among delayed and early pushing groups for episiotomy, perineal lacerations, postpartum haemorrhage and maternal satisfaction with labour care (Table 3). Only one study reported maternal pelvic floor morbidity at three months finding non-significant increases in faecal incontinence and dyspareunia (Table 3). No study reported rates of urinary incontinence.

Few of the studies reported infant outcomes, with admission to NICU the most commonly reported outcome from four trials (Table 3). There were no significant differences following delayed or early pushing in Apgar scores, PPV resuscitation, umbilical artery pH, admission to NICU, infant trauma or perinatal death.

Subgroup analyses were limited to the primary outcome, instrumental deliveries (Table 4). The only statistically significant reduction in instrumental deliveries was in the pre-specified subgroup of studies where the epidurals contained local anaesthetic (bupivacaine) alone.26,27 In the post hoc analyses, the relative risk of instrumental delivery associated with delayed pushing was lower in the subgroup with high forceps rates,23,26–28,30 but this finding was not significantly different to either the overall result or the low forceps rate subgroup (Table 4).

Table 4.  Summary of subgroup analyses for any instrumental delivery.
Subgroup analysesStudies*No. of subjectsRelative risk (95% CI)
  • *

    Studies contributing to the pooled result.

  • Post hoc analyses.

Overall12,23–3029530.92 (0.84 to 1.01)
Nulliparas12,24–27,29,3026930.92 (0.83 to 1.01)
Continuous infusion epidural12,24,25,27,2926410.92 (0.83 to 1.02)
Intermittent bolus epidural23,26781.07 (0.71 to 1.61)
Local anaesthetic epidural26,272830.60 (0.40 to 0.90)
Local anaesthetic + opioid epidural12,24,25,2923950.94 (0.85 to 1.04)
Immediate pushing control group12,23–2928180.92 (0.84 to 1.02)
Exclude pushing ≤1 hour12,23,25,27–3027380.92 (0.83 to 1.01)
Relatively low forceps use24,2520400.93 (0.83 to 1.04)
Relatively high forceps use23,26–28,305580.84 (0.68 to 1.04)


This systematic review of nine randomised controlled trials comparing delayed versus early pushing in nearly 3000 women with epidural analgesia indicates a significant reduction in rotational or mid-pelvic instrumental deliveries. There was also a tendency towards reductions in overall instrumental deliveries and in second stage caesarean sections. However, the benefits come at the expense of an increase in the duration of the second stage of labour, attributable to an increase in the passive phase. The findings were consistent across meta-analysis of all trials, the single large trial alone and meta-analysis of the small trials.

The reduction in overall instrumental deliveries (our primary outcome) was not statistically significant and is of arguable clinical significance. For example, if the instrumental delivery rate among women with epidural analgesia is 38% (combined rate from the control groups), a policy of delayed pushing might reduce this rate to 35% (95% CI 32% to 38%). The only pre-specified subgroup where the reduction of instrumental deliveries was statistically significant and marked (RR 0.60) was where the epidural contained bupivacaine alone (generally indicating a higher dose of bupivacaine). Randomised trials comparing high dose bupivacaine (0.25%) with low dose bupivacaine (≤0.125%) and opioid epidurals found that the low dose bupivacaine/opioid epidurals reduced the risk of instrumental deliveries.31,32

There was also a trend to a greater reduction in instrumental deliveries in the post hoc analysis of studies with high forceps rates compared with vacuum rates. The vacuum to forceps ratio has been recommended as a useful tool for comparing operative delivery practices across different populations as it remains stable when denominators vary.9 This ratio could be determined for four studies overall24,25,27,30 and another three studies reported only forceps use for instrumental deliveries.23,26,28 Of three studies24,25,30 where the vaccum/forceps ratio could be calculated by delayed and early pushing groups, two had similar ratios for the delayed and early pushing arms.25,30 However, in one trial it differed, 1:1.3 in the delayed pushing group and 1:0.9 in the immediate pushing group, suggesting that in this trial the intervention may have been associated with other changes in obstetric practices.24 Unblinded studies are subject to this type of contamination.

The greatest potential benefit of a policy of delayed pushing for women with epidural analgesia is the decrease in rotational and mid-pelvic instrumental deliveries without evidence of an increase in second stage caesarean sections. Indeed, there is a possibility that delayed pushing may decrease second stage caesarean sections, although this tendency was not statistically significant and a small increase could not be ruled out. The potential impact of delayed pushing on mid-pelvic procedures needs to be considered in the context of declining use of rotational forceps deliveries.9 However, the single large trial in this systematic review found delayed pushing more likely to reduce mid-pelvic vacuum extractions, although mid-pelvic forceps were more commonly used than mid-pelvic vacuum extraction.25 A greater impact in reducing vacuum extraction may explain the small, non-significant effect of delayed pushing on perineal tears and episiotomies as these outcomes are more strongly associated with forceps deliveries.33

A potential downside of a policy of delayed pushing among women with epidural analgesia is the statistically significant and clinically important 58 minute increase in the duration of second stage of labour and consequent increase in the time spent in the staff-intensive delivery suite. As there was no significant difference in the duration of pushing, this increase was in the passive phase of the second stage. Pooled analyses of both of these findings showed significant heterogeneity which may reflect differences in other aspects of second stage management. Although one or both of these outcomes were reported by eight12,23–27,29,30 of the nine trials, data could only be pooled for four studies for the duration of the second stage and for three studies for the duration of pushing. The other studies reported median durations. While the median may be the most appropriate measure for outcomes which exhibit skewness (such as the duration of second stage), additional reporting of the mean would allow the opportunity of pooling in a meta-analyses.

Injury to the pelvic floor and anal sphincter in childbirth can be mechanical, neurological or both.34 In addition to instrumental deliveries, observational studies assessing predictors of injury have identified a prolonged second stage as a risk factor for pelvic floor morbidity.35,36 However, when differentiated into passive and active phases of the second stage, it is a prolonged active stage that is associated with increased risk of pelvic floor trauma.34,37 The randomised controlled trials in this systematic review provided an opportunity to answer the question of whether a reduction in mid-pelvic instrumental deliveries offset by a prolonged passive phase of second stage of labour results in any increased or decreased risk of longer term pelvic floor morbidity. Such an outcome may swing the balance on the relative harms and benefits of delayed pushing. Only one trial of 178 women assessed pelvic floor outcomes—dyspareunia and faecal incontinence at three months postpartum and the data are insufficient to draw conclusions with a non-significant increase in both outcomes.24 The power this study had to detect the observed increases in dyspareunia (RR 1.15) and faecal incontinence (RR 1.47) were 7% and 41%, respectively. Importantly, the latter trial had a 1 hour time limit for the duration of pushing, had a policy of no rotational forceps and found a non-significant increase in the risk of instrumental delivery associated with delayed pushing.24 It is remiss that eight of the nine trials involving almost 3000 women did not address the clinically important question of long term pelvic floor morbidity.

Based on data from three trials, there was no evidence of increased risk of postpartum haemorrhage, an important finding given that mid-pelvic delivery and prolonged second stage are risk factors for postpartum haemorrhage.38 Only one trial reported maternal satisfaction with care and found no difference between delayed and early pushing groups.24 Another reported maternal fatigue (a non-specified outcome) and found that nulliparous women with delayed pushing were less likely to be fatigued than those in the immediate pushing group but there was no difference in fatigue for multiparas.27

Only two studies assessed intrapartum maternal fever and the results were heterogeneous with Fraser et al.25 finding a significantly increased risk (RR 1.88 95% CI 1.31 to 2.71) while Plunkett et al.29 found no difference (RR 0.93 95% CI 0.54 to 1.61). The pooled result was a non-significant increase in maternal fever (Table 3). Both studies used the same definition of fever (Fraser = 38°C, Plunkett = 100.4°F) but had very different baseline rates of intrapartum fever in the control group (Fraser 4.5% and Plunkett 21%), suggesting that the heterogeneity may be explained by different methods of monitoring maternal temperature during labour.25,29 Fraser et al. suggest that the increased risk of intrapartum fever associated with delayed pushing is mediated through an effect of epidural analgesia on thermoregulation rather than through infection.25,39 They found no differences between treatment groups in the frequencies of maternal postpartum fever, neonatal investigations for suspected infectious morbidity or maternal and neonatal systemic antibiotic therapy.25 However, they note that it is possible that in some clinical settings delayed pushing could lead to increased investigations for infectious morbidity, particularly in the neonate.25

We found no evidence of adverse outcomes for infants in terms of Apgar scores, resuscitation, umbilical artery pH, trauma or perinatal death when delayed pushing rather than early or immediate pushing was used in women with epidural analgesia. Data on almost 2000 infants contribute to these results. The lack of association between adverse infant outcomes and duration of second stage is supported by two large cohort studies each of over 6000 women–infant dyads.40,41


Interventions to minimise the unwanted effects of epidural analgesia for women in labour are urgently required in a context of globally increasing rates of epidural analgesia. For women in hospitals with high rates of rotational or mid-pelvic instrumental deliveries, a policy of delayed pushing for those with epidural analgesia would be beneficial. However, for hospitals with low rates of these procedures, a small reduction in rotational or mid-pelvic instrumental deliveries may not justify the increased duration of the second stage and time spent in the delivery suite. Trials with sufficient power to assess longer term pelvic floor morbidity and whether delayed pushing can reduce second stage caesarean sections are required.


Christine Roberts is supported by a National Health and Medical Research Council (NHMRC) of Australia Public Health Practitioner Fellowship and Siranda Torvaldsen is supported by a NHMRC Public Health Fellowship. The authors would like to thank Charles Algert for assistance with statistical analyses.