All topics › Labor delivery postpartum recovery
What are the proven benefits and risks of an epidural for pain relief in labour?
What are the proven benefits and risks of an epidural for pain relief in labour — and what does the best evidence actually say about the common worries (caesarean section, backache, breastfeeding, effects on the baby)?
Epidurals are the most effective form of labour pain relief we have: a 2018 Cochrane review of 40 trials (over 11,000 women) found lower pain scores and more women rating their pain relief "excellent or very good" than with opioids or other methods [1]. They do not increase the chance of a caesarean section, and they do not cause long-term backache [1]. The real trade-offs are a higher chance of an assisted (forceps or ventouse) birth in older studies — though this seems to have faded with modern low-dose techniques — plus common short-term side effects: low blood pressure, fever, difficulty passing urine, and a second stage of labour about 15 minutes longer on average [1]. Whether epidurals affect breastfeeding is genuinely unclear: the studies are all observational and point in different directions [2].
The Cochrane review by Anim-Somuah and colleagues (2018 update, 40 randomised trials, over 11,000 women — mostly comparing epidural with injected opioids such as pethidine) is the backbone of this topic [1]. On pain, the findings were:
A note on quality: the review authors rated this evidence low-quality — not because the trials were tiny, but because blinding is impossible (you know whether you have an epidural), pain is subjective, and results varied a lot between trials [1]. The direction, though, is consistent across decades of research: nothing else touches an epidural for labour pain.
Two of the most persistent worries don't hold up:
Overall, women with epidurals were more likely to have an assisted vaginal birth — forceps or ventouse (risk ratio 1.44; 30 trials, 9,948 women; low-quality evidence) [1]. In absolute terms, that's about 99 in 1,000 with opioids vs about 142 in 1,000 with an epidural — roughly 43 more assisted births in 1,000. But the review's post-hoc subgroup analysis tells an important story: in trials conducted after 2005, the effect disappeared (risk ratio 1.19, confidence interval 0.97–1.46 — crossing 1.0, i.e. no statistically significant difference) [1]. The authors' interpretation: modern practice uses much lower concentrations of local anaesthetic (often with patient-controlled top-ups), which preserves the ability to push — and the older, denser blocks are what drove the instrumental-birth signal [1].
Caveat: this was a post-hoc analysis (decided after seeing the data), which makes it weaker evidence than the headline result. But it matches what obstetric anaesthetists have observed as practice changed, and it is the finding most relevant to an epidural you'd be offered today.
Epidurals come with a well-documented set of short-term effects (2018 Cochrane review pooled estimates, all epidural-vs-opioids trials) [1]. The absolute figures below are crude rates from pooling the trial event counts — they show how common these were in the trials, where definitions and techniques varied; individual trials differed substantially (high heterogeneity), so treat them as approximate:
On the other side of the ledger, women with epidurals had less nausea and vomiting — roughly 97 vs 149 in 1,000 (risk ratio 0.62, 95% CI 0.45–0.87; 15 trials) — and less respiratory depression needing oxygen (risk ratio 0.23), with their babies less likely to need naloxone (the drug that reverses opioid effects) [1].
These figures update the pooled estimates previously taken from the 2011 version of the review; the 2018 update gives somewhat smaller effect sizes for hypotension and fever (11.34 vs 18.23, 2.51 vs 3.34) but the same pattern of direction.
One practical knock-on effect of epidural fever: fever in labour usually triggers a sepsis evaluation for the baby (blood tests, sometimes antibiotics) because infection can't be ruled out at the time. An older US study reported that babies of mothers who had an epidural were evaluated for sepsis much more often [3]. A large 2021 Chinese cohort study (37,786 full-term vaginal births at one Shanghai hospital, propensity-score matched) went further and found epidural use was associated with diagnosed neonatal infection: an absolute risk difference of 2.6 percentage points, or about 26 extra babies with a recorded infection per 1,000 (relative risk 2.43; 95% CI 2.11–2.78) [4]. Most of that was "uncharacterized infection" (a diagnostic category in that hospital); confirmed neonatal sepsis was rare, with an absolute difference of 0.1 percentage points, or about 1 extra per 1,000 (RR 3.50; 95% CI 1.73–7.07) [4]. In plain terms: the fever often isn't infection, but your baby will likely be investigated as if it might be — and in this large cohort, slightly more babies had a diagnosed infection too. This evidence is observational — women who choose epidurals may differ in other ways — the authors call for further investigation, and practice and diagnostic categories differ outside China, so treat these numbers as indicative rather than directly transferable to a UK birth.
Trials are far too small to measure rare harms, so these figures come from anaesthetic audit and surveillance data, as summarised for counselling by the Royal College of Anaesthetists / Obstetric Anaesthetists' Association (reproduced in a 2022 peer-reviewed narrative review) [2]:
| Complication | Frequency |
|---|---|
| Epidural doesn't work well enough without further attention | 1 in 8 |
| Not working well enough for a caesarean — a spinal or general anaesthetic is needed instead | 1 in 20 |
| Significant drop in blood pressure | 1 in 50 |
| Severe headache (post-dural-puncture headache) | 1 in 100 |
| Temporary nerve damage (e.g. numb patch or weak leg) | 1 in 1,000 |
| Permanent nerve damage | 1 in 13,000 |
| Epidural abscess (infection) | 1 in 50,000 |
| Meningitis | 1 in 100,000 |
| Epidural haematoma (blood clot pressing on nerves) | 1 in 170,000 |
| Severe injury including paralysis | 1 in 250,000 |
These are the figures UK anaesthetists use when consenting women — small enough to be reassuring, real enough to deserve a mention [2].
No randomised trial has directly tested epidural vs no epidural for breastfeeding outcomes (randomising women to labour pain relief this way isn't really feasible), and the observational studies conflict: a 2016 systematic review of 23 studies found no association in 10, negative associations in 12, and a positive association in one — with only 3 small randomised studies, all comparing different epidural drug doses rather than epidural vs no epidural [5]; a 2021 update of 15 studies was similarly split (6 no difference, 6 lower breastfeeding with neuraxial analgesia, 3 mixed) [6]. A 2022 narrative review concluded the literature is "largely limited to observational studies, or small RCTs, but is reassuring with limited evidence of a detrimental effect" [2]. Honest summary: there is no reliable evidence of harm, and no reliable evidence of no harm either.
For the woman giving birth, the decision is mostly about pain vs medicalisation. An epidural reliably takes away most labour pain, and women rate their pain relief more highly — but it turns labour into a more medical experience: expect IV fluids, regular blood pressure checks, continuous monitoring of the baby (which usually limits you to the labour ward rather than home or a midwife-led unit and restricts movement), and often a bladder catheter, with legs that may feel heavy or numb for a while afterwards [1][2]. (Exactly what's standard varies by unit — worth asking your midwife what's routine where you'll give birth.) The second stage is a little longer, and there's a somewhat higher chance of forceps or ventouse in the older data (about 142 vs 99 in 1,000) — though with modern low-dose techniques that increase may no longer apply [1].
For partners and birth companions, the evidence is thin — trials didn't measure partner outcomes. Practically, an epidural shifts the partner's role: less hands-on pain support (breathing, massage, position changes matter less when pain is blocked) and more advocacy and communication — keeping track of what's happening, asking questions when interventions are suggested, and being the steady presence while the woman can't move freely. If that's you, it's worth talking through in advance what "support" looks like when pain isn't the main event.
Mental health: the Cochrane review found no clear difference in postnatal depression between epidural and non-epidural groups [1]. That's a genuine evidence-based reassurance against the occasional claim that epidurals harm bonding or mood.
The fever knock-on is the parental experience most people don't expect. If you develop a temperature with an epidural, your baby is likely to have blood tests and possibly antibiotics for suspected infection, even though the fever is often not infectious [3][4]. Parents describe this as one of the more distressing parts of an otherwise straightforward birth — worth knowing about in advance, not to put you off, but so it doesn't come as a shock.
Sleep, work, and finances: no trial measured these for epidural vs other pain relief. The honest answer is that any effect on recovery, sleep, or time off work runs through the mode of birth (vaginal vs assisted vs caesarean) rather than the pain relief method itself — and on caesarean rates, epidurals make no difference [1].
| Outcome | Effect | Evidence rating |
|---|---|---|
| Pain relief | Better than opioids/other methods (SMD −2.64); far less need for additional pain relief (RR 0.10) [1] | B |
| Maternal satisfaction ("excellent/very good") | Higher: ~500 vs ~735 in 1,000 rate pain relief excellent/very good (RR 1.47) [1] | B |
| Caesarean section | No difference: ~114 vs ~122 in 1,000 (RR 1.07; 33 trials, 10,350 women) [1] | B |
| Assisted vaginal birth (forceps/ventouse) | Increased overall: ~99 vs ~142 in 1,000 (RR 1.44); no significant increase in post-2005 trials of modern techniques (RR 1.19) [1] | B (C for modern practice) |
| Long-term backache | No difference: ~585 in 1,000 in both groups (RR 1.00) [1] | B |
| Second stage of labour | About 15 minutes longer on average; somewhat more oxytocin use (RR 1.12) [1] | B |
| Low blood pressure | More common: ~129 vs ~15 in 1,000 crude trial rate (RR 11.34) — monitored and treated as routine [1] | B |
| Fever | More common: ~145 vs ~56 in 1,000 crude trial rate (RR 2.51); can trigger neonatal sepsis evaluation [1][3][4] | B |
| Urinary retention | More common: ~179 vs ~5 in 1,000 crude trial rate (RR 14.18) — a bladder catheter is commonly used alongside the epidural [1] | B |
| Temporary motor blockade | ~184 in 1,000 with epidural vs none in opioid groups (RR 31.71; older higher-dose techniques) [1] | B |
| Nausea/vomiting, respiratory depression | Less than with opioids: ~97 vs ~149 in 1,000 for nausea/vomiting (RR 0.62) [1] | B |
| Apgar <7 at 5 min; NICU admission | No clear difference: Apgar ~17 vs ~12 in 1,000; NICU ~204 vs ~210 in 1,000 [1] | B (for the nulls) |
| Postnatal depression | No clear difference: ~130 vs ~205 in 1,000 in the single trial reporting it (RR 0.63, CI crosses 1.0) [1] | B (for the null) |
| Severe headache / nerve injury / abscess / haematoma / paralysis | 1 in 100 down to 1 in 250,000 (counselling figures from audit data) [2] | C |
| Breastfeeding | Unclear — conflicting observational studies, no trials of epidural vs no epidural [5][6] | D |