MedSysEvidence-graded reference

MedSys / Emergencies / Obstetric / Preterm labour

Emergencies & first aid · Within chapter 25, Obstetric emergencies

Preterm labour protocols

The whole protocol exists to buy 24 to 48 hours, and to use them for three things: steroids, magnesium where indicated, and transfer to a unit with a cot. Every decision is judged against those.

Preterm labour is contractions with cervical effacement and dilatation between 24 and 37 completed weeks. What is done about it depends on two things: how many weeks, and whether the membranes have ruptured.

What the whole protocol is for. Not to prevent preterm birth — no tocolytic reliably does that. It is to buy 24 to 48 hours, and to use them for three specific things: antenatal corticosteroids, magnesium sulfate for neuroprotection where indicated, and transfer to a unit with a neonatal cot. Every decision below is judged against whether it serves those three.

This is a summary of UK practice anchored on NICE NG25, not a local protocol. Individual units differ on gestational thresholds, on magnesium, and on which second-line agents they hold. Where a local guideline exists, it governs.

1. Terminology, which is doing real work here

TermMeans
Suspected preterm labourSymptoms, and an assessment that confirms the possibility but rules out established labour
Diagnosed preterm labourSymptoms plus a positive diagnostic test
Progressive preterm labourPersistent painful contractions with evidence of cervical change — effacement and dilatation
Established preterm labourRegular painful contractions with progressive dilatation from 4 cm

The distinction matters because it determines what is offered. Painless contractions less often than every 10 minutes may be observed. Established labour beyond 4 cm is generally past the point at which tocolysis is attempted.

2. Gestational bands

GestationApproach
22 weeks or lessOutlook is poor and treatment to stop established labour is not usually given. An individualised decision with senior obstetric and neonatal input
22 to 29+6The group of greatest concern. Steroids, magnesium sulfate, GBS prophylaxis, tocolysis considered
30 to 33+6Managed similarly. Steroids recommended; tocolysis considered
34 weeks or moreOutlook is good and treatment to stop labour is not usually given. Steroids between 34 and 37 weeks are a consultant decision

3. Initial assessment — membranes intact

  • History: when the contractions started, how long they last, how strong, how often, and how painful — pain is part of the diagnosis, not incidental. Ask about vaginal loss: show, liquor or blood.
  • Observations: pulse, blood pressure, temperature, and urinalysis. Urinary infection is a common and treatable trigger.
  • Abdominal examination: fundal height, lie, presentation, and palpation for contractions.
  • Speculum examination, with fetal fibronectin where indicated. Digital vaginal examination only if the extent of dilatation cannot be assessed otherwise.
  • Investigations: mid-stream urine, and a low vaginal swab if preterm labour is confirmed. Full blood count, and CRP if indicated. Ultrasound for presentation if needed.
  • Review the antenatal record and assess risk factors before deciding anything.

Two points of technique that change the answer

  • Contractions are best assessed by palpation, not by the cardiotocograph. The CTG monitors the fetus well and measures uterine activity poorly. A hand on the abdomen for ten minutes tells you more about whether these are real contractions than the tocodynamometer trace does.
  • Observe for two hours before concluding. If contractions persist, repeat the examination. If the cervix effaces and dilates within that window, that is progressive preterm labour — and it is the change over two hours, rather than any single finding, that makes the diagnosis.
  • Do not take routine vaginal swabs. Swab if preterm labour is confirmed or there is a clinical indication.
  • Transvaginal cervical length is worth measuring only if the result would change management. Discuss rather than reflexively order.

4. Antenatal corticosteroids

Betamethasone 12 mg intramuscularly, two doses 24 hours apart. If delivery looks imminent, 12 hours apart after consultant discussion. Give the first dose as early as possible: benefit is greatest between 24 hours and 7 days after the second.

ContraindicatedUse with great caution in
Systemic infection, including tuberculosis
Placental abruption is a contraindication to tocolysis and a reason to deliver — it is not a reason to withhold corticosteroids, which are still given where there is time
Insulin-dependent diabetes — steroids cause secondary hyperglycaemia and can precipitate ketoacidosis. An insulin sliding scale may be needed, and this is the interaction most often underestimated
Suspected chorioamnionitis

Repeat courses are not routine and should never be given weekly. A single rescue course may be considered where the first was given some time earlier and delivery again looks imminent — a senior decision rather than a protocol step.

5. Magnesium sulfate for neuroprotection

NICE offers magnesium sulfate between 24+0 and 29+6 weeks where birth is established or planned within 24 hours, and says to consider it between 30+0 and 33+6.

This is the point where local protocols most often diverge from NICE. Some units confine magnesium to under 30 weeks and do not offer it above that; others follow the full NICE range. If you are working to a local guideline that stops at 29+6, that is a deliberate local position rather than an oversight — but it is worth knowing that NICE goes further, because the question comes up at 31 weeks and someone has to answer it.

6. Group B streptococcus prophylaxis

Everyone in confirmed preterm labour is offered intrapartum antibiotic prophylaxis, regardless of carrier status and regardless of whether the membranes have ruptured. Preterm labour is itself the indication, which is a common source of confusion — a negative or unknown GBS swab does not remove it. Agent and regimen follow local policy; intravenous benzylpenicillin is usual. Gentamicin has no place in maternal intrapartum prophylaxis, and neither does co-amoxiclav where membranes have ruptured — erythromycin is the agent there. It is a different matter in treatment: broad-spectrum regimens for established chorioamnionitis commonly do include gentamicin, and that is a treatment decision for a suspected intrauterine infection rather than prophylaxis for a well woman in labour. Conflating the two is how the drug acquires a reputation for being given “by default”.

Gentamicin, hearing loss, and the test that takes 26 minutes

Where gentamicin does appear is at the neonatal end, and it is worth understanding properly because the popular version of this is wrong in both directions.

The claimWhat is actually the case
“Gentamicin causes deafness, so it should be avoided unless there is no alternative” True for a specific group and not in general. NICE NG195 recommends benzylpenicillin plus gentamicin as first-line for suspected early-onset neonatal sepsis, and the reason is instructive: not that the combination is more effective per baby, but that it has a narrower spectrum and so drives less resistance. Antibiotics are to be given within an hour of the decision to treat.
“It is a small risk” For most babies, yes; for around 1 in 500, no. Carriers of the mitochondrial variant m.1555A>G can suffer profound, irreversible sensorineural deafness from a single standard exposure. Roughly 1,200 infants a year are born with it in England and Wales, and with about one newborn in ten screened and treated with antibiotics — it is the treating rather than the screening that exposes them — an estimated 120 a year meet that risk unnecessarily.
“You cannot know in advance” You can now. A point-of-care test detects m.1555A>G from a buccal swab in about 26 minutes, and an implementation trial across UK neonatal units integrated it without delaying antibiotics — only 3.3% of admissions went untested. Where the variant is found, a cephalosporin-based regimen is used instead.

Why this belongs in a preterm labour chapter rather than a neonatal one

  • The inheritance is mitochondrial, so the mother's genotype predicts the baby's. In one large series the variant was found in every mother of an affected child and in none of nearly 3,000 mothers of unaffected infants.
  • Which makes maternal testing during preterm labour a way of knowing before the baby needs the drug — and preterm babies are exactly the group most likely to be screened for sepsis and treated. It has been proposed on that basis, and it is not yet routine practice anywhere in the UK.
  • A family history of hearing loss on the maternal side is the clinical flag — deafness in the mother, her siblings, her mother, or her mother's siblings. It is worth asking, it takes ten seconds, and it is not on any standard proforma.
  • Counselling runs down the maternal line. A positive result is relevant to every maternal relative, and current guidance is that anyone with m.1555A>G, m.1494C>T or m.1095T>C should avoid aminoglycosides unless the severity of infection and the absence of a safe alternative outweigh a high risk of permanent hearing loss.

7. Tocolysis

Before starting: speak to the consultant on call and inform the neonatal unit. Tocolysis without a cot and a plan buys time that nobody is using.

Contraindications to tocolysis of any kind

  • Lethal congenital or chromosomal abnormality
  • Intrauterine infection
  • Severe pre-eclampsia or eclampsia
  • Antepartum haemorrhage, placental abruption, or placenta praevia
  • Advanced cervical dilatation — 4 cm or more
  • Evidence of fetal compromise or placental insufficiency

Nifedipine — first line

A calcium channel blocker. Initial dose 20 mg orally. Provided there is no hypotension or fetal distress, followed by 10–20 mg orally three to four times daily, adjusted to uterine activity.

  • Do not use the slow-release preparation.
  • Do not give it sublingually.
  • Main complications are maternal hypotension and fetal distress. Caution if systolic pressure is below 100 mmHg, or falls by more than 20%.
  • Contraindications include porphyria and significant cardiac disease.

Atosiban — where nifedipine is contraindicated

An oxytocin receptor antagonist, given intravenously.

StepDoseRate
Bolus6.75 mg intravenously over 1 minute
Then, for 3 hours300 micrograms/minute24 mL/hour
Then, for up to 45 hours100 micrograms/minute8 mL/hour

Infusion preparation: two 5 mL ampoules at 7.5 mg/mL added to 90 mL of sodium chloride 0.9%, giving 0.75 mg/mL. At that concentration 24 mL/hour delivers 18 mg/hour and 8 mL/hour delivers 6 mg/hour.

Side effects: nausea and vomiting are common; tachycardia, hypotension, headache, dizziness and hot flushes occur; hyperglycaemia and injection-site reaction are described; pruritus, rash, fever and insomnia are less common.

Indomethacin — below 32 weeks only

100 mg rectally, 12-hourly, for 48 hours — maximum four doses.

Above 32 weeks the fetal risk rises sharply through anti-prostaglandin effects: premature closure of the ductus arteriosus, renal failure, oligohydramnios, necrotising enterocolitis and fetal haemorrhage. Maternal contraindications: dyspepsia, asthma, NSAID allergy, and significant renal disease.

Combinations

Speak to the consultant before combining tocolytics. In idiopathic preterm labour it should generally be possible to inhibit contractions even at full dilatation; where membranes have ruptured, or there is antepartum haemorrhage or chorioamnionitis, it may be neither possible nor desirable. If one agent at maximum dose does not work, a combination may be used.

CombinationConditions
Atosiban and indomethacinAtosiban infused at 18 mg/hour; gestation must be under 32 weeks
Atosiban and nifedipineAtosiban infused at 18 mg/hour. Systolic pressure must be above 100 mmHg before nifedipine is added

8. How well do these actually work?

The answer has two halves, and almost every summary gives only the first.

They delay birth. Whether that produces a better baby is not established. The 2022 Cochrane network meta-analysis — the largest synthesis of this literature, covering every tocolytic class — concluded that all of them are probably or possibly effective at delaying birth by 48 hours and by 7 days against placebo or no treatment. And that the effects on neonatal and perinatal mortality were uncertain, as were the effects on maternal and neonatal infection. Delay is a surrogate endpoint, and the same caution applies here as everywhere else on this site: see how to read a trial.

That is not an argument against using them. It is the reason the protocol above is built around what the delay is for — steroids, magnesium and transfer, all of which do have outcome evidence — rather than around the delay as an end in itself.

Typical pregnancy prolongation

061218243036Nifedipine5.4–26.7, best 18Nitric oxide donors9.5–34, best 9.5Atosiban10Magnesium sulfate4.6Indomethacinnot reported in days — see the tableBetamimeticsnot reported in days — see the tableDays — range of reported study averages, not a head-to-head comparison. Tick marks the most reliable single estimate
Read that as a range of study averages, not of individual outcomes, and not as a league table. The bar spans the lowest and highest average reported across studies; the tick marks the single most reliable figure, which is the randomised or individual-participant one rather than the most flattering. The studies are not measuring the same thing — some report days from trial entry to delivery, some prolongation from the start of treatment, some only the proportion reaching 48 hours or 7 days — so comparing one bar against another is weaker than it looks. Indomethacin and betamimetics have no bar because their trials report proportions rather than days, and inventing a number to fill the gap would be worse than leaving it visible.

AgentLowest reported averageBest single estimateHighest reported averageWhat the figures rest on
Nifedipine5.4 days18 days to delivery26.7 days The 18-day figure is from an individual-participant-data meta-analysis against atosiban — the strongest design here. The 5.4 is a Cochrane maintenance-tocolysis estimate and the 26.7 a single trial. One randomised study reported a median of 20 days with an interquartile range of 2.5 to 51, which gives a better sense of the spread than any average does
Nitric oxide donors transdermal GTN9.5 days9.5 days34 days 9.5 is the RNOTT randomised mean and matches the Cochrane network estimate of 1.35 weeks. 34 is the mean from the original 1994 consecutive series; a 2024 retrospective cohort of 100 women reported 28.6. Individual prolongations beyond 45 days are described
Atosiban10 days to delivery From the same individual-participant-data meta-analysis, which is why it is directly comparable with the nifedipine figure and the others are not
Magnesium sulfate as a tocolytic4.6 days A single head-to-head against nifedipine, which reported 6.1 days. Not used as a tocolytic in the UK
IndomethacinNot reported in days Trials report proportions: delivery delayed beyond 48 hours in roughly 90–94%, and beyond 7 days in about 75%. Effective and fast; the constraint is fetal risk above 32 weeks, not efficacy
BetamimeticsNot reported in days Trials report the 48-hour endpoint only. Abandoned on tolerability rather than on failure to delay

The result that tests whether more days is the same as a better baby

  • That individual-participant-data meta-analysis found nifedipine prolonged pregnancy nearly twice as long as atosiban — 18 days against 10. If days gained were the outcome, that would settle it.
  • It also found no difference in the composite of neonatal morbidity and perinatal mortality, and a numerically higher neonatal mortality in the nifedipine arm (OR 1.4, 95% CI 0.60–3.4) — not statistically significant, and the authors flagged it as warranting further investigation. NICU admission, meanwhile, was less frequent after nifedipine.
  • So the agent that bought eight more days did not produce better babies in that analysis, and possibly the reverse. That is the clearest available demonstration that prolongation is a surrogate and not the thing being aimed at — and it is why the chapter is organised around what the days are spent on rather than how many there are.
  • None of which is an argument for a shorter delay. It is an argument for judging an agent on tolerability, route, speed of onset and the ability to get steroids in — because on the outcome that matters, the evidence does not currently separate them.

How long does the delay actually last? The nitric oxide donor case

Worth taking separately, because the reported durations for transdermal GTN span an unusually wide range and the figure you pick changes the impression substantially.

SourceDesignProlongation reported
Lees et al., Lancet 1994Consecutive series, 13 women, 20 episodes at 23–33 weeks. All 20 episodes respondedMean 34 days — a mean, not a maximum
Bisits et al., RNOTT, AJOG 2004Randomised trialMean about 9.5 days in the GTN arm
Cochrane network meta-analysis, 2022Synthesis across the classGestational age 1.35 weeks more advanced — about 9.5 days — low certainty
Gaikwad et al., Cureus 2024Retrospective cohort, 100 women at 27–35 weeksMean 28.6 days, with 90% reaching 48 hours and contractions inhibited in 92%
Parveen et al., 2012Quasi-experimental, 65 women at 28–34 weeks, 5 mg/12 h patch to the anterior abdominal wallEffective and safe; pregnancy prolonged with improved neonatal outcome
Lees et al., Obstet Gynecol 1999International multicentre randomised, against ritodrineGestation prolonged by 74% of the time remaining to 37 weeks, identical between arms — with fewer preterm deliveries and fewer side effects on GTN
Systematic review and meta-analysis, AJOG 2013Pooled comparisonNo significant difference against placebo for delivery within 48 hours or before 28, 34 or 37 weeks. Superior to β2-agonists for birth before 34 and 37 weeks, NICU admission and ventilation; no different from nifedipine or magnesium
How to hold those together, because they look contradictory and are not. The continuous measures — mean days gained, mean gestational age at delivery — consistently favour GTN, and cluster from about 9.5 days in randomised work up to a mean of 28.6 days in the largest retrospective cohort, with the original 1994 series reporting a mean of 34 days. The binary endpoints tell a more divided story: the 2013 pooled analysis found no significant difference against placebo for delivery within 48 hours or before 28, 34 or 37 weeks, while the 2022 Cochrane network meta-analysis found nitric oxide donors probably or possibly effective at both 48 hours and 7 days. So the null on binaries belongs to one analysis rather than to the literature. Both can be true: an agent can shift the average substantially while failing to move a threshold that most women in both arms were going to cross or avoid anyway. It is also the pattern you would expect from small trials measuring the harder endpoint. There is also a gradient by design: the largest figures come from the weakest designs — an uncontrolled consecutive series and a retrospective cohort — and the smallest from the randomised work. That is the pattern you would expect if selection and the absence of a comparison group were inflating the observational numbers, and it is a reason to weight the randomised figure even though it is the least flattering. A prolongation in double figures is well within what the literature describes, and the 9.5 days on the chart is the randomised central estimate rather than an upper bound on what any individual might achieve.

On small numbers. A handful of cases cannot establish a duration — three women who all did well is equally consistent with a good drug, a favourable selection of patients, or chance, and there is no comparison group. What it can legitimately do is prompt the question, which is what happened here: the chart originally carried a lower figure than the published central estimate, and checking it against the sources moved it up. Observation is a reason to go and look; the published range is what the number rests on.

The comparison in full

AgentDelays birth?Effect on the babyMain risksPractical verdict
Nifedipine first line Yes, at 48 hours and 7 days. Moderate certainty Uncertain for mortality. Neonatal outcomes similar to comparators in head-to-head trials Maternal hypotension and fetal distress are the ones that stop treatment. Headache and flushing common. Above 60 mg/day, headache and hypotension rise roughly fourfold Oral, cheap, familiar, best tolerated of the effective options. The reason NICE puts it first
Atosiban Yes, on the network analysis. But the class review found it no better than placebo or betamimetics on the outcomes that matter Uncertain. One trial reported more infant deaths, in a study with a gestational-age imbalance between arms Best tolerated of all — nausea and vomiting common, little cardiovascular effect. Hyperglycaemia described The alternative where nifedipine is contraindicated, chosen for tolerability rather than potency. Intravenous, and expensive
Indomethacin Yes. Effective and fast Fetal risk is the limiting factor above 32 weeks: premature ductal closure, renal impairment, oligohydramnios, necrotising enterocolitis, fetal haemorrhage Maternal effects mild — dyspepsia the main one. The problem is fetal, not maternal Useful below 32 weeks, particularly rectally when the oral route is difficult. Not above 32 weeks
Nitric oxide donors transdermal GTN Yes. The longest central estimate of any class at about 9.5 days, and the widest reported range — see the panel above The most interesting signal in the review, and low certainty: birthweight higher by around 425 g, fewer babies under 2500 g (RR 0.40), and gestational age about 1.35 weeks more advanced. No class showed better neonatal signals — and no class has confirmed them either Headache in roughly a quarter to a third, often severe. Hypotension. Both settle quickly on removing the patch Not in NICE. Transdermal, so no cannula and no pump — see the protocol
Magnesium sulfate as a tocolytic Weakly, and it is not used for this in the UK Uncertain as a tocolytic. Its neuroprotective benefit is a separate and much better established thing — section 5, not this table Flushing, nausea, hypotension, reduced reflexes. Toxicity needs monitoring Do not confuse the two indications. Magnesium earns its place in preterm labour for the brain, not for the uterus
Betamimetics ritodrine, terbutaline, salbutamol Possibly, at 48 hours No demonstrated improvement The reason they were abandoned. Against placebo: dyspnoea RR 12.1, palpitations RR 7.4, tachycardia RR 3.0, vomiting RR 1.9. Most likely of any class to have treatment stopped for side effects, and pulmonary oedema is described Historical. Included because they appear in older protocols and in the literature you will read
Combinations Yes Possibly fewer babies under 2500 g (RR 0.74), low certainty Adverse effects and treatment cessation both rise. Consultant decision For failure of a single agent at full dose, not as a starting point

The intervention in this chapter with the strongest outcome evidence is not a tocolytic

  • Antenatal corticosteroids reduce respiratory distress syndrome, intraventricular haemorrhage and neonatal death. In the late preterm trial, severe respiratory complications fell from 12.1% to 8.1% — a relative risk of 0.67 — and the effect at earlier gestations is larger.
  • Magnesium sulfate for neuroprotection reduces cerebral palsy. Again, an outcome rather than a surrogate.
  • Transfer to a unit with a neonatal cot changes outcomes by more than any drug in the table above.
  • So the value of tocolysis is almost entirely instrumental. It creates the window in which the three things that work can happen — which is a real contribution, and a different claim from "it improves outcomes".

9. Where the membranes have ruptured

Rupture changes the calculation: the infection risk rises, tocolysis becomes less appropriate, and the balance shifts toward delivery. Erythromycin is the antibiotic of choice in preterm prelabour rupture of membranes; co-amoxiclav is specifically avoided because of the association with neonatal necrotising enterocolitis. Chorioamnionitis is a reason to deliver, not to inhibit.

10. Transdermal glyceryl trinitrate

A separate protocol, for a specific situation. Transdermal GTN is a nitric oxide donor with a thirty-year evidence base and it does not appear in NICE NG25 — nifedipine and atosiban above are what UK units use. Its interest is practical rather than pharmacological: it is a patch, so it needs no cannula, no pump and no swallowing; it is stable at room temperature and can be held in advance; and it acts within about an hour. That combination makes it worth considering where a woman with identified recurrent risk is some distance from a unit and a plan has been made ahead of time by the team who know the pregnancy.

The evidence is Contested — a real effect on uterine contractility that is not in dispute, attached to trial evidence that is limited and mixed, with recent comparisons favouring oral nifedipine. The full protocol — entry criteria, dosing, the patch schedule, monitoring and contraindications — is set out on the transdermal GTN page.

The GTN tocolysis protocol →

11. If the baby arrives

Preterm delivery outside a unit is covered in obstetric emergencies, and the thermal care differs from a term birth in a way that matters: a preterm baby's body should go into a clear polyethylene bag or wrap without being dried, with the head dried and covered. Drying a very preterm baby wastes the heat you are trying to keep.