You know how to treat anaphylaxis. You probably do it every week. That is not what the exam is testing — the exam is in the small details: what happens beyond the algorithm once two doses have not worked, the numbers in the refractory pathway, how long the patient actually stays afterwards, and the conditions that look exactly like anaphylaxis but will not respond to adrenaline at all.

The 2026 update incorporates the changes the Resuscitation Council made in 2021, so most of it is not new thinking. The algorithms and the emergency treatment are unchanged. What has been added is the fine print afterwards, principally around discharge and disposition.

What is actually new:

• A 2-hour discharge route, with four criteria that must all be met. It has been in the resuscitation guidance since 2021, but it's new to NICE: CG134 had no 2-hour route.
• Clarified 6- and 12-hour observation periods, and the clock defined — it runs from resolution, not from arrival.
• Under-16s who do not meet the 2-hour criteria are admitted under an inpatient paediatric medical team. There is no adult equivalent.

Each of these is worked through below.

Anaphylaxis is a severe, systemic hypersensitivity reaction — an allergic reaction at the dangerous end of the spectrum. Three things define it: sudden onset and rapid progression; a life-threatening problem in the airway, breathing or circulation; and, usually but not always, skin or mucosal changes.

That word usually is where marks go. Skin changes are absent in up to 20% of reactions, so waiting for a rash will cost you the diagnosis. Equally, urticaria, angioedema or rhinitis on their own is not anaphylaxis — without an A, B or C problem it does not meet the definition. The rash is neither necessary nor sufficient.

The life-threatening features:

• Airway — pharyngeal or laryngeal oedema, hoarse voice, stridor.
• Breathing — bronchospasm, tachypnoea, wheeze, fatigue, cyanosis, SpO₂ below 94%.
• Circulation — hypotension, tachycardia, shock, confusion, reduced consciousness.

Timing and speed of onset matter, because they point to the trigger. Parenteral exposure — an intravenous drug, an injection, a sting — comes on faster than something eaten. That is why you are advised to record the time of onset and the circumstances immediately before it: together they are usually the best clue to what caused it. The onset time is also what every tryptase sample is timed from.

The presentation varies by trigger too, which is a separate point. Food is the commonest cause in children and more often presents as a breathing problem, with a less rapid onset and a background of asthma or atopy. Drug and venom reactions are commoner in older patients, come on faster, and more often present with circulatory compromise. In 5–10% of cases no trigger is ever identified.

These are genuinely hard to separate at the front door, which is exactly why they get tested. The discriminators are the drug history, the pattern of the attacks, and what happens after adrenaline.

The ones adrenaline will not touch

ACE inhibitor-induced angioedema is bradykinin-mediated rather than histamine-mediated, which is the whole reason the anaphylaxis drugs do nothing. It presents as isolated swelling of the lips, tongue or face, and the giveaway is that there is no urticaria and no itch. It can begin after years of uneventful treatment, which is exactly why the drug gets overlooked. It can also be life-threatening, and no drug reliably reverses it, so the airway is the emergency: involve anaesthetics and ENT early, before the anatomy makes intubation impossible, and stop the drug.

Hereditary angioedema, from C1 esterase inhibitor deficiency, is bradykinin-mediated too. It causes recurrent attacks over years, usually with a family history and again without urticaria, and it can present as abdominal pain rather than swelling. It is treated with C1 esterase inhibitor concentrate or icatibant, and it will not respond to adrenaline, steroids or antihistamines.

Inducible laryngeal obstruction, formerly called vocal cord dysfunction, causes recurrent stridor that comes on and settles abruptly, without a rash and without hypotension. Management is supportive, and speech and language therapy is the definitive treatment.

The ones that look the same

Scombroid poisoning follows a meal of oily, dark-fleshed fish that has been poorly refrigerated — tuna, mackerel, sardines, herring, mahi-mahi and marlin among others. Bacteria convert the histidine in the flesh into histamine, and cooking does not destroy it. Within minutes the patient develops flushing, headache, palpitations and urticaria, often with a peppery or metallic taste, and they have no history of fish allergy. Because this is histamine poisoning rather than an allergic reaction, antihistamines are the treatment.

Life-threatening asthma can look exactly like anaphylaxis, and the overlap is commonest in children — which is unsurprising, given how many paediatric reactions are food-triggered and present as a breathing problem.

A panic attack is easiest to mistake in someone who has had anaphylaxis before and believes they have been re-exposed. They describe a sense of impending doom and hyperventilate, and anxiety flushing muddies the picture further. The same list also includes a vasovagal faint, breath-holding in a child, and spontaneous non-allergic urticaria or angioedema.

Position first, and position by the problem.

• Hypotensive — lie them flat, with or without the legs raised.
• Struggling to breathe — semi-recumbent. Sitting up makes breathing easier, and they will usually choose it themselves.
• Unconscious but breathing normally — recovery position, monitored continuously.
• Pregnant — put her into the left lateral position, to relieve aortocaval compression. If she has to be supine for the airway or for CPR, displace the uterus manually to the left and use head-down tilt rather than raising the legs.

Remove the trigger if you can — stop the infusion — but never delay adrenaline to hunt for one. Then IM adrenaline into the anterolateral thigh, middle third. Every dose in the algorithm is intramuscular; intravenous adrenaline is for experienced specialists in an appropriate setting. Airway, high-flow oxygen and monitoring follow: ECG, oxygen saturations and blood pressure.

At five minutes, repeat the IM dose if there has been no improvement. Peak absorption is 5–10 minutes after an intramuscular dose, so judge the response rather than the clock alone. Some guidance suggests the contralateral thigh for further doses, though the evidence for that is uncertain. Give a fluid bolus at the same point: 500–1000 mL crystalloid in an adult, 10 mL/kg in a child. Persisting breathing or circulatory problems after two doses take you into the refractory pathway.

Not as part of the emergency treatment. Both came out of the algorithm in 2021, and the reasoning is examinable.

Corticosteroids are no longer advised for the routine emergency treatment of anaphylaxis. There is little evidence they shorten symptoms or prevent biphasic reactions, and emerging data associate early steroid use with an increased risk of intensive care admission, even after adjusting for severity — though these recommendations rest on very low-certainty evidence. Where they may still be used is after initial resuscitation, for refractory reactions or ongoing asthma or shock, and never in preference to adrenaline. Hydrocortisone 200 mg IV in an adult; 4 mg/kg to a maximum of 200 mg in a child.

Antihistamines are third line, with no role in respiratory or cardiovascular symptoms. They are for skin symptoms once the patient is stabilised, and a non-sedating oral antihistamine such as cetirizine is preferred to chlorphenamine, which sedates. If the oral route is not available, chlorphenamine can be given intramuscularly or intravenously — but a rapid intravenous bolus can itself cause hypotension.

Refractory anaphylaxis is anaphylaxis requiring ongoing treatment — persisting respiratory or cardiovascular symptoms — despite two appropriate doses of IM adrenaline. That is the whole definition, and meeting it is itself the trigger to leave the standard algorithm.

Fewer than 1% of reactions are refractory, so why does it get its own pathway? Because a third of deaths from food anaphylaxis occur despite a timely first dose. The contributor is usually insufficient adrenaline delivery rather than an avoidable delay, which is why the emphasis is on recognising it early and escalating.

The infusion

Mix 1 mg (1 mL of 1 mg/mL) of adrenaline in 100 mL of 0.9% sodium chloride and run it by pump through a dedicated line — not on the same side as a blood pressure cuff, because the cuff interferes with the infusion and risks extravasation. Start at 0.5 mL/kg/h, or 1 mL/kg/h if the patient is hypotensive or hypoxic. Steady state is reached 5–10 minutes after a rate change, so titrate and reassess rather than making repeated adjustments. Keep giving IM adrenaline every five minutes until the infusion is ready. Intravenous boluses are not recommended unless the patient is in cardiac arrest, though those experienced in titrating vasopressors may use one while the infusion is set up. Monitoring is mandatory: continuous ECG and oxygen saturations, and non-invasive blood pressure at least every five minutes.

Then work A, B and C

For stridor or partial upper airway obstruction, nebulised adrenaline, 5 mL of 1 mg/mL; for total obstruction, expert help and the difficult airway algorithm. For breathing, oxygenation matters more than intubation — titrate to SpO₂ 94–98%, and add nebulised salbutamol and ipratropium for persistent bronchospasm. For circulation, glucose-free crystalloid, titrated to response, and 3–5 litres may be needed in an adult.

Still not responding — and this is what the exam asks about

Once the infusion is running and the patient is still shocked, you are into the part of the pathway people are least sure of, which is precisely why it gets tested.

• Add a second vasopressor — noradrenaline, vasopressin or metaraminol. Follow your local protocol; if there isn’t one, use the drug you are most comfortable with. There are no human data favouring one over another.
• If the patient is beta-blocked, consider glucagon — 1 mg IV, repeated or followed by an infusion of 1–2 mg/h. Adrenaline is less effective in these patients, and the drug history is the clue that puts glucagon in play.
• Steroids. The algorithm says consider them after initial resuscitation, so if you want to add hydrocortisone at this point, add it. By now you are fighting a losing battle for this patient’s life, and the steroid is not what will decide it either way — just never at the expense of the adrenaline infusion and the fluids.
• Consider extracorporeal life support. Realistically that means a younger patient with a clearly reversible trigger who is in refractory shock despite an infusion and a second vasopressor, or who arrests and is likely to need prolonged resuscitation. If your hospital does not have it, the question is whether you can activate the nearest ECMO team — and that is a call made early or not at all.
• In cardiac arrest, follow ALS with aggressive fluid resuscitation, and consider prolonged resuscitation or extracorporeal CPR.

Critical care support is essential throughout, and resuscitation may run for hours to days.

Weaning

Weaning is part of the pathway too: as symptoms improve, reduce to 50% of the starting rate; one hour after resolution of all symptoms and signs, reduce progressively over 30 minutes and stop, watching for recurrence.

Tryptase is a follow-up test, not part of recognition or treatment, and taking it must never delay either. Critically, a normal tryptase does not exclude anaphylaxis.

Worth being clear on who gets it, because the two sources differ. The current recommendation is to take timed samples after any suspected anaphylaxis in someone aged 16 or over — it is not conditional on the diagnosis being in doubt. The older resuscitation guidance framed it more narrowly, as a test for when the diagnosis is uncertain, which is why you will still see it described that way. In practice, uncertainty is when it changes anything: in a clear-cut reaction it confirms what you already knew and gives the allergy clinic a baseline to work from.

The kinetics explain the timings. It may not rise significantly until 30 minutes or more after symptom onset, peaks at 1–2 hours, has a half-life of about two hours, and can be back to normal within 6–8 hours.

Three samples, all timed from onset:

• The first as soon as treatment has started — without delaying treatment to take it.
• The second at 1–2 hours, and no later than four hours.
• A baseline at least 24 hours after complete resolution, or in convalescence at the allergy clinic. Some people have a raised baseline and are at greater risk from certain triggers.

Serial samples have better sensitivity and specificity than a single one. Serum, in a liver function test tube, or plasma is acceptable in most laboratories.

Who gets it depends on age. In an adult or young person aged 16 or over, take the timed samples as standard.

The first thing to get right is when the clock starts: from resolution of airway swelling, resumption of normal breathing, and a stable blood pressure. Not from arrival, and not from the adrenaline.

The 2-hour route is an early discharge pathway, for patients who do not warrant a 6-hour observation period on the 6-hour criteria. It is not a default — it is a way out for a narrow group, and all four conditions have to hold.

Consider discharge at 2 hours — all four must be present:

1. A good response within 5–10 minutes to a single IM dose given within 30 minutes of onset.
2. Symptoms have completely resolved.
3. The person already has two in-date auto-injectors and knows how and when to use them.
4. There is adequate supervision from an appropriate adult afterwards, if needed.

There is one exception: after a supervised allergy challenge, discharge at two hours from resolution can be considered even if two doses were given.

A minimum of 6 hours if two doses of IM adrenaline were needed, or there is a history of biphasic reaction.

A minimum of 12 hours if any one of: more than two doses were needed; severe asthma or anaphylaxis involving severe respiratory compromise; possible continuing absorption of allergen, such as a slow-release medicine; the person presents out-of-hours; the person may not be able to respond to a deterioration; or discharge is to an area where access to emergency care is difficult.

Under-16s who do not meet the 2-hour criteria are admitted under an inpatient paediatric medical team. If a child does not qualify for the 2-hour route — for any reason at all, including something as ordinary as not already carrying auto-injectors — then they are admitted. There is no adult equivalent.

Behind all of it is the reason the periods exist: there is no reliable way of predicting who will have a biphasic reaction. The risk factors are a more severe initial presentation, needing more than one dose of adrenaline, and a delay in giving adrenaline of more than 30–60 minutes from onset. Reassuringly, fatal outcomes from biphasic reactions are very rare.

A child who fails any one of the 2-hour criteria is admitted. For an adult in exactly the same position there is no stated period and no stated destination, so this comes down to clinical judgement.

The useful thing to notice is what they do not have. They did not need two doses of adrenaline and they have no history of a biphasic reaction — the two things that would have earned six hours in the first place. On that basis they sit at lower risk than the patient the 6-hour rule was written for, so six hours of observation is comfortably enough.

The practical constraint settles the rest. Six hours will overrun a four-hour emergency department stay, so this is an onward referral rather than a wait in the department.

What is written is that every patient should be reviewed by a senior clinician, and that because there is no reliable way of predicting a biphasic reaction, decisions about discharge must be made for each patient by an experienced clinician. That delegation is explicit — which is another way of saying the gap is deliberate.

A discharge bundle, and it is a list you can be asked to complete:

• Two in-date adrenaline auto-injectors, via a prescription if needed, and they know when and how to use them — unless the anaphylaxis was due to a drug allergy and the drug can easily be avoided.
• A brand-specific demonstration: different devices, different technique.
• Advice to carry both at all times.
• Information on the risk of a biphasic reaction, and on avoiding the trigger where it is known.
• A referral to a specialist allergy service, age-appropriate where possible, and an explanation of how that referral works. Every trust should hold separate adult and paediatric pathways.
• Information about patient support groups.

This is a consultant-level exam, and it is written on the assumption that you can already treat anaphylaxis. That does not make the doses safe ground — the paediatric ladder by age has been asked before and is worth knowing cold. But what separates candidates is the finer detail, and more than that, whether you know the criteria well enough to pick them up without effort while you are reading the stem.

That is the real skill being tested. When a stem gives you an arrival time of 02:40, or two doses of adrenaline, or a patient who lives an hour from the nearest hospital, those details are there deliberately — each one is a criterion. Know them cold and you register them as you read, rather than going back to hunt for them once you have seen the options. Know them vaguely and the question looks like a judgement call when it is actually a rule.

So learn the numbers properly: the four conditions for early discharge, the two triggers for six hours, the six for twelve, what defines refractory anaphylaxis, and the sampling times for tryptase. Then learn the mimics well enough to recognise one from the history, because that is the question where knowing the treatment for anaphylaxis will not save you.

Anaphylaxis is a good example of why revising from a guideline you read last year is risky — the treatment half was stable and the disposition half was rewritten underneath it. If you want the method behind keeping guideline revision current, see how I rebuilt my guideline revision for the FRCEM Final and our list of the guidelines worth knowing for the FRCEM Final. For the arrest end of the same pathway, see cardiac arrest in special circumstances.

Go to the source

• Read the emergency treatment guideline in full — Emergency treatment of anaphylactic reactions, or the anaphylaxis guidance landing page.
• Print the algorithms — the anaphylaxis algorithm (PDF) and the refractory anaphylaxis algorithm (PDF). Worth having both on the wall of resus.
• Read the observation and discharge criteria yourself — NG258, observation and discharge, or the guideline overview. This is the part that changed in 2026, and it is short enough to read in one sitting.

Doses and thresholds here are a revision guide — always confirm against current guidance and your local protocol.