After the Epinephrine Works: Whether a Steroid Actually Prevents the Second Reaction
Standard practice has long added a steroid before discharge after anaphylaxis to prevent a second wave of symptoms — but the evidence behind it, and the drug's own timing, don't line up the way the practice assumes.
Doug F., a 42-year-old man, had shrimp for the first time in his adult life at a work dinner two hours ago, at a colleague's insistence that he “just try it” after years of politely declining — and within twenty minutes developed hives, lip swelling, and audible wheeze that progressed quickly enough that a coworker drove him straight to the emergency department rather than waiting for the reaction to pass. He received intramuscular epinephrine, antihistamines, and nebulized albuterol on arrival, and his symptoms resolved within thirty minutes. He has no prior history of any food allergy, no asthma, and no other chronic illness — this appears to be a genuinely new-onset shellfish allergy, not a known allergy that finally caught up with him. His wife, reached by phone from the department, mentioned that his father developed a similar shellfish sensitivity in his fifties after decades of eating it without issue, a detail nobody had thought relevant enough to mention before tonight.
The team's standing practice has been to give a dose of IV corticosteroid before discharge, on the long-held assumption that it prevents a biphasic reaction — a recurrence of symptoms hours after the initial reaction resolves, without any new exposure. More recent evidence complicates that assumption directly: Grunau's 2015 cohort found no reduction in biphasic reactions among patients given corticosteroids, and Alqurashi and Ellis's systematic review found only one of twenty-two studies supporting the practice. What does not exist anywhere in that literature is a randomized trial — the 2020 practice parameter graded the whole body of evidence very low quality and concluded only that steroids are not a reliable way to prevent biphasic anaphylaxis, which is a weaker and more honest claim than a proven negative. The mechanistic mismatch helps explain why: corticosteroids work through delayed, genomic anti-inflammatory pathways that generally take several hours to produce a measurable effect, while most biphasic reactions that do occur cluster within the first several hours after the initial reaction resolves — a window the drug's own onset of action may not reliably cover regardless of whether it works at all for this purpose. Doug's own reaction, moderate-to-severe with wheeze and a rapid twenty-minute onset, places him in a higher-than-average risk category for a biphasic recurrence by symptom severity alone — which makes the actual disposition decision, not the medication question, the one that matters most for what happens to him tonight.
Deciding what actually prevents a second reaction hours from now
Giving a dose of IV corticosteroid before discharge for a reaction like his has been standard practice for years, specifically to prevent a biphasic reaction hours from now when he's no longer under observation. I'd like to continue that practice here — it's low-risk, it's what most of us trained to do, and the downside of a missed biphasic reaction at home is much worse than the downside of an unnecessary steroid dose.
I recognize 'this is what we've always done' isn't itself a strong argument — I'm open to being wrong about this if the evidence has actually moved.
It has moved, and I think it's worth taking seriously rather than defaulting past. Grunau's cohort found corticosteroids did not reduce biphasic reaction rates, and Alqurashi and Ellis's systematic review found twenty-one of twenty-two studies pointing the same way. I'd rather state the strength of that honestly than overstate it: it is all observational, the 2020 practice parameter grades it very low quality, and its conclusion is that steroids are unreliable for this purpose, not that they have been disproven. That was still enough for the parameter to drop routine administration in favor of risk-stratified observation length — based on reaction severity, response to initial treatment, and access to care if symptoms recur — over a medication the evidence doesn't support for this specific purpose. I'd recommend skipping the steroid and instead keeping him for an appropriately extended observation period given how significant his initial reaction was.
'Low-risk and what we've always done' was a reasonable basis for this practice when nobody had looked — it's a weaker one now that everybody who has looked has found nothing, even granting that none of them ran a randomized trial to do it.
I want to add why those negative results make pharmacologic sense, not just report them. Corticosteroids work through genomic, delayed anti-inflammatory pathways that generally take several hours to produce a measurable effect — they're not a fast-acting drug the way epinephrine or even antihistamines are. Most biphasic reactions that do occur cluster within the first several hours after the initial reaction resolves, a window the drug's own onset of action may not reliably cover even if it worked for this purpose at all. The mismatch between how the drug actually works and when the risk it's meant to prevent actually occurs is a real, mechanistic reason the clinical trials came back negative, not just a statistical curiosity.
Agreed: no corticosteroid given. Doug was kept for an extended observation period, six hours rather than the standard two, given the significant severity of his initial reaction, with clear return-precautions and a prescription for two epinephrine autoinjectors before discharge given his new diagnosis. He remained symptom-free through the full observation window and was discharged without incident.