PFAPA at Two Years In: When a Steroid That Works Is Still the Wrong Long-Term Plan
The abortive steroid dose has never failed to end an episode within hours. The actual argument is whether that success is quietly making the next fever arrive sooner.
Maya S. is four years old and has had a fever every three to five weeks since she was two, each episode following the same script closely enough that her parents can often predict it a day before it starts: a sore throat, then tender, swollen glands on both sides of her neck, then small ulcers inside her cheeks, then fever climbing to 39-40°C for two to five days before resolving completely on its own. Between episodes she is entirely well — eating, growing, and developing normally, with no fevers, no joint symptoms, and a completely reassuring physical exam and CBC drawn during a well interval six months ago. The diagnosis, made clinically after two other pediatricians and a normal infectious workup, is PFAPA, and it has held for eighteen months without a single episode breaking the pattern.
A single dose of betamethasone given at the very start of an episode has, every time it has been tried, ended the fever within six to twelve hours — a response so reliable her parents now request it by name at urgent care visits rather than waiting to see if the episode resolves on its own. What complicates treating that response as an unambiguous win is a consistent observational finding, reported by Kyvsgaard and colleagues and again in Butbul Aviel's 365-child cohort, where 46 percent of steroid-treated children had their inter-episode interval shorten: the drug appears to trade a longer, milder course for a shorter, harder-stopped one. No randomized trial has tested that trade directly. Her own episode log over the past six months shows exactly that pattern — intervals of five weeks before steroid use, closer to three weeks in the months since starting it, though eighteen months of episodic data in one child cannot distinguish a true drug effect from ordinary variation in an already-cyclical illness. Tonsillectomy, the other well-studied option, has shown real effectiveness in the only two randomized trials that exist, Renko's and Garavello's, but carries its own surgical and anesthetic risk in a four-year-old whose disease, left alone, is expected to eventually resolve on its own within a few more years regardless of what is done now.
Her parents have missed a combined nine days of work this year, and daycare has twice asked whether she should be excluded during febrile weeks even though PFAPA is not contagious. That accumulated cost, not any single fever, is what has actually brought the three of them into the same room to decide what happens next.
What a steroid that never fails is actually costing
I want to be honest about what we actually know here: the steroid has ended every single episode within hours, which is a remarkably clean, reproducible response. The shortened-interval effect is real in cohort data — roughly half of steroid-treated children in the largest series — and her own log is at least consistent with it, but it has never been tested in a randomized design, and eighteen months in one child is not enough to separate a drug effect from the ordinary variability PFAPA already has cycle to cycle.
I'd rather keep using a therapy that reliably works than move to surgery in a four-year-old on the strength of an interval pattern we can't yet call causal.
I take the uncertainty point, but I'd frame the cost differently: nine missed workdays and repeated daycare exclusion questions over a single year is not a small burden just because each individual fever resolves quickly. Renko's trial randomized children to tonsillectomy versus follow-up alone and the episodes ceased in every child who had the surgery — not a partial symptomatic fix, an end to the pattern. I'll say the uncomfortable half of that result too: half the control children remitted on their own inside six months.
I'd also push back gently on treating 'the steroid always works' as reassuring on its own. It works for the fever in front of us. It hasn't yet been shown to work for the actual question this family is asking, which is when this stops being their life every three weeks.
That last point is the one I'd build on, and it cuts against surgery tonight. If half of untreated children in Renko's own control arm remitted within six months, we are considering a general anesthetic for a disease that may be ending on its own. Daily prophylaxis — colchicine or cimetidine — has real, if modest, evidence behind it: Feder and Salazar found cimetidine interrupted attacks in about a quarter of children while colchicine lengthened intervals in a majority, and Butbul Aviel's small randomized trial supports colchicine specifically. It's reversible in a way neither of the other two options is.
If it meaningfully lengthens her interval, it may resolve both concerns at once without ever forcing tonight's choice. If it doesn't, we're back exactly where we started, having lost very little by trying.
A three-month trial of daily colchicine was started, with the abortive betamethasone kept on hand for any breakthrough episode rather than discontinued — the family's explicit request, since giving up a treatment that has never failed felt riskier to them than adding a second one.
Not agreed, and stated as an open branch point rather than smoothed over: if the prophylaxis trial fails to lengthen her interval, the Otolaryngologist and Pediatric Rheumatologist still disagree about what should happen next — one favoring tonsillectomy as the definitive answer to two years of documented burden, the other wanting a second prophylactic agent tried first given surgery's irreversibility in a child this young. Both agreed only that the three-month mark, not tonight, is when that specific disagreement needs resolving.