A Letter to JAAD Asks What a Retrospective Review of Switching Alopecia Drugs Can and Cannot Prove

A letter published in the October 2026 issue of the Journal of the American Academy of Dermatology does something journals rarely make room for: it interrupts the momentum of a widely covered finding to ask whether the finding can actually support the conclusion everyone drew from it. The letter, by Ke Xu, is formatted as a comment on a separate paper, "Switching between Janus kinase inhibitors for treatment of alopecia areata: A multi-center retrospective review," which ran in JAAD in January 2026 and tracked 108 patients with severe alopecia areata who switched oral JAK inhibitors after at least six months on a prior drug.

The original study reported that roughly two-thirds of patients saw some improvement on their first JAK inhibitor, and that after switching to a second agent, about 48.8% reached a SALT score of 20 or below, a standard measure of scalp hair loss. A third switch, tried in a small subgroup, produced a response in roughly half of patients. Coverage of the study, including write-ups from the National Alopecia Areata Foundation and HealthDay, framed these numbers as evidence that a drug that fails one patient may still work when swapped for another.

What a retrospective chart review of this kind cannot establish, by design, is why the second drug worked when the first did not. Patients who switched had already been on treatment for six months or longer before the swap, and alopecia areata is a disease known for delayed response kinetics: JAK inhibitors can take many months to show their full effect on hair regrowth. A patient who improves shortly after switching drugs may be responding to the new agent, or may simply be reaching the point in a slow response curve where the first drug would have worked anyway, given more time.

The disease's own natural history adds a second confound that a single retrospective cohort has no way to rule out. A prospective 52 week natural-history study of alopecia areata found a 31.8% spontaneous relapse rate by week 52, with baseline severity and prior relapse history, not treatment choice, as the strongest predictors of what happened next. Separately, clinical data on tofacitinib discontinuation found relapse in 45.7% of patients within one to two months of stopping the drug and 38.9% within four to six months, underscoring how much this disease moves on its own schedule independent of which JAK inhibitor a patient happens to be taking.

The January study's own authors acknowledged the core limitations: retrospective design, treatment heterogeneity across sites, and selection bias in which patients were offered a second or third agent in the first place. Those caveats are standard boilerplate in a methods section, but the invited comment format exists precisely to press on what boilerplate caveats mean in practice. A retrospective review can describe what happened to 108 patients who switched drugs. It cannot, on its own, separate a true drug-specific effect from the combined pull of longer cumulative treatment time, delayed onset, and the disease's tendency to wax and wane regardless of intervention.

The finding that did hold up under a more rigorous statistical test is narrower than the headline numbers suggest: the original study reported that a patient's response to their first JAK inhibitor predicted their response to the second, with an odds ratio of 3.33 (95% CI 1.22 to 9.68, p=.022). That is a real, quantified association between two measured outcomes in the same patients, distinct from the uncontrolled observation that switching drugs was followed by improvement. One finding describes a correlation within the data; the other describes a sequence of events that a retrospective design was never built to explain.

Two smaller, independently reported studies on the same question sit in the background of this exchange without resolving it. A single-center study of patients switched from baricitinib to ritlecitinib found SALT scores improving by roughly 20 percentage points across 16 patients, and a separate Yale-led case series followed about 13 patients who switched agents between 2014 and 2023. Neither is large enough to adjudicate the confounding problem the JAAD comment raises, but their rough consistency with the 108-patient cohort is itself a data point: whatever is driving the improvement after a switch, it shows up across several small, differently constructed retrospective samples, not just one.

None of this settles whether switching JAK inhibitors is the right clinical move for a patient who has stalled on their first agent. It settles only what kind of evidence currently exists to answer that question, and what kind does not. A prospective trial that holds treatment duration constant across arms, something closer to what the FDA required before approving baricitinib, ritlecitinib, and deuruxolitinib for alopecia areata in the first place, would be needed to tell a true drug effect apart from a patient simply running out the clock on a slow-acting first drug. Until that trial exists, the 108-patient review remains a description of what clinicians observed in practice, filed eight months later alongside a letter asking what, precisely, that observation is evidence of.