Alopecia Areata: What Actually Works in 2026

Eighty percent of the baricitinib responders who stopped taking it had lost the benefit again within two years. That’s from a randomised withdrawal arm of forty patients, so hold it loosely, but it reframes the decision: these aren’t cures, they’re suppression, and the question in front of a patient isn’t whether to try one but whether to take one indefinitely.

I write almost entirely about androgenetic alopecia, so let me be clear at the top: alopecia areata is a different disease. Not a variant, not a subtype. Different mechanism, different course, different treatment, and a completely different prognosis. If you’ve come here because your hair is receding at the temples, this article isn’t about you, and I have written elsewhere about how pattern hair loss actually works.

Why it is a different disease

In androgenetic alopecia, the follicle itself shrinks over years under androgen signalling. It’s a slow structural change in the organ.

In alopecia areata, the follicle is intact and healthy and is being attacked. Work published in 2014 showed that CD8-positive NKG2D-positive T cells are both necessary and sufficient to induce the disease in mouse models, and that blocking the relevant cytokine signalling prevented it.^[1] The same paper reported three patients treated with an oral JAK inhibitor achieving near-complete regrowth within five months, which is the observation that started the entire modern treatment era.

The difference is worth a picture. In pattern baldness the factory is being slowly downsized. In alopecia areata the factory is fully staffed and working, and someone has locked the doors from outside. Unlock them and production resumes, which is why regrowth here can be total.

The follicle normally protects itself from the immune system through immune privilege: it presents almost no MHC class I or II and it signals actively that there is no danger. Alopecia areata begins with the collapse of that privilege in the anagen hair bulb, after which immune cells move in.^[2]

What is approved, verified today

Three JAK inhibitors are approved in the United States for severe alopecia areata. I checked the current labels directly rather than relying on secondary sources, because this area has moved fast and the approved populations differ.

One live divergence worth knowing if you’re outside the US. The European Union expanded baricitinib to adolescents aged 12 and over in 2026, following a positive committee opinion in February and a Commission decision in April. The US label as of mid-2026 still reads adults only. So in Europe a 14-year-old has two options and in the US only one.

A fourth drug, upadacitinib, published two phase 3 trials in August 2026 with the highest response rates yet reported. It is not approved for alopecia areata,^[15] and you should treat anyone presenting it as available with suspicion.

How well they work

The standard endpoint is SALT 20 or below, meaning 80% or more scalp coverage.

Baricitinib delivered SALT 20 or below in 38.8% and 35.9% at 4 mg across its two replicate trials.^[3] Ritlecitinib reached 23% at week 24 on the 50 mg dose that is actually approved, and 31% in the trial’s 200 mg loading arm, which is not the marketed regimen.^[4] Deuruxolitinib reached 41.5% and 38.3% at 12 mg twice daily.^[5][6] With longer treatment the numbers improve: at three years on ritlecitinib, 47.1% had reached SALT 20 or below on the conservative analysis, and 22.5% had complete regrowth.^[7]

The number nobody leads with

A randomised withdrawal study took patients who had responded to baricitinib at week 52 and re-randomised them to continue or to stop.

Zero percent lost benefit at four weeks, 10 to 11% at eight weeks, and 80% by week 152, against 7% in those who kept taking it.^[8] This is suppression of an ongoing autoimmune process, not a cure of it.

The safety conversation, done properly

All three drugs carry the same boxed warning: serious infections, mortality, malignancy, major adverse cardiovascular events and thrombosis.

Read the labels closely and they say something important. Those risks were observed with another JAK inhibitor compared against TNF blockers in rheumatoid arthritis patients. The warning is a class-wide extrapolation from a trial in a different drug, a different disease and an older, sicker population. Two of the three labels explicitly state the drug is not approved for use in rheumatoid arthritis patients.

What has been observed in alopecia areata itself: across 1,294 patients and 2,091.7 patient-years of ritlecitinib exposure, opportunistic infections occurred at under 0.1%, herpes zoster at 1.5%, malignancy excluding non-melanoma skin cancer at 0.5%, and major adverse cardiovascular events at 0.2%.^[9] A real-world cohort of 17,068 propensity-matched patients with skin immune-mediated disease found JAK inhibitors associated with lower all-cause mortality and lower cardiovascular events than methotrexate or cyclosporine.^[10]

Now the counterweight. Median exposure in the safety dataset was 624 days, in a young and generally healthy population. That’s far too short and too low-risk a cohort to detect the long-latency harms the warning is actually about. The real-world study can’t refute a boxed warning, only fail to reproduce it over two years. Absence of a signal at two years in young patients is reassuring, and it isn’t the same as safety established over decades.

None of these are drugs you simply start, and the monitoring differs by drug more than most summaries admit. All three labels call for tuberculosis evaluation and viral hepatitis screening before the first dose, and blood counts before starting and during treatment. After that they diverge: baricitinib wants baseline liver and kidney function with a lipid check about twelve weeks in, deuruxolitinib wants lipids at baseline and periodically, and ritlecitinib’s label has no lipid requirement at all, asking instead for lymphocyte and platelet counts plus liver enzymes. Live vaccines are to be avoided during treatment, and immunisations brought up to date before you start. Deuruxolitinib alone requires CYP2C9 genotyping, and it is formally contraindicated in poor metabolisers and in anyone taking a moderate or strong CYP2C9 inhibitor. If a prescriber offers you one of these with none of that, get a second opinion.

Two corrections about the natural history

The figure everyone quotes is that 34 to 50% of alopecia areata resolves spontaneously within a year. That figure applies to limited patchy disease, and applying it to severe disease is badly wrong.

A 2025 meta-analysis pooled the placebo arms of 15 trials, covering 893 patients with mostly severe disease. Spontaneous regrowth to SALT 20 or below occurred in 4.68%. Only 1.59% achieved 90% improvement.^[11] The long-horizon picture is the same story told over decades: at a mean 17.7 years of follow-up, 68.3% of patients who started with limited patchy disease were disease-free, against under one in ten of those who started extensive.^[14]

The second correction concerns the treatment dermatologists have injected for fifty years. In the Cochrane network meta-analysis of 63 studies, baricitinib against placebo for 75% or more regrowth gave a risk ratio of 7.54 with a confidence interval of 3.90 to 14.58, rated high certainty. Intralesional corticosteroid against placebo gave a risk ratio of 13.84 with a confidence interval of 0.87 to 219.76, rated very low certainty, from a single study.^[12]

Read those confidence intervals again. The expensive new drug has the tighter, more certain estimate. The first-line standard of care rests on one trial with an interval spanning almost three orders of magnitude. That isn’t an argument against steroid injections, which do work for patchy disease and remain sensible. It’s an argument for noticing that familiarity isn’t evidence.

Where I would start

Get the severity assessed properly by a dermatologist, because that single classification determines almost everything: the odds of spontaneous recovery, whether the JAK evidence applies to you, and whether the boxed warning trade-off is worth making.

If it is severe, take the drug conversation seriously and go into it understanding two things. Roughly a third to a half of patients reach 80% coverage, so it isn’t a certainty. And when you stop, most people lose it again, so the realistic commitment is ongoing.

Be sceptical of anything sold for alopecia areata that isn’t one of these. The Cochrane review looked at 47 treatments across 63 studies. Almost all of it was very low certainty evidence from trials averaging 78 participants. The gap between the marketing volume around this condition and the evidence underneath it is enormous, and it always has been, which is precisely why the arrival of drugs with real trials behind them matters.

One last thing worth saying plainly. Alopecia areata affects roughly 2% of people over a lifetime.^[13] It isn’t rare, it isn’t your fault, and it isn’t caused by stress in the way the folklore insists. It is an autoimmune condition with a known mechanism and, for the first time, treatments built on that mechanism.