After remission in atopic dermatitis, oral drugs relapse early when stopped

Design

Systematic review and network meta-analysis (55 studies; NMA on 9)

Population

Patients with atopic dermatitis in remission on systemic therapy (2,585 observations, 13 regimens)

Primary outcome

Maintained EASI75 at 16 to 52 weeks after withdrawal, reduction or interval extension

Effect

Early relapse after stopping oral drugs; lebrikizumab withdrawal or spacing had the lowest risk of losing EASI75 in the network

A systematic review and network meta-analysis included 55 studies in patients with atopic dermatitis who reached remission on systemic therapy and then withdrew treatment, cut the dose or extended the interval. The primary outcome was keeping a 75% reduction in EASI (EASI75) 16 to 52 weeks later.

Raw pairwise data showed that stopping oral treatments led to early relapse, while most active treatments gave similar EASI75 maintenance in the unadjusted data. In the network analysis, built on 9 studies, 26 pairwise comparisons and 2,585 observations across 13 regimens, withdrawal or interval extension of lebrikizumab carried a lower estimated risk of losing EASI75 than the other agents.

The network rests on few studies and, with industry-led trial designs, cannot rank agents confidently. The practical message is about route: biologics appear more forgiving of spacing than oral drugs, and planned breaks, such as around vaccination or pregnancy, are safer to plan around that difference.

• Expect earlier relapse when an oral systemic agent is stopped in a patient in remission.
• Biologic interval extension kept remission more often in the pooled analysis.
• Plan review soon after any withdrawal, especially of oral treatment.
• Do not use the single-agent ranking to choose a drug; the network is small.

Why it matters

It gives a basis for different planning when a patient wants a treatment break, depending on whether they take an oral drug or a biologic.

Don't overread it

The agent-level ranking is built from few studies and should not guide drug choice.

The statistics, in plain English

A network meta-analysis estimates comparisons between treatments that were never tested head to head, so its rankings are only as strong as the links between trials. Nine studies across 13 regimens leaves each comparison thinly supported.

Read the rest in the app

You have read your two free briefings this month. The app carries all 27 specialties, every morning, free — and this finding is waiting in it.

Get Daily Doctor — free

Scan to keep reading on your phone. No account needed to start.