Can Systemic Immunomodulatory Therapy Induce Remission of Atopic Dermatitis? A Systematic Review

To the Editor,

Atopic dermatitis (AD) is a chronic inflammatory condition associated with substantial impairment of quality of life. ¹ Disease control ‘on drug’ is now achievable with effective systemic immunomodulatory therapies, including interleukin-4 (IL-4)/IL-13 pathway antagonists and Janus kinase inhibitors (JAKi). ² However, it is unclear whether control is sustainable ‘off drug’ – this is important given immunomodulatory therapies carry risks of adverse events. We conducted a systematic review to evaluate the duration of sustained disease control in AD following discontinuation of systemic immunomodulatory therapy.

This review followed Preferred reporting items for systematic reviews and meta-analyses (PRISMA) guidelines (PROSPERO CRD42024548153). ³ MEDLINE, Embase, Web of Science, and CENTRAL were searched from inception to January 8, 2025 (Supplementary Table 1). Randomised controlled trials (RCTs) and cohort studies enrolling adults (≥18 years) with AD who discontinued systemic immunomodulators after achieving disease control were included. Outcomes were maintenance of ≥75% improvement in the Eczema Area and Severity Index (EASI-75) or Investigator Global Assessment 0/1. Studies investigating conventional immunosuppressants, biologics, and small-molecule therapies were eligible. Studies exploring oral corticosteroids, dual systemic therapy, paediatric populations, and unadjusted cohorts were excluded. Risk of bias was assessed using a modified Newcastle-Ottawa Scale, reflecting the observational nature of post hoc remission cohorts, and certainty was graded using GRADE.^4,5

From 13 397 records, 9 studies (8 RCT, 1 cohort; n = 2558 participants) met inclusion criteria (Supplementary Figure 1 and Supplementary Appendix 1). Therapies included dupilumuab, tralokinumab, amlitelimab, rocatinlimab, rezpegaldesleukin, abrocitinib, baricitinib, and upadacitinib (Supplementary Table 2). No eligible studies assessing conventional systemic immunosuppressants were identified. Treatment duration pre-discontinuation ranged from 12 to 54.8 weeks; follow-up post-discontinuation ranged from 4 to 36 weeks (Supplementary Figure 2). Definitions of disease control and relapse varied considerably, as did concomitant use of topical therapy.

Among IL-4/IL-13 antagonists, approximately one-third of participants who achieved EASI-75 at week 16 maintained EASI-75 at 36 weeks after discontinuation (dupilumab 40.4%; tralokinumab 26.4%), substantially lower than continuation arms (dupilumuab 71.6%; tralokinumab 57.3%). In a small dupilumab cohort treated for a mean 54.6 weeks, 68.2% maintained disease control for a mean 40.5 weeks after cessation, although risk of bias was high (Supplementary Tables 3 and 4).

OX40 pathway antagonists demonstrated greater durability: 61.6% maintained EASI-75 at 28 weeks after amlitelimab withdrawal, and Kaplan-Meier estimates suggested 73% to 96% probability of maintaining EASI-75 at 20 weeks following rocatinlimab cessation. In a small trial of the IL-2 agonist rezpegaldesleukin, 71.4% to 80% of responders maintained disease control at 36 weeks post-withdrawal.

JAKis were associated with more rapid loss of control. At 16 weeks after baricitinib cessation, 45.2% maintained EASI-75, whilst only ~20% maintained EASI-75 after upadacitinib withdrawal. Shorter-term maintenance was observed at 4 weeks after abrocitinib cessation (EASI-75: 42.3%). Retreatment was generally effective, with most participants regaining response.

Overall certainty of evidence was very low due to heterogeneity in treatment duration, remission definitions, follow-up, and incomplete outcome reporting. Despite these limitations, targeted biologics appeared to induce more durable post-treatment control than small molecules. Prospective studies evaluating withdrawal and dose-reduction strategies, alongside non-drug patient-level predictors of remission (such as disease duration and comorbidities), are needed to inform shared decision-making in AD.