To compare the relative effects of different treatment strategies for Systemic Juvenile Idiopathic Arthritis (sJIA) on clinical inactivity (CID)/remission and safety using a Bayesian network meta-analysis and to rank the treatment options. PubMed, Embase, Web of Science, the Cochrane Library, and Scopus were systematically searched from inception to October 2025 according to PRISMA 2020 and PRISMA-NMA specifications. Randomized controlled trials (RCTs), prospective cohorts, and retrospective/registry studies were included. The primary outcome was the CID/remission rate, and the secondary outcome was the incidence of adverse events (AEs) (normalized per patient-year). Random-effects and Bayesian hierarchical network models were used. Risk of bias was assessed using ROB-2 and NOS, and the certainty of evidence was assessed using GRADE. A total of 2,408 records were retrieved. After deduplication, 1,088 records were screened for titles/abstracts, and 200 were reviewed in full text. Fifteen studies (total sample size 1,548) were ultimately included. A pooled analysis showed that approximately two-thirds of patients achieved CID/remission during follow-up (random-effects model). Stratification suggested that an early IL-1/IL-6 strategy had a more favorable response rate. A network meta-ranking analysis (SUCRA) showed that early IL-1/IL-6 blockade ranked first, followed by anakinra and canakinumab; tocilizumab was in the middle; and conventional care and mixed biologic strategies ranked last. The overall AE rate was low and consistent across studies; IL-1/IL-6 targeted drugs had a favorable safety profile. Multiple sensitivity analyses and cumulative evidence analysis supported the robustness of the conclusions. In sJIA, immediate intensive IL-1/IL-6 therapy at diagnosis is more effective than a step-up approach in achieving and maintaining CID/remission, while also having an acceptable safety profile, supporting the initial “target-to-treat (T2T)” strategy. Future prospective studies and biomarker-stratified validation of each strategy are needed.
Liu et al. (Mon,) studied this question.