Risk-updating nomogram for refractory systemic JIA in children: multicentre development and external validation

To develop and externally validate a pragmatic prognostic model for identifying children at risk of refractory systemic JIA (rSJIA) by combining baseline disease activity with early treatment-response information.

In a multicentre retrospective cohort of newly diagnosed sJIA, we prespecified baseline predictors (systemic Juvenile Arthritis Disease Activity Score, sJADAS27; refractory rash, rerash) and an early response marker (time to sustained clinical inactive disease, tCID). We developed an update model (sJADAS27 + rerash + tCID) and a baseline-only model (sJADAS27 + rerash + CRP ≥75 mg/l), and assessed discrimination and calibration in training, internal testing and pooled external validation cohorts.

Among 128 children (training n = 55; internal testing n = 25; external validation n = 48), 42 (32.8%) met the prespecified composite rSJIA outcome during follow-up. In the update model, sJADAS27 (OR 1.34 per 1-point increase, 95% CI 1.06-1.69) and rerash (OR 7.38, 95% CI 1.07-51.16) were independently associated with rSJIA; tCID showed a positive but non-significant association (OR 1.04 per day, 95% CI 0.99-1.08). The update model showed excellent discrimination in internal testing (AUC 0.919, 95% CI 0.803-1.000) and external validation (AUC 0.907, 95% CI 0.818-0.996). The baseline-only model discriminated well in internal testing (AUC 0.882, 95% CI 0.746-1.000) but less well in external validation (AUC 0.803, 95% CI 0.679-0.926). The update model outperformed sJADAS27 alone in both cohorts (DeLong P = 0.040 and 0.019).

A nomogram-based approach integrating baseline disease activity with early response information enables risk updating after initial treatment and may support timely treatment individualization for rSJIA.