Exploration of a simple-indicator prediction model for omalizumab efficacy in chronic inducible urticaria

Chronic inducible urticaria (CIndU) shows variable response to omalizumab, and clinicians lack a parsimonious, clinic-friendly tool to predict benefit from routine data and early symptom change.

We developed and validated a practical prediction model using easy-to-obtain indicators.

We conducted a single-centre, non-interventional cohort at The First Affiliated Hospital of Soochow University, including consecutive provocation-test-confirmed CIndU patients starting omalizumab. The primary outcome was week-12 response, defined as UCT ≥ 12; non-response (UCT < 12) was the modelled event to prioritise early identification of patients unlikely to benefit. Pre-specified predictors were demographics, disease duration, subtype, baseline UCT, log₂(total IgE), basic blood indices, CRP, anti-TPO, and provocation thresholds. Penalised logistic regression with multiple imputation and bootstrap validation was used. A dynamic model added week-4 change in UCT-7 (∆UCT-7).

We analysed 557 patients; 343 (61.6%) were week-12 responders and 214 (38.4%) non-responders. The baseline model showed good performance (AUROC 0.78, Brier 0.19) and near-ideal calibration (calibration slope 0.98; intercept 0.01). Adding week-4 ∆UCT-7 improved discrimination (AUROC 0.84) and accuracy (Brier 0.16). Performance gains were consistent across major CIndU subtypes and in a temporal test set. Decision curve analysis indicated higher net benefit than treat-all or treat-none strategies at clinically relevant non-response risk thresholds. No anaphylaxis occurred and five mild adverse events were recorded.

A simple indicator model, strengthened by an early week-4 update, provides accurate and clinically useful prediction of week-12 omalizumab response in CIndU and supports external validation and prospective impact studies.