Early inhaled corticosteroid reduction as a dynamic indicator of treatment response in children receiving house dust mite subcutaneous immunotherapy

Background: Clinical response to house dust mite subcutaneous immunotherapy (HDM-SCIT) in children is heterogeneous. Existing prediction approaches rely mainly on baseline characteristics and do not capture early treatment-related change. Objective: To evaluate whether early change in inhaled corticosteroid use within 3 months (ΔICS) provides additional predictive information beyond baseline characteristics for HDM-SCIT response stratification. Methods: Two ordinal regression models were developed. The base model included age, sex, baseline ICS dose (ICS0), FEV1baseline, eosinophils, FeNOcat25, tIgEbaseline, HDM_max, and ratio_max. The extended model additionally included early ICS change at 3 months (ΔICS), scaled per 100 μg/day (ΔICS_100). Model performance was assessed using AIC, discrimination across four classification tasks, and calibration. Restricted cubic spline analysis was used to examine the association between ΔICS and high-response probability. Results: A total of 470 children were included, including 159 high responders, 151 middle responders, and 160 low responders. Adding ΔICS_100 significantly improved model fit (AIC 762.5 ± 3.3 vs. 871.2 ± 4.1; pooled Wald test, P < 0.001). The extended model showed better discrimination for high versus non-high response (AUC 0.738 to 0.878), middle versus non-middle response (0.613 to 0.725), low versus non-low response (0.820 to 0.842), and high versus low response (0.856 to 0.928). Restricted cubic spline analysis demonstrated a significant non-linear association between ΔICS and high-response probability (overall P < 0.001; non-linearity P < 0.001), with an inflection region around -80 to -50 μg/day. Conclusion: Early change in inhaled corticosteroid use (ΔICS) may provide additional predictive information beyond baseline variables in children receiving HDM-SCIT. The observed non-linear association suggests that ΔICS may be a useful exploratory early treatment-related indicator, but external validation is required before clinical application.