Development and internal validation of a prediction model for 6-month response in children with infantile hemangioma selected for propranolol and polidocanol injection

To develop and internally validate a baseline model for estimating 6-month response in children with infantile hemangioma (IH) already selected to receive oral propranolol combined with intralesional polidocanol, and to explore baseline factors associated with additional-injection burden in the full treated cohort. Children aged 24 months or younger at initiation of combined therapy, treated during the study period of 2012–2025, were assessed. Missing baseline data were addressed using 20 multiply imputed datasets. LASSO was repeated across imputations with a 200-resample stability comparison; the final 10-predictor specification was locked before a separate 1,000-resample conditional bootstrap validation. Additional injections during 6 months were analyzed in all 675 children using robust Poisson regression. Among 675 children, 560 (83.0%) had an effective 6-month response. The locked Stable10 model included lesion location, premature birth, age at combined-therapy initiation, lesion depth, lesion type, baseline echo pattern, treatment pathway, maximum diameter, thickness, and blood-flow grade. For the pooled, shrunken coefficient formula implemented in the calculator, the apparent AUC was 0.920 and the bootstrap optimism-corrected AUC was 0.896 (conditional location-shifted 95% bootstrap interval, 0.886–0.926). The corrected Brier score was 0.095 (0.075–0.102), calibration intercept 0.023 (− 0.028 to 0.064), and slope 0.949 (0.914–1.018). Surface elevation was associated with more additional injections in the whole cohort (adjusted incidence rate ratio, 1.27; 95% CI, 1.01–1.58; P = 0.039). This internally validated model estimates continuous 6-month response probability in children already selected for combined treatment. The injection-burden association is exploratory. Prospective external validation is needed before routine clinical use.