Development and validation of a prediction model for 6-month early death in pediatric H3K27M-mutant diffuse intrinsic pontine glioma (DIPG).

Background: H3K27M-mutant diffuse intrinsic pontine glioma (H3K27M-DIPG) is among pediatric brain tumors with the poorest prognosis. Early identification of high-risk patients may facilitate timely, individualized clinical management. This study aimed to identify independent predictors of early death (ED), defined as death within 6 months of diagnosis, in pediatric patients with H3K27M-DIPG and to develop and validate a clinical prediction model to estimate individualized ED risk in this population. Methods: We retrospectively collected clinical data from 147 pediatric patients in the Surveillance, Epidemiology, and End Results (SEER), including sex, race, tumor grade, surgical approach, radiotherapy, chemotherapy, tumor size, age, and time from diagnosis to treatment. Death within 6 months of diagnosis was used as the primary outcome. To minimize the risk of omitting critical confounders while preventing overfitting, variables with P<0.15 in the univariable analysis were selected for multivariable logistic regression to construct a streamlined nomogram. The discriminatory ability of the model was evaluated by the area under the curve (AUC). Internal validation was performed using 500 bootstrap resamples to calculate the optimism-corrected C-index and verify model stability. Results: Following the univariable screening, four variables (radiotherapy, chemotherapy, time to treatment, and tumor grade) were incorporated into the multivariable model. Multivariable logistic regression analysis showed that non-receipt of radiotherapy was the sole independent risk factor for ED [odds ratio (OR) =8.00, 95% confidence interval (CI): 2.26-28.30, P=0.001]. The developed nomogram demonstrated excellent discrimination, and the apparent C-index (AUC) was 0.732 (95% CI: 0.607-0.854). After internal validation with 500 bootstrap resamples, the optimism-corrected C-index remained stable at 0.702. Conclusions: This study successfully developed and internally validated a streamlined 4-variable prediction model to estimate the risk of 6-month ED in children with H3K27M-DIPG. Radiotherapy status emerged as the singular most critical factor associated with ED. This model may assist clinicians in early identification of high-risk patients, facilitate transparent clinician-family communication, and support individualized treatment planning and counseling.