Reviews
Prognostic Nomogram for Supratentorial H3 K27-Altered Diffuse Midline Glioma: IPD Meta-Analysis
Li et alBMC Neurology

Supratentorial diffuse midline glioma (DMG), H3 K27-altered, presents distinct clinical, molecular, and therapeutic characteristics compared to brainstem disease, yet previous prognostic models have pooled anatomical locations. To address this, investigators performed a systematic review and individual patient data (IPD) meta-analysis comprising 354 patients across tier-A per-patient data and tier-B digitized Kaplan–Meier reconstructions to develop the first supratentorial-specific prognostic nomogram.
Because treatment intensity violated proportional hazards, investigators developed a 6-month landmark Cox model from 133 patients (80 events) who were alive 6 months following diagnosis. Predictor selection via backward AIC and LASSO-Cox identified a final four-variable model incorporating treatment intensity (HR 0.684 per level, p = 0.075), Ki-67 labeling index (HR 1.017 per 1%, p = 0.051), ATRX loss, and sex to predict conditional 12-, 24-, and 36-month overall survival.
Internal bootstrap validation indicated modest discrimination with a C-index of 0.570 and time-dependent AUCs of 0.610 at 12 months, 0.711 at 24 months, and 0.835 at 36 months. Calibration was substantially improved, showing bootstrap-corrected slopes of 1.03, 1.13, and 1.19. Nomogram risk tertiles effectively stratified post-landmark median overall survival at 29.9, 18.0, and 8.7 months (p = 0.0137), showing positive net benefit at intermediate-to-high risk thresholds. External assessment showed reasonable generalization in adult thalamic cohorts but identified systematic overestimation of pediatric 24-month survival, indicating that prospective multicenter validation and pediatric recalibration are required prior to clinical implementation.
Key Takeaways
- In patients alive 6 months after diagnosis, nomogram tertiles stratified median post-landmark overall survival into 29.9, 18.0, and 8.7 months (p = 0.0137).
- The four-variable model yielded a bootstrap-corrected C-index of 0.570, with time-dependent AUCs of 0.610 at 12 months, 0.711 at 24 months, and 0.835 at 36 months, and calibration slopes of 1.03, 1.13, and 1.19.
- External validation using aggregate-level pseudo-IPD revealed reasonable short-term and adult thalamic predictions but systematic overestimation of 24-month survival in pediatric cohorts.
Read the original sourceBMC Neurology
