Development and aggregate-level external corroboration of a prognostic nomogram for supratentorial diffuse midline glioma, H3 K27-altered: an individual patient data meta-analysis

Purpose Supratentorial diffuse midline glioma (DMG), H3 K27-altered, differs from brainstem disease in demographics, molecular composition, and therapeutic feasibility, yet all published prognostic models pool anatomical sites. We aimed to develop and externally corroborate the first supratentorial-specific prognostic nomogram using individual patient data (IPD). Methods We performed a systematic review with IPD meta-analysis, assembling published supratentorial DMG series with per-patient data (tier A) and digitized Kaplan–Meier reconstructions (tier B), with cross-cohort de-duplication. Missing covariates were handled by multiple imputation (m = 20). Predictors were selected via backward AIC and LASSO-Cox; because treatment violated proportional hazards, a 6-month landmark Cox model was developed and translated into a nomogram predicting conditional 12-, 24-, and 36-month overall survival (OS). Internal validation used bootstrap correction; external generalization assessment used tier-B pseudo-IPD at the cohort and subgroup level. Results The cohort comprised 354 patients; 190 tier-A patients (129 events) formed the survival-analysis set, and 133 patients (80 events) alive 6 months after diagnosis formed the landmark modeling cohort. The final four-variable model included treatment intensity (borderline; HR 0.684 per level, p = 0.075), Ki-67 (continuous; borderline association, HR 1.017 per 1%, p = 0.051), ATRX loss, and sex. The bootstrap-corrected C-index was 0.570 (modest discrimination); time-dependent AUCs were 0.610, 0.711, and 0.835 at 12, 24, and 36 months; calibration was substantially improved, with bootstrap-corrected calibration slopes of 1.03, 1.13, and 1.19. Nomogram tertiles stratified median post-landmark OS at 29.9, 18.0, and 8.7 months ( p = 0.0137), with positive net benefit at intermediate-to-high risk thresholds. Aggregate-level external generalization showed reasonable short-term and adult thalamic predictions but systematic overestimation of pediatric 24-month survival. Conclusion We present the first supratentorial-specific prognostic nomogram for H3 K27-altered DMG, providing risk stratification with modest discrimination but substantially improved calibration for patients alive 6 months after diagnosis. Prospective multicenter validation and pediatric recalibration are required before clinical implementation. Clinical trial number Not applicable.