Answers are grounded in this paper and cite their sources.
Population registries support large-scale studies of glioma outcomes, but historical histology codes and incomplete molecular data preclude direct alignment with modern WHO-defined diffuse glioma entities. We examined treatment patterns, survival, and early-mortality prediction among adults with registry-coded malignant glioma in a SEER export covering 2000–2021. We included adults with positive histologic confirmation, malignant behavior, ICD-O-3 9380/3, 9381/3, or 9382/3, and non-missing survival time. Zero-month survival records were retained and assigned 0.5 months for time-to-event analysis. Treatment was classified into eight patterns defined by combinations of surgery, radiotherapy, and chemotherapy. Analyses included Kaplan-Meier estimation, multivariable Cox models, inverse probability of treatment weighting among surgically treated patients, a 3-month landmark analysis, a 2018–2021 grade-adjusted analysis, and machine-learning models for 12-month mortality among patients with ascertainable 12-month status. The analytic cohort included 4,555 patients, including 217 with zero-month survival. Overall, 2,522 deaths occurred. Reverse Kaplan-Meier median follow-up was 127.0 months, whereas the raw median time to death or last contact was 37.0 months. The most frequent treatment patterns were surgery only (1,254 patients) and triple therapy (1,219 patients). With surgery only as the reference, adjusted HRs were 1.48 (95% CI 1.32–1.67) for triple therapy and 1.97 (95% CI 1.69–2.30) for radiotherapy plus chemotherapy; these estimates were interpreted as observational associations because of substantial case-mix differences. IPTW reduced measured covariate imbalance, although residual confounding remained likely. Treatment estimates differed in the 2018–2021 grade-adjusted subset, consistent with confounding by disease severity, calendar era, and subset composition. Gradient boosting yielded an AUROC of 0.839 (95% CI 0.814–0.860) and a Brier score of 0.134 for baseline-available 12-month mortality prediction. A further 324 patients censored alive before 12 months were excluded because their endpoint status was indeterminate. This SEER export supports observational analyses of registry-coded malignant glioma, but not causal treatment comparisons or inference for modern WHO-defined molecular subgroups. The findings define the interpretive limits of historical registry-coded glioma data and indicate that baseline registry variables can support internally validated early-mortality risk stratification.
No takes yet. Share an insight, caveat, or question.
Chen et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: