Designing a web-based platform for dynamic estimation of individualized conditional survival in grade 3 gliomas

Grade 3 gliomas, including anaplastic astrocytoma and anaplastic oligodendroglioma, are aggressive brain tumors with heterogeneous prognoses. Existing survival models often fail to account for time-dependent changes in risk, limiting their utility in long-term survivorship planning. This study aimed to characterize dynamic survival trends and construct a conditional survival (CS)-based nomogram to facilitate individualized risk prediction.

Data from 7278 patients with histologically confirmed grade 3 glioma were extracted from the SEER database (2000-2021). CS and annual hazard rate analyses were conducted to assess time-varying survival probabilities. Prognostic variables were selected using best subset regression (BSR), least absolute shrinkage and selection operator (LASSO), and stepwise regression methods. A multivariable Cox model incorporating key predictors was used to construct a CS-nomogram, which was validated using calibration curves, time-dependent ROC analysis, and decision curve analysis (DCA). A risk stratification system was developed, and an interactive web-based survival calculator was created for clinical application.

CS analysis revealed that 10-year survival probability increased from 29% at diagnosis to 67% after surviving 5 years. The annual hazard rate declined from 29.06% in the first year to 1.8% by the tenth year. Seven independent prognostic factors-age, tumor site, histology, tumor stage, surgery, radiotherapy, and chemotherapy-were incorporated into the final CS-nomogram, which demonstrated strong discrimination (AUCs for 3-, 5-, and 10-year survival > 0.75) and excellent calibration. Risk scores derived from the model effectively stratified patients into high- and low-risk groups with significantly different survival outcomes (P < 0.001). An online calculator was developed to enable real-time, patient-specific survival prediction.

This study provided a comprehensive assessment of dynamic survival patterns in grade 3 glioma and introduced a validated CS-based nomogram and web-based tool for personalized prognostic evaluation. These resources offer valuable guidance for treatment planning and long-term follow-up in clinical practice.