10169-CS-17 Long-term Survival Modeling in Recurrent Glioblastoma Patients Treated with Accelerator-based BNCT

Abstract An extended follow-up analysis was conducted for patients with recurrent glioblastoma who participated in a Phase II clinical trial (JG002) evaluating accelerator-based boron neutron capture therapy (AB-BNCT). The objective was to explore long-term survival patterns and assess the applicability of parametric survival models for future trial planning. Kaplan-Meier estimates revealed a 1-year survival rate of 79.2% (95% CI: 63.3 - 88.7), with median overall survival of 19.2 months (95% CI: 13.1 - 24.8). Parametric modeling using both exponential and Weibull distributions was performed. The exponential model estimated a median survival of 17.2 months with a constant hazard rate of 0.0403/month, assuming time-independent risk. In contrast, the Weibull model, characterized by a decreasing hazard over time, yielded a median of 19.4 months with a shape parameter of 0.85 and scale parameter of 21.0. These findings highlight a potential discrepancy between simplistic constant-hazard assumptions and actual patient survival patterns under active and persistent medical care. While exponential models are commonly employed in early-phase trials due to simplicity and ease of interpretation, the Weibull model may offer a more accurate reflection of clinical behavior in certain populations, particularly in recurrent glioblastoma where patient engagement and therapeutic response may evolve over time. This analysis underscores the importance of model selection when designing prospective studies and interpreting long-term outcomes. Use of the Weibull model should be considered when designing trials that aim to capture realistic survival expectations in populations with complex disease trajectories.