Abstract
Despite decades of effort, outcomes for recurrent high-grade glioma remain poor. A central challenge in therapy development is that early-phase trials often fail to demonstrate whether an investigational agent meaningfully engages its molecular target within the central nervous system. Retrospective analyses have greatly advanced our understanding of glioma heterogeneity and evolution 1–4 but are inherently susceptible to sampling bias, temporal evolution between surgeries, and immortal time bias that distorts associations between molecular features and clinical outcomes5 (Figure 1A). Traditional Phase I-IV clinical trials (Figure 1B), while essential for establishing safety and efficacy, typically follow standardized workflows from resection through postoperative therapy and longitudinal follow-up but provide limited opportunity for paired, mechanistic tissue sampling. Window-of-opportunity trial designs (Figure 1C) address a long-standing gap by administering therapy preoperatively and analyzing tumor tissue from resection, allowing measurement of drug penetration and target modulation in situ. These studies can leverage the interval between biopsy and planned surgery to obtain paired pre- and post-treatment samples, providing a framework for causal inference. Phase 0 trials, which typically use short course, microdosed or limited therapeutic exposure to assess pharmacokinetics and early pharmacodynamic target engagement, offer a complementary but distinct strategy for interrogating mechanism of action. Although the two designs can overlap, each address different mechanistic questions in specific clinical and scientific contexts.