Glioblastoma research presents us with an unusual paradox. Clinically, the last two decades have yielded little progress. The disease remains highly lethal, and the treatment backbone is largely unchanged. Scientifically, however, a lot has happened. We have since redefined how we classify and conceptualize the disease, uncovering molecular complexity that was barely visible twenty years ago. Yet this expansion of knowledge has not led to a comparable expansion of therapeutic outcomes.
Perhaps because clinical outcomes are so central to how we judge progress, the scientific advances themselves can be easy to overlook. For example, for decades the field has been introduced with the same four premises:
- Glioblastoma is the most aggressive primary brain cancer and carries a dismal prognosis.
- The standard of care has remained largely unchanged since the Stupp protocol was introduced in 2005.
- Its pronounced intratumoral heterogeneity makes it difficult to identify robust and effective therapeutic strategies.
- Therefore, we urgently need to understand the disease better to develop more effective treatments
In this way, we are told that the status quo has not changed. What typically follows is a list of molecular features – mutations, chromosomal alterations, and signaling pathways – relevant to the research hypothesis but which are often presented out of context. These features are individually well characterized, but their broader significance is rarely articulated. We accumulate knowledge about increasingly specific aspects of the tumor without integrating the findings into a more coherent understanding of the disease.
This creates a striking mismatch: How we write about the field does not necessarily reflect what the field actually knows. Clinical outcome has become a proxy for scientific progress. Because the treatment has barely improved, the field is repeatedly being presented as though it has little to show for years of research. New studies are being framed as attempts to fill another gap in our understanding. But maybe the more pressing question is no longer simply what we do not know, but how to make better use of what we already do.
The introductory premises are a good and prominent example of the problem. While they present genuine clinical challenges, they also obscure a lot of progress that has been made since:
01 – Standard of Care
Although the treatment backbone has remained unchanged, subsequent trials have refined the standard of care for defined patient groups. For example, tumor treating fields were approved for use alongside temozolomide (showing some of the highest increase in overall survival!), and lomustine was added as an option for patients with MGMT-methylated tumors.
These are not inherent changes to the treatment paradigm but demonstrate instead that meaningful improvements can emerge when treatment is matched more carefully to the patient and disease context. Treatment failure should not simply be interpreted as evidence that a treatment mechanism failed. It matters who receives the treatment.
02 – Disease Definition
Molecular classification has fundamentally changed the definition of the disease. Since 2021, glioblastoma refers specifically to IDH-wildtype tumors meeting defined histopathological criteria. Additionally, the MGMT promoter methylation status has become a recognized predictive and prognostic biomarker.
This enabled more refined molecular stratification in clinical trials and more targeted treatments. However, it also made comparisons across trials more complicated because the patients currently being recruited are different from historical patient cohorts. Contemporary and historical survival estimates are not measurements of the same entity.
03 – Heterogeneity
Multiple molecular states coexist within the same tumor and shift across space and time. Heterogeneity therefore is not simply a set of differing mutations but a dynamic system of cellular states and signaling programs that adapts. The tumor we measure is consequently dependent, at least in part, on where and when we measure it.
All of these are not separate observations. They point to the same problem at different scales: The units we average over do not always match the biology we are trying to understand. At the level of patient populations, treatment effects became visible when we stopped treating them as one group and began identifying relevant subgroups. Across time, the disease definition has changed and became more precise, altering the population to which treatment outcomes refer. Within individual tumors, we realized that what appeared to be a single tumor entity contained cellular states that coexist and change across space and time.
The introductory argument usually ends with an urgent call for better understanding and better treatment. But the need it names is not necessarily a need for more information. It is that our units of observation fail to preserve the biological context that gives it meaning. And if our observations do not reflect the biological structure of the disease, accumulating more measurements within those same units will not resolve the problem.
In other words, our bottleneck has changed. We should not stop generating knowledge, but design studies that allow us to distinguish causal relationships from mere correlations, integrate molecular observations across different biological layers and develop testable models of the disease. The challenge is no longer to add another (“the missing!”) piece to our collection but to meaningfully connect the biological dots.
Leave a Reply