Beating the odds: extreme long-term survival with glioblastoma

See the article by Gerber et al, on pages 1186–1195.

Amongst all cancers, the diagnosis of glioblastoma multiforme (GBM) portends one of the most malevolent clinical prognoses. Median survival is only nine months, rising to 15–16 months for those receiving standard of care surgery and adjuvant chemoradiation.^1–3

Nevertheless, GBMs exhibit a substantial degree of heterogeneity in both their molecular and clinical profiles. Several dozen somatic genetic events have been identified as likely “drivers” of GBM tumorigenesis, and individual GBMs can contain countless combinations of these.⁴ GBM methylation profiles tend to divide into two major classes which themselves associate with the presence or absence of IDH1/2 mutations.⁵ GBMs can also be sorted into at least four classes based upon their expression profiles, and each of these is enriched for specific genetic and epigenetic features.⁶ Beyond this, individual GBMs exhibit substantial intratumoral genetic heterogeneity, and the degree of this heterogeneity itself varies between tumors.⁷ Clinically, GBMs can exhibit varying responses to treatment, with 12% of patients who receive standard chemoradiation surviving four years after diagnosis.⁸

Initial investigations into factors influencing differential survival in GBMs identified advanced age, pre-existing neurologic impairment, and poor functional status with worse outcome in patients.⁹ Furthermore, the extent of resection, at either initial diagnosis or on tumor recurrence, has been clearly shown as an important predictor of survival.^10,11 These features have been integrated into a glioblastoma prognostic model.¹² More recently, several molecular signatures have been identified to predict favorable patient response to treatment as well as overall outcome in gliomas. These include chromosomal arm-level events such as loss of heterogeneity at 1p and 19q,¹³ epigenetic silencing of MGMT through methylation,¹⁴ the presence of IDH1/2 mutations,¹⁵ and favorable expression signatures.¹⁶ Regions of chromosomal instability have also been credited with worse prognosis in GBM patients.¹⁷

However, few studies have focused on features that predict extreme long-term survival. These patients exhibit a disease course that is qualitatively different from the vast majority of GBM patients. For example, half of patients who have survived four years will survive four more.¹⁸ Do these patients simply exhibit all the features associated with good prognosis in the overall GBM population, or does their extreme survival relate to a different set of features of the disease? A detailed answer to this question has the potential not only to enable refinement of prognostic classes, but also to indicate therapeutic strategies to convert more patients into extreme long-term survivors.

In their article “Transcriptomal diversity of long-term glioblastoma survivors”, Gerber and colleagues begin to address this question. Their study evaluated molecular patterns associated with glioblastoma patients with survival past four years, using data from patients at a single institution (MSKCC) and extending their results to the larger TCGA and REMBRANDT databases. They found that extreme survivors frequently exhibit methylation of MGMT, suggesting that responsiveness to temozolomide is a major contributor to extreme survival. Surprisingly, however, they did not identify a preponderance of IDH-mutant tumors among long-term survivors. They also found no relationship between long-term survival and membership of one of the four expression-based subclasses of GBM as determined by TCGA. Similarly, the study defining these subclasses found no association with increased survival in any of these groups.⁶

Although an important start, this study leaves several questions unaddressed. First, because extreme long-term survivors are by definition rare, only a small number could be included in this study. It will be important to generate data on larger cohorts to obtain sufficient power to detect robust differences between extremely long-term survivors and the general GBM population. Additionally, features of these tumors that were not assessed may be important in determining extremely long-term survival. Chief among these are intratumoral heterogeneity, both genetic and epigenetic, which may determine the uniformity of treatment response among cells within the tumor.^19,20 Moreover, all prognostic features will need continued reassessment as our treatment strategies evolve. The inability to show improvements in survival with targeted therapeutics, despite a wealth of druggable targets in GBMs, has been a great failing of the field. Hopefully, a greater understanding of why some patients survive so long will enable us to determine why, despite our best efforts, most do not.