The siren song of bevacizumab: swan song or clarion call? Free

Bevacizumab burst onto the neuro-oncology scene in 2005 based upon dramatic responses in recurrent glioblastoma patients receiving this agent with the ineffective drug irinotecan. Acceptable toxicity and superior radiographic response rates compared with standard salvage therapy, confirmed in pivotal uncontrolled phase II trials,^1,2 led to accelerated FDA approval in 2009. Moreover, many patients were able to reduce their corticosteroid dose, suggesting steroid-sparing properties.³ Nonetheless, these reports, which focused on progression-free survival (PFS) and radiographic response, did not suggest seismic improvement in overall survival (OS). Thus, concerns persisted that the agent’s apparent activity might principally or wholly reflect anti–vascular endothelial growth factor mediated stabilization of the blood–brain barrier, creating a “pseudoresponse.”

With the initial FDA accelerated approval based upon a surrogate imaging endpoint, additional studies were mandated to confirm clinical benefit to warrant full FDA approval. Subsequently, 2 placebo-controlled phase III trials adding bevacizumab to standard chemoradiation in newly diagnosed glioblastoma failed to demonstrate improved OS despite unequivocal benefits in PFS.^4,5 These 2 trials yielded divergent results on bevacizumab’s impact upon quality of life and neurocognition, further obscuring the role of this agent. A limitation of both trials was crossover from the placebo arm to bevacizumab at tumor recurrence, potentially masking survival benefit. Against this backdrop, a randomized phase II trial (“BELOB”) in the Netherlands (where bevacizumab was unavailable, minimizing crossover) suggested improved survival in recurrent glioblastoma with the combination of bevacizumab and lomustine compared with either single agent.⁶

With bevacizumab’s survival benefit still unproven, it was recognized that the recurrent glioblastoma setting provided such an opportunity. The European Organisation for Research and Treatment of Cancer (EORTC) trial 26101, originally a randomized phase II trial exploring sequencing of lomustine and bevacizumab, was converted to an unblinded phase III study comparing lomustine with lomustine/bevacizumab based upon the BELOB results. The study population was glioblastoma at first relapse, with OS as the primary endpoint. Quality of life, neurocognition, and steroid dosing were monitored, and tumor tissue and MR scans were collected for methylation profiling and central radiology review. Treatment at progression was at the investigator’s discretion. Investigators randomized 437 patients in a 1:1 ratio to the 2 study arms.

Unfortunately, EORTC 26101 failed to demonstrate an OS benefit from the addition of bevacizumab to lomustine (median OS, 9.1 mo) compared with lomustine alone (8.5 mo) despite an improvement in PFS from 1.5 to 4.2 months.⁷ There was no tail to the survival curve in either study arm, and no subgroup or factor (including older age, increasing tumor diameter, or steroid requirement at study entry) predicted benefit from combination therapy. Approximately half the patients on each arm were not on corticosteroids at study entry, and the time until patients required steroid institution did not differ between arms. There were also no major differences in cognitive function or quality of life between the 2 arms. Thus, there was no apparent clinical benefit from the addition of bevacizumab to lomustine.

The marked PFS benefit but convincingly negative OS results raise the question of whether postprogression crossover of patients in the lomustine monotherapy arm to bevacizumab might have masked OS benefit—a concern similarly raised with the Radiation Therapy Oncology Group (RTOG) trial 0825⁴ and AVAglio,⁵ where crossover was 48% and ~30%, respectively. Of note, a reanalysis of the results of AVAglio limited to patients who received no treatment for recurrent disease demonstrated that the bevacizumab arm survived 3.6 months longer than the placebo control arm.⁸ In EORTC 26101, 35.5% of patients on the lomustine monotherapy arm received bevacizumab at progression. Complex models do exist that attempt to control for the effects of crossover on OS in oncology studies.⁹ Whether such studies are planned for EORTC 26101 is unknown. However, it is of interest that in EORTC 26101 there is apparent modest initial separation of the OS curves favoring the combined therapy arm with subsequent convergence of the curves, suggesting possible modest effect of crossover on OS.

Interesting data emerged on the role of nitrosoureas in recurrent glioblastomas. Although all patients had previously received temozolomide and thus presumably had tumors with some degree of temozolomide resistance, O⁶-methylguanine-DNA-methyltransferase (MGMT) status was nonetheless a powerful prognostic factor. The median PFS in MGMT-methylated patients was 5.7 months compared with 2.8 months in unmethylated patients, with a hazard ratio for progression or death of 0.37. Overall survival was similarly superior in patients with methylated tumors (median 13.5 vs 8.0 mo, hazard ratio 0.48), and as would be predicted, the favorable effect of MGMT status was independent of bevacizumab. While lomustine monotherapy was ineffective against unmethylated tumors (median PFS 1.5 mo), it was more effective against methylated ones (median PFS 3.0 mo). With a 6-month PFS of 30%, lomustine monotherapy was a modestly effective agent in hypermethylated tumors. These findings are highly reminiscent of the activity of dose-dense temozolomide in recurrent glioblastoma in the DIRECTOR study.¹⁰ Thus, EORTC 26101 adds to the data that alkylating agents that penetrate the blood–brain barrier, like nitrosoureas and temozolomide, are reasonable options in recurrent hypermethylated glioblastomas, and consideration should be given to stratifying randomized recurrent glioblastoma trials on the basis of MGMT methylation status.

Despite the lack of OS benefit in several carefully controlled glioblastoma trials, most neuro-oncologists continue to believe there is a role for bevacizumab. How do we ascertain which patients will benefit? The search for biomarkers continues, and while EORTC 26101 has not yet been mined, other studies have raised possibilities. For example, AVAglio trial investigators using a NanoString gene expression platform to classify glioblastomas according to their subtype in The Cancer Genome Atlas found a survival benefit in the isocitrate dehydrogenase wildtype proneural subtype following treatment with bevacizumab.¹¹ The BELOB trial found a trend toward benefit from bevacizumab in patients with the classical glioblastoma subtype based upon gene expression profiling and RNA-seq.¹² Other studies using imaging biomarkers have suggested that apparent diffusion coefficients on diffusion MRI,¹³ relative cerebral blood volume on perfusion MRI,¹⁴ and radiomic signatures¹⁵ may potentially select patients who benefit from treatment with bevacizumab.

Regardless of the apparent lack of survival benefit with bevacizumab in glioblastoma patients, the reduction in peritumoral edema, and in some studies the reduction in corticosteroid use, is a real benefit to patients, contributing to an improved quality of life. The FDA has recognized this, granting bevacizumab full approval in recurrent glioblastoma on December 5, 2017. Nonetheless, the missteps in the development of bevacizumab for this indication, including the failure to include a standard therapy control arm in the early studies, which EORTC 26101 has now addressed, provide an important lesson for the field.

References