Glioblastoma Clinical Trials: Current Landscape and Opportunities for Improvement

Glioblastoma is the most common malignant primary brain tumor in adults and remains incurable (1, 2). Despite considerable progress in understanding the biology of glioblastoma (3–9), there have been no regulatory drug approvals since bevacizumab in the United States in 2009 (10), and a multitude of recent phase III clinical trials did not meet their prespecified primary endpoints (11–16). This disconnect between the urgent unmet need, growing scientific understanding of the disease, and lack of translation into effective novel therapies can be attributed to many causes, some of which are related to the difficult biology and clinical challenges inherent to glioblastoma. However, glioblastoma clinical trial design has also attracted significant recent scrutiny and may represent an important barrier to progress (17–25). Key problems include a paucity of control arms in phase II trials (18, 25), overly stringent clinical eligibility criteria and insufficient patient and provider participation (19, 20), inadequate assessment of brain penetration (26, 27) and pharmacodynamics (28, 29) of new drugs, inefficient clinical trial infrastructure (30), and limited development and implementation of prognostic and predictive biomarkers (31–37). Concerted efforts have started to address some of these, including a push toward adaptive platform trials (38–40) and joint publications from neuro-oncology professional societies and clinical trial consortia (19, 20).

A Think Tank was held by the Society for Neuro-Oncology (SNO) in 2020 to better define these barriers and to outline a road map for improving translation of discoveries into effective glioblastoma therapies. Participants included scientists, representatives from the Food and Drug Administration (FDA) and pharmaceutical industry, patient advocacy groups, and clinical trialists from both within and outside the field of neuro-oncology. The concluding session, Thinking Outside the Box, served as a summary session to cover the key issues and is the focus of this review. Issues covered in the other sessions are reviewed in separate articles: one on combination immunotherapy is published jointly with this article, and one on combination therapies in general is already published (41). Here, we first summarize key features of currently recruiting and soon-to-be recruiting glioblastoma clinical trials listed on ClinicalTrials.gov, including the type of experimental therapy, key study design parameters, primary endpoints, and eligibility criteria, providing a current snapshot of the glioblastoma clinical trials landscape. We then review the relevant literature and summarize the SNO 2020 Think Tank with regard to (1) optimizing eligibility criteria to increase enrollment and generalizability of trial results, and (2) optimizing study designs to provide an accurate assessment of efficacy before conducting a phase III study.

This section includes an analysis of key study design elements for glioblastoma clinical trials listed on ClinicalTrials.gov as “recruiting” or “not yet recruiting.” These data provide a real-world context to frame key discussion points that emerged from the Think Tank. Similar analyses have been performed by Vanderbeek and colleagues (25) and Cihoric and colleagues (42), although these studies only included trials initiated through 2016 and 2015, respectively. ClinicalTrials.gov was queried on February 10, 2021, for interventional trials with therapeutic intent enrolling patients >18 years old with glioblastoma.

One hundred fifty-seven clinical trials were identified. Key characteristics are listed in Table 1. The majority (92%) were phase I, I/II, or II, and over half had primary outcome measures related to toxicity, safety, and/or dose finding. Of the 76 phase I/II or phase II trials, 66 (87%) did not have a control arm and 63 (83%) were nonrandomized. Among phase I/II and II trials with an efficacy endpoint listed as the primary outcome measure (n = 60), the primary outcome measures were overall survival (OS) in 24 trials (40%), progression-free survival (PFS) in 16 (27%), objective response rate (ORR) in 13 (22%), and disease control rate (DCR) in two (3%). The types of therapy being evaluated in systemic therapy trials (n = 136) are displayed in Fig. 1 and key inclusion/exclusion criteria in Fig. 2. In trials for recurrent glioblastoma (n = 84), 31 (37%) specified that both standard-dose radiotherapy to 60 Gy and temozolomide were required as prior therapy to be eligible, 31 (37%) specified that any dose of prior radiotherapy with or without temozolomide was acceptable, and 22 (26%) did not specify this one way or the other.

Although restrictive eligibility to clinical trials is only one barrier to participation, unnecessarily restrictive inclusion and exclusion criteria limit the number of patients who can enroll and potentially benefit from investigational therapy, a critical factor in glioblastoma given that only 21% of patients with primary brain tumors participate in clinical trials (20). Overly selective eligibility criteria also reduce the generalizability of results (43), presenting a potential problem when “positive” phase II results from a single academic center are put to the test in a randomized phase III trial. With wider recognition of these issues, there has been a recent push to broaden eligibility criteria, both across oncology (43) and in neuro-oncology specifically (19). In 2016, the American Society of Clinical Oncology (ASCO), Friends of Cancer Research, and the FDA formed a collaboration to address overly restrictive cancer clinical trial eligibility criteria, leading to published recommendations for more inclusive eligibility criteria (44) and updated recommendations in 2021 (45).

Certain eligibility criteria are particularly relevant in neuro-oncology. In recurrent glioblastoma trials, details related to patients' prior therapies are important for determining eligibility. This includes the type of first-line therapy received, number of prior relapses, and previous use of bevacizumab (19). Following maximal safe surgical resection, the widely accepted standard of care for newly diagnosed glioblastoma includes radiotherapy to 60 Gy with concurrent temozolomide followed by six maintenance temozolomide cycles (46). However, variations of this regimen are used in clinical practice and supported by National Comprehensive Cancer Network (NCCN) and other guidelines (1), including hypofractionated radiation (e.g., 40 Gy delivered in 15 fractions) in elderly patients (47) and omission of temozolomide in patients with MGMT unmethylated tumors (48). As a result, requiring previous receipt of traditional standard of care is seldom necessary from a scientific standpoint and may unnecessarily exclude a large percentage of patients who would otherwise be eligible (19). In addition, the types of therapies received for prior relapses of glioblastoma and their impact on the ability to respond to the experimental treatment should play a greater role in determining eligibility than the number of prior relapses, which itself may have little effect on treatment response (19). A key example is use of bevacizumab, as bevacizumab-refractory disease may harbor unique biology (49) and rarely responds to further salvage therapy (50). Thus, it has been suggested that for phase I trials with primary endpoints related to dose finding and safety, an unlimited number of prior relapses and therapies (including bevacizumab) should be allowed (19). Conversely, phase II/III efficacy studies might allow a limited number of relapses (e.g., up to two) in bevacizumab-naïve patients and need to determine on a trial-by-trial basis whether to exclude or stratify by prior bevacizumab (19).

Another particularly relevant eligibility criterion for glioblastoma clinical trials is the performance score (PS). In patients with glioblastoma, poor PS directly attributable to a fixed neurologic deficit may not accurately represent the patient's appropriateness for receiving experimental therapy. This unique feature of brain tumor patients requires that each glioblastoma clinical trial carefully considers the cutoff point for PS that will be used to establish eligibility. Others have recommended (19), and we agree, that patients with primary brain tumors with lower PS (as low as Karnofsky performance status of 60) due to fixed neurologic deficits can be accrued to phase I trials without affecting the study's integrity.

We also found that approximately one-third of recurrent glioblastoma clinical trials required that patients previously completed radiation to 60 Gy with temozolomide, regardless of age and MGMT methylation status. This points to one area for improvement because in most early-phase recurrent glioblastoma trials, there is no scientific rationale to eliminate patients from participation on the basis of having previously received less than 60 Gy of radiation and/or no temozolomide. Moreover, it seems unfair to exclude otherwise eligible patients from participation simply because they received a variation on standard of care that was tailored to their individual circumstances (e.g., 40 Gy in the elderly or omission of temozolomide in MGMT unmethylated patients) and has been widely viewed as acceptable first-line treatment across the field of neuro-oncology. At the same time, we found that other key clinical inclusion/exclusion criteria are less stringent, a sign that investigators may be more carefully considering the issue of generalizability when designing studies. These include allowance for a Karnofsky Performance Scale (KPS) of 60 in nearly half of all glioblastoma trials and allowance for prior bevacizumab use in about two thirds of recurrent glioblastoma trials. Our analysis also found that 47% of current glioblastoma clinical trials are for newly diagnosed disease. This reflects a recent shift toward more clinical trials being offered in the newly diagnosed setting as opposed to recurrence, as a previous study noted that only one third of all glioblastoma trials between 2005 and 2016 were for newly diagnosed disease (25). Although this may allow enrollment of patients earlier in the disease course when they are generally healthier and better able to tolerate treatment, it also carries important implications for clinical trial design (18), including eligibility criteria. For example, the majority of newly diagnosed glioblastoma trials make no mention of whether tumor-treating fields (TTFields) are allowed or excluded. Although TTFields are approved by the FDA based on a previously demonstrated OS benefit (51), this therapy is not used in a substantial proportion of patients with newly diagnosed glioblastoma (52). The question of whether to allow use of TTFields in upfront glioblastoma clinical trials needs to be addressed individually for each study based on its specific scientific and clinical concerns. However, both patient and clinician decision-making would benefit from greater clarity regarding whether or not TTFields are allowed in a given trial.

Compared with randomized studies, single-arm phase II trials require significantly fewer patients and resources and can be opened and completed fairly rapidly at a single institution. However, it is generally agreed upon that the abundance of single-arm efficacy studies with primary endpoint analysis based on historical controls has generated an inadequate phase II program in glioblastoma (17, 18, 21, 22, 25). Issues with single-arm phase II designs for glioblastoma trials include a lack of reliable historical controls for most protocols (ref. 53; particularly since the WHO revised classification of central nervous system tumors in 2016 and subsequent updates from the Consortium to Inform Molecular and Practical Approaches to CNS Tumor Taxonomy [cIMPACT-NOW] working group; refs. 54, 55), selection bias (23, 56), and patient temporal drift (57, 58). As a result of these limitations, single-arm phase II studies have a high risk of leading to incorrect go/no-go decision-making for phase III (25, 59). In 2018, two studies were published analyzing the phase II–III transition in glioblastoma (22, 25). Mandel and colleagues examined the ability of positive phase II studies in glioblastoma to predict positive phase III studies. The authors identified seven phase III clinical trials in newly diagnosed glioblastoma and four phase III trials in recurrent glioblastoma over the prior 25 years, all of which were preceded by positive phase II studies; only one (9%) phase III study was subsequently positive (22). Of the seven phase III trials in newly diagnosed glioblastoma, 12 prior positive phase II studies were cited as justification for phase III. Of the 12 phase II studies, PFS or OS were the primary endpoints in nine studies (75%), and historical control groups were used for comparison in nine studies (75%). Moreover, the authors found that more than half (12/23) of all phase III trials in glioblastoma proceeded without a prior phase II study being performed or with a negative phase II study. Around the same time, Vanderbeek and colleagues analyzed glioblastoma clinical trials conducted from 2005 to 2016 and found that of the eight completed phase III trials, only one reported positive results (25). Although 58% of the phase III trials were supported by positive phase II data with a similar endpoint, only 25% of these phase II trials were randomized.

Since these publications, six additional negative phase III trials have been reported (11–16). The first of these was a randomized, double-blind phase III trial of rindopepimut with temozolomide for patients with newly diagnosed, EGFRvIII-expressing glioblastoma (13). This trial was preceded by two small single-arm phase II trials (60, 61), as well as a larger phase II trial that was initially randomized was but was converted to a single-arm design after nearly all of the first 16 patients enrolled to the standard-of-care temozolomide arm voluntarily withdrew (62). The patients enrolled on these single-arm studies showed encouraging median PFS/OS compared with contemporary patient cohorts receiving standard of care. The second trial was a randomized, placebo-controlled trial of personalized peptide vaccination in patients with recurrent glioblastoma (14), which followed a phase I study demonstrating feasibility and encouraging OS (although median PFS was only 2.3 months; ref. 63). The third trial, presented but not yet published at the time of this review, was a randomized, double-blind, placebo-controlled trial of depatuximab mafodotin (ABT-414) in EGFR–amplified, newly diagnosed glioblastoma (15). This study came on the heels of phase I data in both the newly diagnosed (64) and recurrent settings (65), with single-arm efficacy data demonstrating potentially encouraging PFS rate at 6 months (PFS-6) and a small number of objective responses (64–66). The fourth study was a randomized trial of nivolumab versus bevacizumab in patients with recurrent glioblastoma (11). This trial followed only a phase I study demonstrating the safety of nivolumab in this population (67), without supporting phase II efficacy data. The fifth study was a seamless phase II/III randomized-controlled trial of vocimagene amiretrorepvec (Toca 511) with flucytosine versus standard-of-care physician's choice chemotherapy for recurrent glioblastoma (12), which followed phase I data demonstrating a subset of patients experiencing multiyear durable responses (68, 69). The sixth and final study was a randomized trial of temozolomide-based chemoradiotherapy with versus without marizomib in newly diagnosed glioblastoma (16). This study, presented but not yet published, was launched based on the results of a phase I trial of marizomib in combination with bevacizumab in patients with recurrent glioblastoma (70), which demonstrated a PFS-6 (39%) that compared favorably with the expected PFS-6 from bevacizumab alone (16%; ref. 71).

Although the reasons for these trials not meeting their endpoints are numerous and complex, these recent negative phase III studies, taken together with those of the prior 2 decades, should generate further caution about proceeding to phase III studies based on only single-arm phase I/II data. Interestingly, this point is underscored even by one of the few phase III trials in glioblastoma that actually met its primary endpoint: CeTeG/NOA-09, a phase III randomized trial of lomustine-temozolomide versus temozolomide alone in newly diagnosed, MGMT-methylated glioblastoma (72). This study, like many of those described above, was launched based on the results of a single-arm phase II trial (73). Although it was technically a positive study, the difference in OS between the two arms barely reached statistical significance. Moreover, the 95% confidence interval for relative risk of death was wide and crossed 1.0, and there was no difference in PFS between the two arms. Due to these limitations, it is unclear whether the study's findings are clinically meaningful, and it is unlikely to change practice unless the results are reproduced in a larger study.

In our analysis of currently enrolling glioblastoma clinical trials, only 11% included a control arm of any sort, and just 18% were randomized. This is in part related to the fact that nearly half of all glioblastoma clinical trials are phase I and thus expected to be nonrandomized and lack control arms. However, 87% of all the phase I/II or phase II trials also did not have a control arm, and 83% of these were nonrandomized. Although this finding is not new, it is remarkable that, despite widespread recognition of this problem and frequent calls for change, nearly all glioblastoma phase II studies remain single arm. Although this may be less of a concern for phase II trials using objective response as the primary endpoint (assuming the experimental agent does not have antiangiogenic properties that may cause “pseudo-response”; ref. 74), approximately two-thirds of the phase II trials identified in our analysis used time-to-event outcomes, which can have more variance unrelated to the experimental treatment.

Despite the clear benefits to control arms in phase II studies, there are multiple potential barriers to their use. The first is patient and clinician dissatisfaction with randomization to a standard-of-care arm that can be perceived as ineffective or marginally effective. This is especially pertinent in glioblastoma, as agents with meaningful activity in the relapsed setting are generally lacking (75). Second, randomized studies often require larger, multicenter patient cohorts and longer timeframes. Although there are cooperative group mechanisms, there remains, at least in the United States, a need for more robust and interconnected clinical trial infrastructure across brain tumor centers to facilitate collaboration and multicenter studies. Adaptive platform trials, such as GBM AGILE or INSIGhT, have been implemented as potential solutions to some of these problems. GBM AGILE is a global trial for both newly diagnosed and recurrent glioblastoma that involves testing novel therapies in an initial Bayesian adaptively randomized screening stage to detect signals of efficacy based on comparing OS with the experimental agent to that achieved by patients enrolled on a common control arm (38). Highly effective therapies seamlessly transition to a secondary confirmatory stage in the identified population of interest, wherein fixed randomization is used to confirm the findings from the first stage. Such a design allows for numerous therapeutic arms to be added and deleted from the study over time, potentially speeding up the drug development process for glioblastoma. In addition, control therapy is assigned to only 20% of the patients throughout the trial, which may make the study more appealing for patients compared with a fixed 50% chance of being randomized to a control arm on most trials. Beyond adaptive platform trials, another possible path forward is to increase the use of hybrid randomized-controlled/externally controlled trials (76). We also recommend that cooperative groups work together to set national goals and priorities regarding the specific trials to be conducted within in a specific time frame (e.g., 5 years); each cooperative group would lead at least one of these trials and participate in all of them. The NCI National Clinical Trials Network provides a ready mechanism for this type of cross-cooperative group collaboration and clinical trial participation (77).

Many biological factors of glioblastoma have contributed to the relative lack of progress over the past 2 decades. However, successful development of new therapies also requires that clinical trials open and accrue efficiently, are adequately designed and powered to answer relevant scientific questions, and yield generalizable results. In this review, we found that phase II glioblastoma trials continue to be conducted largely in single-center settings and with single-arm designs, placing the field at risk for continued late phase trial failures and beleaguered drug development. To develop effective therapies and avoid wastage of precious patient and financial resources, investigators should carefully consider eligibility criteria to enhance trial enrollment and generalizability. In addition, most phase II studies, in particular those for newly diagnosed glioblastoma or for recurrent glioblastoma with PFS or OS-based endpoints, should be randomized, controlled, and sufficiently powered. However, there remain many unanswered questions that will need to be addressed by the neuro-oncology community to achieve tangible progress in the field (Table 2).

S.J. Bagley reports grants and personal fees from Novocure, grants from Incyte, GlaxoSmithKline, and Eli Lilly, and personal fees from Bayer and Sumitomo Dainnipon outside the submitted work. E.Q. Lee reports other support from Wolters Kluwer (Up to Date, Inc.), MedLink, Inc., Medscape, and Continuum outside the submitted work. G.P. Dunn reports personal fees from ImmunoGenesis and Ziopharm Oncology outside the submitted work. E. Galanis reports personal fees from Gradalis, Inc., Kiyatec, Inc., and Karyopharm and grants from Servier Pharmaceuticals LLC (formerly Agios Pharmaceuticals Inc.), Celgene, MedImmune, Inc., and Tracon Pharmaceuticals outside the submitted work. S.M. Chang reports other support from Agiios outside the submitted work. L.B. Nabors reports personal fees from Karyopharm and Penn outside the submitted work; in addition, L.B. Nabors has a patent for International Patent Application No. PCT/US2021/028460 pending. M.S. Ahluwalia reports grants from AstraZeneca Incyte, Pharmacyclics, Merck, and Bristol Myers Squibb; grants and personal fees from Bayer, Novocure, and AbbVie; grants, personal fees, and other support from Mimivax; personal fees from Kiyatec, Varian Medical Systems, Insightec, GlaxoSmithKline, Xoft, Nuvation, Cellularity, SDP Oncology, Flatiron, VBI Vaccines, Apollomics, Prelude, Janssen, Forma Therapeutics, Tocagen, Karyopharm, and Viewray; personal fees and other support from Doctible, Cytodyn, and MedInnovate Advisors LLC outside the submitted work; and is a stock shareholder in Doctible, Mimivax, Cytodyn, MedInnovate Advisors LLC. R. Stupp reports nonfinancial support from CarThera, Celularity, CranioVation/Alpheus, Hemispherian, Insightec, GT Medical Technologies, Northwest Biotherapeutics, and TriAct outside the submitted work. M.P. Mehta reports personal fees from Karyopharm, Sapience, Zap, Mevion, Xoft, Tocagen, Oncoceutics, and Chimerix outside the submitted work and is an NRG Oncology Brain Tumor Committee Chair. D.A. Reardon reports personal fees from AbbVie, Advantagen, Agenus, Agios, Amgen, AnHeart Therapeutics, Bayer, Boston Biomedical, Boehringer Ingelheim, Bristol Myers Squibb, Celldex, Deciphera, Del Mar Pharma, DNATrix, Ellipses Pharma, EMD Serono, Genenta, Genentech/Roche, Hoffman-LaRoche, Ltd., Imvax, Inovio, Kintara, Kiyatec, Medicenna Biopharma, Inc., Merck, Merck KGaA, Monteris, Neuvogen, Novartis, Novocure, Oncorus, Oxigene, Regeneron, Stemline, Sumitono Dainippon Pharma, Pyramid Biosciences, Inc., Taiho Oncology, Inc., and Y-mabs Therapeutics and grants from Acerta Pharmaceuticals, Agenus, Bristol Myers Squibb, Celldex, EMD Serono, Enterome, Epitopoietic Research Corporation, Incyte, Inovio, Insightec, Novartis, Omniox, and Tragara outside the submitted work. E.P. Sulman reports grants, personal fees, and nonfinancial support from Novocure, Zai Lab, and Merck outside the submitted work. J.H. Sampson reports consulting and licensing fees from Celldex Therapeutics and Medicenna Therapeutics and is on the board of directors with equity interest at Annias Immunotherapeutics and holds equity in Istari Oncology, Neuronium, and Immunomic Therapeutics. M. Weller reports grants and personal fees from Apogenix, Merck (EMD), and Adastra, grants from Quercis, personal fees from BMS, Medac, Merck Sharp & Dohme, Nerviano, Novartis, Orbus, Philogen, and y-Mabs outside the submitted work. T.F. Cloughesy reports personal fees from Global Coalition for Adaptive Research, Gan & Lee, BrainStorm, Katmai, Sapience, Inovio, DNATrix, Tyme, SDP, Novartis, Roche, Kintara, Bayer, Merck, Boehringer Ingelheim, VBL, Amgen, Kiyatec, Odonate, QED, Medfield, Pascal Biosciences, Tocagen, Karyopharm, AbbVie, and VBI outside the submitted work; in addition, T.F. Cloughesy has a patent for 62/819,322 pending and licensed to Katmai Pharmaceuticals and a patent for 63/179,769 pending and licensed to Katmai Pharmaceuticals; and is a cofounder, major stock holder, consultant and board member of Katmai Pharmaceuticals, member of the board for the 501c3 Global Coalition for Adaptive Research, holds stock option in Notable Labs, holds stock in Chimerix and receives milestone payments and possible future royalties, and is a member of the scientific advisory board for Break Through Cancer and Cure Brain Cancer Foundation. P.Y. Wen reports personal fees and other support from Agios, AstraZeneca/Medimmune, Merck, Vascular Biogenics; other support from Beigene, Celgene, Eli Lily, Genentech/Roche, Kazia, and MediciNova; and personal fees from Bayer, Boston Pharmaceuticals, CNS Pharmaceuticals, Elevate Bio Immunomic Therapeutics, Imvax, Karyopharm, VBI Vaccines, Voyager, QED, Celularity, and Sapience; and receives other support from Novartis, Nuvation Bio, Oncoceutics, and Vascular Biogenics outside the submitted work. M. Khasraw reports personal fees from Roche, grants from BMS and AbbVie during the conduct of the study; personal fees from Janssen, Jax lab, Ipsen, and Pfizer; and grants from Specialized Therapeutics outside the submitted work. No disclosures were reported by the other authors.

The authors thank the Society for Neuro-Oncology and their staff for their tremendous contributions to the arrangements and coordination for the Think Tank meeting.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.