The trials of neuro-oncology clinical research

See the article by Vanderbeek et al., pp. 1034–1043.

Glioblastoma (GBM) remains one of the most challenging and intractable of cancers. Despite being the initial focus of the national effort to identify the molecular driver(s) of solid tumors, these data failed to lead to the anticipated subgroup identification that would guide future clinical trials and therapeutics. Thus, the rationale and design of clinical trials for GBM have not changed that much in more than 50 years. If one examines the initial Brain Tumor Study Group trial (#69-01), which demonstrated the effectiveness of radiotherapy, one can see a sample size (303 patients) and endpoint (overall survival) that look frighteningly similar to modern studies of patients with newly diagnosed GBM.¹ In this issue of Neuro-Oncology, Vanderbeek et al examine our current clinical trial processes and how they are failing to advance only convincingly promising agents forward to phase III studies.² They do not suggest that the failure to identify better therapeutics rests solely with our flawed clinical trial system, and they state at the outset that only further scientific research will open new therapeutic avenues likely to be most helpful. However, they do highlight the wastefulness of our current clinical trial efforts—of time, effort, and patients—that ultimately fail to advance our mission toward improved outcomes.

Vanderbeek et al provide data that highlight the many obstacles currently hampering clinical trial function. They estimate that only 8%–11% of patients with newly diagnosed disease enroll in a therapeutic trial and approximately 10% of patients overall enroll in any therapeutic study annually. This does not account for those patients who enroll in more than one study, a relatively frequent occurrence. Only 4% of all clinical trials were phase III studies but they enrolled 26% of all study patients, a sobering fact considering that only 1 of the 8 completed phase III studies reported positive results. Seventy-five percent of the phase III trials had preceding phase II studies and only 58% of those employed the same endpoint in the phase III trial. Not highlighted, but perhaps even more amazing, is that 25% of phase III studies did not even have a phase II trial before launching a presumed definitive study. The other major area of failure is in the publication of results. Only 12% of all trials and 34% of completed studies had linked publications 100 months after trial initiation. Of course, negative trials suffer from lack of publication more than positive studies, which enhances bias in the literature; publication deficiency is nothing new but remains disturbing.

How then to address these concerns and improve our ability to conduct therapeutic trials in patients with GBM? Patient recruitment remains challenging. Implementing a comprehensive program designed to increase enrollment in oncology trials can have a dramatic impact; however, this is primarily an effort confined to academic institutions.³ A further challenge is the relative rarity of GBM, and the fact that the disease is often handled locally, preventing the tumor specimen from entering a biorepository. Many communities are not close to a large academic center with neuro-oncologic expertise where trials are conducted; however, even when such expertise is close at hand, patients are frequently not referred for consideration of a clinical trial. The reasons for this are multiple, including physician nihilism, but a major modern driver is economic competition. As large hospital systems consolidate, they are more inclined to retain every patient. The ability to deliver high-quality current care is often confused with the novel therapeutic strategies that will develop the next standard of care. In addition, the availability of non-approved drugs for “compassionate” use can often thwart clinical trial enrollment. This was seen when bevacizumab was first approved by the FDA for recurrent disease but was rapidly assimilated into upfront treatment in many communities. This compromised trial enrollment 2-fold: (i) such patients did not want to enroll in the phase III studies with a 50% chance of receiving placebo, and (ii) upfront bevacizumab made it impossible for those patients to be eligible for trials employing an anti-angiogenic agent at recurrence. Almost uniformly, such patients were never informed that using bevacizumab upfront would compromise their future clinical trial eligibility. We see this same phenomenon happening again with checkpoint inhibitors. Despite weak efficacy data of these agents in GBM, we continue to see many such patients given the drug off study at diagnosis and at recurrence, again compromising eligibility for a series of important clinical trials that incorporate biomarkers which may help interpret either positive or negative results. While most physicians are giving these drugs with the intention of helping a desperate patient facing a fatal disease, the unconscious biases of economic gain, retention of the patient in their own health system, and the appearance of doing something “new” cannot be underestimated. Personally, I have seen any number of patients persuaded against clinical trial participation by someone promising them the active drug off label if they were treated at another institution. To be fair, who among us facing such a diagnosis would not follow such an opportunity regardless of potential toxicity? Thus, as professionals, we must commit to advancing the clinical science and not forfeit our objectivity and critical thinking. Furthermore, our poor track record of converting promising phase II data into definitive benefit in the phase III setting should give all of us pause to consider the potential for added toxicity with off-label use.

Vanderbeek et al provide the sobering data on our failure rate when taking a drug from phase II to phase III. Recent data suggest that even early phase trials may be based upon incomplete preclinical data that overemphasize potential benefit.⁴ A major challenge is the ability to detect a meaningful signal of efficacy in a phase II trial. Often historical control data are used to show presumed superiority of a new approach. However, the critical criteria for inclusion in the phase II study are frequently not considered when trying to select the historical “control” cohort. This was illustrated more than 25 years ago by Florell et al, who demonstrated that the presumed benefit of brachytherapy in GBM that was touted in multiple phase II trials was completely due to patient selection bias, and if an appropriate control population was assembled, the negative phase III study results that were seen ultimately were totally predictable.⁵ This was shown for intra-arterial chemotherapy trials as well,⁶ and continues to be an issue today.

The authors correctly point out that there is also a lag and lack of publication of many studies. This not only deprives the community of the knowledge regarding the results of a study, but is also a fundamental breach of our contract with the participants, many of whom have altruistic considerations when they decide to enroll. We ask a lot of trial participants, including extra visits and testing which often cost them time, money, and occasionally distress. A consent form is a contract of trust, and we violate that trust when the study results remain unpublished.

Publication may lag or never occur for any number of reasons, but negative results are particularly problematic. It can be difficult to publish a negative clinical trial, and there are ample studies demonstrating that negative oncology results disproportionately remain unpublished.⁷ High-profile journals rarely accept such studies and often do so only when there is controversy or some glimmer of positivity. Although publications are important for academic advancement, negative studies are typically published in lower-impact journals which carry less weight toward an academic career. Thus, even academic achievement may be a poor motivating factor toward publication of every study, not to mention the reduced enthusiasm of any investigator toward publishing a negative trial. In some modern analyses, there may be indication of less publication bias than described previously, but we still have a long way to go.⁸

Academic achievement deserves a mention here. Vanderbeek et al have restricted themselves to presenting data on clinical trial design and efficacy of the current system. However, it is worth noting that criteria for promotion often work against the greater goal of advancing the care of patients with malignant glioma. The need to define a “leadership” role in studies, which often carries the largest sum of money to conduct the trial and concentrates the control of drug development and clinical trial design into the hands of a few, may be counterproductive to overall progress. First, most trials beyond phase I need to be multicenter studies, and we need a better way to acknowledge the critically important contributions of the many institutional participants. Academic reward for such participation is critical and often rests with the senior leaders in the field advocating for our younger colleagues, including offering first and senior authorship positions which could help reduce the publication gap. Second, we all tend to become invested in our own perspectives and approaches, which may make us less critical of our efforts when our career is on the line. All of us prefer to see a glass half full than half empty. This is human nature, and while there is no eliminating this bias, conscious efforts to evaluate early study results in a critical manner may help us make better decisions regarding which agents should move forward.

All investigators involved in clinical trials for patients with GBM are seeking real benefit in patient outcome and survival. Everyone opens a new therapeutic study hopeful that it will improve the current standard of care, which itself was advanced by earlier studies. Despite the best intentions, we must still do a better job designing, executing, and completing these studies so physicians and patients alike can reap the full benefit of this considerable effort.

Funding

This work was supported by Craig Thompson, MD (grant P30 CA008748).

References