Correcting the drug development paradigm for glioblastoma requires serial tissue sampling

Abstract

A breakthrough in drug discovery for glioblastoma requires serial collection of tissue from the central nervous system via window of opportunity trials

Therapeutic drug development in glioblastoma has remained static despite numerous clinical trials, advances in tumor biology, and substantial public interest. Since 2005, more than 1,250 interventional glioblastoma trials have been registered on ClinicalTrials.gov. Of these, 1,100 are phase 1 or 2 trials, most of which are dose-escalation studies. Unfortunately, these often provide little insight into whether drugs reach the tumor, exert a biological effect, or can overcome mechanisms of resistance. Trials often evaluate drugs that show promise in early development but subsequently fail at later stages¹. Biological understanding of these failures requires tissue samples before and after treatment, but the collection of these has not been a routine element of either neurosurgical practice or drug development in neuro-oncology. Window of opportunity (WoO) trials can help to determine the mechanisms of drug failure, but expanding their use requires a clearly defined nomenclature, serial access to tissue, and standardized design principles and outcomes. To achieve the clinical success seen for other malignancies, a culture change within the neurosurgical and neuro-oncology community is necessary to support repeated tissue collection from the central nervous system (CNS).

Misleading pre-clinical models

For a therapeutic agent to reach the clinic, it typically undergoes extensive pre-clinical profiling. However, many in vitro and in vivo glioblastoma models poorly recapitulate the key barriers of therapeutic efficacy such as intra-tumoral and intra-patient heterogeneity². Even worse, these models can actively mislead investigators. Drugs designed with favorable blood–brain or blood–tumor barrier entry characteristics were initially found to be non-penetrant in xenograft models of glioblastoma³. These in vivo findings were subsequently overturned by a WoO study (NCT02207010) that detected the drug in clinical tumor samples and at a concentration that could produce pharmacological activity³. In this example, the drug (AZD1775) was initially described in 2010, brain-penetrance studies were completed in 2015³, and the final determination of drug behavior in clinical disease was published in 2018⁴.

Given the uncertain fidelity of pre-clinical models, tissue-based trials should be considered for primary assessments of pharmacokinetic (PK) and pharmacodynamic (PD) characteristics. PK and PD findings from clinical disease can then be reverse engineered back into an intentional drug-development process. Intentional drug development is vital for glioblastoma given its unique therapy challenges, which include heterogeneity, an immunosuppressive tumor micro-environment, and the presence of the blood–brain or blood–tumor barrier. Re-purposed therapies (immune or otherwise) that have shown efficacy in other malignancies have so far yielded disappointing results in glioblastoma⁵. These failures further highlight the necessity for a greater understanding of glioma biology and the effects of treatments.

Clarifying tissue-based trials

Tissue-based trials have been described for several decades, but their use remains limited owing to challenges such as a lack of appropriate controls for biological endpoints (Box 1). The collection of control samples is hindered by current neuro-surgical/oncology practices of not routinely collecting tissue before therapy. Furthermore, the nomenclature around such trials is often confusing. The terms phase 0, neoadjuvant, WoO or surgical window of opportunity (SWOOPP) are used interchangeably despite having subtle but important differences. Traditionally, phase 0 has been used to refer to micro-dosing studies, in which a small number of participants (10–12) are exposed to a single course of a low, non-toxic dose of drug for a limited duration (<7 days). Although these studies seek to assess the PK and PD characteristics of a drug on the tissue of interest, they are performed without therapeutic intent.

Box 1. Barriers for window of opportunity trials in glioblastoma.

1. Lack of appropriate controls

2. Small numbers

3. Pharmacodynamic assays

4. Pre-operative timing

5. Pharmacokinetic assays

6. Multidisciplinary coordination

7. Funding

8. Uncertain dosing regimen

9. Recruitment

Note that these barriers were ranked by 35 participants at the 27th SNO annual meeting, with ‘1’ being the most important challenge in window of opportunity trial development, and ‘9’ being the least important of the obstacles listed

By contrast, a WoO study evaluates the PK and PD characteristics of an intervention but at doses and schedules that are hypothesized to result in a therapeutic effect. Cross-over from phase 0 to WoO studies can occur, such as the inclusion of a PK–PD ‘trigger’ as part of a trial protocol. If a drug shows promise in a phase 0 micro-dosing study (that is, meets trigger conditions), treatment can then be evaluated at higher, therapeutic doses (Fig. 1). Both phase 0 and WoO studies can occur before, during or after early phase studies. However, WoO studies with therapeutic intent should generally follow phase 1 studies to ensure that the maximum tolerated dose and PK are understood before constructing a treatment regimen. WoO studies fit best either alongside or immediately after phase 1 studies, with the goal of determining go/no-go decisions before proceeding to larger phase 2 studies.

Fig. 1 |. Differentiating between phase 0 and window of opportunity studies.

Although phase 0 micro-dosing studies have been established for some time, the term is often used interchangeably with window of opportunity (WoO) studies. Phase 0 studies evaluate therapies via a micro-dosing non-therapeutic approach, but WoO studies seek to evaluate pharmacokinetic and pharmacodynamic endpoints using treatment doses or schedules. Studies can switch from phase 0 to WoO if the agent of interest meets certain pharmocokinetic/pharmacodynamic ‘triggers’. The timing of phase 0 or WoO studies in relation to phase 1 or 2 trials can vary, occurring before, alongside or after these investigations.

Serial biopsy for controls

Acquiring tumor tissue before treatment for biological controls is vital for understanding the tumor landscape before therapy. Although tumor tissue from biobanks is often available for patients with recurrent disease, these specimens are typically collected weeks, months or even years earlier, and before exposure to genotoxic therapies. During this time, the immune profile, microenvironment and cell states of a tumor can, and (based on increasing new evidence) will, evolve substantially⁶. Instead, the role of pre-resection biopsies should be expanded, enabling the assessment of immune, environmental and cellular states of a tumor immediately before intervention.

This will require a change in current practice with checks to ensure that the additional invasive procedures do not expose patients to harm. However, serial biopsies are a common practice across many other cancers. Stereotactic needle biopsies (SNBs) are routine in neurosurgery, and although they carry low risks such as hemorrhage, appropriate patient stratification enables SNBs to be performed as an outpatient procedure⁷. Analyses of immediate complications after outpatient SNBs find that they present 4–6 hours after the procedure, meaning that patients can safely leave the hospital on the same day. With appropriate patient selection, complication rates are similar to inpatient biopsies (5–7%)⁸. A well-optimized outpatient SNB pipeline can therefore be safe and minimally disruptive to the patient’s treatment pathway, enabling tissue sampling up to a week before resection. Longitudinal tissue follow-up after treatment may also be possible via this approach. These types of analysis will be essential to understand how long a therapeutic remains at the target and how the tumor and microenvironment change after treatment.

Consideration must also be given to the timing of pre-treatment biopsy relative to resection. Moving rapidly from biopsy to resection may not leave sufficient time for an investigational agent to reach peak bioavailability. Large retrospective analyses have found no difference in the outcomes of patients undergoing surgery 3 days or 1 month after the initial diagnostic scan⁹. These outcomes included the extent of resection, residual tumor volume, postoperative performance changes, and overall survival⁹ — in the context of appropriate patient stratification. Delayed surgery led to inferior outcomes in patients with poor performance status and tumors >50 ml in volume⁹. However, the tumor location was not associated with worse outcomes, even in more eloquent, high-risk settings. In a WoO trial, a pre-treatment biopsy should therefore be reserved for patients with adequate performance status and tumors <50 ml. Analysis of tumor growth rates finds a median doubling time of 21–22 days⁹. This window allows for biopsy followed by administration of a study drug and then a planned resection 3–4 weeks later in suitable patients. Median assessment of growth will not reflect all patients, and some may experience more rapid increases in tumor volume. Strategies to identify patients with rapidly growing tumors who may require prompt surgery may include serial diagnostic imaging or cellular proliferation markers.

Given that intra-tumoral heterogeneity in glioblastoma is substantial, a single-region biopsy is an imperfect representation of heterogeneous tumors. Previous work seeking to understand how biopsy samples could truly reflect a heterogenous tumor showed correlation between greater amounts of tumor material available and more accurate histological feature detection¹⁰. Tumors need to be large enough to allow for acquisition of sufficient biopsy material to be representative, but not so large as to necessitate more urgent resection. Tissue samples before and after treatment should also be compared from similar regions of the tumor to ensure that the observed changes are not due to intra-tumoral heterogeneity. Conveniently, the recent advice of the RANO resect group (encouraging maximal resection of non-contrast-enhancing tumors) enables greater potential matching of biopsy sites before treatment to resection tissues after treatment¹¹.

Recruitment and stratification

Despite only seeking to enroll small numbers, the complex trial logistics and expertise required to perform surgery and subsequent tissue analysis means that WoO studies will likely remain concentrated in tertiary healthcare systems. This will lead to an ongoing mismatch between available patients at these centers and continuing historical lack of representation from community practices. As WoO trials become more widely used, a trial platform akin to GBM AGILE (NCT03970447) with an open Institutional Review Board protocol for several centers should be considered. This coordinated approach to expand enrollment and pool expertise may also drive pharmaceutical industry engagement.

Umbrella trial designs, in which patients are stratified according to molecular subtypes, should also be considered for WoO studies. Of value would be the ability to explore treatment effects in MGMT (O⁶-methylguanine-DNA methyltransferase) unmethylated patients who are known not to benefit from temozolomide. Trial designs such as PARADIGM-2 (ref. 12), in which MGMT methylation status is used to both stratify patients and determine treatment schedules, are one such example. This stratification enables the evaluation of the effect of therapy on tumor biology, immune infiltrate and microenvironment with and without the presence of chemotherapy.

Other designs of interest include parallel collection of serial liquid biopsies (blood and cerebrospinal fluid) and tumor samples after treatment to characterize longitudinal multi-omic and immune markers. To expand this evidence base, liquid biopsies should be routinely included alongside resection and potential biopsies in future WoO studies. Sequential specimens can also be correlated with genome datasets such as The Cancer Genome Atlas (TCGA) and The Glioma Longitudinal AnalySiS (GLASS) consortium as well as follow-up magnetic resonance imaging (as in iGLASS). Over time, these approaches may enable the development of predictive patterns for PK/PD outcomes and reduce the reliance on biopsies.

Driving culture change

Obtaining several biopsies over the course of an individual’s disease to assess recurrence and/or treatment responses is widely accepted across many tumor types. This is not the case for glioblastoma, in which current guidelines suggest either biopsy or resection (but not both) at diagnosis¹³. This limits tissue acquisition at recurrence to clinical scenarios in which there is greater than average diagnostic uncertainty or other indications for resection. There are many reasons for this, not least of which is that the risk of biopsy can be perceived as unacceptably high. However, as discussed above, outcomes after stereotactic biopsy are generally excellent at high-volume referral centers.

There is also concern about how biopsies before or after resection would be funded as staged procedures. These must be balanced against the enormous societal cost of ineffective and inappropriate therapy received by patients with glioblastoma. Although liquid biopsy holds promise to avoid repeated brain biopsies, developing these for patients with glioblastoma remains experimental. Thus, there is a need for a concerted effort to change practice patterns to obtain repeated biopsies. This will require demonstration of the safety, feasibility and effect on management as well as incorporation into standard guidelines and acceptance by payors.

We recognize that making such changes in current clinical practice requires not just clinician and scientific input but also engagement from regulatory agencies and funders. Of note, the recent establishment of Project Optimus and Project FrontRunner by the US Food and Drug Administration (FDA) demonstrates regulatory awareness and support for changing the approach to diseases in which progress has been limited^14,15. These initiatives strongly encourage the development of innovative approaches to therapeutic dosing, schedules and evaluation of therapeutic effect. WoO clinical trials are an excellent medium for such programs.

Intrinsic biological factors are the greatest barrier to therapeutic success, and well-defined tissue-based trials and analyses should be expanded. To do this effectively, the role of biopsies before resection must be increased, and clinicians must be open to the concept of serial tissue acquisition encompassing both the pre- and post-treatment setting. Fortunately, these approaches are already used in clinical practice in many centers of excellence. SNO and the relevant stakeholders will expand this effort across the neuro-oncology community, understanding that transformative change requires time, investment, and a collective modification to the way that the neuro-oncology community practices.