A platform for efficient early evaluation of biomarker-associated therapies in newly diagnosed IDH wild-type, MGMT unmethylated glioblastoma
Year: 2018
Type: article
Abstract: See the article by Wick et al, pp. 95–107 The impact of the precision medicine era has unfortunately not resulted in a substantial improvement in survival of glioblastoma (GBM) patients. Despite extensive molecular knowledge and biomarker subtyping from multiplatform profiling of large numbers of GBMs, O6-methylguanine-DNA methyltransferase (MGMT) promoter methylation is the only clinical biomarker used to (sometimes) alter treatment for newly diagnosed GBM.1 In contrast to other solid tumors, no effective biomarker-based targeted therapies have been identified for GBM to date. While differences in biology and drug penetration in GBM compared with other solid tumors is an important contributor, other factors also play a role. One complicating factor is the time expense, and large number of patients required for traditional phases I/II/III drug development associations in a relatively rare and heterogeneous disease such as GBM.2 In particular, the standard randomized 2-arm studies for phases II/III evaluate only a single treatment in a single group of patients. While this study design provides results with a high degree of statistical confidence, it is inefficient for evaluation of many potential therapies in different diseases or biomarker-defined subtypes of a disease. In order to address some of these challenges, statisticians, clinical trialists, and regulators have developed and are utilizing novel study designs that can improve efficiency and answer multiple clinical questions within a single clinical trial.3–5 These trial designs often include a master protocol that governs multiple substudies, and these designs can have various aims, including: to evaluate a single drug in multiple tumor or biomarker groups (eg, basket study) and/or to evaluate multiple therapies in a specific disease or subtype (eg, umbrella study). A platform study can utilize either or both elements as well as allow for addition or deletion of substudies depending on outcomes during the trial. Statistical design in a platform study can incorporate standard or Bayesian/adaptive designs in which information gleaned during the operation of the study is used to modify patient randomization and/or continuation or closure of individual substudies.5,6 More efficient study designs to simultaneously evaluate a wide variety of targeted therapies in biomarker-defined GBM patients in early phase II could significantly improve the field’s ability to prioritize the most promising drug-biomarker target combinations for confirmatory phase III testing. In the current issue, Wick et al describe a platform/umbrella trial called NCT Neuro Master Match (N2M2) that seeks to address some of these issues in isocitrate dehydrogenase (IDH) wild-type, newly diagnosed GBM with unmethylated MGMT promoter.7 This platform study utilizes an interesting combination of both biomarker-based and biomarker-independent substudies, along with a concurrent temozolomide/radiation control group. Patients initially enroll in the discovery portion, in which tissue from initial GBM resection is subjected to multiplatform profiling, including: whole exome and low coverage whole-genome DNA sequencing, RNA sequencing, methylation profiling by array, and gene expression arrays, and relevant biomarkers are confirmed by immunohistochemistry or Sanger sequencing. Based on results of real-time tumor profiling, patients are assigned to one of 7 treatment arms based on the presence (match) or absence (no-match) of predefined biomarkers. Individual subtrials include an efficient (Bayesian) phase I component (when necessary) followed by a phase IIa study. The primary endpoint for each substudy is progression-free survival at 6 months. Targeted agents in subtrials that have an associated “match” biomarker include: asinercept, alectinib, idasanutlin, atezolizumab, vismodegib, palbociclib, and temsirolimus. Other subtrials where evidence for associated biomarkers was not considered strong enough for treatment assignment “no-match” include asinercept and atezolizumab. “Match” patients will be assigned to the treatment arm associated with that specific biomarker, and will receive the targeted therapy along with radiation. When a tumor harbors multiple “matches,” a previously described algorithm will be used to determine treatment assignment,8 or to the best performing relevant subtrial (if known at the time of enrollment). “No-match” patients are randomly assigned to receive asinercept, atezolizumab, or temozolomide along with radiation. The final decision regarding treatment and subtrial assignment will be made by a molecular tumor board consisting of trial steering committee and enrolling site members. The temozolomide arm serves as a contemporaneous control for all subtrials. This study design has the potential to efficiently evaluate the efficacy of multiple therapies in biomarker-defined and biomarker-independent GBM subpopulations in a single platform trial. While improving efficiency, this approach does have some potential drawbacks that will require careful analysis and consideration when evaluating the results of individual subtrials. First, only patients who are “no-matches” will be eligible for the control (temozolomide/external beam radiotherapy) substudy, so the control population does potentially differ from the biomarker-defined “match” substudies. Second and related, since there is not true balanced randomization of patients among all treatment and control, there is the potential that an individual biomarker has prognostic impact itself, or is correlated with another prognostic factor (eg, age). This eventuality could make it hard to deconvolve prognostic from therapeutic effects within an individual substudy. However, to date no biomarkers beyond MGMT and IDH have shown a clear prognostic effect in GBM.9 Thus the design tradeoffs utilized in this protocol are appropriate, particularly in a phase IIa, signal-finding study where additional confirmation in randomized phase III studies are expected. It will, however, be very important to carefully evaluate retrospectively any differences in study populations between treatment substudies to ensure that there are no unexpected biases in treatment assignment. Overall, the N2M2 study represents a significant advance in study design for evaluation of drugs in the early phase II space for newly diagnosed GBM. It will be of great interest to the field to see how this design plays out in “real life,” and whether any of the substudies show evidence of efficacy in their target populations. Furthermore, the practical experience of using extensive multiplatform biomarker data in a large, prospective platform study will aid other platform efforts in the field, including adaptive studies such as INSIGhT9,10 and GBM AGILE,11 which target the later phase II and phase III areas of drug evaluation for both newly diagnosed and recurrent GBM.
Source: Neuro-Oncology
Author: Howard Colman
Institutions: University of Utah, Huntsman Cancer Institute
Language: English
Subfield: Genetics (clinical)
Field: Medicine
Domain: Health Sciences
Open Access status: bronze
OpenAlex ID: W2906344255