Dendritic cell vaccine trials in gliomas: Untangling the lines

Abstract

Glioblastoma is a deadly brain tumor without any significantly successful treatments to date. Tumor antigen-targeted immunotherapy platforms including peptide and dendritic cell (DC) vaccines, have extended survival in hematologic malignancies. The relatively “cold” tumor immune microenvironment and heterogenous nature of glioblastoma have proven to be major limitations to translational application and efficacy of DC vaccines. Furthermore, many DC vaccine trials in glioblastoma are difficult to interpret due to a lack of contemporaneous controls, absence of any control comparison, or inconsistent patient populations. Here we review glioblastoma immunobiology aspects that are relevant to DC vaccines, review the clinical experience with DC vaccines targeting glioblastoma, discuss challenges in clinical trial design, and summarize conclusions and directions for future research for the development of effective DC vaccines for patients.

Glioblastoma is the most fatal of primary central nervous system (CNS) tumors in adults and despite aggressive therapies, the median survival remains dismal, especially at recurrence.^1,2 While immunotherapy has changed the treatment landscape for many cancers, it has not yet demonstrated benefit in glioblastoma.

Glioblastoma, which is considered a “cold tumor” because of its immune suppressive microenvironment, has a predominance of myeloid cells over lymphoid lineage cells. The immune component surrounding glioblastoma contains macrophages and microglia that form 30% to 50% of the total cellular composition,³ whereas infiltrating endogenous dendritic cells (DC) make up less than 1%.⁴ The fact that immunotherapy has clinical benefits in other cancers including brain metastases from solid tumors coupled with evidence that the brain is no longer regarded as “immune privileged” provides a rationale to explore immunotherapy in glioblastoma.⁵ Other reviews have addressed immunotherapeutic approaches in gliomas including immune checkpoint inhibitors and their combinations with other immunotherapeutic modalities.^6–9 Vaccines to treat cancer comprise a range of vehicles including synthesized peptides and co-administered adjuvants to elicit antitumor immunity.^10–13 Another potent vehicle for cancer vaccination is professional antigen-presenting cells (APCs) known as DC. In this review, we discuss DC vaccination, the pivotal clinical trials evaluating DC vaccines for glioma. DC vaccines are one form of immunotherapy. There is no reason that this specific therapy should be considered differently from other immune-directed therapies. In this review, we use DC vaccine as a model and framework to outline pertinent challenges in immunotherapy development and then clinical trial design elucidating aspects relevant to immunotherapy and DC vaccines in neuro-oncology. We conclude with recommendations for future DC vaccine trials.

What is a DC Vaccine?

Conventional DC are derived from bone marrow progenitors and possess the unique machinery required to phagocytose, process, and present antigen to T-cells via major histocompatibility complex (MHC) assembly, thereby initiating a cascade of antigen-specific T-cell responses.¹⁴ Type 1 and 2 conventional DC (cDC1, cDC2) are superior antigen presenters to T-cells, with cDC1 regarded as the ultimate DC vaccine.¹⁵ However, given their natural origin from the bone marrow, development of cDC vaccines as a translatable platform for cancer patients has not been feasible or even scalable. Instead, cancer immunologists have relied on the ex vivo differentiation of monocytes isolated from peripheral blood into inflammatory DC (iDC) via coculture with cytokines granulocyte-macrophage colony-stimulating factor (GM-CSF) and interleukin (IL)-4, as these differentiated iDC have repeatedly demonstrated the capacity to take up tumor antigen and activate T-cells.^16,17

For these reasons, DC are also categorized as APCs, linking innate and adaptive arms of the immune system. Using inflammatory and chemotactic signals to guide their infiltration into antigen-laden tumors, their function is to reach the tumor microenvironment (TME), process antigen, and migrate to tumor-draining lymph nodes to prime T-cell responses.¹⁸ This sequence ultimately should result in T-cell infiltration into the TME with effector T-cell mediated killing. Importantly, peripheral DC vaccination has resulted in increased tumor-infiltrating lymphocytes in preclinical glioma models with resultant extended survival.^19,20

DC-mediated tumor antigen recognition and processing are bypassed by loading DC with the tumor antigen of interest prior to vaccination. With patient DC vaccines, autologous ex vivo generated DC are loaded with tumor-associated antigens (TAAs) by whole-tumor lysate, antigenic peptide coculture, viral transfection, or messenger RNA (mRNA) electroporation.^21,22 Then, these matured, antigen-loaded DC are returned to the patient in the form of a vaccine where they must reach draining lymph nodes to transfer antigen and prime T cells (Figure 1). By this method using TAAs specific to the patient’s tumor, DC vaccines can facilitate the patient’s own immune system to identify, target, and eliminate cancer cells.

Figure 1.

Overview of dendritic cell (DC) vaccine strategies. Patients undergo leukapheresis to isolate peripheral blood mononuclear cells which are cultured with tailored cytokines and growth factors to differentiate monocytes into monocyte-derived inflammatory DC (iDC). These DC can be loaded with antigens taken directly from resected tumors such by either (A) coculturing iDC with tumor lysate, (B) pulsing tumor-derived or antigen-encoding mRNA, (C) viral vectors, or (D) loading iDC with tumor-derived peptides. Antigen-loaded iDC are then administered back to patients to generate robust immune responses against selected tumor antigens.

Strategies to improve the immunogenicity of DC vaccines for various cancers have been explored, including but not limited to enhancing antigen presentation by maturing and activating DC with factors such as IL-6, IL-1β, interferon-gamma (IFN)-γ and tumor necrosis factor alpha (TNFα).²³ Creating a more migratory DC phenotype has been trialed with prostaglandin E2 (PGE2)²⁴ and preconditioning the DC vaccine site²⁵ to generate more effective lymph node homing. Lastly, vaccination with various adjuvants to stimulate endogenous DC populations has also been investigated.^26–28

Clinical Trials with DC Vaccines in Gliomas

DC vaccines can stimulate immune responses against specific antigens and have demonstrated this activity in clinical trials including those for glioblastoma.^29–32 The first DC vaccine clinical trial was conducted in the late 1990s by Nestle et al.³³ This trial enrolled 16 patients and was aimed at evaluating safety and efficacy of cytokine-treated, tumor lysate, or peptide-pulsed DC vaccines in patients with metastatic melanoma. The trial showed that the DC vaccine was safe and well-tolerated. Antigen-specific immunity was induced during DC vaccination with objective responses achieved in 5 of the 16 evaluated patients. Investigators concluded that autologous DCs generated from peripheral blood was a safe and promising approach in the treatment of metastatic melanoma.

These results spurred further research in translatable DC vaccines, and many more clinical trials have been conducted since then to evaluate their safety and efficacy in various cancers, including gliomas (Table 1). The first DC vaccine tested in glioblastoma was published as a single case report in 2000,³⁴ and one of the first patient trials was conducted by Yu et al. In this study, DC vaccines pulsed with MHC-I peptides eluted from autologous glioma cells were able to elicit cytotoxic activity in four of seven patients. Additionally, of the four that underwent repeat resection, 2 demonstrated robust CD8 + and memory (CD45RO+) TILs, but because of the small sample size survival analysis was not confirmed.³⁵ The investigators then evaluated the safety of MHC-I peptide-pulsed DC combined with checkpoint therapy, imiquimod, in a follow-up phase I study (NCT01792505). DC vaccines in glioma were proving safe albeit with no ability to assess efficacy in the context of poorly powered sample sizes. These studies also uncovered one of the key parameters in generating the most immunogenic DC vaccine, which is the method of antigen selection and loading covered in the next section.

Table 1.

Clinical Trials That Have Been Reported to Date, Evaluating Dendritic Cell Vaccines in Glioma

Reference or NCT#	N	Trial Phase	Population	Vaccine Target	Survival Data
Yu et al36	8	I	Glioblastoma, anaplastic astrocytoma	Autologous tumor peptide	OS: 133 weeks
Kikuchi et al37	8	I	5 glioblastoma, 3 anaplastic astrocytoma	Tumor cells fused with the DC	Unclear, 1/8 patients had SD after 12 months
Rutkowski et al38	6	I	Recurrent malignant glioma	Tumor lysates	OS in 2 patients was >35 months
Yamanaka et al39	24	I/II	Recurrent malignant glioma, III Recurrent malignant glioma, IV	Tumor lysates	OS: 480 days
Okada et al40	22	I/II	Glioblastoma, Anaplastic astrocytoma, Anaplastic Oligodendroglioma	EphA2, IL-13Rα2, YKL-40, gp100	9/22 patients were progression-free ≥12 months
Prins et al41	23	I	Glioblastoma	Tumor lysate	OS: 31 months
Sakai et al42	10	I	Glioblastoma, anaplastic astrocytoma, Anaplastic Oligodendroglioma	WT-1	Unclear
Wen et al43	81 on ICT-107 vs. 43 control in a 2:1 randomization	II	Glioblastoma	ICT-107	mOS: 17.0 months; mPFS: 11.2 months
Liau et al44	331	III	Glioblastoma	Tumor Lysate	OS of 19.3 months
Parney et al45	21	II	Glioblastoma	Lysates, covering EGFRvIII, erbB2, gp100, MAGE-A3, IL13Ra2 plus more.	OS 25% at 2 years
Batich et al46	11	I	Glioblastoma	CMV pp65 DC with ddTMZ	mOS from diagnosis 41.1 months (21.6–113.3)
Batich et al46	12	II	Glioblastoma	CMV pp65 DC with stdTMZ	mOS from diagnosis Td arm 41.4 months (20.6–∞). mOS from diagnosis unpulsed DC arm 18.5 months (13.8–41.3)

Abbreviations: ∞, undefined; ddTMZ, dose dense 21 day temozolomide 100 mg/m2; mOS, median overall survival; mPFS, median progression-free survival; stdTMZ, standard 5 day temozolomide 200 mg/m².

Antigen Loading and Selection

The best antigen for a DC vaccine targeting glioblastoma is not well-established, and research is ongoing to identify the most effective TAAs for use in these vaccines. One of the earliest described approaches to generating DC vaccines is peptide-pulsed DC, wherein short peptides are synthesized that are either fragments of TAAs or encompass tumor-specific mutations. Of note, prior to peptide-pulsed DC vaccines, peptide vaccines consisted of injecting peptides directly in vivo along with an immune adjuvant to promote APC migration, phagocytosis, and presentation. While promising results were generated in glioblastoma in phase I and phase II clinical trials using rindopepimut (CDX-110) (Celldex Therapeutics), a synthetic mutated EGFRvIII neoantigen-specific peptide administered with GM-CSF, this approach was ultimately not successful in phase III trials (NCT01480479).^47,48

Rather than relying on endogenous DC to migrate, phagocytose, and present antigen, peptide-pulsed DC are made by culturing DC with peptides ex vivo so that they can process and present peptides of interest in an efficient and controlled manner. By presenting antigens of interest on the surface of DC via MHC molecules, tumor antigen-loaded DC can expand T-cell populations capable of recognizing tumor-specific targets. The activity of DC vaccines may be explained by the fact that DC vaccines allow for the natural processing and presentation of antigens. DC vaccines have been loaded with antigens through peptide pulsing or electroporation with mRNA encoding for tumor targets. Electroporation involves applying a short, high-voltage electric pulse to a cell to create temporary pores in the cell membrane, where mRNA encoding for TAA is passively taken up by DC and assembled with MHC machinery.^21,22,49 This mRNA approach has been extensively used with the Cytomegalovirus (CMV) immunodominant protein pp65.^46,50

DC Vaccines Targeting a Single Antigen

One of the earlier trials of DC vaccines in glioblastoma reported a significant intratumoral cytotoxic killer CD8 + T-cell infiltration in 3 of the 6 patients who underwent reoperation.³⁶ In this study, the median survival for patients with recurrent glioblastoma (n = 8) was 133 weeks. To date, several peptide-pulsed DC cell vaccines have been generated and explored in clinical trials (Table 1) that target TAAs such as WT1,^42,51 HER2,^43,51,52 IL-13Rα2,^52,53 Survivin,⁴³ and MAGE-A3⁴³ as well as tumor-specific antigens such as EGFRvIII⁵⁴ and IDH1 R132H (NCT02771301). A phase I study tested EGFRvIII-specific peptide (PEPvIII)–KLH pulsed DCs in 12 patients with newly diagnosed glioblastoma after surgical resection and radiotherapy.⁵⁴ Median progression-free survival (PFS) and overall survival (OS) were 6.8 and 18.7 months, respectively, and no life-threatening adverse reactions or toxicities were observed.⁵⁴ Another phase I trial injected WT-1-loaded DC into seven patients with high-grade glioma. Five patients exhibited stable clinical responses after the first DC vaccine with an OS of 12.3 months.^42,55

In addition to peptide-pulsed DC vaccines, an alternative approach has been described wherein DC cells are electroporated, with mRNA encoding targets such as CMV antigen pp65.^46,50 Like peptide-pulsed DC vaccines, this approach allows DC to express specified antigens via MHC molecules. CMV targeting in glioblastoma represents another avenue for single antigen targeting with DC vaccines. Antigens such as CMV pp65 expression are commonly expressed in > 90% of sampled glioma specimens^56–59 and not surrounding normal brain. Although the pp65 protein is heterogeneously expressed within tumors, importantly their antigen expression is retained in both primary and recurrent glioblastoma. Thus, pp65-encoding mRNA has also been serially tested in DC vaccine trials as a potent immunotherapy.⁶⁰

A small randomized single-blinded phase II trial (NCT00639639) evaluated pp65-specific DC along with preconditioning the vaccine site with tetanus-diphtheria toxoid (Td) in patients with newly diagnosed glioblastoma showed an encouraging PFS of 15.4–47.3 months and OS of 20.6–47.3 months.²⁵ The Td arm revealed long-term survivors in 3 of the 6 treated patients, with follow-up data showing an unprecedented OS of 5–13 years from original diagnosis.⁴⁶ Since this initial small study, a validation double-blinded randomized phase II clinical trial has been conducted in the same patient population with the same DC vaccination and preconditioning scheme (NCT02366728), and results are pending. Another multi-institutional randomized phase II trial involving a CMV pp65 DC vaccine is currently recruiting (NCT02465268).

DC Vaccines Targeting Multiple Antigens

To overcome limitations due to heterogeneity, multiple DC vaccine strategies have moved towards multi-epitope vaccine-based approaches to provide additional antigenic coverage. A DC vaccine strategy involved pulsing DC with several TAA epitopes (included HER2, TRP-2, gp100, MAGE-1, IL13Rα2, and AIM-2) in HLA-A1- and/or HLA-A2-positive patients with glioblastoma. The trial reported an OS of 38.4 months (NCT01222221).⁵² Although this approach may seem efficacious and synergistic, it should be noted that many candidate tumor antigens are not particularly immunogenic and fail to elicit adequate T cell clonotypic expansion. These strategies might extend DC cell vaccines to patients with various MHC haplotypes. Vaccines encoding multiple epitopes increase the likelihood of an epitope being presented by a patient’s MHC molecule.⁶¹

There has also been investment in identifying the optimal non-viral tumor-specific antigen to encode via mRNA and deliver to DC.^62–64 Lin et al. evaluated the expression profile of glioblastoma antigens using the Gene Expression Profiling Interactive Analysis to determine their influence on clinical prognosis.⁶⁴ Patient survival rate and infiltration of APCs were associated with six overexpressed and mutated tumor antigens (ARHGAP9, ARHGAP30, ARPC1B, CLEC7A, MAN2B1, and PLB1) in glioblastoma.⁶⁴ These promising targets will be the foundation of several clinical trials (NCT02649582, NCT02709616).

ICT-107 is a multi-antigenic glioblastoma polypeptide DC vaccine. DCs have been developed and loaded with 6 peptides selected based on a gene-overexpression comparison between glioblastoma cells and nonmalignant tissues: melanoma-associated antigen 1 (MAGEA1), HER2, interferon-inducible protein AIM-2, I-dopachrome tautomerase (DCT), melanocyte protein (PMEL), and IL-13 receptor subunit-α2 (IL-13Rα2). In a phase I clinical trial, ICT-107 DC vaccines were administered to 15 patients with newly diagnosed glioblastoma. Median PFS was 16.9 months, and median OS was 38.4 months. Six patients showed no evidence of tumor recurrence after 40.1 months.⁵² A randomized phase II trial in 124 newly diagnosed glioblastoma patients (77 HLA A2-positive patients) showed encouraging results in the experimental group, especially for HLA-A2-positive patients, including a statistically significant elevation in the PFS by 2.2 months in the ICT-107 cohort with preservation of the quality of life in addition to a considerable therapeutic benefit with ICT-107 for MGMT promoter methylated and unmethylated HLA-A2 patients.⁴³ The trial was underpowered, which disabled subgroup conclusions. Financial difficulties disrupted the phase III trial of this vaccine, and progress in the ICT-107 program has stalled (NCT02546102).

Although algorithms like NETMHC have been developed to predict peptides that are most likely to bind patients’ MHC class I and class II molecules, these algorithms are imperfect and routinely miss or falsely predict MHC binders.⁶⁵ Additionally, there is a low frequency of tumors that can be targeted using these vaccination strategies given the infrequency and heterogeneity of specific antigens in glioma. Given this limitation, many DC vaccine approaches have attempted to derive personalized antigens directly from patients’ tumors using tumor lysates,³⁶ irradiated tumor samples,¹⁹ and pulsing DC with peptides that have been eluted from tumors.²⁰

Using the described approaches, investigators have been able to take antigens directly from the tumor without needing to predict or identify the antigens themselves. In doing so, it can generate immune responses specific to a patient’s HLA and against antigens that are expressed within a patient’s tumor. DCs that are pulsed with tumor derivatives have been shown to generate both conventional class II responses and class I responses through cross-presentation.⁶⁶ One advantage of using these whole-tumor cell approaches to loading DC with antigens is that they contain a wide range of antigens, including both proteins and non-protein antigens such as lipids and carbohydrates. This can provide a more comprehensive immune response compared to using isolated protein antigens. Additionally, these approaches allow patients to generate immune responses against multiple antigens thereby making it less likely for a patient’s tumor to evade the immune system due to heterogeneity. Lastly, these approaches allow DC to display non-mutated yet post-translationally modified antigens specific to the tumor that would not normally be predicted using conventional peptide-based vaccination strategies.⁶⁷ Preclinical studies by Grauer et al. have found that, compared to peptide-pulsed DC vaccine approaches, tumor lysate-loaded DC vaccines were superior at generating protective immune responses against gliomas.⁶⁸

Despite these noteworthy advantages, one major challenge is that whole-tumor-derived antigens can be difficult to purify, prepare, and quantify. Whole-tumor lysates contain a variety of antigens, some immunogenic and some not. This can make it difficult to accurately measure the dose of antigens being delivered to DC, to standardize across patients, and to ensure the safety and efficacy of each DC vaccine. Novel screening and immunogenicity assays need to be developed to ensure DC vaccines loaded with full tumor lysate can develop potent T cell responses. Clinical trials with DC vaccines loaded with known antigens, like mRNA CMV pp65 (NCT00639639),⁵⁰ are able to validate the vaccine by quantifying vaccine antigen-specific T cells responses in peripheral blood. Unfortunately, experiments attempting to quantify immune responses through stimulation with classical tumor antigens (Survivin, EPHA2, IL-13Rα2, etc) lead to variable results between patients and healthy controls, thus are not reliable measures of response. This is a major limitation in validating successful DC vaccine treatment and predicting potential success outside of survival outcomes. While unlikely, the lysate samples may also contain antigens from healthy tissue and lead to potential autoimmune reactions that should be monitored.

All things considered, multi-antigenic DC vaccine platforms remain preferred over single antigen targeting, particularly given the majority of these antigens are TAAs with relatively low expression and avidity for endogenous immune recognition.

DC Vaccines Targeting Glioma Cancer Stem Cells

GSCs reside in gliomas and contribute to the highly aggressive and recurrent nature of this tumor. They are known for their self-renewal capability, resistance to chemoradiotherapies, and implicated in tumor recurrence.^69–71 It is presumed that targeting glioma cancer stem cells (GSC) elimination can allow a greater antitumor response provided by DC vaccines. In a recently published preclinical study, Sy Do et al. used CD133 mRNA-loaded DC targeting humanized mouse CD133 + GSCs, leading to a robust and long-lasting immune response along with inhibition of CD133-positive glioma stem cell propagation and tumor growth.⁷² One approach to retain the benefits of tumor-derived antigens while increasing the standardization and reproducibility of DC vaccines has been to extract RNA-encoding antigens from tumors and deliver these RNA molecules using approaches including but not limited to electroporation. When comparing DC electroporated with whole-tumor cell RNA to pulsing DC with UV-irradiated tumor cells, Benencia et al. found RNA-electroporated DC to be superior to tumor lysate-pulsed DC in preclinical studies.⁷³ An additional advantage of this approach is that RNA can be extracted from subsets of cells such as CD133 + GSCs (NCT00890032) to selectively eliminate cells that are most likely to contribute towards immune resistance. Given the benefits of RNA-electroporated DC and relative ease of manufacturing, this methodology has gained significant popularity to treat gliomas. An important consideration; however, in manufacturing DC vaccines is that gliomas are constantly evolving particularly after treatment with therapies such as temozolomide. As such, in gliomas that have undergone significant immunoediting following chemotherapy or radiation, it may not be appropriate to use the initial surgical lysate or biopsy to generate DC vaccines to treat these gliomas over an extended period.

DC Vaccines Targeting Combined Tumor Populations

In a trial by Parney et al., 21 patients were enrolled and received multiple vaccine doses (≥15) following a single leukapheresis for each patient.⁴⁵ Investigators used a panel of GBM cell lines derived from patient biopsies that grew in culture using a previously described method,⁷⁴ lysed the tumor cell lines, and loaded DC with the mixture of lysates. This study showed how it is possible to vaccinate more patients more often with DC vaccines using an optimized manufacturing strategy. Antigens identified in their mixed lysate platform include EGFRvIII, erbB2, gp100, MAGE-A3, IL13Ra2, and several others. No vaccine toxicities that were dose-limiting have been reported. One patient developed symptomatic pseudoprogression that was proven histologically. Median PFS was 9.7 months. Median OS was 19 months. OS was 25% at 2 years and 10% at 4 years. Specific CD8 + T cell responses for the TAA gp100 were observed post-vaccination. Interestingly, when compared to healthy donors, patients began the trial with a leukocyte deficit that partly normalized over the course of DC therapy.

DC Vaccines Developed in Combination With Immune Checkpoint Inhibitors

Combination of immune checkpoint inhibitors to reverse the immunosuppressive microenvironment is currently an area of investigation as a strategy to enhance the efficacy of DC vaccines. The challenge in primary CNS tumors is to have sufficient and non-exhausted T cells reaching the TME. While several trials are either completed (NCT02529072) or still enrolling patients (NCT04201873), there are no reported data on the efficacy of DC vaccines in combination with immune checkpoint inhibitors. While vaccine generating protocols and antigen selection methods have evolved, DC vaccines still fail to produce a robust and complete antitumor effect in glioblastoma even when paired with checkpoint inhibitor therapy.

Challenges in DC Vaccine Trials in Neuro-Oncology

Many features of clinical trial design can influence validity of the trial results and this may diminish the meaningfulness of the findings. These neuro-oncology trial design issues have been addressed extensively elsewhere in the literature.^75,76 The issues include a lack of control arms in an overwhelming majority of past and current studies, poor accrual rates, patient selection biases, and poor statistical design that may not adequately answer the research questions. Here we discuss two key features (patient selection and lack of controls) that are highly pertinent to DC vaccine trials.

Patient Selection

Generalization is one aim of a clinical trial. Shared biology is the strongest basis for generalization, but patients enrolled should be representative of those suitable for treatment. In a randomized trial, the set of all randomized patients is known as the “intention to treat population” (ITT) population, reflective of who might be offered treatment in clinical practice. The ITT population should be the primary focus for conclusions about treatment efficacy. This can be a problem if DC vaccination is used for patients with recurrent glioblastoma in tumors that are more heterogeneous and immunosuppressed compared with newly diagnosed glioblastoma.

Selection of Control Group

A control group is a group of clinical trial subjects who are not treated with the experimental therapy but instead receive a standard of care regimen with or without a placebo. A control group is a fundamental explicit component of any randomized clinical trial but can also be an implicit element of a single cohort trial. The most reliable results come from trials with internal control groups. If participants in the active study arm have a better outcome than those in the control group, then the intervention is deemed effective. Randomization and treatment masking are methods to minimize the risk of bias. Randomization balances known and unknown confounders across arms and ensures that the experimental therapy and control groups are similar. Masking eliminates observer bias so that participants are managed the same way during the trial.

Randomized controlled trials (RCTs) are the gold standard design for efficacy, particularly in late development testing. However, RCTs are expensive, require larger sample size, and more time to complete than many other trial designs. The issue with trials with sample size is not unique to studies evaluating DC studies but it extends to all trials in neuro-oncology. RCTs can also be less appealing to patients and providers, knowing that they may not receive the experimental therapy. The best control group for a clinical trial depends on the specific research question being addressed and the type of brain cancer being studied. Additionally, the choice of control group might be discussed with regulatory authorities and the ethics review board before the start of the trial to ensure the study design is appropriate. There are several types of control groups that are commonly used in neuro-oncology clinical trials:

• Placebo control: Participants in the control group receive a placebo (an inactive treatment) instead of the experimental treatment. This type of control group is used to evaluate the effectiveness of the experimental treatment by comparing the outcomes of the 2 groups.

• Active control: Participants in the control group receive a standard of care treatment that is currently used to treat brain cancer, rather than a placebo. This type of control group is used to evaluate the effectiveness of the experimental treatment by comparing it to the standard of care.

• Concurrent control: Participants are randomly assigned to either the experimental group or the control group, this way the characteristics of the group are similar, and it reduces the bias of the study. Also randomized controls are usually active control (or placebo).

• Historical control: Participants in the control group are selected from a recent historical database of patients who have been treated for brain cancer. The outcomes of the historical control group are then compared to the outcomes of the experimental group.

• Externally augmented clinical trial (EACT) controls: To overcome challenges with RCTs, trials leveraging external data, termed EACT designs, have received increasing attention in the field. EACTs could use external (synthetic) or hybrid control arms.

• External (Synthetic) Control: Clinical trials may employ external data for interim decision making or for final analysis. An external control may include clinically annotated pretreatment data and outcomes from previously completed studies.⁷⁷

• Hybrid control: this type of control integrates randomization to an internal control arm, but it is supplemented by an external control.⁷⁸ An external control arm can supplement an existing internal control in a hybrid design to lower the number of patients randomized to the control arm. The recently published phase II trial of MDNA55 (that targets the IL-4 receptor) in recurrent glioblastoma used an external synthetic control arm.⁷⁹ A registration trial evaluating MDNA55 in recurrent glioblastoma is planned to use a hybrid design. Accessing high-quality data and potential risks of bias of unmeasured confounders with external data remains a major challenge.

Almost all published DC vaccine trials (Table 1) including studies with the DC vaccines targeting pp65 CMV antigen and the DCVax-L study illustrate the importance of appropriate controls. The pp65 CMV antigen DC vaccine trial²⁵ suggested a link between increased vaccine migration resulting from Td preconditioning of vaccine site and greater overall survival. However, while blinded and randomized, this clinical trial was small (n = 11) and not powered for survival. Additional findings from this trial focused on elevated levels of systemic inflammatory proteins in patients receiving Td compared to control patients treated with unpulsed DCs at vaccination site. While the difference between treatment and control groups are interesting our trial did not have a pretreatment sample to determine the specific effect of Td preconditioning treatment. The combination of small sample size and lack of essential control samples only allowed for trends to be determined and new questions developed for future trials. The DCVax-L study reported a median OS for newly diagnosed GBM patients (n = 232) of 19.3 months from randomization (22.4 months from surgery). While the study was initially designed as a randomized trial, investigators reconstituted the primary analysis to compare DCVax-L patients with an external control arm in newly diagnosed and recurrent glioblastoma cohorts. As investigators lacked patient-level data, they drew matched external control cohorts of patients treated with standard of care based upon 14 criteria accessible in already published RCT data. In their updated analysis and published report, DCVax-L was associated with better survival as compared to external control (median 16.5 months; HR = 0.80, P = .002).⁴⁴ In this trial, 48 months survival measured from randomization was 15.7% versus 9.9% and at 60 months was 13% versus 5.7%. In patients with recurrent GBM (n = 64), mOS was 13.2 months from time of relapse versus 7.8 months (HR = 0.58, P < .001). Survival at 24- and 30-month post-recurrence was 20.7% versus 9.6%, and 11.1% versus 5.1%, respectively. In patients with newly diagnosed GBM who have methylated MGMT, mOS were 30.2 months from time of randomization (33 months from surgery) with DCVax-L (n = 90) versus 21.3 months in controls (n = 199) (HR = 0.74, P = .027) (NCT00045968). The authors reported an improvement in survival in DCVax-L-treated patients, but there is a need to better understand the external control arm that was obtained from patient data from other reported trials. It is critical to determine if the external data could have included patients who were likely to have worse survival than the patients who received the vaccine in this trial. For example, the DCVax-L trial excluded patients with residual disease or progression during chemoradiation, which are known prognostic factors, but all the included trials in the external control arm did not necessarily use that criterion. Unfortunately, this was also an issue in other glioblastoma randomized clinical trials eg, the rindopepimut ACT IV trial that only enrolled patients with glioblastoma who received maximal resection of the tumor and have completed standard radiation and chemotherapy without progression. Another concern in interpreting the efficacy of this particular DC vaccine is that in recurrent patients, the lysate used in these DCVax-L vaccines was manufactured from the tumor at diagnosis and then given to patients at the time of recurrence. As such, while further study is needed, using lysate generated from the initial surgical resection may not be ideal for making vaccines to treat tumors at recurrence.

It is well recognized that there are major challenges with manufacturing DC vaccines, particularly when loaded with tumor lysates. The number of relevant antigens in tumor lysates is lower compared to peptide-loaded DCs. This has not been adequately quantified and typically, antigen-specific responses are not monitored in clinical trials. Unfortunately, there is no appropriate quality control of the product prior to administration of the vaccine, neither with respect to quality of the tumor tissue nor with respect to antigenicity. This is one of the central challenges that face standardized testing in clinical trials.

Emphasis should be placed on trial design of using DC vaccines.⁸⁰ The single-arm uncontrolled DC vaccine trials, often from a single center (eg, the DC vaccine trial targeting pp65 CMV),^46,50 have been commonly used in treatments developed for glioblastoma. Even in two-arm vaccine trials²⁵ patient numbers are exceedingly small and consist of highly stringent patient selection criteria biased towards more favorable prognoses and not generalizable to most of the glioblastoma population. As a result, these DC vaccine trials overestimate benefits and are not reliable in predicting efficacy.

Neoadjuvant, Window of Opportunity, and Trials With Primary Biologic Endpoints

Immunotherapies can facilitate endogenous antitumor immunity and hold great promise but there are unanswered questions about sequencing and ideal schedules of administration. Neoadjuvant and window of opportunity trials focused on neoadjuvant (presurgical) immunotherapy provide biospecimens for analyses and have the potential to identify mechanisms, and pathways that can inform further development of efficacy trials.⁸¹ Clinical trials employing a pharmacokinetic (PK) or a pharmacodynamic (PD) assays as primary endpoints could improve the validity and efficiency of early development of immunotherapeutic agents.⁸² These trials may include PK and PD assessments but also other biologic measurements such as DC migration or immunologic measurements specific to DC vaccine trials. While these biological assessments can be very useful in earlyphase development, they cannot be used as primary endpoints in later development where improvement in clinical outcomes is sought. Caution is needed when relying on biologic and surrogate markers that might lead to oversimplification of complex biological processes and obscure important insights into the mechanisms of disease. Biological and surrogate markers can be efficient interim surrogates, appropriate before moving to a larger, time- and resource-consuming trial. However, these markers may not accurately reflect vaccine impact on patient outcomes. A balance is needed between the use of biological and surrogate markers and clinical outcomes in clinical trials and especially for immunotherapies such as DC vaccines in glioblastoma where the interpretation of immunological measures may not be straightforward. Lastly, there is a need for validation of surrogate markers/ endpoints before using them to inform important interim or final analyses in trials.

Conclusion: Lessons for Future DC Vaccine Trials

We have summarized past and ongoing efforts to develop DC vaccines in glioblastoma. The existing literature on DC vaccine trials consists of many underpowered, uncontrolled, or not adequately controlled clinical trials. Ongoing careful and critical evaluation of the strengths and weaknesses of the studies is needed to leverage these insights to design future trials in this area. However, several lessons can be learned from the current state of DC vaccines in glioblastoma.

Concerns regarding clinical trial design in neuro-oncology are well recognized. Limitations also extend to aspects of accrual, and data reporting.^75,76 Low rates of randomization, underuse of controls, and overestimation of benefit/ effect size, particularly pronounced in earlyphase trials, are all likely to limit the wide applicability of results. Suboptimal designs may be driven by accrual challenges, highlighting the need for more collaborative efforts and to create incentives across these areas to be able to conduct larger trials, ideally with contemporaneous controls but when appropriate unequal (eg, 2:1 or 3:1) randomization can be considered.⁸³ Methods such as sequence balance minimization could be used⁸⁴ because they have favorable factor-balancing of relevant properties and is useful for small trials seeking to achieve balance across several prognostic factors, which is very relevant in an area like DC vaccine development. When designing clinical trials, the main goal should be the generation of meaningful and interpretable results that contribute to the scientific understanding in this field but more importantly to improve patient outcomes. A well-designed control group is critical for meaningful comparisons. External, synthetic, and hybrid controls might be appropriate in certain situations. They are more appropriate in early-phase clinical trials, but they need validation and extra scrutiny especially when used in late-stage development.

Another area that needs improvement is misrepresentation of endpoints, which can lead to misleading results and undermine validity. Using historic data creates biases in patient selection, as a result, it has been recommended that single-arm phase II studies that are also well-controlled, in the setting of recurrent glioblastoma should aim for an objective response rate of > 25% to convincingly demonstrate antitumor activity.⁸⁵ Response rate is not an ideal endpoint in the setting of DC vaccines. Similarly, progression-Free Survival (PFS) is often used, but PFS has the same issues with response rate relying on MRI evaluation that can be inaccurate because of the not uncommon occurrence of pseudoprogression and pseudoresponse in glioblastoma. Pseudoprogression is likely to be much more of an issue with vaccine and other immunotherapies trials relative to cytotoxic and targeted therapies. The Response Assessment in Neuro-Oncology (RANO)⁸⁶ criteria for radiographic response assessment in glioblastoma clinical trials have been developed to standardize the evaluation of imaging and determination of PFS. Although specific immune RANO⁸⁷ have been developed but, there is still debate about the ideal criteria in the setting of immunotherapy in neuro-oncology. The most robust clinical endpoint is OS. However, OS has its own challenges, including the need for longer follow-up and the impact of patients crossing over to the experimental arm.

Significant focus should be placed on improving the design and implementation of trials by evaluating potential benefits to patients, hypothesis-driven scientific questions, safety, and feasibility both in terms of logistical aspects as well as the likelihood that the trial will lead to interpretable clinical insight. By considering these metrics, trial sponsors and regulatory agencies can make informed decisions about whether to proceed with DC vaccine trials and ensure that trials are designed and conducted in a safe, effective, and feasible manner for patients.

Limited patient accrual and challenges in selecting appropriate control cohorts in glioma DC vaccine trials highlight the need for more collaborative efforts, not only to increase sample size and statistical power of clinical trials but also to ensure the use of optimal clinical trial designs, expedite data reporting, harmonization, and generalizability of trial results. Through these improvements, the field of DC vaccine clinical trials may help the field to achieve the ultimate goal ie, improving outcomes of patients with both lower grades of glioma as well as glioblastoma.

Funding

This work did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.

Conflicts of interest statement

KB reports equity interest in Annias Immunotherapeutics, which has licensed intellectual property from Duke related to the use of CMV DC vaccine with tetanus-diphtheria (Td) toxoid in the treatment of glioblastoma. She is an inventor on patents related to Td + DC vaccine and CCL3 in the treatment of glioblastoma. MK reports research funding to institution: BMS, AbbVie, Daiichi Sankyo, BioNTech, Immorna Therapeutics, Celldex, Astellas, CNS Pharmaceuticals. Honoraria: GSK, Novocure, JAX lab for genomic research, Johnson and Johnson, Voyager therapeutics and George Clinical. The other authors have not reported any disclosures.

Authorship statement

Design and its implementation: KH, KB, MK. Acquisition, analysis, or interpretation of data: all authors Writing of the manuscript, revision, approval of the final version: all authors