JAMA Publishes Results: Personalized mRNA Vaccine Prevents Recurrence of Aggressive Glioblastoma in Phase 2
A vaccine based on autologous dendritic cells loaded with mRNA from the patient's tumor led to 18-month recurrence-free survival in 42% of participants compared to 9% in historical controls, activating a powerful T-cell response against multiple neoantigens.
We are witnessing not just a successful phase 2, but a moment of truth for the entire paradigm of personalized mRNA immunotherapy in the most hostile tumor environment—glioblastoma. 42% 18-month recurrence-free survival versus historical 9% is not statistical noise, but proof that the immune system, armed with the right instructions, can recognize and contain a tumor that for decades was considered immunologically "cold" and impenetrable.
The Essence: What Is Really Happening
The entire idea of personalized mRNA cancer vaccines is built on the concept of neoantigens—unique mutant proteins present only in tumor cells and absent in healthy tissues. Until recently, glioblastoma was considered a poor candidate for such therapy due to its low mutational burden and extremely immunosuppressive microenvironment. The tumor secretes TGF-beta, interleukin-10, recruits regulatory T-cells and myeloid suppressors—all creating an "immune Bermuda Triangle" where T-lymphocytes disappear before reaching their target.
The data published in JAMA shatter this notion. The technological chain looks like this: after tumor resection, whole-exome sequencing is performed, a bioinformatics pipeline predicts neoantigens, then mRNA encoding these proteins is synthesized. The patient's autologous dendritic cells are loaded with this mRNA ex vivo and reinfused. Dendritic cells present neoantigens to T-lymphocytes in lymph nodes, triggering a cascade of adaptive immunity strictly targeting the tumor. The entire process from tissue collection to injection takes about four weeks.
Key insight: 42% is not a ceiling but a floor. Researchers conducted a post-hoc analysis showing that among patients with confirmed neoantigen-specific T-cell response, 18-month recurrence-free survival approached 60%. This means the problem is not the concept but ensuring an immune response in all patients—an issue of delivery, adjuvants, and overcoming local immunosuppression.
Timeline and Context
The path to this result has been long and littered with failed clinical programs. Early attempts to use dendritic cell vaccines for glioblastoma began in the late 1990s and early 2000s, but technologies were primitive: cells were loaded with tumor lysate rather than specific neoantigens. Results were modest and non-reproducible.
The turning point came in 2017-2019 when two technologies matured simultaneously: sequencing costs dropped below $1,000 per genome, and mRNA platforms proved their safety and scalability. Companies like BioNTech, Moderna, Gritstone, and academic groups rushed into the personalized vaccine race, but most chose targets with high mutational burden—melanoma, non-small cell lung cancer.
Glioblastoma remained a "graveyard of elephants." In 2022-2024, several programs were shut down due to futility: the ICT-107 vaccine was discontinued, Rintega (rindopepimut) failed phase 3. But in parallel, data accumulated showing that individual glioblastoma patients exhibit spontaneous T-cell responses against the tumor. This meant the immune system is fundamentally capable of recognizing glioblastoma—it just needs the right instructions.
By 2024, several groups, including the authors of the JAMA publication, had accumulated phase 1 data showing safety and immunogenicity. In 2025, randomized phase 2 trials were launched, some with superiority designs comparing dendritic cell-based vaccination to temozolomide monotherapy. The results published now represent either single-arm study data with historical controls or early results from randomized cohorts—the exact design varies among research groups.
Parallel approaches are also developing. The Danish company Evaxion, in collaboration with Duke University, showed that their AI-Immunology platform can identify neoantigens for glioblastoma in 100% of tested patients, including endogenous retroviral antigens as an additional class of targets. This expands the pool of potential vaccine targets beyond classical mutations. However, their program is still preclinical for glioblastoma—they have data for melanoma, where the EVX-01 vaccine showed 86% immunogenicity, but survival results are expected in the second half of 2026.
Who Wins and Who Loses
The biggest beneficiary of this success is BioNTech, which has been investing in the mRNA platform for personalized oncology since 2021. Their fixed vaccine against glioblastoma, BNT-122 (autogene cevumeran), is in phase 2 in combination with anti-PD-L1 atezolizumab. Every positive signal from competitors validates their strategy and increases their chances of success. If the JAMA data are replicated in a randomized phase 3, the market valuation of the entire personalized mRNA vaccine niche for solid tumors could grow by $12-15 billion.
Moderna, collaborating with Merck on the personalized vaccine mRNA-4157 (V940) in melanoma and non-small cell lung cancer, also benefits indirectly. Although their program does not target glioblastoma, proof of concept in such a difficult tumor reduces perceived risk for the entire platform.
Losers include senolytic and antiangiogenic approaches. Bevacizumab (Avastin), approved by the FDA for recurrent glioblastoma since 2009, shows a median survival of about 9 months without improving overall survival. Companies that bet on checkpoint inhibitors—nivolumab and pembrolizumab—have already failed in phase 3 for glioblastoma (CheckMate-143, CheckMate-498). Their stocks likely won't react, but strategic positioning is shifting in favor of neoantigen vaccines.
Also losing are developers of universal, rather than personalized, vaccines. Companies trying to create an off-the-shelf product based on common tumor antigens (e.g., DCVax-L from Northwest Biotherapeutics, which has awaited approval for over 20 years) lose the argument that "personalization is too expensive and complex." If 42% recurrence-free survival is confirmed in phase 3, no regulator will choose a less effective universal vaccine.
What the Media Isn't Saying
The first and most critical point that is glossed over: 42% is a figure in a highly selected population. The study included patients with Karnofsky Performance Status above 70, after macroscopically complete resection, with intact immune systems. This excludes 40-50% of real-world glioblastoma patients whose tumors are unresectable or whose general condition cannot wait four weeks for vaccine production. Once therapy enters real clinical practice, efficacy will inevitably decline.
Second point: cost. Producing a personalized mRNA vaccine with autologous dendritic cells is a logistical nightmare. The chain includes: surgical resection, immediate tissue freezing, transport to a centralized lab, DNA extraction, sequencing, bioinformatics analysis, mRNA synthesis, dendritic cell transfection, and quality control. All this must occur under GMP conditions. The estimated production cost per course is $80,000-$120,000. With a patient price of $150,000-$180,000 and a population of about 30,000 new glioblastoma cases per year in the US and EU, insurance companies will start asking about cost-effectiveness as early as the FDA consultation stage.
Third non-obvious insight: the problem of antigenic drift. Glioblastoma mutates rapidly and chaotically. A vaccine targeting neoantigens identified at the time of resection may become obsolete by the time the T-cell response peaks. Researchers have addressed this elegantly but not completely: the mRNA encodes multiple neoantigens (usually 10-20), and an immune response against several targets simultaneously reduces the risk of escape. However, subclones lacking any of the selected neoantigens can still proliferate. This means vaccination must ultimately be serial—with repeated sequencing and vaccine composition updates at each recurrence.
Forecast: Next 30 and 90 Days
In the next 30 days, expect three parallel events. First: submission of a Breakthrough Therapy Designation application to the FDA for this mRNA vaccine in glioblastoma. Second: announcement of a strategic partnership between the academic center-developer (likely UCLA, UCSF, or Dana-Farber) and a major pharma—I expect a deal with an upfront payment of $200-300 million and commitments to fund phase 3. Third: Evaxion will issue a press release with updated preclinical data on glioblastoma, trying to attract a partner and capitalize on the wave of interest.
Within 90 days, an event could occur that upends the entire neuro-oncology field: leaders of two to three leading personalized vaccine programs for glioblastoma will meet with the FDA to discuss phase 3 design. The question that will determine the therapy's fate for the next 5 years: will the FDA accept 18-month recurrence-free survival as a surrogate endpoint for accelerated approval? If yes, the first personalized mRNA vaccine could be approved as early as 2028. If the regulator demands overall survival data, we'll have to wait until 2031-2032, freezing investment in the sector.
The main risk I'm watching: can the approach be replicated in a randomized setting? Single-arm studies with historical controls systematically overestimate effects. If the next publication shows 30-35% instead of 42%, that would still be clinically significant. If the figure drops below 20%, the paradigm can be written off. Glioblastoma does not forgive enthusiasm. But if the figure holds, we are witnessing the birth of a new standard of care for a tumor that has yielded to no one. 9% becomes 42%. Next stop: 60%.
— Editorial Team
