CAR-T cell therapy changed what seemed possible in cancer immunotherapy. In B-cell leukemias and lymphomas, and later in multiple myeloma, genetically engineered T cells produced deep and sometimes durable responses in patients whose cancers had progressed through multiple previous treatments. Translating the same principle to solid tumors has proved considerably more difficult.
The reason is not simply that solid tumors are more resistant to immunotherapy. Blood cancers and solid tumors present fundamentally different problems for engineered immune cells. In hematologic malignancies, highly useful lineage-associated targets such as CD19 and BCMA can be accessed relatively easily. In solid tumors, an ideal antigen is rarely expressed uniformly by malignant cells while remaining absent from essential normal tissues. T cells must also reach the tumor, penetrate its architecture, remain functional in a hostile metabolic environment and continue recognizing a cancer that can evolve under immune pressure.
Recent developments suggest that some of these barriers can be overcome. Tumor-infiltrating lymphocyte therapy and TCR-engineered T cells have already reached regulatory approval in selected solid tumors. CAR-T studies are producing increasingly convincing signals in gastric cancer and CNS tumors, and in 2026 the first CAR-T therapy for a solid tumor received regulatory approval in China.
The question is therefore becoming more specific: which principles that made cellular immunotherapy successful in blood cancers can be transferred to solid tumors, and which need to be redesigned entirely?
Why Did CAR-T Work So Well in Blood Cancers?
The extraordinary success of CAR-T therapy in B-cell malignancies was made possible by a fortunate biological feature: the existence of accessible and relatively homogeneous surface antigens.
CD19 is expressed throughout much of B-cell development and across many B-cell malignancies. A CAR can recognize CD19 directly on the cell surface without requiring antigen processing or presentation through HLA molecules. Loss of normal B cells is clinically manageable, making CD19 an unusually favorable target even though it is not cancer-specific.
BCMA provides a similar opportunity in plasma-cell malignancies.
This combination is difficult to reproduce in solid cancers. The ideal CAR target would need to be highly and uniformly expressed on malignant cells, accessible on the cell surface and absent from essential healthy tissues. Few solid-tumor antigens satisfy all of these conditions.
Solid tumors also introduce a physical problem. A leukemia cell circulating in blood does not need to be located inside a dense tumor mass before a CAR-T cell can encounter it. In a solid malignancy, therapeutic T cells must leave the circulation, cross abnormal tumor vasculature, migrate along the appropriate chemokine gradients and penetrate tumor tissue before they can begin killing malignant cells.
The challenge is therefore not simply engineering a T cell capable of recognizing cancer. It is ensuring that the cell can find, enter, survive within and repeatedly attack a solid tumor.
CAR T-Cell Therapy Death Rate: Treatment-Related Mortality, Causes of Death, and Risk Factors
The Antigen Problem Comes First
Solid tumors rarely provide an equivalent of CD19.
Many potential targets, including HER2, mesothelin, EGFR, GD2, B7-H3, GPC3 and CLDN18.2, are tumor-associated rather than strictly tumor-specific. This creates a narrow therapeutic window because CAR-T cells can attack normal tissues expressing the same antigen.
At the same time, expression within the tumor itself can be heterogeneous. Some malignant cells may strongly express the target, others weakly, and some not at all.
This becomes an evolutionary problem once treatment begins. CAR-T cells eliminate antigen-positive cells, creating selective pressure favoring malignant clones that lack or downregulate the target. A treatment can therefore be biologically active and still fail because the tumor changes its antigenic composition.
Next-generation strategies are attempting to address this through dual-target and multi-target CARs, tandem receptors, logic-gated systems and circuits that require specific combinations of antigens before full T-cell activation occurs.
The objective is no longer simply stronger recognition. It is more selective and evolution-resistant recognition.
CLDN18.2 Has Provided the Strongest CAR-T Evidence So Far
One of the most important developments has come from gastric and gastroesophageal junction cancer.
CLDN18.2 is a tight-junction protein whose normal expression is largely restricted to gastric epithelial cells, where much of the protein is structurally sequestered within tight junctions. During malignant transformation, disruption of normal tissue architecture can expose CLDN18.2 on the tumor-cell surface, making it an attractive therapeutic target.
Satricabtagene autoleucel, or satri-cel, is an autologous CAR-T therapy directed against CLDN18.2.
The randomized phase II CT041-ST-01 trial, published in The Lancet in 2025, compared satri-cel with physician’s choice of treatment in previously treated CLDN18.2-positive advanced gastric or gastroesophageal junction cancer.
The trial represented an important transition for solid-tumor CAR-T development because it moved beyond small single-arm feasibility studies into randomized clinical testing.
Patients receiving satri-cel experienced significantly longer progression-free survival than those receiving physician’s choice therapy, providing evidence that CAR-T therapy could produce clinically meaningful benefit in an epithelial solid tumor.
Then came an even larger milestone.
In June 2026, China’s National Medical Products Administration approved satri-cel for previously treated CLDN18.2-positive advanced gastric or gastroesophageal junction cancer.
It became the first CAR-T cell therapy approved anywhere in the world for a solid tumor.
For more than a decade, the division seemed clear: CAR-T belonged to hematologic oncology, while solid tumors remained experimental territory. Satri-cel does not erase the biological differences between the two settings, but it demonstrates that those differences are not an absolute barrier.
CARs Are Not the Only Way to Redirect T Cells
One limitation of CAR technology is frequently overlooked.
CARs primarily recognize molecules expressed on the cell surface.
Yet most proteins produced by cancer cells are intracellular.
T-cell receptors operate differently. Intracellular proteins are degraded into peptides, which can then be presented on the cell surface by HLA molecules. TCR-engineered T cells can therefore recognize targets that would remain invisible to conventional CARs.
This dramatically expands the potential antigen space.
The trade-off is HLA restriction. A TCR therapy recognizing a peptide presented by one HLA allele will not necessarily work in a patient with another HLA genotype.
Nevertheless, TCR therapy has already crossed an important regulatory threshold.
Synovial Sarcoma Became the First Solid Tumor With an Approved Engineered TCR Therapy
In August 2024, the FDA granted accelerated approval to afamitresgene autoleucel, or afami-cel, for selected adults with unresectable or metastatic synovial sarcoma after prior chemotherapy.
Afami-cel is fundamentally different from CAR-T therapy.
Patient T cells are genetically engineered to express a TCR recognizing a peptide derived from MAGE-A4, a cancer-testis antigen, presented by specific HLA-A*02 molecules.
The approval was based on the SPEARHEAD-1 trial and represented the first FDA approval of an engineered TCR T-cell therapy.
Its importance extends beyond synovial sarcoma.
TCR engineering provides a strategy for targeting intracellular cancer proteins, including cancer-testis antigens and potentially mutation-derived neoantigens. The accessible target universe is therefore considerably larger than the surface proteome available to conventional CARs.
The limitation is precision. The tumor must express the relevant antigen, and the patient must carry the appropriate HLA allele.
Cell therapy in solid tumors may therefore increasingly resemble precision oncology, with treatment determined simultaneously by tumor antigen expression and host HLA genotype.
TIL Therapy Took a Completely Different Route
CAR-T and TCR-T therapies begin by engineering recognition.
Tumor-infiltrating lymphocyte therapy begins with immune recognition that has already occurred naturally.
A tumor is surgically removed, lymphocytes that have infiltrated the tumor are isolated and expanded ex vivo, and billions of these cells are reinfused after lymphodepleting chemotherapy, traditionally followed by IL-2 support.
The biological logic is compelling. If a T cell has already entered the tumor, it may already recognize something relevant within that cancer.
In February 2024, lifileucel became the first FDA-approved cellular therapy for a solid tumor and the first approved TIL therapy, initially for adults with advanced melanoma after progression on appropriate previous therapies.
Long-term follow-up has strengthened the evidence that some responses can be remarkably durable.
The five-year analysis of the C-144-01 study reported an objective response rate of approximately 31%, with a median duration of response of 36.5 months. Some responses deepened over time, and approximately one-third of responders who reached the five-year assessment continued to have ongoing responses.
TIL therapy therefore demonstrates something fundamental: effective cell therapy in solid tumors does not necessarily require inventing an artificial receptor. Sometimes the tumor has already selected the relevant T cells.
The challenge is identifying, expanding and reinvigorating the right ones.
Not Every TIL Is Equally Useful
The success of TIL therapy in melanoma has not automatically translated to epithelial cancers.
This has led investigators to ask whether the problem lies not with TIL therapy itself but with the composition of the cells being infused.
A particularly interesting example came from the 2025 Nature Medicine phase II study “Neoantigen-Specific Tumor-Infiltrating Lymphocytes in Gastrointestinal Cancers.”
The study treated patients with refractory mismatch repair-proficient metastatic gastrointestinal cancers.
When patients received conventional bulk, unselected TILs, no objective responses were observed among the first 18 patients.
The investigators then changed the strategy.
Instead of expanding TILs indiscriminately, they screened the cells for recognition of patient-specific tumor neoantigens and selected those populations for treatment.
With neoantigen-selected TILs, objective responses began to appear. When pembrolizumab was subsequently added, 8 of 34 patients achieved objective responses, corresponding to an ORR of 23.5%.
This study illustrates an important evolution in adoptive cell therapy.
The future may not depend simply on producing more T cells.
It may depend on identifying which T cells recognize the biologically relevant targets within each patient’s cancer.
Neoantigen-Specific Cell Therapy Moves Personalization Even Further
Tumor sequencing can now be integrated directly into adoptive cell therapy.
Somatic mutations are identified from an individual cancer, candidate neoantigens are predicted, and the patient’s lymphocytes can be screened for T cells capable of recognizing those mutations. Relevant TCRs can then be isolated, expanded or engineered into new lymphocytes.
This creates a form of personalized immunotherapy in which both the antigen and the cellular product are individualized.
A 2026 Nature Reviews Immunology review, “Hallmarks and Correlates of Effective Adoptive Cell Immunotherapy for Cancer,” highlights how immunogenomics, systems immunology, genome editing and computational biology are increasingly converging around this strategy.
Neoantigen-directed TIL and TCR approaches have already produced tumor regression in patients with several metastatic solid tumors, including melanoma, breast cancer and gastrointestinal malignancies.
The concept is particularly attractive because mutation-derived neoantigens can provide genuine tumor specificity.
The challenge is complexity. Every patient’s tumor can contain a different set of mutations, every patient carries a different HLA repertoire, and not every predicted neoantigen generates an effective T-cell response.
The therapy therefore begins to look less like a conventional drug and more like an individualized immune-engineering process.
What Happens If CAR T-Cell Therapy Fails? Relapse, Refractory Disease, and Next Treatment Options
Glioblastoma Has Become an Unexpected Laboratory for CAR-T Innovation
Few cancers illustrate the difficulty of solid-tumor immunotherapy better than glioblastoma.
The tumor is highly heterogeneous, profoundly immunosuppressive and located within the central nervous system. Early systemic CAR-T studies demonstrated that engineered cells could reach glioblastoma but rarely produced durable clinical responses.
Recent trials have begun changing the design rather than simply increasing the dose.
One approach is locoregional delivery.
A 2024 Nature Medicine phase I trial evaluated IL-13Rα2-targeted CAR-T cells administered directly into the tumor cavity or cerebrospinal fluid in 65 patients with recurrent high-grade glioma.
Among evaluable patients, 50% achieved stable disease or better, including partial and complete responses. Treatment was accompanied by increases in inflammatory mediators such as IFNγ, CXCL9 and CXCL10 within the CNS, providing evidence of local immune activation.
Another 2024 Nature Medicine trial tested intrathecally delivered bivalent CAR-T cells targeting both EGFR and IL13Rα2 in recurrent glioblastoma.
All six patients in the initial report demonstrated early reductions in tumor enhancement and size, although none met formal objective response criteria. CAR-T cells and inflammatory cytokines were readily detected in cerebrospinal fluid.
The strategy is important because it addresses two barriers simultaneously: physical access and antigen heterogeneity.
Instead of asking systemically administered CAR-T cells to traffic efficiently into the brain and recognize a single antigen, investigators delivered the cells directly into the CNS and designed them to recognize more than one tumor-associated target.
Diffuse Midline Glioma Produced Some of the Most Striking Early CAR-T Responses
Another important signal has emerged from H3K27M-mutant diffuse midline glioma, including diffuse intrinsic pontine glioma.
These tumors express high levels of GD2, creating a potential CAR target.
In a phase I study published in Nature, patients received GD2-directed CAR-T cells intravenously after lymphodepleting chemotherapy, followed in eligible patients by intracerebroventricular infusions.
Among the treated patients, several experienced substantial tumor reductions. Four patients had major volumetric reductions of 52%, 54%, 91% and 100%, and one complete response remained ongoing for more than 30 months at the time of publication. Neurological improvements were also documented in multiple patients.
The treatment was not without risk. Tumor inflammation-associated neurotoxicity required intensive monitoring and management.
These results remain early and require validation in larger studies, but they provide an important demonstration that engineered T cells can produce profound biological effects even within an anatomically difficult CNS solid tumor.
Delivery May Be Part of the Therapy
The CNS studies illustrate a broader principle.
For conventional systemic drugs, route of administration is often primarily a pharmacological consideration. For living cell therapies, delivery can determine whether the therapeutic cells ever reach their biological target.
Intravenous CAR-T cells must migrate through the circulation, recognize appropriate endothelial signals, extravasate and navigate the extracellular matrix before reaching malignant cells.
Locoregional administration can bypass several of these barriers.
Intratumoral, intracavitary, intrapleural, hepatic arterial and intracerebroventricular delivery strategies are being explored depending on tumor type.
A 2025 Nature Reviews Clinical Oncology review, “Optimizing CAR T Cell Therapy for Solid Tumours: A Clinical Perspective,” highlights locoregional administration alongside early apheresis, optimized lymphodepletion, repeat infusion and improved toxicity management as increasingly important components of solid-tumor CAR-T development.
This suggests that successful cell therapy may require engineering not only the cell but also the route by which the cell reaches cancer.
Reaching the Tumor Is Only the Beginning
Even after engineered T cells reach a solid tumor, they encounter conditions fundamentally different from blood.
Tumor vasculature is abnormal. Extracellular matrix and cancer-associated fibroblasts can restrict movement. Oxygen and glucose may be limited. Lactate and other metabolites accumulate. Suppressive myeloid populations and regulatory T cells alter local signaling.
Tumor cells and surrounding stromal populations can produce inhibitory mediators including TGF-β and other immunosuppressive factors.
CAR-T cells entering this environment can progressively lose proliferative capacity, cytotoxic function and persistence.
A 2025 Nature Reviews Immunology article on microenvironmental regulation of solid-tumor resistance to CAR-T therapy emphasizes that trafficking, function and persistence are all affected by the tumor microenvironment.
The engineering problem therefore extends beyond antigen recognition.
A successful solid-tumor cell therapy must function as a living drug capable of adapting to an environment actively designed by the cancer to suppress immune attack.
Armored CARs Are Being Designed to Fight Back
One strategy is to engineer CAR-T cells that do more than recognize an antigen.
So-called armored CAR-T cells can be designed to release immunomodulatory molecules after activation. Experimental platforms have incorporated cytokines such as IL-12, IL-15, IL-18 and IL-7, chemokines, checkpoint-blocking molecules and other factors intended to enhance T-cell persistence or remodel the surrounding tumor microenvironment.
Other approaches modify inhibitory signaling itself.
A CAR-T cell can be engineered with a dominant-negative TGF-β receptor, altered checkpoint pathways or synthetic receptors that convert normally suppressive signals into stimulatory ones.
Metabolic engineering is another emerging direction. T cells can potentially be modified to tolerate hypoxia, nutrient deprivation or suppressive metabolites more effectively.
The goal is no longer to create a cell carrying only a receptor.
It is to create a cell capable of sensing, surviving and modifying the environment in which it operates.
CAR-T Cells May Also Change the Tumor Around Them
This concept is becoming increasingly important.
The conventional view of solid-tumor CAR-T therapy focuses on what the tumor microenvironment does to the infused T cells.
A 2026 Nature Reviews Bioengineering article, “CAR-T Cells Reshape the Tumour Microenvironment,” emphasizes the reciprocal interaction.
Activated CAR-T cells release cytokines and chemokines, recruit endogenous immune populations and can alter local inflammatory signaling. Tumor-cell killing releases additional antigens that may become available to endogenous antigen-presenting cells.
The therapeutic effect of CAR-T therapy may therefore extend beyond direct CAR-mediated killing.
An engineered T cell could potentially act as a local immune organizer, converting the environment around the tumor in ways that recruit parts of the patient’s endogenous immune system.
This is particularly attractive in heterogeneous tumors where not every malignant cell expresses the CAR target.
If CAR-T activation against one population can trigger broader endogenous immunity against additional tumor antigens, engineered cells may be able to produce effects beyond the antigen encoded in their receptor.
Whether this can be generated consistently in patients remains an important research question.
Persistence Is Another Major Difference Between Blood and Solid Tumors
CAR-T expansion and long-term persistence are strongly associated with durable responses in several hematologic malignancies.
Achieving similar persistence in solid tumors has been difficult.
Patients with advanced solid malignancies are often heavily pretreated, and their harvested T cells may already show impaired fitness. Manufacturing itself can further alter differentiation state.
This has generated interest in collecting T cells earlier in the treatment pathway, shortening manufacturing times and preserving less differentiated T-cell populations with greater proliferative potential.
The optimal cell product may therefore not be the one with the greatest immediate cytotoxicity. A population capable of self-renewal, expansion and sustained function may ultimately provide better tumor control.
The 2025 clinical perspective on CAR-T therapy in solid tumors specifically highlights early apheresis and rapid manufacturing as potentially important strategies for preserving T-cell fitness.
Cell quality is becoming part of treatment selection.
Combination Therapy Must Solve a Specific Barrier
It is tempting to combine CAR-T cells with checkpoint inhibitors, cytokines, targeted therapies or radiation simply because each has independent immunological activity.
Clinical experience suggests that this is not sufficient.
A phase I glioblastoma study combining repeated peripheral EGFRvIII CAR-T infusions with pembrolizumab demonstrated biological activity but no convincing clinical efficacy. Analysis of recurrent tumors revealed substantial evolution of the immune microenvironment, including exhausted and regulatory T-cell states and interferon-associated changes.
This is an important negative result.
PD-1 inhibition cannot compensate for every mechanism limiting CAR-T therapy.
A rational combination should address the dominant biological failure of the cellular product. If trafficking is limiting, treatment should improve trafficking. If antigen heterogeneity is driving escape, broader recognition may be required. If TGF-β dominates the local environment, resistance to that pathway may be more relevant than adding another systemic checkpoint inhibitor.
The future of combination therapy in solid tumors may therefore depend less on adding more immunotherapy and more on identifying why the engineered cells are failing in each tumor.
Solid Tumors May Require Several Different Types of Cell Therapy
The progress of the past several years suggests that there may never be one cellular platform that reproduces the success of CD19 CAR-T across all solid cancers.
Different biological problems may require different solutions.
CAR-T cells are particularly attractive when a sufficiently selective surface antigen exists. TCR-engineered cells expand the target space to intracellular proteins presented through HLA molecules. TIL therapy exploits naturally occurring tumor recognition and can potentially target multiple patient-specific antigens simultaneously. Neoantigen-selected TILs and individualized TCR therapies add another level of precision by identifying exactly which tumor mutations the transferred cells should recognize.
These approaches are converging with advances in genome editing, synthetic biology, single-cell sequencing and computational neoantigen discovery.
The future may therefore be less about finding the solid-tumor equivalent of CD19 CAR-T and more about developing a toolbox of cellular therapies matched to the biology of each tumor.
The Regulatory Landscape Is Already Changing
The transition is no longer theoretical.
Several milestones have occurred within only a few years:
- Lifileucel established TIL therapy as an approved treatment for advanced melanoma.
- Afamitresgene autoleucel established engineered TCR therapy for MAGE-A4-positive, HLA-selected synovial sarcoma.
- Satri-cel became the first approved CAR-T therapy for a solid tumor in 2026, targeting CLDN18.2-positive gastric and gastroesophageal junction cancer.
- Early CAR-T studies in glioblastoma and diffuse midline glioma have demonstrated that locoregional delivery, multi-antigen targeting and carefully selected tumor antigens can generate biological and clinical activity in diseases previously considered exceptionally difficult for cellular immunotherapy.
These therapies remain far from reproducing the broad success of CAR-T in hematologic malignancies, but they show that the barrier between blood cancers and solid tumors is becoming less absolute.
What Still Needs to Be Solved?
Several questions now define the next stage of the field:
- Can truly tumor-selective surface antigens be identified across common epithelial cancers?
- How can multi-antigen targeting prevent escape without increasing toxicity to normal tissues?
- Which patients should receive CAR-T, TCR-T or TIL therapy?
- Can neoantigen-specific T cells be identified and manufactured rapidly enough for routine treatment?
- How can engineered cells penetrate tumors and persist within suppressive microenvironments?
- When should locoregional delivery replace systemic administration?
- Can cellular therapies remodel the tumor microenvironment strongly enough to recruit endogenous immunity?
- How can T-cell fitness be preserved during collection and manufacturing?
- Can allogeneic or off-the-shelf products reduce manufacturing time without introducing unacceptable immune complications?
- Which combinations address genuine mechanisms of resistance rather than simply adding additional immune stimulation?
These questions reflect an important change in the field. The central problem is no longer whether T cells can kill solid-tumor cells. They clearly can. The challenge is building a therapeutic system that allows them to do so safely, repeatedly and across the heterogeneous biology of human solid cancers.
