The 2026 Solid-Tumor CAR-T Map: What the Data Actually Shows

For most of the past decade, CAR-T therapy in China meant one thing: a ¥1.2 million infusion for a blood cancer. That is still the price of the best-known product, and it is still what most patients arrive asking about.

In June 2026 that changed at the edges. China approved satri-cel, sold as 恺力美, the first CAR-T therapy anywhere in the world approved for a solid tumor — Claudin18.2-positive gastric and gastroesophageal junction adenocarcinoma, at ¥990,000 per infusion. It is a narrow indication with a modest survival benefit. It is also the first time any regulator has accepted that a CAR-T can treat a cancer that grows as a lump rather than in the blood.

Behind that single approval sits an unusually deep bench of early-phase data, most of it read out in the past twelve months across brain, lung, gastric, liver, colorectal, kidney, ovarian and prostate cancer. This article walks through it organ by organ, and states the caveats as plainly as the numbers.

The One Approved Product, and What It Delivered

Satri-cel, also known as CT041, targets Claudin18.2, a protein exposed on the surface of many gastric and pancreatic cancers but largely hidden in normal tissue. Chinese regulators approved it for advanced gastric or gastroesophageal junction adenocarcinoma that is Claudin18.2-positive, HER2-negative, and has progressed after at least two prior lines of therapy.

The registrational phase II study reported a median overall survival of roughly 9.5 months in a population that had already failed standard treatment. That number deserves to be read carefully. It is a real benefit for patients with very few options left, and it is not a cure: most responders achieved tumor shrinkage or disease control rather than complete disappearance of disease.

This distinction matters for expectation setting. When a patient hears that CAR-T now works for stomach cancer, the honest version of that sentence is: one product, one biomarker, a meaningful but limited survival gain, at a price most families pay themselves.

How to Read the Numbers That Follow

Almost everything below comes from phase I or phase I/II studies with five to forty patients, no control arm, and response assessed by the investigators who ran the trial. Four habits make those figures usable.

• Response rate is not survival. A high objective response rate in fifteen patients can sit alongside a median survival measured in months.
• Disease control is not shrinkage. Stable disease counts toward disease control and does nothing for a tumor that needs to shrink.
• Small samples produce dramatic outliers. A response rate of one hundred percent in five patients means five patients.
• Severe toxicity still happens. At least one renal-cell program reported fatal adverse events, and any estimate of benefit has to carry that weight.

With those four rules in hand, what follows is a map of where the science stands. It is not a menu of options a patient can buy.

The Map in One Table

Brain Tumors: Delivery Beat Design

Glioblastoma is the most heavily explored solid tumor in CAR-T research, and for obvious reasons. The blood-brain barrier, extreme antigen heterogeneity and a locally immunosuppressive microenvironment all stack against cell therapy. After a year of readouts, one conclusion is now widely shared: how the cells are delivered has mattered more than how cleverly they were engineered.

The clearest example is CART-EGFR-IL-13Rα2, a dual-target product given by intraventricular injection straight into the cerebrospinal fluid. The phase I study published in Nature Medicine in 2025 enrolled 18 patients with recurrent glioblastoma; 62 percent of evaluable patients showed tumor regression, and median overall survival had not been reached at a median follow-up of 8.1 months. Longer follow-up presented at ASCO 2026 put median overall survival at 12.0 months, with three patients alive beyond 18 months. Apart from one previously reported grade 1 persistent neurotoxicity, no delayed or off-target toxicity appeared, and transient neurological decline after infusion recovered quickly. A second infusion did not reproduce the fluctuation seen after the first, which suggests repeat dosing is feasible.

A Chinese team took a different route at ESMO 2025, reporting the first-in-human results of a bispecific CAR-T aimed at CD44 and CD133, two markers associated with glioma stem cells that drive recurrence and resistance. Ten patients with recurrent grade IV glioma received intraventricular or intracavitary injections. Three achieved durable partial responses, disease control reached 70 percent, and only low-grade adverse events occurred, with no cytokine release syndrome, no dose-limiting toxicity and no off-target effects.

A third design, targeting B7-H3 and IL-13Rα2 with an armored construct, was reported at ASCO 2026 in just two patients. Given as low-dose repeated infusions, it produced no dose-limiting toxicity, cytokine release syndrome no worse than grade 1 resolving within three to five days, and CAR-T cells detectable in cerebrospinal fluid for at least three weeks. One patient had more than 30 percent tumor shrinkage at three months. With two patients, this is a signal and nothing more.

The most conceptually interesting product is CARv3-TEAM-E, published in the New England Journal of Medicine in 2024. It targets EGFRvIII and simultaneously secretes a T-cell engaging antibody against wild-type EGFR, so one product covers two antigen forms and partly answers the problem of antigen loss. Three patients received a single intraventricular infusion with no grade 3 or higher events, and tumors shrank markedly within days. Those responses were short-lived, which led to repeat dosing in one case at day 37. The ASCO 2025 update added seven patients: no dose-limiting toxicity, best response of stable disease in five, and a 33 percent reduction in tumor diameter after two infusions in one. Two patients who did not receive lymphodepleting chemotherapy survived beyond 12 and 20 months.

Two off-the-shelf products point at the next stage. MT026 targets IL-13Rα2 and uses CRISPR-Cas9 to knock out the endogenous T-cell receptor and HLA class I molecules, preventing graft-versus-host disease and host rejection respectively. Its phase I study, published in Nature Communications in January 2026, treated five patients with recurrent high-grade glioma and very short expected survival by intrathecal injection, and reported an objective response rate of 80 percent, comprising one complete and three partial responses, with no grade 3 or higher adverse events. Its successor MT027, aimed at B7-H3, has been cleared by the US FDA for a phase II study in recurrent glioblastoma.

Thoracic Tumors: DLL3, CEA, and a Logic Gate

Chest cancers are hard for a structural reason: there is no target as clean as CD19. Antigen heterogeneity and an immunosuppressive microenvironment arrive together, and the past year's data are best read as several partial answers rather than one.

DLL3 has become the busiest target in small cell lung cancer. LB2102 targets DLL3 and co-expresses a TGF-beta receptor blocking protein, an attempt to disarm the immunosuppressive microenvironment rather than simply increase dose. In the phase I data reported at ASCO 2026, 20 patients with recurrent small cell lung cancer or large cell neuroendocrine carcinoma received a single infusion with no dose-limiting toxicity and low, manageable rates of cytokine release syndrome and neurotoxicity. Among 17 evaluable patients, the objective response rate was 18 percent with disease control at 76 percent, and the median duration of response was 208 days. At doses of 2 million cells per kilogram or above, the response rate rose to 27 percent and disease control to 91 percent, a clear dose-response relationship. A separate case report described one patient with extensive-stage disease who remained free of progression 12 months after infusion, with complete regression of the target lesion, a healed compression fracture and no need for pain medication.

A second DLL3 product, SNC115, treated eight patients with recurrent small cell lung cancer. There was no dose-limiting toxicity and no neurotoxicity, three patients had grade 1 to 2 cytokine release syndrome, and the objective response rate was 37.5 percent with disease control at 75 percent. Both patients in the high-dose group reached partial response.

For non-small cell lung cancer, the CEA program presented at ASCO 2025 treated 15 patients who had failed at least two prior lines. A single infusion produced no dose-limiting toxicity and no grade 4 or higher cytokine release syndrome or neurotoxicity. At a median follow-up of 5.7 months the objective response rate was 47 percent and disease control 87 percent. Patients whose tumors stained at least 30 percent for CEA and who had no brain metastases did better, and CAR-T cells peaked at a median of 10 days and persisted beyond three months.

The most elegant design in this review is A2B694, a logic-gated autologous product that pairs a mesothelin-targeting activator with an HLA-A*02 blocker. Tumor cells are killed; normal tissue that shares the target but expresses HLA-A*02 is spared. In the EVEREST-2 study, one patient with advanced metastatic non-small cell lung cancer refractory to first-line therapy reached a complete response at day 90, confirmed at day 180. That is the first reported complete response to CAR-T therapy in lung cancer. No dose-limiting toxicity or cytokine release syndrome occurred in the study.

Breast cancer remains the weakest of the thoracic signals. The HER2-directed product C406 treated eight patients intravenously, with disease control in 75 percent, six of them stable disease, and one progression-free beyond eight months. No partial responses were observed. Every patient had treatment-related adverse events, mainly hematological, two of the eight had grade 1 to 2 cytokine release syndrome, and no toxicity led to discontinuation or death. Disease control without shrinkage is real but limited.

Malignant mesothelioma produced one of the more surprising results. A non-viral CAR-T that secretes its own PD-1 nanobody and is manufactured in about 30 hours was given to seven patients with advanced disease as a single intravenous infusion. Overall response rate was 57 percent and disease control 86 percent. Both patients in the second dose level responded, one partially and one completely, and that complete response has now lasted beyond 12 months. There was no neurotoxicity, cytokine release syndrome was manageable, and one dose-limiting toxicity occurred at the second dose level.

Digestive Tumors: Claudin18.2 Leads, and the Numbers Get Loud

Gastric and gastroesophageal cancer carry the heaviest disease burden in China and now also the deepest CAR-T pipeline. This is where the most striking percentages appear, and therefore where the caveats matter most.

At ASCO 2026 the CT041-CG4006 investigators reported an extended follow-up of a cohort that received satri-cel as sequential therapy after first-line treatment. Five patients with Claudin18.2-positive advanced gastric or gastroesophageal junction cancer were enrolled, and they were a difficult group: 60 percent had diffuse-type histology, 80 percent had signet ring cells and 80 percent had peritoneal metastases. The confirmed objective response rate was 100 percent. Median progression-free survival measured from first-line treatment was 20.9 months and median overall survival 22.1 months, with no grade 3 or higher cytokine release syndrome, no neurotoxicity of any grade and no treatment-related deaths. Five patients produced those numbers. They are a hypothesis, not a benchmark.

The more substantial dataset belongs to IMC002, a Claudin18.2 product built on a single-domain VHH nanobody. In the phase I/IIa study, 16 patients with multi-line treatment failure received infusions with no dose-limiting toxicity, no grade 3 or higher cytokine release syndrome or neurotoxicity, and no treatment-related deaths. Among 15 evaluable patients the objective response rate was 66.7 percent, disease control 93.3 percent and median progression-free survival 7.0 months, with one complete response lasting beyond 48 weeks and one partial response progression-free at 60 weeks. IMC002 has since entered phase III, making it the first VHH-based CAR-T in the world to reach a phase III trial in a solid tumor.

A fourth-generation design, RD07, does not rely on dose alone. It secretes interleukin-7 and XCL1, the first to sustain CAR-T survival and proliferation, the second to recruit dendritic cells and endogenous T cells, together reversing local immunosuppression. The phase I study published in Signal Transduction and Targeted Therapy in 2026 treated 10 patients with advanced digestive tract tumors, of whom seven showed tumor regression; among three with mid-to-high Claudin18.2 expression, disease control reached 100 percent. The product is now in phase II.

LB1908 reported at ASCO 2026 enrolled 13 patients with advanced gastric or gastroesophageal adenocarcinoma, 10 of them evaluable. Nine, or 90 percent, had tumor shrinkage and three reached partial response. Low-grade cytokine release syndrome occurred in 76.9 percent, there was no neurotoxicity, and upper gastrointestinal toxicity was controlled with prophylactic medication.

Liver cancer has its own flagship in Ori-C101, a GPC3-directed armored CAR-T. The BEACON phase Ib study reported at ASCO 2026 enrolled 19 patients with GPC3-positive advanced hepatocellular carcinoma after at least two systemic lines, 18 of them evaluable. Objective response rate was 50 percent, disease control 77.8 percent and median overall survival 14.4 months; in the third dose level the response rate reached 66.7 percent with disease control at 88.9 percent, and one patient remains in complete response beyond 20 months. The safety profile demands its own sentence: cytokine release syndrome occurred in every patient, grade 3 or above in 42.1 percent, the highest rate in this review. There was no neurotoxicity and no delayed toxicity in extended follow-up.

Colorectal cancer is being approached from three angles at once. For metastatic disease, a GUCY2C-directed phase I trial enrolled 20 patients with no dose-limiting toxicity during dose escalation; overall response rate was 26.3 percent, all responses occurred at higher dose levels, the third dose level reached 40 percent with median progression-free survival of 7.0 months, and patients with high GUCY2C expression reached 50 percent with 9.0 months. The cost was gastrointestinal: 55 percent had grade 3 diarrhea. In the adjuvant setting after curative resection of colorectal liver metastases, a CEA-directed study treated 12 patients with no serious adverse events, and 57.1 percent of the high-dose group remained recurrence-free at two years. For peritoneal metastases, a rapidly manufactured interleukin-9-secreting CAR-T aimed at CEACAM5 produced a response rate of 57.1 percent, disease control of 100 percent and maximum tumor shrinkage of 85.6 percent in 14 evaluable patients. And PC13, a hypoxia-responsive CEA product published in Nature Cancer in 2026, treated 43 patients with multi-line CEA-positive solid tumors through both intraperitoneal and intravenous routes, reaching disease control of 82.4 percent intraperitoneally and a response rate of 57.1 percent in CEA-high peritoneal disease.

Pancreatic cancer appears mainly through KD-496, a bispecific aimed at Claudin18.2 and NKG2DL. The phase I study reported at ASCO 2026 enrolled seven patients with advanced gastrointestinal and pancreatic cancer. Response rate was 75 percent in gastric cancer and disease control 66.7 percent in pancreatic cancer, and at the middle dose level both reached 100 percent. No dose-limiting toxicity or serious adverse events occurred, and the dual-antigen design appeared to prevent off-tumor toxicity.

Genitourinary Tumors: Small Cohorts, Wide Target Map

The genitourinary data are the thinnest in this review, and they are also the most candid about cost.

Renal cell carcinoma is being pursued through CD70. The allogeneic product ALLO-316 treated 39 patients, 38 of them evaluable, with an objective response rate of 20 percent overall and 33 percent in those with high CD70 expression. Grade 3 or higher cytokine release syndrome occurred in 2 percent and there was no grade 3 or higher neurotoxicity, but three grade 5 events were reported. Those are the only fatal events in this review, and they belong in any discussion of benefit. A second program, UCL70802, is autologous and co-expresses a LACO-Stim costimulatory molecule; in six patients with metastatic clear cell disease there was no dose-limiting toxicity, cytokine release syndrome was grade 1 to 2 in four and grade 3 in two, recovering to grade 2 within a day. At the 2 million cells per kilogram level, two of three patients reached partial response and one had confirmed shrinkage of 68.9 percent. One low-dose patient with low CAR-T exposure nonetheless had shrinkage lasting beyond six months, which hints that the LACO-Stim molecule may be activating endogenous antitumor immunity.

Ovarian cancer data center on mesothelin. KT127 completed a phase I study in 11 patients with platinum-resistant advanced disease, reporting disease control of 80 percent, median progression-free survival of 84 days and median overall survival of 129 days. One lesion shrank by half at nine months and a nodal lesion by 70 percent over three months, with no dose-limiting toxicity, no cytokine release syndrome and no neurotoxicity. The survival figures are sobering and should be quoted alongside the response rate. A second product, UCMYM802, is prepared by mRNA electroporation and co-expresses LACO-Stim; in nine patients with advanced mesothelin-positive tumors, including ovarian, pancreatic and peritoneal mesothelioma, response rate was 28.6 percent with disease control at 57.1 percent.

Prostate cancer has no single flagship yet, but a meta-analysis presented at ASCO 2026 pooled seven studies and 69 patients with metastatic castration-resistant disease treated with PSMA, PSCA or STEAP1-directed CAR-T. Objective response rate was 39.0 percent and PSA50 response 47.8 percent. Cytokine release syndrome occurred in 45.0 percent, grade 3 or above in 9.2 percent, and neurotoxicity in 13.2 percent, grade 3 or above in 4.4 percent.

What a Patient Needs Before Asking About Any of This

The gate is not interest. It is a pathology report. Nearly every program described above is biomarker-selected, and the test required differs by target.

• Claudin18.2 requires immunohistochemistry reporting both the percentage of positive cells and staining intensity, performed on gastric or gastroesophageal tissue. HER2 status must be negative for satri-cel.
• GPC3 requires immunohistochemistry on liver tissue, and DLL3 on small cell or large cell neuroendocrine tissue.
• Mesothelin, GUCY2C, B7-H3, CD70 and HER2 are assessed by immunohistochemistry or expression profiling depending on the protocol.
• PSMA is assessed by PSMA PET imaging or immunohistochemistry, and prostate programs also require a documented PSA trajectory.

Beyond the marker, four practical questions decide eligibility: how many prior lines a patient has received, whether organ function and performance status allow leukapheresis and lymphodepletion, how long the washout period from the last therapy must be, and for international patients whether the protocol's follow-up schedule can realistically be met. Many trials require a patient to remain near the treating center for several weeks after infusion.

Why China Is Where This Is Happening

China has nine CAR-T products approved for blood cancers, priced between ¥900,000 and ¥1,290,000, which is roughly a quarter to a third of US list prices. It is also the only country with an approved solid-tumor CAR-T. And a large share of the programs in this review are Chinese-sponsored: satri-cel, IMC002, RD07, LB1908, KD-496, Ori-C101 and MT026 among them.

That combination of approval depth, trial density and price is why patients from Southeast Asia, the Middle East and Central Asia increasingly come here for evaluation rather than for a guaranteed product. Evaluation is the honest word. What a center can offer depends entirely on the biomarker, the prior lines and the open protocols at the time.

What This Review Does Not Claim

Trial data are drawn from publicly presented results at ASCO, ESMO and AACR between 2025 and 2026 and from the primary publications cited, together with a Chinese-language clinical review by Dr. Qiu Lixin of Fudan University Shanghai Cancer Center, published in October 2026.

Send Us Your Pathology Report