Lung cancer has sat at the top of China's cancer burden for years. National figures for 2024 put new cases at roughly 1.18 million and deaths at about 740,000, still the highest of any cancer. Five-year survival sits between ten and twenty per cent. Targeted drugs and immune checkpoint inhibitors helped subsets of patients with driver mutations, but most eventually run out of effective options or develop resistance. For that large, hard-to-treat majority, cell therapy is the obvious next frontier.
The list of lung-cancer targets being chased by CAR-T is already long: EGFR, MSLN, MUC1, DLL3, GPC3, CEA and B7-H3 have all been flagged as promising. Among them, MSLN, EGFR, MUC1 and CEA have moved into clinical studies. Two datasets reported in 2025 show why the field is paying attention.
A single infusion, a confirmed complete response
At the 40th annual meeting of the Society for Immunotherapy of Cancer in November 2025, researchers presented early safety and efficacy data from EVEREST-2, a phase 1/2 study of a CAR-T product called A2B694. A2B694 is a split, or logic-gated, design: it carries an activator aimed at mesothelin and a blocker aimed at HLA-A*02, so it should engage only cells that are both mesothelin-positive and present that common tissue type.
By September 11, 2025, nine patients had enrolled across several tumour types: three ovarian, two pancreatic, and one each of non-small-cell lung cancer, colorectal, gastroesophageal and mesothelioma. Doses ranged from one hundred million to four hundred million cells. The lung-cancer case is the one that drew attention.
The patient had progressive, metastatic non-small-cell lung cancer with co-mutation of KRAS G12V and STK11, a subtype that resists chemotherapy and immunotherapy and carries a poor outlook. Standard first-line treatment with carboplatin, pemetrexed and pembrolizumab had already failed. After a single infusion of A2B694, the tumour reached complete response at day 90, the result was confirmed at day 180, and a day-190 PET-CT showed no evidence of disease. A local recurrence in the central nervous system appeared at day 243; after dexamethasone and gamma-knife radiosurgery, a later PET-CT showed no disease, and the twelve-month scan showed no new lesions. Across the study, no dose-limiting or serious toxicity was observed.
Disease control in almost nine of ten
The second dataset came at the 2025 American Society of Clinical Oncology meeting, from a Chinese group running a phase 1 trial of a hypoxia-responsive CEA-targeted CAR-T in relapsed or refractory non-small-cell lung cancer. Between August 2023 and July 2024, eighteen patients were screened and fifteen were treated. Their median age was sixty, they had received a median of four prior lines of therapy, more than half had spread to three or more organs, a fifth had brain metastases, and forty per cent had tumours at least five centimetres across.
With a median follow-up of 5.7 months, seven patients achieved partial response, six stable disease and two progression. That is a disease-control rate of 87 per cent, and the lung target lesions were controlled in 93 per cent. The one-year overall survival was 71.1 per cent and the objective response rate was 47 per cent. Patients whose tumours expressed CEA at 30 per cent or higher did better, a hint that biomarker selection matters. Safety was manageable, with most reactions resolving after symptomatic care and no treatment-related deaths.
| Programme | Target and design | Population | Headline result |
|---|---|---|---|
| EVEREST-2 (A2B694) | Mesothelin activator + HLA-A*02 blocker (split) | 9 across tumour types; 1 NSCLC | Confirmed complete response, one infusion |
| CEA phase 1 | Hypoxia-responsive CEA CAR-T | 15 heavily pretreated NSCLC | Disease control 87%; ORR 47%; 1-yr OS 71.1% |
Read these numbers carefully
The complete response is one patient, in a study of nine, and a brain recurrence later appeared. Dramatic, but not yet a pattern.
The 87 per cent disease control comes from fifteen patients with short median follow-up of under six months. Stable disease counts toward control and is not the same as shrinkage.
The biomarker signal, CEA expression of 30 per cent or higher, is exactly the kind of selection marker that could decide who is offered the therapy next.
Why lung cancer is hard for CAR-T
Solid tumours in the lung pose the same obstacles CAR-T faces everywhere outside blood cancers: a hostile tumour microenvironment, physical barriers to infiltration, and targets that are also expressed on healthy tissue, which raises the risk of on-target, off-tumour toxicity. Mesothelin and CEA are both expressed at low levels on some normal tissues, so the engineering challenge is to sharpen specificity. Logic-gated designs like A2B694, and tumour-selective switches like hypoxia responsiveness, are two answers to that problem.
What it means for patients considering China
China is one of the few places where CAR-T is both approved for blood cancers and actively studied in solid tumours, and its centres are central to several of the trials referenced here. For an international patient with advanced lung cancer who has exhausted standard options, these datasets are not yet an available treatment, but they define the questions a consultation should ask: which target fits my tumour, what biomarker selects for response, and whether a trial is recruiting patients with my history.
A note on expectations.
Both studies are early-phase and small. Complete response in one patient and disease control across fifteen are signals worth following, not proof of benefit at scale. Anyone weighing a lung-cancer CAR-T trial should discuss realistic response rates, the burden of lymphodepletion and infusion, and the realistic chance of central-nervous-system or other relapse with an accredited oncology team.