In Vivo CAR-T Has Its First Clinical Playbook

The promise of in vivo CAR-T has always been economic rather than clinical. Instead of harvesting a patient's T cells, shipping them to a laboratory, modifying and expanding them for three to five weeks, and shipping them back, you infuse a vector that edits the T cells where they already are. No apheresis. No culture. No waiting list. A product that behaves like a drug rather than a procedure.

What has been missing, until now, is any real sense of whether the edited cells behave like CAR-T cells once they are made inside a body rather than inside a cleanroom. The haematology department at Wuhan Union Hospital, part of Tongji Medical College at Huazhong University of Science and Technology, has now published follow-up data from the first cohort treated this way, and alongside it a complete clinical protocol the team calls the Wuhan protocol. It is the second document, not the first, that is likely to matter historically.

Four Patients, and What Each Number Means

The cohort is small enough that every individual result changes the aggregate, which is why the aggregate should be read with care. The headline figure that circulated in Chinese coverage was the fifteen-month remission. The figure that determines whether the platform is competitive is the median progression-free survival of four months.

A median progression-free survival of four months in a group of four, one of whom is in remission at fifteen months, describes a distribution rather than a central tendency. Three patients progressed; one did not. In a cohort this size the median is essentially telling you that most patients did not get durable disease control, and it would be a mistake to dress that up. It would be an equal mistake to conclude the platform has failed, because the dose, the vector, the target and the patient selection were all being explored for the first time simultaneously.

The comparison that matters is with the first generation of ex vivo trials, which also produced a small number of dramatic remissions alongside many early progressions. What moved the field forward then was not the dramatic cases but the accumulation of a manufacturing and monitoring discipline that made the dramatic cases reproducible. That is what the Wuhan group appears to have built.

A Toxicity Category That Did Not Have a Name

In vivo CAR-T does not produce exactly the same adverse events as its laboratory counterpart, and the difference is structural. When you infuse a vector rather than cells, the vector itself is a pharmacological agent with its own immunology: capsid or envelope proteins, dose-dependent innate immune sensing, and a distribution profile that includes organs the T cells never occupy in numbers.

The Wuhan team defined and characterised what it calls a vector-related acute inflammatory reaction, distinguished it from classical cytokine release syndrome, and built a multi-factor panel to describe the early immune activation signature that precedes it. They then proposed prevention and intervention strategies. Naming a toxicity is not a bureaucratic act. Until a pattern has a name, it cannot be graded, reported consistently across centres, or prevented by protocol — and every centre invents its own response.

This is the kind of contribution that does not travel well in press coverage and travels extremely well between hospitals. For a patient, the practical consequence is narrower but real: if a centre offering in vivo CAR-T cannot tell you how it distinguishes vector-related inflammation from CRS, and what it does differently for each, the protocol is probably not mature.

The Immunogenicity Problem Nobody Was Measuring

Every in vivo platform faces a problem that ex vivo products largely avoid. The vector is a foreign object, and the immune system can respond to it. If a patient has pre-existing antibodies against the vector's envelope protein, or develops them after the first dose, the second dose may be neutralised before it reaches a T cell.

The Wuhan group established assays for anti-VSVG antibodies and anti-CAR antibodies, and used them to evaluate immunogenicity in treated patients. This is unglamorous work with a large strategic consequence: repeat dosing is central to most in vivo platforms' commercial logic, and repeat dosing is exactly what anti-vector immunity threatens. A platform that cannot measure the antibody response cannot know when redosing will fail.

Running Both Platforms in the Same Patient

The most unusual element of the series is that the team treated at least one patient with both in vivo and ex vivo CAR-T, describing the two routes as complementary rather than competing. That framing is worth taking seriously, because the economics point the same way.

Laboratory-made CAR-T delivers a defined cell dose with controlled phenotype, which is why it produces deep responses. In vivo delivery cannot yet match that control, but it can be repeated, it needs no apheresis, and it costs a fraction. A plausible division of labour therefore emerges: ex vivo for the deep remission, in vivo for consolidation, maintenance or re-treatment when disease returns and the patient cannot wait five weeks for another manufacturing run.

None of that has been demonstrated. It has merely been made clinically testable, which is the correct description of where this platform now sits.

The Protocol Is the Deliverable

The Wuhan group describes its output as a complete clinical implementation and evaluation framework covering four areas: cell kinetics monitoring, immunogenicity assessment, genomic integration site analysis, and end-to-end management of clinical complications. That list is a fair summary of everything a centre needs before it can run in vivo CAR-T responsibly, and none of it existed as a standard before.

For patients outside China considering this route, the implication is direct and slightly deflating. In vivo CAR-T is not available as a commercial therapy anywhere. It is a clinical trial category, in a small number of centres, with cohorts in the single digits and follow-up measured in months. The realistic question is not whether to travel for it, but whether a trial exists for your disease at a centre whose monitoring you can trust — and whether entering that trial forecloses other options.

What Would Have to Be True

Three things, in order of difficulty.

First, the response depth has to improve. A platform that produces transient responses in most patients is a bridging strategy, not a therapy. Getting from four months of median progression-free survival to something competitive will require better vectors, better dosing, and probably better patient selection — the same three levers that took ex vivo CAR-T from 2010 to 2017.

Second, redosing has to work despite anti-vector immunity. The assays now exist; whether the biology permits repeated effective dosing is unanswered.

Third, the long-term safety record has to hold. Polyclonal integration in four patients is a good beginning and not a substitute for a decade of follow-up, which is precisely the kind of evidence the HIV trials provided for the ex vivo route.

What has changed as of this publication is that all three questions can now be asked against real data instead of preclinical models. That is a smaller claim than the headlines made, and a more durable one.