Skip to main content


CAR-T Deaths Cannot Be Good

Late yesterday afternoon, Matthew Herper broke the news that Juno Therapeutics had had their CAR-T leukemia trial put on hold after three patient deaths. That’s bad news no matter how you look at it, even for their competitors.

It’s too early to say for sure if Juno will be slowed in its path to market. In an interview, Bishop and Juno chief financial officer Steven Harr said that they think they understand why the deaths occurred, that they think they can get the trial back on track quickly, and that they do not think this will affect the development of the other 8 CART programs the company is pursuing.

Bishop says that the culprit appears to be a new drug, the chemotherapy fludarabine, that Juno recently added to the trial, called ROCKET. In the CART treatment, patients are first given a chemo cocktail that kills their existing T-cells. This gives the new T-cell, genetically re-engineered to attack cancer, room to grow. Juno has previously presented work showing that adding a drug called fludarabine to the chemotherapy makes the CART cells take root faster. For this reason, JUNO decided to start adding fludarabine to patients partway through.

As that second link above (from Adam Feuerstein) says, though, later in the evening Juno said that they’re going to be delayed for at least a year. The deaths were due to cerebral edema, and if that’s due to the fludarabine treatment, well, others in this field are apparently using it as well. Herper’s article says that the three deaths seem to be out of fewer than ten patients who have seen this combination thus far, so that’s very, very bad indeed. Since we’re dealing with the immune system, though, the potential for very specific, very dramatic effects (both good and bad) is always there, and it may be that Juno’s particular CAR-T approach interacts particularly badly.

But we don’t know that yet. And until we do know something, the whole field is going to be moving cautiously. Or so you’d think – as Feuerstein notes, Kite Therapeutics took the opportunity last night to issue a press release about how they themselves had completed enrollment in their latest CAR-T trial. Classy. They need to go read up on hubris and Nemesis and consider reining it in a bit, because guess what? They don’t know what’s going to happen next, either. No one does. A little humility might be appropriate when dealing with therapies that we don’t completely understand and which can either pull patients practically out of the grave or kill them without warning.

Update: the FDA lifted the clinical hold after about three days, so onward we go . . .

45 comments on “CAR-T Deaths Cannot Be Good”

  1. mikeb says:

    Messing around with the immune system, who’d ever think that it’d be harder than we thought?

  2. medchemist314 says:

    Does it make sense that fludarabine pretreatment is at the root of the edema? No judgement, just asking.

  3. bhip says:

    One of the complications associated with fludarabine is a progressive and severe neurotoxicity (Cancer Res. 1986 Nov;46(11):5953-8, amoung others). Usually noted with high dose therapy (fludarabine >40mg/m2/day), the onset of neurological symptoms typically begins weeks after the final dose, and is often irreversible and fatal. As far as I know, the cause is unknown.

  4. ADCchem says:

    The verdict still out if it’s correlated with fludarabine. It could just be a wild chain reaction where you first get lymphodepletion, then more CAR-T expansion, then more inflammation of the brain, then more CAR-T infiltration.

    It could be the fludarabine is causing severe thrombocytopenia and brain swelling causing inflammation and then a runaway immune response.

    Ultimately it makes more sense to have a CAR-T that could be shut off with the treatment of a small molecule like rimiducid, Bellicum’s platform technology.

  5. steve says:

    One of the major side effects of CAR-T therapy is cytokine storm. Fludarabine was used as it allows for more rapid T cell expansion but this might very well have been associated with increased cytokine storm. The issue is changing a clinical protocol in the middle of a trial. This occurred with the IL-12 trials, which originally had a priming dose followed by a larger infusion. Someone just decided they didn’t need the priming dose and the trial switched to just the larger infusion resulting in patient deaths. It turned out that the priming dose down-regulated the response to the larger infusion. People shouldn’t change clinical protocols assuming they’re better unless safety of the new protocol has been demonstrated.

  6. Rene Noergaard says:

    CAR-T has been a major step forward in relation to cancer treatment, understanding the potentials, showing that a cure for cancer may be possible, but also reminding us that we have to take care when playing with our bodys immune system with such potent “weapons”, a system which is a more complex challenge than we are ready to solve yet. Some technologies in the CAR-T area are safer than others, maybe its time to take a closer look at those upcoming platforms (ie using NK targets instead). Bi specific antibodies also show promise – but still are limited to few targets which restrict efficacy and population in the target group – is it time for multispecific antibodies to come into focus? Or are we ready to take the step into multitarget approaches, in vivo approaches using the bodys own cells to activate multiple pathways of the immune system. A lot of other promising and more safe (and cheaper) platforms are already in development, ready to take immunotherapy to the next level. Its just a matter of opening our eyes and search.

    1. Gwen says:

      Rene, “Some technologies in the CAR-T area are safer than others, maybe its time to take a closer look at those upcoming platforms (ie using NK targets instead).”

      NK cells are part of the innate immune system so may be even more indiscriminate than T cells, hence NK cells may be less safe. Time will tell.

  7. luysii says:

    [ Ann. Int. Med. vol. 118 pp. 114 – 116 ’93 ] Fludarabine (Fludara) in high doses ( 40 – 125 mg/meters2/day x 5 – 7 days) in patients with acute leukemia have caused life threatening neurotoxicity occurring 3 – 8 weeks after therapy was complete, with optic neuritis, cortical blindness, altered mental status, seizures and paralysis. Neurologic deterioration was often progressive and fatal. There was demyelination and vacuolar changes of the white matter throughout the CNS.

    Neurotoxicity is unusual in lower doses (18 – 25 mg/meters2/day x 5 days). This paper shows reversible white matter demyelination (impressive pictures given) in 2 patients given the drug for mycosis fungoides.

    1. steve says:

      Interesting. Fludarabine is routinely used in stem cell transplants so its use should be pretty standardized by now. My guess is still that the problem was too rapid expansion with consequent cytokine storm but time will tell. To the other commenters, it’s very sanguine to say that the immune system is complicated and we shouldn’t play with it but that’s been true of every medical advance every made. The idea that we can’t use a modality until it’s completely understood only results in stagnation. We still don’t know even how statins work for goodness sake.

      1. Derek Lowe says:

        Agreed – we shouldn’t let the complexity and power of the immune system scare us off, but we should be ready for bad and/or unexpected things to happen when we target it.

  8. Eric says:

    Juno is clearly positioning this as a fludarabine issue and one that can be readily handled with a change in the chemotherapy pretreatment regimen. I wonder, is there a mechanism behind the fludarabine induced edema? I haven’t seen anything convincing. It leads me to believe that they are basing this solely on the observation that 3 deaths (out of 10 subjects) occurred in the fludarabine group while zero deaths (out of 10 subjects) occurred in the non-fludarabine group. That certainly looks bad for fludarabine, but from a statistical point of view – this isn’t a highly improbable outcome. We readily complain on this blog about underpowered studies and poorly designed statistical analyses leading to unsubstantiated efficacy claims. It shouldn’t be any different when looking at adverse events. It would make me very nervous to continue this trial.

    1. tangent says:

      Good point. To the extent this is empirically driven, it’s betting that a one-in-eight chance wouldn’t have come up. The subjects are at risk like 1/8 * 3/20 = 2% (yes it’s more involved than that), which they may find acceptable in their circumstances, but I hope they’re fully informed.

      This all must be awful for the patients in these trials.

      1. NJ BugHunter says:

        Using a two-way Fisher’s exact test the probability of that outcome (3/10 in one group vs. 0/10 in another) occurring due to random chance would be 21%. So yes, blaming the chemo at this point is premature (yet understandable from a marketing POV).

  9. Barry says:

    Juno keeps talking about “the new T-cell, genetically re-engineered to attack cancer” but the CAR successes reported so far to my knowledge are wiping out the whole host tissue (e.g. B-cell repertoire) rather than targeting a cancer per se.

    1. UudonRock says:

      That is true; the subjects have been leukopenic post treatment almost consistently. The modified T-cells are indiscriminant in the onset. The hope is this will “reboot” the immune system and eliminate proliferation of the B-cell tumor in the process. The CAR-T modification doesn’t have a mechanism that allows it to reproduce so it is a one-time treatment. In theory at least as there are some instances of re-treatments.

      1. Steve says:

        This is incorrect on many levels CAR-T cells certainly reproduce and too rapid expansion may be the issue here. Carl June has shown that they can persist for over a year in a patient. They are not “indiscriminate”; they attack a specific antigen. The one that has been used the most is CD19, which is on B cells. Other blood cells are not attacked so the idea that patients are leukopenic from the CAR-T treatment is incorrect. What may leave them leukopenic is the conditioning regimen, which can deplete bone marrow in an effort to leave room for CAR-T expansion. That’s why they added the fludaribine.

        1. Barry says:

          to target CD19 is to wipe out the entire B-cell repertoire (the host tissue) rather than to target the cancer as Juno would spin it. If not “indiscriminate” that’s certainly less discriminate than one would want.

          1. steve says:

            Not indiscriminate, it is specific. What you’re referring to is that CD19 is expressed on both leukemic and normal cells. One of the best selling drugs of all time, Rituxan, works in a similar manner (targeting CD20 instead of CD19) but I don’t think people would call it indiscriminate. One of the major thrusts of CAR-T research is to identify cell surface markers that can distinguish normal from malignant cells but this has been very difficult to do.

          2. Mol Biologist says:

            Barry, do not spill your broth. Immune system is a reservoir and can not be in a debit and let patients die. So, I accept “cytokine storm” as a joke or unprofessional statement.

  10. cynical1 says:

    I’m no expert in the CAR-T approach and there’s not much specifics here but I wonder about this whole thing. The ablative treatment protocol they are using sounds like a perfect setup for JCV reactivation leading to PML which then leads to cerebral edema and ultimately death in immunocompromised patients. And it’s well documented that even standard dose of fludarabine can lead to PML. So this may have nothing to with CAR-T and everything to do with the chemotherapy protocol but three separate cases would be a high rate of PML occurrence. So adding on CAR-T protocol may accentuate reactivation of the virus for some reason. I wonder if they did a spinal tap of these patients and looked for JC virus……….just a thought.

    1. Barry says:

      why would the JCV have such prevalent among this leukemia clinical population?

      1. luysii says:

        Proc. Natl. Acad. Sci. vol. 94 pp. 14542 – 14546 ’97 — Initial infection with JCV occurs early in childhood and eventually reaches a seroprevalence of between 70 -80% in the adult population. At any given time, 5% of the population is actively excreting JCV in the urine.

  11. UudonRock says:

    I have been working on one of the CAR-T trials for over a year now. I won’t mention which one, but it was not the problem child. We are all closely watching and waiting as this trial is also using flubarabine, albeit in lower dosage. There have been anomalous neurologic blips, but nothing as of yet that would indicate concern. One hopes this continues to be the trend, but I’m inclined to believe the FDA may step up their audit time tables on the other CAR-T studies once they have concluded with Juno.

  12. tangent says:

    They’re dropping fludarabine from this specific CAR-T but continuing it for others? I hope to hell they have enough toxicology-in-hindsight to say why it’s the intersection of the two that’s the risk. Sounds suspiciously like hubris though…

    “Juno says it will continue to use fludarabine in trials of its other CART therapies, because the drug seems to make its other treatments, JCAR014 and JCAR017, more effective without this level of toxicity.”

  13. Gegenwind says:

    Dear Juno, house of CAR-Ts, continue to make sure that there’s always a small molecule that can be blamed. Sincerely, Gegenwind.

  14. Barry says:

    re: Steve
    Of course we all want to identify markers that distinguish malignant cells from normal. Because TCRs recognize fragments from the cytosol displayed on MHC1 on the cell surface, the distinctive proteins don’t even have to be surface proteins in their nature. There is therefore reason to hope. But to call a CAR-T that wipes out a whole tissue (rather than the malignancy on that tissue) “specific” is just to deny how far we still have to go. There are a few tissues we could afford to delete entirely and call it a bargain to trade e.g. a lethal pancreatic cancer for manageable Type-1 diabetes. But the goal must be to discriminate cancer cells from normal.

    1. steve says:

      True enough but that’s true of almost all cancer drugs. As I said, rituxan also wipes out the B cell population. Most traditional chemotherapeutics target proliferating cells and therefore have bone marrow and gut toxicity. The commentators on this thread, which often have a cynical bent for some reason, make it sound like Juno is being deceptive about CD19 CAR-T cells because they aren’t highly specific for transformed cells. But the simple fact of the matter is that very few (if any) cancer drugs are. If we had clearcut targets that distinguished the normal from the malignant state then cancer wouldn’t be such a hard problem. Given a choice, most people would rather lose their B cells than die of lymphoma or leukemia.

  15. Gegenwind says:

    Dear Steve, also cynical the NCI paper revealing CD19 expression on CNS neurons? Now this doesn’t make CAR-Ts a no-brainer, does it? Sincerely, Gegenwind.

  16. steve says:

    I never said that CAR-T was a no brainer; just that some of the comments in this thread were. Sometimes people on this blog feel a need to express an opinion (often cynical) even when they are ignorant of the field. The article you cite is interesting and has been commented on before by the CAR-T community. However, if the problem was simply that CD19 is expressed on brain cells then you would think brain edema would be a lot more common – i.e., not confined to the small subset from Juno but also in patients treated by Kite, Bellicum, Celectis and academic institutions like Penn, MD Anderson, Sloan Kettering, Fred Hutch, NIH, etc. It might be that fludarabine somehow allows better access to the CNS and that’s why there were more side effects with that conditioning regimen but time will tell. Certainly there could be toxicities other than cytokine release syndrome that affect subsets of patients (perhaps those that for some reason have high level CD19 expression on neural cells). FDA is monitoring toxicities across this class of immunotherapy so this will become more evident over time. Regardless, one always has to weight the potential for toxicity against the risk of death when considering any cancer treatment. However, as an immunologist not directly involved with CAR-T therapies, if I was dying of ALL (or even pancreatic now that the Penn group is developing a CAR-T against mesothelin) is looking at I sure as hell would give CAR-T a try

  17. Gegenwind says:

    Dear Steve, dying of ALL, I might consider Blincyto first or a clinical trial with newer bispecific antibodies: Dying of pancreatic cancer, I would look very closely for safety and any evidence of activity of the mesothelin or any other CAR-T that may become available. Sincerely, Gegenwind.

  18. steve says:

    Silly. Blincyto also causes cytokine release syndrome. It causes neutropenia that can lead to life-threatening illnesses as well. As I said before, there is no panacea when it comes to cancer therapy, whether it’s traditional or immune-based.

  19. steve says:

    BTW, just FYI, Blincyto treatment put about 40% of adult ALL patients into complete remission (CR), while JCAR015 (Juno’s CAR-T) has a CR rate of 87%.

  20. Mol. Biologist says:

    Silly? When patients died? In case of a relapse you have to pay the debit first.
    “Grasp all, lose all”.

  21. Gegenwind says:

    Dear Steve, silly to chose an FDA-approved drug over an experimental treatment the same agency just put on hold? Silly to only look at complete responses and ignore the fatalities? Silly to be concerned about this acute toxicity just being the tip of the iceberg of a mountain of chronic toxicity? Silly to have more trust in a drug that can be controlled over a treatment that gets eternally reactivated by reconstituting B cells? Look, as you know very well, the CAR-T treatment is championed and run by transplanters with a high-risk-high gain attitude that, bless their hearts, has cured leukemia and lymphoma patients. Don’t get me wrong, these are devoted pioneers of cancer immunotherapy. But transplants traditionally also cause a substantial portion of treatment related deaths and cures often come at the price of horrible chronic graft versus host disease. What do we know yet about the price of cures following CAR-T treatment? Is a cure with high morbidity better than managing the cancer? Sincerely, Gegenwind.

  22. Gwen says:

    Could patients be treated with Etanercept (aka Enbrel, an anti-Tumor necrosis factor “antibody”) to help reduce inflammation and potential for edema?

  23. steve says:

    The “silly” referred to the wording of the post; I’m not dying of ALL. Sure, taking Blincyto would be an option for patients who have the disease but I would still go with a CAR-T if I could. You are looking at Juno’s product in isolation and an SAE that is attributed to adding fludaribine to the conditioning regimen. There are extensive data on CAR-T from a number of groups worldwide. Carl June at Penn showed that 27 of the 30 ALL patients he treated achieved a complete remission by a month after receiving CAR-T against CD19, and 78 percent of the patients were alive six months after treatment. These results are much better than Blincyto. So yes, I probably would look at CAR-T over Blincyto even though it’s still experimental.

  24. David Young, MD says:

    As an oncologist, I know that Fludarabine is a drug where the total dose must be divided over 3 to 5 days. A somewhat out-of-date treatment for chronic lymphoid leukemia uses Fludarabine at 25 mg per meter squared daily for five days, repeated every 4 weeks. A still used regimen for low grade lymphoma gives Fludarabine 25 mg per meter squared daily for three days. If one tries to give 100 mg per meter squared as a single infusion on one day, the patient gets into trouble….. .with CNS side effects. If the folks at Juno where giving all the Fludarabine on one day the one can easily imagine CNS side effects. If they are giving it all on one day, they did not take the time to read the package insert. But perhaps I am thinking too much ahead here… maybe they did divide the dose over several days and the CNS side effects happened anyhow.

    1. steve says:

      It’s a bit curious because others, like the Penn group, also use fludarabine in their conditioning regiment. As I mentioned, it’s pretty standard for stem cell transplant conditioning and many have adapted the regimen for CAR-T. As you say, the devil is in the details. According to Steve Grupp at CHOP in Philly, ““We have not seen significant cerebral edema in any of our pediatric ALL patients on the CHOP/Penn study treated with CTL019, many of whom received fludarabine. CNS-related side effects have been seen, including confusion, aphasia and seizures in a small number of patients, but these effects have all resolved.”

      See for more details.

  25. Gegenwind says:

    Makes sense for the FDA to lift the hold quickly with expanded CAR-Ts ready to be infused and patients already conditioned. Make no mistake, this is not an endorsement.

  26. Barbara says:

    What do we know about this study with Car-T for AML?

    Genetically Modified T-cell Immunotherapy in Treating Patients with Relapsed or Refractory Acute Myeloid Leukemia|eabf0ca4&ni=10&pn=1

Comments are closed.