Skip to main content


A Startlingly Good Leukemia Trial

You’ve probably seen the headlines about a new experimental treatment for leukemia. For once, the excitement seems justified – this is a remarkable and very promising result, and it’s worth taking a close look at it.
As reported in the New England Journal of Medicine, a patient in this study had been diagnosed with chronic lymphoid leukemia (CLL) since 1996. In this condition, B cells proliferate uncontrollably, piling up in the bone marrow and the lymph nodes. This patient had run through several courses of chemotherapy over the years. He would go for periods with no signs of disease, but it would always come back (in harder-to-treat form, naturally). By the time of this study, he was in bad shape and running out of options. Those, frankly, are the patients who are appropriate to enroll in a trial like this one – you want to treat cancer with what we know can treat it before going to something that might well not work at all (or might even make things worse).
And this particular idea had not shown as much promise in the past as everyone had hoped, despite being immunologically reasonable. The idea is to take T-cells from the patient and modify them to express a new antigen receptor, then infuse them back in and let them go to work on the tumor cells. But previous attempts to do this (against lymphoma, ovarian cancer, and neuroblastoma) hadn’t had much effect, since the modified T cells had apparently not proliferated once back in the patient. Without the cells taking off on their own, it really doesn’t seem feasible to infuse enough of them from outside to show a significant effect.
In this case, the chimeric antigen receptor (CAR) was designed to go after CD19, a surface protein found on all B-cells. That’s as solid a target as you could find for treating CLL, but without something new, trying to have engineered T cells clear them out would very likely fall short in this case as well. But this time, the T-cells were outfitted (via a lentivirus vector) not only with the anti-CD19 CAR, but with signaling domains from CD3-zeta and CD137. These are known to be involved with (respectively) coupling surface antigen recognition to intracellular processes and with T-cell proliferation in general. Animal studies had suggested that this combination could deliver a more robust response from the T-cells after being sent back.
And a robust response is what happened. Before treatment, the patient was given a drug regiment to deplete his lymphocytes (in order to give the new T cells a clear field to work in), and at that point his bone marrow was found to be widely infiltrated by cancerous B cells. He then went through three consecutive days of infusions with his own T cells, 5% of which had been modified. Nothing untoward happened during this stage. And in fact, it doesn’t appear as if much at all happened for a couple of weeks, which must have had everyone wondering.
But on day 14, the patient started experiencing chills and fever, followed by nausea and enough severe flu-like symptoms to send him into the hospital. Blood work showed no evidence of infection, but large increases in uric acid, lactate dehydrogenase, and other factors, with signs of kidney damage as well. But this was actually good news. Because at this same time, more than20% of his circulating lymphocytes turned out to be the engineered T cells, which had indeed proliferated and were vigorously going after the B cells of the leukemia. (At this point, it wouldn’t surprise me if the folks running the study were beginning to wonder what they’d turned loose). The patient’s kidneys were, in fact, having a hard time keeping up with the amount of cellular debris that they were being asked to sweep out of the blood stream; he lost over a kilo of cancerous cells.
On day 23, there was no evidence of CLL in the patient’s bone marrow. The swollen lymph glands had resolved, and a CT scan confirmed that the masses seen before treatment had disappeared. None of the cancerous B-cell types that were present before the therapy (two clones, both with mutations in p53) could be detected. Ten months later, they still can’t. As far as can be told, this case of refractory leukemia has been completely cured.
Two of the three patients treated in this fashion showed this effect – the third still shows signs of leukemia in the bone marrow, but appears to be asymptomatic. Most interestingly, it appears that the T-cell effect is persistent, and may continue as a “surveillance” mechanism in the treated patients.
Now, this is all excellent news, because this sort of therapy can be adapted to a wide variety of tumors. The main requirement is that there is some sort of surface antigen that’s specific to the tumor type, but that still leaves you with a wide field to work in. It’s important to note, though, that in one way this experiment did something quite strange: it worked much better than anyone expected. The dose of engineered T cells was much smaller than used in previous trials, and was deliberately chosen to be on the low side because no one was quite sure what to expect. Given the response, that was certainly a good move. I’ve no idea what would have happened if the therapy had been more aggressive, but it couldn’t have been good.
I hope, though, that everyone involved is enjoying this as much as possible, because this is a rare event indeed. Having things go suddenly, crazily right in a clinical trial is a once-in-a-career thing, if ever. The field of immunological cancer therapy has been given a huge boost, and now all the other groups working in the area have a huge motivation to spur them on. This is potentially some of the best oncology news in years, so let’s hope that it continues to work out.

44 comments on “A Startlingly Good Leukemia Trial”

  1. MDACC Student says:

    I have a class taught by Laurence Cooper. Really enthusiastic guy about his research. Even thought I think it will be hard to help a large number of people…I still wish them the best.

  2. John Schilling says:

    Seems to me that if you make your curative agent aggressively self-replicating, you can reasonably expect crazy-spectacular results on a general basis. Including spectacularly bad results, and from what I’ve seen this one came perilously close to “too much of a good thing” territory.
    The question, and it is as much ethical and regulatory as scientific, is how we manage that risk. There’s probably a fair bit of low-hanging fruit in this and related areas, startlingly good cures for intractable cancers and the like, if we don’t mind killing every tenth test subject while we work out the details. But people tend to frown on that.
    Anyone know a good way to include a robust kill switch on engineered T-cells?

  3. Chris Franco says:

    It’s certainly an exciting finding and I’m always thrilled to see immunological advances in oncology. While the authors mention their hope for long-term tumor surveillance with the persistence of memory T cells, they gloss over the issue of long-term immunodeficiency that results. Granted, surviving CLL is a far better outcome! However, I hope we can ultimately reach the holy grail of remission without any lasting immunological issues seen with this method or in most BMT procedures.

  4. DC says:

    Care to elaborate about the “issue of long-term immunodeficiency that results”? Does this method have adverse effects on other normal T or B lymphocytes?

  5. Chris Franco says:

    @4 These CAR T cells target CD19+ cells indiscriminately, whether they are cancerous or not. Thus, any memory cells that persist not only provide surveillance against cancerous B cells, but will respond to any non-cancerous CD19+ B cells. This adds an extra wrinkle to John Schilling’s comment (#2) asking about a kill switch in that it can be for more than just safety during the initial tumor response. On this note, it will be interesting to see the long-term follow-up on these patients.

  6. Chris Franco says:

    @4 These CAR T cells target CD19+ cells indiscriminately, whether they are cancerous or not. Thus, any memory cells that persist not only provide surveillance against cancerous B cells, but will respond to any non-cancerous CD19+ B cells. This adds an extra wrinkle to John Schilling’s comment (#2) asking about a kill switch in that it can be for more than just safety during the initial tumor response. On this note, it will be interesting to see the long-term follow-up on these patients.

  7. AI Disease Sufferer says:

    If they can stick a new antigen detection region on lymphocytes to hunt down and kill cancer cells, what about removing or shutting down an antigen detection site to stop fighting something good. It seems that the only way to “cure” my autoimmune disease is get my immune system to stop fighting my body.

  8. Anonymous says:

    The paper mentions that “In fact, in our patient, B cells were absent from the blood and bone marrow for at least 6 months after infusion.” The wording seems to suggest that after six months, normal B cells started proliferating again, although I may just be reading into that line what I want to see.

  9. biotechbaumer says:

    The result is exciting but like #6, I am concerned about potential long-term autoimmune-mediated immunosuppression. Additionally, I’d like to see future trials include a control where they inject autologous T-cells with the co-stimulatory molecules without the transgenic TCR to see whether a more non-specific immune reaction is causing the tumors to regress.

  10. Rhenium says:

    8. Anonymous
    It could also be that the results were published after 6 months and they did not yet have additional data.

  11. @8 says:

    They likely took their last data point at the 6 month mark.

  12. Anonymous says:

    I am extremely skeptical of anything called a clinical trial that involved 3 patients.
    Are you kidding me?
    What about the cherry picking of the participants?
    Perhaps they had unique genotypes.
    CLL can be classified in two broad categories -“slow/responsive to chemotherapy” and “resistant to chemotherapy”
    Most patients fall into the slow category. Response and remission to antibody/chemo can occur with very positive results using existing antibody therapies. Of course this is for the slow/responsive individuals.

  13. HFM says:

    @13: I haven’t RTFA (yet), but I would assume they picked highly chemo-resistant patients and screened for robust CD19+. You’re not going to try your crazy stuff when the normal treatment will do, nor when it’s guaranteed to fail.
    I do wonder if something like this could be improved to the point where it would be better than immunosuppressant drugs for various autoimmune diseases. (I also wonder how easy it is to get CD19- leukemia under these kinds of selection pressures, but I suppose that’s something one finds out the hard way.)

  14. Lu says:

    On a side note, is it an immune reaction which causes all the symptoms of flu, not the virus itself?

  15. Morten G says:

    What’s the name of that promoter that they use in knockout mice where the promoter is active as long as you supply a certain compound? It’s pretty clever for getting around lethal knockouts. Would be good to have the engineered elements of the T cells under that kind of control.

  16. GC says:

    @12 According to MSNBC they had only 3 patients because nobody wanted to fund the study (including the National Cancer Institute) so that’s all they had money for.

  17. gippgig says:

    Those engineered T cells wouldn’t by any chance be cancerous, would they?

  18. MO says:

    Any theories on UPN 02?
    The steady climb in ALC after treatment and “His BM showed persistent extensive infiltration of CLL 1 month after therapy despite marked PB cytoreduction” seem to imply something odd.
    Is this a possible result of the corticosteroid treatment?
    (I confess to being disappointed they didn’t spend more space on this in the paper: A partial-response patient would seem to be the interesting one 🙂

  19. PTM says:

    Awesome result, let’s hope the proliferation of the T-cells themselves won’t be the next problem they have to deal with.
    Perhaps some safety measures can be incorporated into those T-cells to try and keep them in check. Though as all it would take is for one such cell to mutate it’s way around those safety circuits, their long term reliability is hard to gouge.

  20. Anonymous says:

    And, unlike Harrison Wintergreen, the patient is still conscious.

  21. Curt F. says:

    I am extremely skeptical of anything called a clinical trial that involved 3 patients.
    Are you kidding me?

    I am trying to understand the basis of your skepticism. Do you think what happened to the patient profiled in the post has been falsely reported? If so, why? Do you think these findings will fail to generalize to other patients? If so, did the authors of the this particular study promise that it would painlessly and obviously generalize to anyone with CLL?
    I agree with Derek. This is great news! I’m thrilled for the scientists involved in this work, and even more for the patients of course!

  22. A Knowing Mess says:

    There’s no mention of the Micromet anti-CD19 BiTE (MT103), a bispecific antibody, which also recruits T cells to kill B cells in a spectacular fashion – look at the clinical results in ALL or DLBCL. Unlike this approach, the effect ends when the antibody is no longer being administered..

  23. Martin Burkle says:

    “clean natural gas” has been advertised millions of times. My wife says that “safe nuclear power” should be advertised millions of times also.
    Insert “safe nuclear power” into every debate question you can.
    Insert “safe nuclear power” into every blog response you can.
    Wouldn’t it be great to insert “safe nuclear power” into all of NEIs TV ads?
    “safe nuclear power” can save the environment if we start building about one reactor per month.
    “safe nuclear power” creates clean air.
    “safe nuclear power” is on when you need it.
    There are over 100 reactors in the United State that generate “safe nuclear power”.
    No one have ever been made sick due to United States “safe nuclear power”.
    Please insert “safe nuclear power” where ever you can.
    I am in favor of safe nuclear power.

  24. Eric Jablow says:

    Perhaps you should write Randall Munroe. His latest, at covers this story, but he might have gotten some of it wrong. Exactly which lentivirus did the researchers use to insert the genes?

  25. Terry says:

    are they reliable scientists?

  26. Terry says:

    can ppl apply patent regarding this therapy? if not, what can they do to apply the patent?

  27. sepisp says:

    Injecting new genes into cells using viruses is the standard practice; there’s nothing special about that. Lentiviruses such as HIV do this to insert their own genes into human cells to multiply. The machinery can be made to inject other genes instead. The virus is then “dead”, an empty cell. In this case, it appears that most of the modified cells had proliferated inside the patient, so there was no viral matter left when the treatment started working. The flu was caused by the toxic debris released from the killed cancer cells, not by the virus.

  28. Dan says:

    To answer question #24, the lentivirus used was VRX496.

  29. Dan says:

    To answer question #24, the lentivirus used was VRX496.

  30. RickW says:

    #28 and 29, I think this is incorrect. The vector is based on Didier Trono’s third generation HIV-based lentiviral packaging system. It has several safety features such as a self-inactivating LTR and a hybrid LTR to make it Tat independent. The rev, pol and gag genes are on separate plasmids to reduce the chance of recombination that could create a live virus during packaging.

  31. T says:

    In response to 12, actually they picked three patients who had failed multiple courses of different chemotherapies, everything had been tried and the only thing left for these patients was bone marrow transplation, which is very risky.
    15, I think your are thinking of the tet (tetracycline controlled) system, where you have to supply the antibiotic tetracycline (or doxycycline) to get expression. The reason they diodn’t do this, or build in some lind of suicide switch is probably that T-cells aredifferentiated cells not stem cells (this also partly adresses the “are they cancerous?” question), they are not immortal, they don’t have unlimited prolifereation capacity and transgenic ones are normally cleared quite quickly by the body (which is also why other similar approaches haven’t so far worked). This means that T-cell therapy is usually considered a transient therapy (that it sometimes isn’t is one of the important findings of this study). One of the protein fragments they used seems to have induced a “memory T-cell” phenotype, meaning that, under constant stimulation from all newly produced B-cells, they are hanging around longer than usual. But they proliferate only in response to stimulation by their antigen, just like normal T-cells, so, like normal T-cells, they’d have to aquire a load of mutations for immortality etc to be cancerous.

  32. T says:

    Regarding being suspicious of the low numbers, this really can’t be compared to a normal phase II clinical trail. It is probably helpful to view these cell therapies as new kinds of transplantations (which is essentially what they are), something quite different to administering drugs. For a start, the phase I trying small doses in healthy volunteers really doesn’t apply, so safety must be evaluated along with efficacy in the target patients. As with the recent facial transplants, this is just a preliminary test, and given the encouraging results (not just that it seems to work, but that it hasn’t killed anyone), they will now be able to think about doing a proper evaluation.

  33. Jonadab says:

    @ 7. AI Disease Sufferer:
    I rather doubt that the technique would be applicable to your situation without EXTENSIVE further development. The engineered T cells here are still only a fraction of the patient’s T-cell population; the majority are still normal T cells. In the trial, the normal T cells just keep doing what they were doing all along (not solving the problem), but the engineered ones do something that wasn’t previously being done, which is where the therapeutic effects come from. *Preventing* the immune system from fighting the patient’s own cells would require that most or all of the T cells stop doing the undesirable thing they’ve been doing. If you introduce a few extra T cells that don’t cause the problem, the regular ones will presumably still do what they’ve been doing, and the problem will remain.
    I’m not saying genetic therapy could never be used to fight autoimmune disorders, but that would be an additional, new, and more advanced development. The technique behind this leukemia therapy is inadequate for that purpose.
    @12. Anonymous:
    You should be skeptical. Three patients is clearly not a large enough population to provide reliable data either on overall effectiveness NOR on side-effects. Clearly they will need to run a trial with a much larger patient population.
    However, the present result will, I imagine, very likely get enough attention from potential investors (and also from potential clinical trial patients, and their families) to allow said larger trial to be run. The danger, scientifically, is that the extant result may be (apparently) _too_ good and thus may discourage proper controls and blinding, which could lead to difficulties with correctly interpreting the results. That could be unfortunate, in the long term.
    @14. Lu:
    I doubt if they’ve nailed down the precise causes of the flu-like symptoms yet (almost *anything* can cause those symptoms), but the virus used to modify the T cells seems unlikely to be the proximate cause.
    16. GC:
    > According to MSNBC they had only 3
    > patients because nobody wanted to fund
    My understanding of human behavior suggests someone will now be willing to fund a follow-up study.
    @18 MO:
    > A partial-response patient would seem
    > to be the interesting one 🙂
    Either you are a scientist, or you should be. Most people are far more interested in the (apparently) completely successful cases.
    @21 Curt F.:
    > I am trying to understand the basis
    > of your skepticism.
    It was an uncontrolled study with a population of three. Nobody who understands the meaning of the phrase “statistically significant” would call this study conclusive.
    It’s interesting, hopeful, promising; but it’s not conclusive.
    (Skepticism is not the same thing as unbelief. Skepticism is the point of view that says “We need to study this matter further before we draw any final conclusions.”)

  34. tegbert says:

    Actually, I think there is a probabilistic way of looking at this. I don’t know what the actual number are, but suppose 1 in 100 patients in other CLL modified T-cell trials went into remission. Using the old concept of drawing balls from urns, and letting the red balls be remissions, drawing 3 out of 3 red balls is unlikely, i.e., one in a million (1,000,000 = 100*100*100). That’s what makes the results interesting and different from other trials from a statistical standpoint.

  35. John Schilling says:

    #31: “But they proliferate only in response to stimulation by their antigen, just like normal T-cells…”
    As I understand it, their antigen is provided by any B-cell, including the healthy ones the patient’s bone marrow will be constantly producing. So these are going to be some rather persistently stimulated, highly prolific T-cells with plenty of opportunities to acquire new mutations.
    I’d still like that robust kill switch, please. And no, a doctor with a vial of smallpox doesn’t count 🙂

  36. Wile E. Coyote, Genius says:

    the flu-like symptoms were probably tumor lysis syndrome (notably the increased uric acid and renal compromise). Occurs when a large mass of cells are rapidly killed all at once. See this with chemotherapy treatment of some lymphomas and leukemias. This is an indicator of a very robust tumor killing over a short period of time. Can itself be life threatening due to the electrolyte imbalances (release of large amount of intracellular K+) and renal failure.

  37. Jonadab says:

    @34. tegbert:
    Your one-in-a-million figure sounds impressive, but the calculation that arrives there assumes several things we don’t know.
    Suppose, just for example, that one in three of patient in the general leukemia population, upon receiving the proposed therapy, will die of organ failure within four months. The probability that the study with only three patients would fail to expose that information would be almost 30% even if the sample population of three was adequately controlled for risk factors, which it in fact certainly could not have been with a sample size of only three, meaning for all we know the probability of the study coming out the way it did purely by luck could in fact be substantially higher than 30%.
    Organ failure, of course, was a completely arbitrary random shot-in-the-dark guess at what a possible side effect we don’t yet know about could be. It could just as well be any number of other unpleasant things. The patients in the study experienced flu-like symptoms but recovered. What if their flu-like symptoms, by luck, were especially mild, and some patients might simply die from fever or whatnot? Or, what if 30% of patients would suffer severe brain damage and be cured of the leukemia only to spend the next fifty years in a vegetative state? How’d you like to pitch those odds to the FDA?
    Yeah, call me a pessimist, but the only way to find out I’m wrong (or not) is to do a larger study.
    Suppose also that only one in five patients in the general leukemia population would show the result that this study indicates for two of the three patients. There would be about an 11% chance, if I calculated correctly, that the study would show the success rate that it did — even if the population were perfectly controlled for various factors that might impact the probability of success, which with a sample size of three is essentially impossible, meaning that the probability could in fact be significantly higher than 11%.
    Putting those together, *if* the therapy in question gives the patient a 20% chance of apparent recovery in six months (to be followed by only God knows what after six months, since the study has not not yet reported data beyond that) and a 33% chance of extreme negative outcome, the probability that the study would come out the way it did could easily be higher than 5%. (Even ignoring the issue of controls, it would be over 3%.) Okay, so that’s still unlikely, but remember that a *lot* of studies get done every year. With sample sizes this small, a few of them are bound to report results that look unrealistically good.
    A study with a sample size of three is just plain too small, period. It’s enough to get our attention and be interesting (especially with a 2/3 apparent success rate, which if it’s real would be a very good result for virtually any cancer therapy), but it is absolutely NOT conclusive.
    A much larger study MUST be done (and with a duration rather longer than six months, I think) before the treatment can even be considered for general use.
    That doesn’t even address the issue of cost (any therapy that relies on making custom modifications to the individual patient’s own cells is going to cost more than a lot of other therapies), but we can cross that bridge if it clears the other hurdles. Let’s see how the larger study turns out.
    A larger study will inevitably be done, and soon. You don’t go to the FDA with a sample size of three, but you DO go to potential investors with that and get more than adequate funding for a larger study, oh yes, you do. Even if the investors are skeptics and figure the odds of the whole thing panning out and going successfully to market are only one in a hundred, the potential payoff is much higher than a hundred to one. I mean, a cure for leukemia? They potential payoff for getting a ground-floor slice of that pie has gotta be thousands to one, at least. Venture capitalists wet their pants over opportunities like that. The larger study will be funded.

  38. tegbert says:

    @39 Jonadab
    1. My simple analysis dealt only with efficacy, within the parameters of the trial, not safety. What makes this study interesting is that the treatment appears so much more efficacious than previous modified T-cell studies, even given its limitations. I assumed there are many patients who have participated in modified T-cell studies, such that a population probability of remission could be determined: I guessed it to be 1.0%. Of course, my numbers could be wildly off since they’re just guesstimates. Given this very rough estimate, the expected number of red balls out of 3 drawn at random (i.e., patients in remission) from this population is 0. Thus, 3 red balls is wildly unexpected.
    2. Safety is a different issue, and I agree that you need trials with more than 3 patients, because safety must be determined only among patients who receive this particular treatment, not among all similarly situated patients, whether treated or not.
    3. “Suppose also that only one in five patients in the general leukemia population would show the result that this study indicates for two of the three patients.” Well, yes, if you change the numbers, the results will be different. The fact is, however, that 3 of 3 very problematic, not “general population” patients went into remission. That can’t happen very often by chance. In a more rigorous analysis, we might hypothesize that these three patients came from the general population of modified T-cell studies and then reject that hypothesis at the p=0.000001 level.
    4. Using balls from an urn for modeling probabilities assumes a lot of things, including lack of systematic bias. Obviously if some kinds of things come to light, the significance of the study could change dramatically.

  39. Ron Hurwitz says:

    In looking at this latest sudy involving the 3 leukemia patients, I wonder how long one could assume the time period would be before approval by FDA. Estimated of course!

  40. Progrest says:

    @14, it looks like flu-like symptoms are indeed primarily determined by a particular immune reaction.

  41. Paul says:

    If you could get these T cells to produce another protein not normally found in the body, I imagine they could be cleared out with a second wave of T cells that would target the first wave.
    Also: if there is a cancer risk from these cells due to the lentivirus inserting into some cancer surpressing gene, would it be possible to load the virus with copies of the most important such genes? Even if it wiped one out in some cell, there would be another copy there ready to take over.

  42. SIRK says:

    The T cells can be made to also express a receptor for a toxin that otherwise may not cause much harm by itself. Once the CLL cells are wiped out and sustained remission/cure has occurred, the toxin can be administered which will bind to its receptor on the engineered T cells and wipe them out selectively.

  43. Dhiran says:

    Can anybody help me I have T Cell lymphoblastic leukeamia?

Comments are closed.