Dr Fonseca, as always, I love reading your posts. I am very glad that cilta-cel is available for first relapse, and there are many patients for whom I believe the benefits clearly outweigh the risks. I will offer five key respectful rebuttals to some of your points above though. All in the spirit of good debate and discussion.
1. Is the 30% progression-free survival (PFS) at 5 years a mirage or reality? Historically, these results from clinical trials have not translated well to real-world data. The MAIA trial of dara/len/dex in newly diagnosed myeloma is a stark reminder—a PFS exceeding 5 years in the clinical trial but only 2-3 years in the real world. Censoring, progression adjudication in trials, patient selection bias, focusing only on intent-to-treat analysis and many other reasons make me think that we are extremely unlikely to see 30% PFS at 5 years in the real world, and to quote this to patients is setting up wrong expectations. I do think that a certain minority (10%, 15%?) will benefit to this exceptional degree, and that in itself is a reason to celebrate for such a heavily pre-treated population.
2. I respect greatly your work on attrition. However, the fundamental argument that undercuts attrition with respect to CAR-T comes from two key observations. In newly diagnosed myeloma trials, PFS1 has often not correlated well with OS, likely because patients can go on to get good treatments in the future. If attrition = death or immense morbidity, we would at least see PFS1 correlate with OS. Talking specifically about CAR-T, the randomized trial for ide-cel (KarMMa-3) showed no overall survival benefit despite a compelling PFS benefit because they allowed for crossover, and the patients in the control arm got ide-cel later. As a result, one way of interpreting the study from an overall survival perspective is that early ide-cel = later ide-cel. We cannot make such conclusions about cilta-cel, because unfairly Janssen did not allow crossover in their study. Nevertheless, the attrition argument is severely weakened given that many patients in the control arm of KarMMa-3 went on to receive CAR-T and ultimately achieve comparable overall survival.
3. The toxicity of cilta-cel is probably worse than we think it is, in part because we are just now recognizing toxicities and paying more attention, and these toxicities may not have been noticed earlier. The real-world study that you quoted with a 2% Parkinson's rate also showed other toxicities including cranial nerve palsy in 11 patients (5%), diplopia in 4 (2%), posterior reversible encephalopathy syndrome (PRES) in 2, dysautonomia in 1 patient, and polyneuropathy in 1 patient. Furthermore, neurological toxicity may not be completely averted by giving in an earlier line—it sure appears that T-cell expansion is robust in patients that are less heavily pre-treated!, and neurological toxicities have been seen even in less heavily pre-treated cohorts of patients (https://ash.confex.com/ash/2024/webprogram/Paper201783.html)
4. Many patients I see in clinic today may choose to defer a toxic CAR-T product today, with the hope of a safer (but still effective) CAR-T product in the future. This remains to be seen and perhaps is wishful thinking, but I do think a safer CAR-T product in the future will change the paradigm and lead to broader acceptance.
5. Many patients relapse today who are sensitive to both daratumumab and lenalidomide. Long-term data from the POLLUX trial (median PFS 44.5 months) is very reassuring, and toxicity is much more predictable than cilta-cel. Such a group of patients very clearly may benefit more from a simpler and safer regimen, even if it does require much more treatment compared to a one-and-done CAR-T. On a side note, the one-and-done nature of CAR-T can be debated given ongoing care needed afterwards, such as infection prophylaxis, managing of blood counts, etc.
So overall, I am very thankful to have CAR-T available at first relapse. There are many patients who are clearly great candidates for it, and there are many patients who after a thorough discussion of risks/benefits wish to proceed. I am all for them having access to it. However, as discussed above there are many reasons to be cautious with universally recommending it at first relapse.
And lastly—thank you for sharing your thoughts and respectfully allowing me to share mine. I remain a great fan of all you do, and look up to you as an inspiration.
I appreciate the time, as well as the insightful and kind comments on this post. One of the pressing issues is that we must make decisions today, at the bedside, with the best available evidence and its corollary interpretation. As they emerge, further data will provide clarity on some of these topics, but the decision as to whether to proceed with a CART in a first relapse is upon us (now an option). The purpose of the post is to posit a combination of arguments that may sway the treating physician towards that decision. I respect that many still prefer a more cautious approach and reserve cilta-cel for a later line. However, my mind still grapples with the balance between risk and gains, with a preliminary conclusion that, despite these risks, the potential benefits outweigh them.
Point 1 – Is the 30% real? This is crucial. Is that 30% real or a mirage? We don’t know. In favor of the argument that it is lower, we have the historical perspective that many clinical trial outcomes are inferior in RWE reports for several reasons discussed below (and as you point out). The percentage is likely to be lower when cilta-cel is used in patients who have received six prior lines of therapy, as observed in the CARTITUDE-1 study. But the question at hand is, “What will that number be when we have 5 years of follow-up for CARTITUDE-4?” Again, historic data shows that in oncology, most treatments have greater benefit when used in earlier lines (and, with all known limitations, as shown by the comparison between CARTITUDE-1 versus 4). We should soon see additional data combining longer follow-up and MRD status in this cohort. There is a high likelihood that the number will be better and could approach or exceed 30%. So, for the very imperfect math of concatenating data supporting the argument of earlier use, a hypothesis of 30% is not unreasonable.
Point 2 – The role of attrition. You and others have reminded us of the limitations of clinical trial design and the importance of comparator arms in Phase 3 trials, as well as how to apply their findings to clinical practice. An obvious one is that by the time the trials were reported, better arms would have been ideal. As an obvious example for CARTITUDE-4, an arm with regimens such as CANDOR or IKEMMA would be more interesting (perhaps even a bispecific arm, as in some countries, that would now be an option). As we know, there are regulatory and practical limitations (like FDA approval of trial design) that encumber how these trials are designed and executed. But to not digress, I will not expand much on this for this post. So, back to attrition.
When we first published the large attrition paper, it was met with significant skepticism. It was not the first paper (Yong, K et al Br. J Haem 2016), but it was much larger and with good data backing the observations. Many reacted with “This does not happen to me.” Fast forward to 2025, and we have six more studies showing the same results with varying percentages. The funnel plots are all of the same shape. The good news is that attrition rates seem to be going down in more recent years (more effective regimens and better tolerated) The percentages of attrition vary depending on its definition, but the effect is real. All these studies support a very conservative rate of 15%. In an ideal world, clinical trials are designed to test various lines of therapy and sequencing approaches would provide a better response. Unfortunately, we know these trials will not happen anytime soon, and if they did, by the time they completed their results would likely be outdated.
A future post is coming regarding attrition studies and changes over time. We could consider attrition rates in our interpretative math as a “mortality rate” if such sequencing studies were ever to be done in between the various lines of therapy.
Kudos to BMS for the crossover design of KARMMA trials, but unfortunately, the effects of ide-cel seem to be less than those of cilta-cel, by all efficacy metrics, and this limits its applicability to a cilta-cel-based approach. Let’s see what numbers continue to come in with anito-cel.
Lastly, we should remain cautious about interpretations of OS without a clinical trial design that is not specifically designed and empowered to conduct that test. A separate post on this here.
Point 3 – accurate description of toxicities with increased awareness. Any toxicity derived from treatment is undesirable and only justified if the benefits outweigh the risks. But among the many possible neurologic toxicities, it is MNT that remains the core of the concern. Once fully established, it is likely irreversible and has major QOL and survival consequences. The tradeoffs for toxicities like facial nerve palsies are mitigated by their reversibility in most cases and the more limited effects on QOL, hence the focus on MNT.
Two vectors will play with regards to MNT and its prevalence.
First, as you point out, increased awareness will increase the reported number (both true positives and also some false positives). More data will be available from RWE reports from multiple groups. The same is true for some of the other more serious toxicities, like enterocolitis, infections, and secondary malignancies. We don’t seem to require long-term follow-up to determine its incidence since it seems to cluster in the first two months after treatment, mostly and rarely occurs (if ever) after the third month.
Second, many factors are being identified that may mitigate, albeit not eliminate, its occurrence such as better selection of patients (don’t do it in high tumor burden), monitoring of lymphocytes and its yet to be proven value of steroid usage, and the arrest of CART effects if incipient symptoms were to occur with systemic or intrathecal chemotherapy. MNT can occur even in carefully selected patients, but it's also clear that these factors and our decisions around them seem to point in the right direction. This becomes feasible when a CART is used for a first relapse. My current heuristic is that “the ideal CART candidate is someone who does not need bridging therapy and when relapsing disease is certain.” Much more to discuss about what that last statement means, and for yet another future post.
Point 4 – let’s wait for a better product or strategy. No argument here. One of the common questions we get in the clinic now is, “Should I wait for anito-cel?” This point is one of the key aspects of my discussion with patients, and I bring up many of the points we have in this conversation.
Point 5 – “classic” regimens can be of great benefit and safer. Indeed, this is an important question too. One of the things I have learned now from the application of deep monitoring (mass spec and clonoSEQ MRD) is that many times (most of the time) these regimens are only of temporary benefit. I have some rare patients who have very long-term control of their myeloma with regimens like POLLUX, but they are unfortunately a rarity. Most achieve a low tumor burden, but relapse is almost inevitable. Even among super responders. Also, the data for these combinations is now showing to what you allude in point 1. The RWE and clinical trials show more limited efficacy. A simple example is the APOLLO and IKARIA (anti-CD38 antibodies plus pomalidomide and dex) clinical trials, where a PFS of only one year was reported. In a world where quads are standard, followed by dara-lenalidomide maintenance, this will probably be the case. So, we can buy time and that is good since we will know more about newer products and better uses. But also, this will bring the risk of attrition and the logistics of therapy (including the so-called time toxicity). Lastly, these regimens also carry a risk of severe complications and death, albeit none as emotionally provoking as MNT – yet death is the ultimate negative outcome.
And you are right about “one and done.” We try to mitigate this now by remote laboratory monitoring, minimizing visits after the third month, and even with home administration of SQ immunoglobulin replacement. This adds up, but is far less than the burden of ongoing therapy. As you know, patient with a successful CART outcome often say “this is the best I have felt. I feel normal again.”
Again, I appreciate the comments and look forward to more data that will clarify these pressing questions.
Dr Fonseca, as always, I love reading your posts. I am very glad that cilta-cel is available for first relapse, and there are many patients for whom I believe the benefits clearly outweigh the risks. I will offer five key respectful rebuttals to some of your points above though. All in the spirit of good debate and discussion.
1. Is the 30% progression-free survival (PFS) at 5 years a mirage or reality? Historically, these results from clinical trials have not translated well to real-world data. The MAIA trial of dara/len/dex in newly diagnosed myeloma is a stark reminder—a PFS exceeding 5 years in the clinical trial but only 2-3 years in the real world. Censoring, progression adjudication in trials, patient selection bias, focusing only on intent-to-treat analysis and many other reasons make me think that we are extremely unlikely to see 30% PFS at 5 years in the real world, and to quote this to patients is setting up wrong expectations. I do think that a certain minority (10%, 15%?) will benefit to this exceptional degree, and that in itself is a reason to celebrate for such a heavily pre-treated population.
2. I respect greatly your work on attrition. However, the fundamental argument that undercuts attrition with respect to CAR-T comes from two key observations. In newly diagnosed myeloma trials, PFS1 has often not correlated well with OS, likely because patients can go on to get good treatments in the future. If attrition = death or immense morbidity, we would at least see PFS1 correlate with OS. Talking specifically about CAR-T, the randomized trial for ide-cel (KarMMa-3) showed no overall survival benefit despite a compelling PFS benefit because they allowed for crossover, and the patients in the control arm got ide-cel later. As a result, one way of interpreting the study from an overall survival perspective is that early ide-cel = later ide-cel. We cannot make such conclusions about cilta-cel, because unfairly Janssen did not allow crossover in their study. Nevertheless, the attrition argument is severely weakened given that many patients in the control arm of KarMMa-3 went on to receive CAR-T and ultimately achieve comparable overall survival.
3. The toxicity of cilta-cel is probably worse than we think it is, in part because we are just now recognizing toxicities and paying more attention, and these toxicities may not have been noticed earlier. The real-world study that you quoted with a 2% Parkinson's rate also showed other toxicities including cranial nerve palsy in 11 patients (5%), diplopia in 4 (2%), posterior reversible encephalopathy syndrome (PRES) in 2, dysautonomia in 1 patient, and polyneuropathy in 1 patient. Furthermore, neurological toxicity may not be completely averted by giving in an earlier line—it sure appears that T-cell expansion is robust in patients that are less heavily pre-treated!, and neurological toxicities have been seen even in less heavily pre-treated cohorts of patients (https://ash.confex.com/ash/2024/webprogram/Paper201783.html)
4. Many patients I see in clinic today may choose to defer a toxic CAR-T product today, with the hope of a safer (but still effective) CAR-T product in the future. This remains to be seen and perhaps is wishful thinking, but I do think a safer CAR-T product in the future will change the paradigm and lead to broader acceptance.
5. Many patients relapse today who are sensitive to both daratumumab and lenalidomide. Long-term data from the POLLUX trial (median PFS 44.5 months) is very reassuring, and toxicity is much more predictable than cilta-cel. Such a group of patients very clearly may benefit more from a simpler and safer regimen, even if it does require much more treatment compared to a one-and-done CAR-T. On a side note, the one-and-done nature of CAR-T can be debated given ongoing care needed afterwards, such as infection prophylaxis, managing of blood counts, etc.
So overall, I am very thankful to have CAR-T available at first relapse. There are many patients who are clearly great candidates for it, and there are many patients who after a thorough discussion of risks/benefits wish to proceed. I am all for them having access to it. However, as discussed above there are many reasons to be cautious with universally recommending it at first relapse.
And lastly—thank you for sharing your thoughts and respectfully allowing me to share mine. I remain a great fan of all you do, and look up to you as an inspiration.
Manni
Manni
I appreciate the time, as well as the insightful and kind comments on this post. One of the pressing issues is that we must make decisions today, at the bedside, with the best available evidence and its corollary interpretation. As they emerge, further data will provide clarity on some of these topics, but the decision as to whether to proceed with a CART in a first relapse is upon us (now an option). The purpose of the post is to posit a combination of arguments that may sway the treating physician towards that decision. I respect that many still prefer a more cautious approach and reserve cilta-cel for a later line. However, my mind still grapples with the balance between risk and gains, with a preliminary conclusion that, despite these risks, the potential benefits outweigh them.
Point 1 – Is the 30% real? This is crucial. Is that 30% real or a mirage? We don’t know. In favor of the argument that it is lower, we have the historical perspective that many clinical trial outcomes are inferior in RWE reports for several reasons discussed below (and as you point out). The percentage is likely to be lower when cilta-cel is used in patients who have received six prior lines of therapy, as observed in the CARTITUDE-1 study. But the question at hand is, “What will that number be when we have 5 years of follow-up for CARTITUDE-4?” Again, historic data shows that in oncology, most treatments have greater benefit when used in earlier lines (and, with all known limitations, as shown by the comparison between CARTITUDE-1 versus 4). We should soon see additional data combining longer follow-up and MRD status in this cohort. There is a high likelihood that the number will be better and could approach or exceed 30%. So, for the very imperfect math of concatenating data supporting the argument of earlier use, a hypothesis of 30% is not unreasonable.
Point 2 – The role of attrition. You and others have reminded us of the limitations of clinical trial design and the importance of comparator arms in Phase 3 trials, as well as how to apply their findings to clinical practice. An obvious one is that by the time the trials were reported, better arms would have been ideal. As an obvious example for CARTITUDE-4, an arm with regimens such as CANDOR or IKEMMA would be more interesting (perhaps even a bispecific arm, as in some countries, that would now be an option). As we know, there are regulatory and practical limitations (like FDA approval of trial design) that encumber how these trials are designed and executed. But to not digress, I will not expand much on this for this post. So, back to attrition.
When we first published the large attrition paper, it was met with significant skepticism. It was not the first paper (Yong, K et al Br. J Haem 2016), but it was much larger and with good data backing the observations. Many reacted with “This does not happen to me.” Fast forward to 2025, and we have six more studies showing the same results with varying percentages. The funnel plots are all of the same shape. The good news is that attrition rates seem to be going down in more recent years (more effective regimens and better tolerated) The percentages of attrition vary depending on its definition, but the effect is real. All these studies support a very conservative rate of 15%. In an ideal world, clinical trials are designed to test various lines of therapy and sequencing approaches would provide a better response. Unfortunately, we know these trials will not happen anytime soon, and if they did, by the time they completed their results would likely be outdated.
A future post is coming regarding attrition studies and changes over time. We could consider attrition rates in our interpretative math as a “mortality rate” if such sequencing studies were ever to be done in between the various lines of therapy.
Kudos to BMS for the crossover design of KARMMA trials, but unfortunately, the effects of ide-cel seem to be less than those of cilta-cel, by all efficacy metrics, and this limits its applicability to a cilta-cel-based approach. Let’s see what numbers continue to come in with anito-cel.
Lastly, we should remain cautious about interpretations of OS without a clinical trial design that is not specifically designed and empowered to conduct that test. A separate post on this here.
https://rafaelfonseca.substack.com/p/improvements-of-myeloma-overall-survival?r=92kzs
Point 3 – accurate description of toxicities with increased awareness. Any toxicity derived from treatment is undesirable and only justified if the benefits outweigh the risks. But among the many possible neurologic toxicities, it is MNT that remains the core of the concern. Once fully established, it is likely irreversible and has major QOL and survival consequences. The tradeoffs for toxicities like facial nerve palsies are mitigated by their reversibility in most cases and the more limited effects on QOL, hence the focus on MNT.
Two vectors will play with regards to MNT and its prevalence.
First, as you point out, increased awareness will increase the reported number (both true positives and also some false positives). More data will be available from RWE reports from multiple groups. The same is true for some of the other more serious toxicities, like enterocolitis, infections, and secondary malignancies. We don’t seem to require long-term follow-up to determine its incidence since it seems to cluster in the first two months after treatment, mostly and rarely occurs (if ever) after the third month.
Second, many factors are being identified that may mitigate, albeit not eliminate, its occurrence such as better selection of patients (don’t do it in high tumor burden), monitoring of lymphocytes and its yet to be proven value of steroid usage, and the arrest of CART effects if incipient symptoms were to occur with systemic or intrathecal chemotherapy. MNT can occur even in carefully selected patients, but it's also clear that these factors and our decisions around them seem to point in the right direction. This becomes feasible when a CART is used for a first relapse. My current heuristic is that “the ideal CART candidate is someone who does not need bridging therapy and when relapsing disease is certain.” Much more to discuss about what that last statement means, and for yet another future post.
Point 4 – let’s wait for a better product or strategy. No argument here. One of the common questions we get in the clinic now is, “Should I wait for anito-cel?” This point is one of the key aspects of my discussion with patients, and I bring up many of the points we have in this conversation.
Point 5 – “classic” regimens can be of great benefit and safer. Indeed, this is an important question too. One of the things I have learned now from the application of deep monitoring (mass spec and clonoSEQ MRD) is that many times (most of the time) these regimens are only of temporary benefit. I have some rare patients who have very long-term control of their myeloma with regimens like POLLUX, but they are unfortunately a rarity. Most achieve a low tumor burden, but relapse is almost inevitable. Even among super responders. Also, the data for these combinations is now showing to what you allude in point 1. The RWE and clinical trials show more limited efficacy. A simple example is the APOLLO and IKARIA (anti-CD38 antibodies plus pomalidomide and dex) clinical trials, where a PFS of only one year was reported. In a world where quads are standard, followed by dara-lenalidomide maintenance, this will probably be the case. So, we can buy time and that is good since we will know more about newer products and better uses. But also, this will bring the risk of attrition and the logistics of therapy (including the so-called time toxicity). Lastly, these regimens also carry a risk of severe complications and death, albeit none as emotionally provoking as MNT – yet death is the ultimate negative outcome.
And you are right about “one and done.” We try to mitigate this now by remote laboratory monitoring, minimizing visits after the third month, and even with home administration of SQ immunoglobulin replacement. This adds up, but is far less than the burden of ongoing therapy. As you know, patient with a successful CART outcome often say “this is the best I have felt. I feel normal again.”
Again, I appreciate the comments and look forward to more data that will clarify these pressing questions.