Imaging Response to PSMA Radiopharmaceutical Therapy "Presentation" - Jeremie Calais
February 12, 2024
At the 2024 UCSF-UCLA PSMA Conference, Jeremie Calais highlights the advanced diagnostic capabilities of PSMA PET/SPECT in monitoring responses to PSMA-targeted radiopharmaceutical therapy, showcasing its superiority over traditional CT and bone scans through patient cases. Dr. Calais discusses the importance of nuanced evaluation methods, including PPP and RECIP criteria, alongside PSA levels and PSMA volume changes, to accurately assess treatment efficacy and explores the potential of flexible dosing schedules in optimizing outcomes, referencing a phase two trial at UCLA.
Biographies:
Jeremie Calais, MD, PhD, Associate Professor, Department of Molecular and Medical Pharmacology, Director, Theranostics Program, Director, Clinical Research Program, Ahmanson Translational Theranostics Division, University of California, Los Angeles (UCLA), Los Angeles, CA
Biographies:
Jeremie Calais, MD, PhD, Associate Professor, Department of Molecular and Medical Pharmacology, Director, Theranostics Program, Director, Clinical Research Program, Ahmanson Translational Theranostics Division, University of California, Los Angeles (UCLA), Los Angeles, CA
Read the Full Video Transcript
Jeremie Calais: All right, now we're going to switch to the topic of imaging response or imaging progression, depending on how we want to see it, under PSMA targeted radiopharmaceutical therapy. So I want to start with conventional imaging and how we can also image with another imaging agent and PSMA targeted PET or SPECT, the response to lutetium PSMA therapy. In all the trials currently, we're still using CT and bone scan to monitor response or progression to lutetium PSMA therapy, and sometimes it works pretty well. I like to show that example; it's a recent patient we had. You can see the baseline bone scan, corresponding PSMA PET scan on the left, two cycles, four cycles, and you see the disappearance of this bone lesion. It's pretty nice. It's not as often; we don't see that that often, but you see it works as well. Another case here where you see a great response, the bone lesions are disappearing. So I was looking for these cases with the bone response visualized by bone scan before, and now I got it this year.
These cases are a bit rare, though, often, and the two plus two rule with the bone flare. Often you have this non-specific bone uptake, that remains very, very long under therapy, so it's very hard to interpret at the end. It's possible, but I think it's hard. In comparison, when you see the accuracy, all the characteristics of PSMA PET imaging, diagnostic efficacy, spatial resolution in comparison to bone scan, look how you can monitor the effects of the therapy in this patient. Here, we have the case of a massive metastatic patient that had a great response on the lutetium PSMA therapy, and you can see how you can visualize that response with PET imaging. So when you have all the parameters that go in the same direction, PSA going down, patient feeling better, no lesions visualized anymore. It's pretty straightforward. The patient seems to benefit from the therapy, and we should continue. On the inverse, when all the parameters go in the other direction the same way, PSA increasing, PSMA volume increasing, new lesions, patient feeling bad, blood count decreasing, the patient is probably not benefiting from the therapy, and it's also straightforward.
But often it's not like that. Often you have some parameters that go in one way and some others that go in another way. So new lesions, tumor volume, PSA response. How do we sort this out? I think these remain unanswered questions and should be still individualized patient-based decisions. Let's look at some criteria. The first criteria that were put out there was a proposal from Dr. Fanti and Ken Herrmann. I think it's the PPP criteria, mimicking the two plus two rule in the bone scan, looking essentially at the appearance of new lesions. You can see that patients in blue are the ones without new lesions. Patients in red are the ones with new lesions, and they do much worse. And you can see the separation of the curve seems to be better in this little data set than with the PCWG criteria. It should take these 29 patients that had no PSA progression and then you reapply the PPP criteria; it seems that the PPP criteria were able to re-separate nicely different response profiles even in this.
And then, as Dr. Eiber introduced yesterday, we tried to go one step beyond just the new lesion. When you see this MIP, you see this overall burden of disease evaluating over time, increasing, decreasing. So you really see that the volume of disease matters. And here, for example, we tried to go further in a data set of 124 patients, all with a PET scan done after two cycles of lutetium PSMA therapy. And you can see that in the data set in blue, patients with no new lesions, they do better than patients with new lesions. And you can see the volume. So a 30% volume decrease in green. They're doing better than the patients with an increase of the volume with a threshold of 20%. But what if you combine the two? This graph was shown already yesterday. It's a waterfall plot. Usually, we show PSA changes. Here, we look at PSMA volume changes. All the patients going down, the volume is decreasing. You can see that most of them are blue. There is no new lesion. All the patients going up, volume is increasing. Pretty much all of them have new lesions.
There is, although some red in the decreasing patient, it means that some patients with decreasing volume still see new lesions appearing. So what do you do with that? Is it a flare effect? Is it some resistant clone? What we try to do is to combine these two parameters at least to see how it would behave over time. And the RECIP criteria are: either patients have a decrease in volume and no new lesion, a good response, or patients have an increase in volume and new lesions, that's a clear progression. And then you have all the combinations of both in between, which we can say is stable disease by RECIP criteria. And you can see that these were able to individualize two groups of response profiles pretty nicely with a decent separation of the curves. When you compare these criteria to others, the classical one, PCWG, PERCIST, PPP, or RECIST, this was the most reproducible one. Why is that? I think it's mostly because it was relying on the whole body tumor. So here, it was based on segmentation done always the same. So there was no interpretation of the whole body screening.
So it requires segmentation, which is not optimal for implementation in a routine clinic. So the team wanted to do something that is implementable and going from quantitative to visual RECIP criteria, basically estimating this volume visually or at least the volume changes. Do you have more than a 30% change of volume? Do you have a 20% change of volume visually? This was assessed by five blinded readers and you can see that it was correlating quite well with the gold standard quantitative metrics. Dr. Gafita put together a website. You can see here on the RECIP criteria, I think you have training cases, you can go over that. And at the end, what it does is it really helps you evaluate visually the total amount of burden of disease. You assess if it's really increasing or decreasing, you look at the appearance of new lesions or not, and then you get your RECIP score without the segmentation. This was obtained for interim PET, interim meaning two cycles or 12 weeks, that's classical time points. But we tried to look what would be the value of RECIP criteria at the end of treatment.
Meaning patients get treated, you do a PET scan, no treatment anymore, and you see how this correlates with progression or response. And so this goes a little bit with the response topic that Mike Morris brought up yesterday. Maybe having an end-of-treatment PET like that would give you a baseline of where you are after treatment and then maybe monitor the patient responding to see when we need maybe to trigger something because it seems that these criteria, the RECIP criteria, do correlate with survival. So these little studies and more are needed to really put PSMA PETs in the clinical trial design, maybe to monitor response and define progression in the design of the trials. People have tried to combine the volume and the intensity of the uptake, SUVmean. So with FDG-PET, we call that TLG. Here, we're going to call it TLP. And it seems it correlates pretty well. Of course, I think the main driving parameter is the volume, but when they combine it with the target expression, it seems that the new lesions were not as important anymore than when you just look at the volume.
I think this needs further confirmation. Again, I want to put it out there. If you do volume-based quantitative estimation, it'll be subject to variation between different tracers. So you cannot do whole body tumor estimation based on SUV if you use different tracers. That's important. Here, I show different types of response. You can see with the PSA, each row is a treatment cycle, and then you have the PSA curve. You can see that you have patients directly progressing, patients who have a slowdown in the progression of the disease but really not going down. Patients who initially respond and then progress. Some have a little flare effect. It's pretty rare, but it happens. And some have a dramatic response right away, the one you really want to see and durable over time. So this really reflects many parameters, the underlying biology of the disease, and others. The early PSA response because the PSA still remains one of the key parameters to assess response or at least to have an estimation of what's going on a little bit.
So what can you do, for example, with the PSA at six weeks? Six weeks, it seems that if it goes down, it's a very good sign. But what if it doesn't go down? In that study, they looked at a dataset of 100 patients. The flare was very rare. If the PSA was going up, there was only one or two cases that after two or three cycles finally went down. If it's stable, meaning in the range of 30 to 20%, there are some chances that it'll go down. What we know is that if it goes down rapidly, that's a very good sign. What we don't really know is what if it increases, until when we should continue or not. The PSA was very well correlated, or at least the depth of response in the major trial, VISION therapy. You can see here the PSA decline correlation with the overall survival. The top curve is the one that has more than a 90% decline. And of course, the bottom curves are the ones that did not experience a PSA decline, or at least not in the 50% range.
What about PSMA RECIP criteria and PSA? Here you have a waterfall plot. This time what we see is the PSA changes. It's not the PSA volume changes. And you have in this color code, I'm sorry for the colorblind people, green, orange, and purple, the various types of response. What you see is that in the patient without the PSA response criteria defined by PCWG, meaning 50% or more decline, RECIP was able to catch some patients. Here, 10 out of 100 that still have a good response to the therapy. At the inverse, the patients who did not progress by PCWG criteria, the ones who did not experience a 25% decrease compared to baseline or nadir. You can see that RECIP was able to identify a subset of patients that really are not benefiting from treatment. So I think there is really some complementarity; there are always some patients that you would miss with only one criterion and not the other. This is my transition to the next presenters after me.
SPECT. You can get this volume estimation not only with PET; you can get it with lutetium PSMA SPECT imaging. You are imaging the same target, PSMA. There is just a different type of radiation coming out of it; it's gamma from lutetium instead of the positron from gallium and F-18. But at the end, you get more or less the same information. The SPECT CT scanners are improving, the spatial resolution is improving, the time of acquisition is improving. So I think this is something we can really think about, as Luis Ahmed suggested. After each cycle, in a time window between 4 to, I would say, 24 hours, you can image the patients. And here, you get, without a PSMA PET scan, the same information on the volume of the disease on SPECT, and then here show different profiles of response. You can see the PSA, you can see the SPECT volume, and all these parameters can be merged together to make our decision on how to monitor response to therapy. I think Luis Ahmed is going to talk about that.
I just want to say that there was a group of patients who responded well based on the SPECT tumor volume and the PSA. This one, for example, could benefit maybe from a treatment holiday period. You may not need to re-expose them in this fixed six-weeks or eight-weeks time interval. And by pushing that, you can extend maybe the benefit of the drug or the time period during which you can use the drug. So this was possible because you monitor both with PSA and imaging. You monitor the response and you try to understand what's happening during treatment. Based on that, at UCLA, we will start, I think this month and next month, our new phase two trial. Based on that, the randomization is fixed. That's the control group. Fixed dose schedule mimicking VISION, six cycles every six weeks, 7.4 for everyone versus a flexible dose schedule up to 12 where a patient can get treatment only based on their monitored response.
Dr. Adrien Holzgreve is presenting a poster. The study protocol is presented here. You can go see it. So I hope this will help us to answer how to better use the monitoring of response for a better outcome. And this was my last slide. And transitioning to Dr. Louise Emmett, I think she has a lot of things to say in that space as well. Thank you.
Jeremie Calais: All right, now we're going to switch to the topic of imaging response or imaging progression, depending on how we want to see it, under PSMA targeted radiopharmaceutical therapy. So I want to start with conventional imaging and how we can also image with another imaging agent and PSMA targeted PET or SPECT, the response to lutetium PSMA therapy. In all the trials currently, we're still using CT and bone scan to monitor response or progression to lutetium PSMA therapy, and sometimes it works pretty well. I like to show that example; it's a recent patient we had. You can see the baseline bone scan, corresponding PSMA PET scan on the left, two cycles, four cycles, and you see the disappearance of this bone lesion. It's pretty nice. It's not as often; we don't see that that often, but you see it works as well. Another case here where you see a great response, the bone lesions are disappearing. So I was looking for these cases with the bone response visualized by bone scan before, and now I got it this year.
These cases are a bit rare, though, often, and the two plus two rule with the bone flare. Often you have this non-specific bone uptake, that remains very, very long under therapy, so it's very hard to interpret at the end. It's possible, but I think it's hard. In comparison, when you see the accuracy, all the characteristics of PSMA PET imaging, diagnostic efficacy, spatial resolution in comparison to bone scan, look how you can monitor the effects of the therapy in this patient. Here, we have the case of a massive metastatic patient that had a great response on the lutetium PSMA therapy, and you can see how you can visualize that response with PET imaging. So when you have all the parameters that go in the same direction, PSA going down, patient feeling better, no lesions visualized anymore. It's pretty straightforward. The patient seems to benefit from the therapy, and we should continue. On the inverse, when all the parameters go in the other direction the same way, PSA increasing, PSMA volume increasing, new lesions, patient feeling bad, blood count decreasing, the patient is probably not benefiting from the therapy, and it's also straightforward.
But often it's not like that. Often you have some parameters that go in one way and some others that go in another way. So new lesions, tumor volume, PSA response. How do we sort this out? I think these remain unanswered questions and should be still individualized patient-based decisions. Let's look at some criteria. The first criteria that were put out there was a proposal from Dr. Fanti and Ken Herrmann. I think it's the PPP criteria, mimicking the two plus two rule in the bone scan, looking essentially at the appearance of new lesions. You can see that patients in blue are the ones without new lesions. Patients in red are the ones with new lesions, and they do much worse. And you can see the separation of the curve seems to be better in this little data set than with the PCWG criteria. It should take these 29 patients that had no PSA progression and then you reapply the PPP criteria; it seems that the PPP criteria were able to re-separate nicely different response profiles even in this.
And then, as Dr. Eiber introduced yesterday, we tried to go one step beyond just the new lesion. When you see this MIP, you see this overall burden of disease evaluating over time, increasing, decreasing. So you really see that the volume of disease matters. And here, for example, we tried to go further in a data set of 124 patients, all with a PET scan done after two cycles of lutetium PSMA therapy. And you can see that in the data set in blue, patients with no new lesions, they do better than patients with new lesions. And you can see the volume. So a 30% volume decrease in green. They're doing better than the patients with an increase of the volume with a threshold of 20%. But what if you combine the two? This graph was shown already yesterday. It's a waterfall plot. Usually, we show PSA changes. Here, we look at PSMA volume changes. All the patients going down, the volume is decreasing. You can see that most of them are blue. There is no new lesion. All the patients going up, volume is increasing. Pretty much all of them have new lesions.
There is, although some red in the decreasing patient, it means that some patients with decreasing volume still see new lesions appearing. So what do you do with that? Is it a flare effect? Is it some resistant clone? What we try to do is to combine these two parameters at least to see how it would behave over time. And the RECIP criteria are: either patients have a decrease in volume and no new lesion, a good response, or patients have an increase in volume and new lesions, that's a clear progression. And then you have all the combinations of both in between, which we can say is stable disease by RECIP criteria. And you can see that these were able to individualize two groups of response profiles pretty nicely with a decent separation of the curves. When you compare these criteria to others, the classical one, PCWG, PERCIST, PPP, or RECIST, this was the most reproducible one. Why is that? I think it's mostly because it was relying on the whole body tumor. So here, it was based on segmentation done always the same. So there was no interpretation of the whole body screening.
So it requires segmentation, which is not optimal for implementation in a routine clinic. So the team wanted to do something that is implementable and going from quantitative to visual RECIP criteria, basically estimating this volume visually or at least the volume changes. Do you have more than a 30% change of volume? Do you have a 20% change of volume visually? This was assessed by five blinded readers and you can see that it was correlating quite well with the gold standard quantitative metrics. Dr. Gafita put together a website. You can see here on the RECIP criteria, I think you have training cases, you can go over that. And at the end, what it does is it really helps you evaluate visually the total amount of burden of disease. You assess if it's really increasing or decreasing, you look at the appearance of new lesions or not, and then you get your RECIP score without the segmentation. This was obtained for interim PET, interim meaning two cycles or 12 weeks, that's classical time points. But we tried to look what would be the value of RECIP criteria at the end of treatment.
Meaning patients get treated, you do a PET scan, no treatment anymore, and you see how this correlates with progression or response. And so this goes a little bit with the response topic that Mike Morris brought up yesterday. Maybe having an end-of-treatment PET like that would give you a baseline of where you are after treatment and then maybe monitor the patient responding to see when we need maybe to trigger something because it seems that these criteria, the RECIP criteria, do correlate with survival. So these little studies and more are needed to really put PSMA PETs in the clinical trial design, maybe to monitor response and define progression in the design of the trials. People have tried to combine the volume and the intensity of the uptake, SUVmean. So with FDG-PET, we call that TLG. Here, we're going to call it TLP. And it seems it correlates pretty well. Of course, I think the main driving parameter is the volume, but when they combine it with the target expression, it seems that the new lesions were not as important anymore than when you just look at the volume.
I think this needs further confirmation. Again, I want to put it out there. If you do volume-based quantitative estimation, it'll be subject to variation between different tracers. So you cannot do whole body tumor estimation based on SUV if you use different tracers. That's important. Here, I show different types of response. You can see with the PSA, each row is a treatment cycle, and then you have the PSA curve. You can see that you have patients directly progressing, patients who have a slowdown in the progression of the disease but really not going down. Patients who initially respond and then progress. Some have a little flare effect. It's pretty rare, but it happens. And some have a dramatic response right away, the one you really want to see and durable over time. So this really reflects many parameters, the underlying biology of the disease, and others. The early PSA response because the PSA still remains one of the key parameters to assess response or at least to have an estimation of what's going on a little bit.
So what can you do, for example, with the PSA at six weeks? Six weeks, it seems that if it goes down, it's a very good sign. But what if it doesn't go down? In that study, they looked at a dataset of 100 patients. The flare was very rare. If the PSA was going up, there was only one or two cases that after two or three cycles finally went down. If it's stable, meaning in the range of 30 to 20%, there are some chances that it'll go down. What we know is that if it goes down rapidly, that's a very good sign. What we don't really know is what if it increases, until when we should continue or not. The PSA was very well correlated, or at least the depth of response in the major trial, VISION therapy. You can see here the PSA decline correlation with the overall survival. The top curve is the one that has more than a 90% decline. And of course, the bottom curves are the ones that did not experience a PSA decline, or at least not in the 50% range.
What about PSMA RECIP criteria and PSA? Here you have a waterfall plot. This time what we see is the PSA changes. It's not the PSA volume changes. And you have in this color code, I'm sorry for the colorblind people, green, orange, and purple, the various types of response. What you see is that in the patient without the PSA response criteria defined by PCWG, meaning 50% or more decline, RECIP was able to catch some patients. Here, 10 out of 100 that still have a good response to the therapy. At the inverse, the patients who did not progress by PCWG criteria, the ones who did not experience a 25% decrease compared to baseline or nadir. You can see that RECIP was able to identify a subset of patients that really are not benefiting from treatment. So I think there is really some complementarity; there are always some patients that you would miss with only one criterion and not the other. This is my transition to the next presenters after me.
SPECT. You can get this volume estimation not only with PET; you can get it with lutetium PSMA SPECT imaging. You are imaging the same target, PSMA. There is just a different type of radiation coming out of it; it's gamma from lutetium instead of the positron from gallium and F-18. But at the end, you get more or less the same information. The SPECT CT scanners are improving, the spatial resolution is improving, the time of acquisition is improving. So I think this is something we can really think about, as Luis Ahmed suggested. After each cycle, in a time window between 4 to, I would say, 24 hours, you can image the patients. And here, you get, without a PSMA PET scan, the same information on the volume of the disease on SPECT, and then here show different profiles of response. You can see the PSA, you can see the SPECT volume, and all these parameters can be merged together to make our decision on how to monitor response to therapy. I think Luis Ahmed is going to talk about that.
I just want to say that there was a group of patients who responded well based on the SPECT tumor volume and the PSA. This one, for example, could benefit maybe from a treatment holiday period. You may not need to re-expose them in this fixed six-weeks or eight-weeks time interval. And by pushing that, you can extend maybe the benefit of the drug or the time period during which you can use the drug. So this was possible because you monitor both with PSA and imaging. You monitor the response and you try to understand what's happening during treatment. Based on that, at UCLA, we will start, I think this month and next month, our new phase two trial. Based on that, the randomization is fixed. That's the control group. Fixed dose schedule mimicking VISION, six cycles every six weeks, 7.4 for everyone versus a flexible dose schedule up to 12 where a patient can get treatment only based on their monitored response.
Dr. Adrien Holzgreve is presenting a poster. The study protocol is presented here. You can go see it. So I hope this will help us to answer how to better use the monitoring of response for a better outcome. And this was my last slide. And transitioning to Dr. Louise Emmett, I think she has a lot of things to say in that space as well. Thank you.