Read the Full Video Transcript

Jeremie Calais: Great overview, Mike, as always. I think my talk will just give more images on top of what Mike just said. So my presentation is about the PSMA PET selection criteria for patients for PSMA radiopharmaceutical therapy. We'll review the data, a little bit of the trial, a little bit like Mike just did, just putting more some visuals, I would say.

Just to make sure we're all on the same page and make sure we understand the mechanism of action of that drug. It is radiation therapy delivery to lesions that express the targets of your radiopharmaceutical ligand. And you can see over time, here are images normalized to the radiation count. So what you actually see is really the estimation of the amount of radiation. You can see that after three, four days, pretty much the whole circulating non-used parts of the radiopharmaceutical have been excreted and only remain the radiopharmaceutical trapped either in the normal organs, but most importantly in the tumor lesions delivering continuously the radiation over the coming two, three weeks. And that's the effect of the radiation delivery, which is different than external beam radiation therapy, of course. It's a low-dose rate delivery of radiation.

So of course, the more you will have uptake and retention in these tumor lesions, the more radiation dose delivery you will have and the more likelihood you can induce DNA damage that will maybe lead or not to cell destruction depending on the radiosensitivity profile down the road. And PSMA PET can give you an estimate of this biodistribution of your drug. Imagine you had that for any drug since the beginning of cancer treatments. You could visualize where your treatment will go, or at least get an idea of it. Because, of course, what we see on the left is a gallium PSMA PET scan obtained at one hour after injection, which is very different than what's actually happening over a few weeks when you use Lutetium PSMA.

But still, it gives you that sense of where the drug will go and in which quantity. And when you see these different PSMA PET scans here, different patients, well, it's a targeted treatment. You can guess that the treatment will probably have different reactions, different treatment responses, and toxicity profiles in these four patients, all eligible for the same therapy. Imagine you could have that for any drugs. That would have added a lot of complexity already on all the treatment algorithms that we're now trying to discover. Before you try a drug, you see if it works or not. And if it works, that's a guideline. Now we're trying to understand it better, so of course, it's more complex. So you can use these images to do patient selection, probably personalized estimation of toxicity and side effects, and use it as a predictive or prognostic biomarker mostly for trials.

When we review the PSMA PET criteria that are in the predictor level derived directly from the VISION trial, they're highlighted here, and I'm going to put them here again. In the VISION trial, to have this screen failure rate of 12%, the people who were screened out or the people who were included... I'm going to come back to that. To me, it's really to exclude patients of PSMA PET and not really to include the patient. But let's see the inclusion criteria. The patient must have had at least one lesion with sufficient target expression. Sufficient meaning at least one lesion with uptake visualized above that of the liver, as assessed by visual analysis and not SUV. And at the whole body on the cross-sectional images, the absence of big PSMA negative lesions.

Big meaning measurable by RECIST CT criteria more than one centimeter. Why is that? It's because if you have a small lesion, the uptake may be just underestimated by partial volume effect and it may not reflect the level of expression of the target; it's just a spatial resolution of the PET detector. So these were the criteria, and that led to this 12% screen failure rate. At that time, the 12% when the first results of the Lutetium PSMA trials were out and shed to the light for all the global medical oncology that was maybe not so used to seeing radiopharmaceutical therapy trials or PSMA PETs embedded in that, people question, "But do we really need to do PSMA PET selection? Maybe is it useless or not?"

So we tried a little bit to answer that question by looking at a dataset already established from patients treated across the world before the VISION trial, across various sites. We identified 300 patients like that and reapplied the VISION criteria on this dataset to see if we could find patients who would have been screen failed in the VISION trial. You can see here, for example, an example of a patient with low PSMA expression. It means you don't see really high uptake lesions. You can see a liver PSMA negative lesion, lung PSMA negative lesion, and pelvic right lymph nodes PSMA negative or at least with low uptake.

We were able to find 30 of these patients, 29. So of course, it's super small, it's retrospective. But it was still interesting to see that these patients, when you look at the progression-free survival, they do not do as well as the ones who were eligible by the VISION criteria. So of course, they have a worse outcome. Does it mean we should really not treat them? I think we're doing medicine; there are no hard rules, right? You have a patient in front of you, you have multiple parameters to take into account. But at least you know a little bit better when you take your decision based on little studies like that. And if you have to do a more refined analysis, so here we're talking about eight patients versus 21, so it's super small. But there may be a signal to say that really the patients who do the worst are the ones with big PSMA negative lesions measurable by CT instead of the ones without good PSMA expression that maybe, for example, we were not able to see that because of partial volume effects and not real PSMA target expression level.

I like to support this statement to show that case in illustration. You can see on the left side the baseline scan before Lutetium PSMA, this patient had liver lesions, multiple ones. You can see on the top row, high expression of the target. And on the right side, you see a post-therapy scan after two cycles, a PSMA PET scan. You see a nice response in that lesion. There is no uptake anymore. A good decrease in size on the CT. On the other hand, in the same patient, you have other lesions that were kind of small. On the CT, you see the yellow arrows, and there was no good PSMA target signal. And after two cycles, you can see a dramatic increase of these liver lesions on the CT still without PSMA expression. So these were really lesions without good target expression that progressed under potentially selective pressure of the targeted treatment.

There was a study done by the creator of the VISION criteria, or at least who published a lot on that. And we looked at the inter-reader agreement. If you have to do widespread adoption of the VISION criteria, and there is a good agreement, it's I would say pretty easy to do. You look at the MIP, you see if there is enough uptake or not visually compared to the liver. And then you go slide by slide trying to identify PSMA negative lesions. And you can see that there are still, of course, it's visual assessment. So there are still disagreements. You cannot reach 100%, but at the end, it was very reproducible among readers.

One question that comes often is now that we have all these PSMA ligands available, how would this impact the treatment selection? And I think Dr. Fendler nicely put the frame for that yesterday by comparing various studies. I think there is a study presented here as well, but initially, there was no data. It's impossible to have head-to-head PSMA PET data in patients treated with Lutetium PSMA; it's a study not so easy to do. And at the time of the Lutetium PSMA approval, people really questioned. Of course, they have DCFPyL and there were others. So multiple groups and panels really put the statement that we should consider them equivalent, that the Gallium PSMA-11 from UCLA UCSF, from the kit beds method, from the DCFPyL, they should be considered equivalent and should not prevent getting the treatment. So that was SNMMI NCCN very rapidly.

Then you can see little studies. There is no big trial, but there are some little studies here and there trying to answer those questions. On the left side, you can see the biodistribution level of uptake. In normal organs, you can see that in the liver it's pretty similar. In the salivary gland, it's pretty similar as well. I think the bottom left image is presented right now as a poster. And on the right side, you have a UCSF study looking at patients who had both DCFPyL and PSMA-11, and you can see the outcomes in the PSA waterfall plots that are fairly similar between the two. So, at the end, between PSMA-11 and DCFPyL, there is no good argument to say that it should make a difference to use one or the other. It might be different a little bit for the PSMA-1007 and rhPSMA that have a different excretion pathway a little bit, or at least the liver uptake is higher on a normal basis.

So if you have liver-based criteria and your normal liver is higher, theoretically it can impact a little bit. And you can see nicely on the image on the left how the liver uptake is higher with PSMA-1007 than with DCFPyL. And you can also see the bar chart showing that on the bottom. Dr. Fendler yesterday presented their study. That's the image that is shown here. Where then for 1007 and rhPSMA, maybe we could use the spleen as another organ of reference. And there was a statement like that from Tom Hope very recently, saying that we should use as well rhPSMA for treatment selection. We should use as well PSMA-1007 for treatment selection. We just have to understand a little bit and just keep in mind that the threshold based on the liver, we have to adapt on a patient basis in what we see. I think that's important to know.

Again, there is no strong statement, I would say. It's just knowledge of how it works, understanding the limitations to understand better what we're doing. I think one thing we have to be sure of is that if you start to compare SUV from one time point to another to do treatment response assessment, and we're going to talk about that again later. If you go from one tracer to another, the SUV, there is so much uncertainty already from one time point to another, from one scanner to another. But then from one tracer to another, it's really complex. You can still do visual assessment. You see lesions, you see new lesions or not. But I think SUV from one tracer to another, you cannot use that for treatment response assessment.

FDG, there are a lot of FDG questions. I think in the US it's complex because there is the belief that FDG is hard to do. Although in my experience, in fact, I get FDG PET scans in many patients. Of course, when you see this mismatched lesion, the ones that have aggressive phenotypes seen on FDG with low PSMA expression, these ones are really the aggressive lesions that do not receive a lot of radiopharmaceutical and that will progress under therapy. Whether you see them at baseline or when they appear during therapy, it's never a good sign. In Australia, they do that routinely, at least at Peter Mac, and that's how they screened out their patients for the TheraP trial. And you can see that the screen failure rate, of course with FDG, you see more PSMA negative lesions and the screen failure rate is higher. It's 28%, versus remember 12% in the VISION trial.

These patients, they do worse anyway. Whatever the treatment you do, they have more aggressive disease intrinsically. There is a little study from the SN team as well. They tried to see in the dataset of 90 patients that had both FDG and PSMA PETs if we really need to do the FDG in another setting. And they found that in fact, the FDG/PSMA mismatched lesions, when you strictly look and apply robustly the VISION criteria, in fact, the addition of FDG was maybe not so much. Only 3% had additional lesions that were negative that you didn't see with the VISION criteria. So maybe on a wide scale basis, it's maybe not needed to exclude patients from the treatments. I won't say it's bad. I think there is very, very good information from it. But I wouldn't say it's needed to make sure you can treat the patients.

Again, comparing the VISION versus TheraP criteria, Mike just did a summary just before. You can see here on the left side VISION criteria; it's the liver. In TheraP, they use an SUV threshold. At least one lesion must have had an SUV max of above 20. And then there was the FDG criteria in addition to that instead of the CT for VISION. And so in that study again, they compared two cohorts selected with the TheraP and VISION criteria. And you can see that the VISION, the bar is lower. You don't select as much as in TheraP. And you can see that, for example, you have nine patients that were excluded only with the VISION criteria. And these nine patients, when you see the curve, they do really badly.

In the TheraP trial, you select more patients. And so you can see that you had, I think it's 18 patients that were screened out. But these 18 patients, they're not as bad as the nine that were selected by the VISION. And you can see the curve; they're not doing so badly. It dilutes a little bit. So I think that plays with all the thresholds, how you would recommend a treatment to a patient. And it's always a continuum; that's important to say. You cannot have binary. Maybe these patients that were screened out in TheraP, they would have been treated in VISION and they benefited. And that's why the red curve on the bottom right corner is like that.

And then maybe we can refine a little bit more the criteria because, for example, these three patients, do you think they will have the same response? But they all meet the VISION criteria. So we tried to play around a little bit with other criteria. We used the salivary gland as a reference organ instead of the liver. We asked 10 blinded readers, experts across the world, many of them are in the room, and we asked them to grade patients with high intermediate and low PSG scores based on the salivary gland uptake and the rest of the lesion comparison. So it's a visual score that you do in maybe five seconds, looking at the MIP to kind of get an estimate of the overall tumor expression of the target. You can see that it separates decently well three groups of treatment response. And this maybe we can add in as a criterion in addition to VISION.

Maybe we can say VISION positive for PET selection, and then with a high PSG score or at least no low PSG score. Because again, I think what is really important is finding red flags on your scan that would really say this treatment is really not a good idea for you. And it's more the low, the exclusion criteria that are important rather than the inclusion criteria, I think. The overall level of expression of the target reflects the disease heterogeneity. Of course, the more expression you will have, the more radiopharmaceutical delivery you will have. And the less you have, the less radiopharmaceutical delivery you will have. The whole-body tumor SUV mean gives you an estimate of this target expression level and also an estimation of the heterogeneity. And it has been shown, like Mike just showed before, that the higher the whole-body SUV mean and not a single SUV max lesion, the higher it is, the better response you get. And the lower it is, the worse response you get.

This was confirmed in VISION, like Mike just showed before. I just want to add a new figure on that; there was no cutoff. That's very important. When we put out the data out there with an SUV mean of 10, people say, "What do you do if I have an SUV of nine?" But this question makes no sense. It's a continuum of response. It's just data to understand what you're doing. But you cannot take individual decisions based on binary thresholds. This is not possible. You can see how when the SUV increases on the X axis, how, in fact, the probability of response and the impact of the treatment decrease as well or increase. But there is no cutoff where it's black and white.

And interestingly, the recent update published in the Lancet Oncology paper from Hofman, the follow-up study of the TheraP trial to look at the overall survival data. What was really interesting is that the PSMA whole-body SUV mean has this great prognostic value, as we know, but was also the case in the cabazitaxel group. So it's not only about the radiopharmaceutical delivery and the target expression; it's also an intrinsic value of the PSMA expression on the biological aggressiveness features. Look at the top right corner. You can see in the cabazitaxel group in red, this is the PSMA SUV mean of more than 10. And in blue, of less than 10. So you have the same separation without radiopharmaceutical delivery. So that's also important to keep in mind.

We did the nomogram, and I think there will be others from the VISION trial, but this one is available online. It gives you an estimate of the PSA response rates, the PSA progression-free survival rates, and the overall survival rate based on different features at baseline that you input. It's available; the link is here. Tumor SUV mean was a strong parameter, along with liver involvement and pre-chemotherapy exposure. So, at the end, what we're sure of is that the eligibility criteria for lutetium PSMA therapy are robust. If you have an absence of PSMA negative lesions of more than one centimeter, I think you don't have a red flag. You can try the treatment. It's a good option if all the other parameters are in line. You need to have good uptake in the lesion.

So, I would say in the majority of the lesions above the liver, I think it's still a decent threshold. And then we have many, let's say, signs and other features that can help you to say this one will probably do better than another, but I don't think we can use it as a binary cutoff to include patients for treatment or not. All right, that was my presentation. Now we'll have Dr. Mike Morris, Oliver Sartor, and Tom Hope that will discuss the three main trials ongoing or already completed in terms of recruitment in the pre-chemotherapy setting: PSMAfore, SPLASH, and ECLIPSE. Thank you.