Read the Full Video Transcript

Carlos Uribe: Hi everyone. It's my turn to add another biomarker. We're going to add now the absorbed dose. But I just want to start by talking about why dosimetry, and to motivate the group here on why we should do this. And I wanted to start by comparing chemotherapy with radiopharmaceutical therapies. We know the mechanism for fraction chemotherapy is maybe some enzymes inhibiting something. We know RPT is the radiation, although the delivery method can be the same. It might be intravenous.

Then we see external beam radiation therapy and radiopharmaceutical therapies. Both of them work through radiation, but there are differences here as well. Like, external beam radiation therapy has a high dose rate. Radiopharmaceutical therapies typically have a low dose rate. I attended the EANM last year in Vienna.

Doctor Jean Pierre Pouget had this very nice slide about radiobiology in which we see there are a lot of differences between external beam radiation therapy and radiopharmaceutical therapies.

However, we're adopting some dose limits that come from external beam radiation therapy, and the ones that I'm showing up there.

So let's keep this in mind as I talk, but what are we currently doing in theranostics? This is an example of two patients that we've treated at our institution.

The patient on the top, you see, has a very low tumor burden. The patient at the bottom has a lot more tumors, and what we see is, well, we're injecting exactly the same amount into these two patients, so that probably doesn't make too much sense.

We've seen in the literature what happens, there's some variability in absorbed doses. There's this study done in Germany, and they analyzed several patients. They looked at the absorbed doses in the red marrow, in the kidneys, in the tumors, and you see that there's a variation of orders of magnitude between patients.

We talk about the 23 Gray to the kidneys. There was this study done in Sweden, and what they found is, basically, this is the histogram of the absorbed doses to the kidney. When they go through that fixed injection, you see that most of the patients don't even get to 23 Gray.

We do know there's been some evidence that as we increase the absorbed dose to the tumors, those tumors respond better. And we've seen also these types of studies in which they aim to reach the 23 Gray to the kidneys and they compare it with patients that do not reach the 23 Gray to the kidneys and patients that reach that limit actually did better. Yeah, this is another example. This is Dosisphere. Again, a personalized approach versus a standard dosimetry approach and the personalized approach did better.

So what I'm trying to get at is we know in theranostics and in radiopharmaceutical therapies, this is not chemotherapy. I hope we have it clear that if we inject lutetium PSMA with a cold lutetium isotope, nothing is going to happen. It's the radiation that's causing the effect. However, the pharmacokinetics differ from external beam radiation therapy and we talk about personalized medicine and moving this into the future, if there's a field that can do personalized medicine, it's probably what we do here in Theranostics because theranostic not only enables us to visualize throughout the entire treatment process, but we have the means to quantitatively measure its impact on the radiation dose delivered.

So it looks like we can do better, and if we can probably implement a dosimetry-based approach to determine the personalized administered activities, like, for example, for Lutetium 177 labeled radiopharmaceuticals, we can help these patients achieve what we want them to achieve, but how can we do it?

Okay, well, this is just a summary of what we need to do for that dosimetry workflow. We need to obtain some quantitative SPECT images, and we're going to take those images and we're going to segment the tumors and organs and then, like the physicists, we're going to do some calculations to be able to convert those values into an absorbed dose value in units of Gray. So let me start with quantitative measurement. We're going to go through this workflow, kind of discussing some of the concerns that people have or that we currently have in the field as to not implement this yet routinely in the clinic, but hopefully, I can convince you that there are some solutions to it.

So the first one is, well, we need to calibrate our equipment. We need to understand our activity meter, how much exactly we're going to be injecting into the patient. But there's some concern that we cannot do it because there's not yet a standard source for the calibration of Lutetium 177. However, well, there are ways of going around this. We can actually get a sample that we know very well how much is the activity and we bring it into the dose calibrator, but that also typically comes in a vial that looks like vial A and we inject the patient with something that looks like the syringe on the right.

So we just need to keep in mind those differences in geometry, but it can be done, and the physicists, we know how to do this. We've been able to find values for the different types of geometries, and now we know that we can inject very accurately like that patient even using the different geometries.

Now you saw the two previous speakers. They are also proposing SPECT imaging because I'm talking about dosimetry. I just want to propose that when you do that imaging, you follow this type of protocol. So it's the same procedure in the clinic, but we need to be able to perform scatter correction and attenuation correction. So what I'm asking you guys to do is when you go back home and you want to set up your SPECT scanning protocol, include these three energy windows, because then we can use those scans the same that we were talking a moment ago, but now we can also do some quantification and we can do some dosimetry.

There's this concern that there's not a standard procedure to calibrate and harmonize different systems. Well, we saw an example from Australia a moment ago. We've also done that in Canada, and we found out that a linear acquisition of point-like sources can be used to determine that camera sensitivity. Well, we need to account for scatter and background contribution, so we need to remove that. And then it's very similar to what we do when we have a sphere. So it simplifies the process. But again, and it's just like a simple scan of a point source there, it's very easy to do in the clinic.

We've been comparing different cameras, and we're starting to see that maybe between manufacturers and cameras that are very similar, those sensitivities are close. Maybe there are some differences between cameras that have a different crystal thickness, but we're understanding this better and we're trying to also harmonize in North America with the Society of Nuclear Medicine. We're trying to find procedures that people can just scan scan the point source and then get a certificate to qualify their SPECT scanners to make sure that we're all harmonized, similar to what Australians have been doing.

We've been able to do this in Canada. We've calibrated, well, we're running this clinical trial that has been discussed, the PSMA compound. Well, Lutetium PSMA 617 compared to docetaxel, but we're doing retrospective dosimetry. So we actually went to calibrate the 12 institutions that are participating in this trial.

And Sara, who is here, has been helping me with all this project. We're going to be presenting all the details later this year at a different conference, but we can show that it can be done. Now, there are more concerns about doing dosimetry. Maybe it was too many visits to the hospital, maybe three scanning points. But just keep in mind that external beam radiation therapy brings their patients to the hospital every day for several weeks. Scanning procedures generate too much patient discomfort, or the scan takes way too long. That's not really feasible to do, but the field has been improving. We can always accommodate the patient and make them feel comfortable. We can use pillows, maybe supports for their legs. There are these types of new SPECT systems that now some procedures that used to take 45 minutes to scan three bed positions, maybe now we can do that closer to 10 minutes.

And if you still have a conventional camera, well, in our group, we're trying to work at using AI to generate synthetic projections. That way we can obtain fewer projections and speed up the scanning time as well without losing quantification.

I'm a physicist. So, the idea of performing SPECT scans is that we can measure the biodistribution of the radiopharmaceutical over time. So, we need to collect some points in this curve so that then we can model those points and find a fit that gives us an idea of the biodistribution. And of course, as a physicist, I would like us to think that all our patients are spherical, we're going to scan them continuously for weeks in a vacuum, but that's, of course, not reality. So I understand that there are concerns about scanning the patient four times. I mean, I want to say that scanning them six times is better than scanning them five, which is better than scanning them four, which is better than scanning them three, and so on and so on. But we've also been starting to look into scanning them for a single time point. Now then the question becomes, okay, as we understand more how the population behaves, we can create that red curve that is fluctuating there.

Then the question becomes, when is the optimal time point? If I'm only going to do one scan that gives us the closest quantification to perform dosimetry. And those are studies that we've performed in our group, and some of our conclusions are, for Lutetium PSMA 617, if you, well, considering that we want to spare the healthy organs, if we want to be the most accurate, to have the most accurate quantification, we should do a scan closer to 48 hours post-injection. Well, but we've analyzed different time points, and that plot on the left, so we understand more or less the errors that we get when we perform that scan at different time points. It's just that the least error is closer to 48 hours. So ideally, if you can only do one scan, but you can do it closer to 48 hours, that would be the best.

Now, there are concerns that when we do segmentation, how are we even going to do this in the clinic? That requires a lot of time. It's very difficult. We don't have enough people to do it. Well, we've discussed the role of AI, and there are a lot of AI tools coming now for automatic segmentation that definitely is improving this field and it saves some time. That concern of variability and lack of standardization. So each person in different institutions, maybe we're getting different ways of doing dosimetry. So the values are not comparable.

So when you have this type of problem, then well, we try to go and solve it. And one of the things that we've done is through the SNMMI, we initiated a dosimetry challenge. And what we did is we sent two data sets of patients and we asked the world to perform dosimetry, and then we collected the data and we're trying to go through each step of that dosimetry workflow trying to understand where the variability is coming from and try to make some recommendations that lead us to standardization and reduced variability.

So we've published these two papers. We've submitted now the first two here, which we're now discussing the S values, what's the last step in the calculation? We're discussing the fitting and we're looking also into making recommendations for the segmentation.

Another concern is that this is really hard to implement in the clinic because there's really not software that can allow me to do this easily, but maybe I need to develop my own code, but I don't really have time, or how am I going to validate it? But I created this slide. There are a lot of open-source and commercial tools now available. It gives you guys multiple options that make that procedure very easy.

So I think that shouldn't be a concern anymore. There's no clear pathway for reimbursement. Well, we've seen some cases we know here in the states there are some people doing it. And again, through the dosimetry task force, well, Steven Graves took the lead on writing this paper about how you can get reimbursed for these types of scans, and there are ways of doing it, so we can discuss more. So those are the concerns.

Hopefully, they're not too big of a concern now because I showed you ways to address them. But then we come to, so how is dosimetry going to help us at the end in clinical practice? 'Cause at the end, that's what we want. So I'm showing this. This is a typical dosimetry workflow. At the end, what we want is to use dosimetry to optimize the treatment and evaluate how the treatment is doing. I'm going to add one more step here. Everything that turns purple there is because at least at our institution, we are bringing our technologies to help us with that. So again, they're helping us now a lot with the segmentation. So it's like they're becoming one external beam. Radiation therapies would be called the dosimetrists. Now I'm starting to call them the nuclear dosimetrists and it's helping us a lot in our workflow.

But this is how I see that dosimetry can guide our clinical practice. It can guide us in determining the optimal injected activities for patients. And I think as I was saying, maybe it's just one extra biomarker because it has to be combined maybe with the age of the patient, the genomic predisposition. If the patient has had prior treatments, for example, there's one patient that only has one kidney, but it will help us guide that.

It will help us determine if a patient should or could go to more cycles or if we should discontinue treatment because maybe it's not working. So we need a little bit of tumor response there. Maybe in combination with AI, it can allow us already to predict the outcome and at the end, I think it can help us extend the survival and improve the prognosis of the patients. I can remember the effects of radiopharmaceutical therapies are the radiation, which is the energy that we deposit in tissue.

But of course, we still have some questions that we haven't answered yet. So, for example, what are the absorbed dose limits for radiopharmaceutical therapies? We're extrapolating from external beam. What are the absorbed doses for tumor control? Well, we've seen papers like the one that I'm showing there. They've shown that patients can get up to 40 Gray in the kidneys and they're still fine. So maybe the 23 Gray is too low. But we have a group of people and we're trying to understand, and we're trying to kind of create the gold standard for phase one clinical trial so that we can start collecting data that shows us these types of curves of if I start increasing the dose, when exactly is it that I'm starting to see tissue complications. I don't think there's much data out there. We don't really know where that curve falls. So sometimes when we talk about dosimetry, not adding too much value because I think we are just in the lower region of that S, so we need to try to start going up.

So remember, the radiation that's causing the effects. So I think we need to start finding these curves. And this was also discussed in, actually, this paper came last week, Tommy is one of the authors. They discussed these limitations of extrapolating external beam radiation constraints and the need for us to find those constraints in the type of therapies that we do.

We've also seen this chart quite a while during this meeting. And I bring up, we had a discussion earlier today, so I'm going to go a little bit quicker because right now we're doing this all the way at the end of radiopharmaceutical therapies. But we saw this study in which they're starting to say, okay, PSMA 617 was demonstrated to be safe and not inferior to docetaxel, maybe can be employed earlier in the disease. We talked also about a PSMAfore, which is even earlier.

So if we're starting to bring these therapies even earlier, maybe it's more important to actually understand what's the radiation dose that we're giving to these patients to see what are the long-term effects and to see how we are going to treat the patients if we need to retreat them later on. And lastly, just, I'm almost done. I think I have the vision that as we collect all these different biomarkers, we can start creating a digital version of a patient, and it's what we call theranostic digital twins, in which we include genomic data, patient anatomy, sensitivity to radiation, absorbed doses, and then we can probably start building a model of a patient in which we can play to design the best treatment possible before we even inject the patient.

But this is kind of the long-term view. So in conclusion, we just discussed some evidence on which patients have a better outcome when therapies are personalized using dosimetry.

We discussed some of the concerns regarding the routine implementation of dosimetry, but we also saw some of the solutions to those concerns. We explored how dosimetry can help in clinical decisions, but to answer some of the questions that we have, we need post-treatment, quantitative imaging so that we can measure the dose. So I hope we can not leave this session, not just from what I'm saying but from also what the last two speakers said, and you go back home, try to convince that we should implement post-treatment imaging. Thank you very much. Thanks to the organizers for inviting me.