Yeah, thank you so much, Axel, and UH great to be here this morning. As uh, as Axel said, my name is Gino Blumenthal, uh uh vice president of Global Clinical Development at merk SO. I lead teams that focus on development of novel a d c's targeting Trope two and HER three and small molecules k s G twelve C and previously I was at the UH at the f d A for over a decade. Unfortunately, because of circumstances, we couldn't have uh someone from the government uh on
our panel. Uh unfortunately, but UH I'll try to fill in as best as I know, how you know, with my uh you know, prior prior experience and in current interactions with with Project Optimists.
We also have a great.
Panel in person and we have a hybrid format as well because Andrew Faris Unfortunately there were problems with air traffic airlines at Ronald Reagan Airport, so she's but we were able to quickly pivot and have her.
Join us virtually.
So why don't we start with our panelists and just if you could introduce yourself and maybe give kind of a teaser, what are your impressions of Project Optimists from the last couple of years since its implementation.
Let's start with Greg.
I Hello, every ready, thanks you down. So my name is Greg Goldmacher. I am a diagnostic radiologist by training. I've been at MERK for about ten years, almost ten years, and I lead the Clinical Imaging and Pathology function.
So this is the group that oversees the U use.
Of imaging and pathology assessments UH to evaluate end points in human trials, so from phase one all the way all the way through, and in addition to you know, so in addition to kind of the routine things that we do with imaging for response rates and progression free survival, you know, the kind of things that are typically done in late phase development, we're doing a lot of active exploratory work internally for using more advanced imaging and analysis
methods to try to get closer to the biology of
the tumors. And well, so our team functions across therapeutic areas, but about eighty five percent of the work we do is in oncology, of course, and so we were doing a lot of work to try to get closer to the biology using various advanced analytic methods, which we'll discuss more more here and you know, as a teaser, and I guess the you know, the big idea I think is that you know, this traditional method of you know, you basically dial up the dose until the patient kills over,
and then you back up a couple of steps and use that the maximum tolered does that doesn't really not only does that not necessarily that does that not balance benefit and risk optimally, but that may not even optimize the benefit side of it. And so there's we're going through a lot of efforts to try to get closer to true analyzes of benefit, with the challenge of course being that in early stage development you're working with such small numbers and that that's what we have to do better.
So I'm Ryan Sullivan. I'm a medicaloncologist at mass General Hospital. I specialize in skin cancers, predominantly melanoma. And then I'm also a faculty member and our Tremier Center for Targeted Therapies, which isn't just MO likely targeted therapies but also includes immuna therapies. And I've been a Phase one investigator for
about fifteen plus years in my life. So Project optimists fun fact, I kind of like it, and the reason I like it is for many of the reasons that Greg just said is that it allows us to be more thoughtful about choosing doses. But it has also requires us to probably spend a little bit more money and time as we're designing the trial to end up carrying it out.
So thank you for the invitation to be here. So my name is Wim Voss. I'm the CEO Radiomics dot bio, a Belgian company that does advanced image analysis on radiology images. I'm an aerospace engineer by training, so I have nothing with the biology of cancer. But for the past twenty years I've been working on extracting more information from radiology images to enhance decision making clinical trials, first in the respiratory space and now for the past years in the
oncology space. And I think what we specialize on is really finding those early signals of what's going on very detailed, not missing an individual tumor using all the information that is there to up sample small data sets. And I think very much like Ryan, I like Project Optimism. It's a step in the right direction. I think it really opens the door for personalized combination therapies, which will probably
be the path way to cure for many patients. I think the problem today is that we face this transition period. We're actually we're treating patients even in the early clinical work to the best way possible for the patient as well as these patients are not volunteers, but typically it is their therapy the clinical trial, which makes it hard to go with lower dosages because you might not be optimal for the individual patient. But I think it is
the way to go. I think we need to really understand what's the optimal dose for a given compound to get optimal tumor killing capabilities. Once we know that and we know in which tumors it is doing it, we can also understand the biology of those tumors, true combinations of different biomarkers in the end to find the right
combination therapies. And I think in the coming five to ten years we will probably see project optimists evolve, but we might also see evolutions in the later stages of development, because how do you get a drug to market that might have an optimal efficacy in tumor killing characteristics but might not have optimal characteristics on a patient survival point
of view? And I think that that's going to be a big challenge for the follow up of Project Optimists into the later stages of drug development.
Thank you, and Andrea, Hi, good morning.
Sorry I couldn't be there in person, but I'm glad I could be here virtually. I'm Andrea Ferris. I'm the president and CEO of Longevity Foundation. We are an organization that is focusing on transforming lung cancer. So with respect to Optimists, I think a lot of the speakers have touched on it in lung cancer specifically because there are so many treatment options available now, still not enough, but there are many more than there were a number of
years ago. Overtreatment and toxicity is are a real issue, and cumulative toxicity is are a real issue. And the last speak Girl also touched on it just a little bit, but I think under treatment is also a big concern. So conceptually, I like the idea of finding.
The right dose.
My question is really more the methodology of doing it and whether randomized trials are really the best approach to finding it and how much additional cost, how many additional patients, how much additional time is that going to introduce into this new paradigm. So conceptually, I love the fact that we're looking at not overtreating, especially as I mentioned, many people are going on multiple therapies, so the cumulative toxicity
is a real issue. But it's just I question the how and whether this really is the right approach and what's going to happen when we get into combination therapies.
Great, thanks for the introductions and for the teasers. I think this will be a very robust discussion. So we have a bunch of different stakeholders true to the IO three sixty concepts, who are all you know with the vision of you know, developing you know, better, safer, more effective treatments for patients. We did hear that, you know, cancer therapeutics has evolved so much over the past few decades. We were not developing conventional chemotherapy anymore.
We don't have to take drugs to the maximally tolerated dose.
We're developing immunotherapies that harness the immune system, the t cells, et cetera to attack the cancer. Targeted therapies which you know, switch off a certain gene or protein that the tumor may be addicted to, and we're treating patients. For many, many years, a lot of cancers have been turned into chronic diseases, so we have to think about long term cumulative toxicities. We're trying to move agents into earlier treatment settings,
more curative settings. So I think these are things that the entire community is is struggling with.
I'm wondering.
You know, we heard a little bit from each of you about sort of efficiency and in drug development. You know, it's intensely competitive. Patients can't wait. You know, we need to get good enough therapies, not necessarily perfect therapies, out to the clinics as quickly as possible. So how do you guys think about sort of trade offs in trying to develop you know, early data sets that can be acceptable to regulators, to clinicians versus you know, versus speed to get things into phase three?
Maybe right, yeah, maybe I'll start since I do clinical trial, so I think it really as the I n D submissions are going in and you're actually even before and you're designing your strategy, I think keeping in mind project
optimists is not even ideal, it's a necessity. And you have to begin to think about what's the strategy for dose escalation and dose finding, and then how can we begin to expand patients cohorts at various doses throughout that trial that aren't needlessly treating patients at a dose that's thought to be ineffective, but exposing patients to doses that are thought to have potentially some benefit while we're still
escalating upwards. And I think the strategy that's been used in a lot of phase one clinical trials is the backfill, so as you're escalating upwards and you're clearing safety doses, that there's some metric of success. And I think we'll hear about some of those potential novel metrics of success.
And once there's either a clear clinical response or a reduction in a biomarker that's thought to be associated with that disease, then I think you can begin to say that level is potentially therapeutic and you can expand enrollment at that while you continue to escalate upwards.
On your dose.
I think it's hard not to build in dose limiting toxicities and some statement about we will identify if it's possible to maximum tolerated dose, but ultimately we're going to move forward to an optimal dose that's been informed either by these expanded cohorts at dose levels that seem to be potentially effective, and then also based on the pharmacokinetics, the pharmacode dynamics, and really trying to choose a dose that hits the sweet spot and that maybe you know
some I've seen some drugs that have a very linear PK, but the PD sort of begins.
To level off.
Well, that's that's already telling you something that you probably shouldn't keep pushing the dose if your pharmacodynamic effect is plateauing. And so I think that's one strategy, and then certainly other strategies would be let's find three or four doses that we think are possible to move forward with and then you start randomizing in your dose expansion or face two.
But that's just an initial thought.
Yeah, and we talked about PD.
So pharmaco dynamic markers, Greg and whim, what are your thoughts on, you know, how do you assess what are the pharmacodynamic markets from an imaging standpoint you would look at to help help select the dose, and then what's the future behold start off again.
So where we see it is really focusing in on the concept of using each tumor as its own data point. Of course, with the right statistics behind it. What we typically see is that when a tumor shows some kind of response to therapy, it tends to continued with that response. I would say the majority of tumors either grow or shrink. If you look at individual tumors, there's very little that actually change their course. Of course, if you look at a patient level and you look true resist or total
tumor burden, you see these hockey stick effects. One of the questions I have there for myself and for the space is if a tumor is decaying at a certain dose, does it stop decaying or can we, even with a low dose, get that tumor to zero, because it really changes the whole paradigm of how we deal with things. Another concept for pharmacodynamics, when when we talk about trial design, might be to use many more patients as their own
control using the concept of tumor growth or decay. And I think Greg will allude to that later, where you actually use the temporal change and not just a volume change, but the temporal tendency, which makes it possible to if you look at similar time intervals, even when a tumor is at a different initial volume, to compare the change rate over time, which I would say gives more power to the clinical trials because the patient is its own control.
On the statistical side, it would really uplift that. Plus it would also prevent, I think to the point that was made before, it would prevent the undertreatment of patients because you can step them up to higher dosages.
All right, So to follow up on this, thanks when the distinction that I'm thinking about here is that looking at pharmacodynamics and looking at efficacy markers.
So they're related, but they're not quite the same thing.
So pharmacodynamics, you know, seeing whether the treatment is having some biological effect, and looking at some combination of So in the case of radiomics, of course you know this idea that that biological changes are reflected in histology, and that histology is reflected in pixel level changes that may be actually not visible to the human eye.
So this is something that.
You know, you need AI models to extract from images, but that you can see the biological effects of treatment, and so you can establish some kind of a dose response curve by looking at you know, just biological change. The next level, though, the really challenging bit is translating that into long term efficacy and you know, as everybody who's done oncology DIRUGT development for a length of time knows, what happens is that early on you're looking at response, right,
radiographic response, What does that mean? That means tumor shrinkage. But it's not hard to find things that shrink tumors. And unfortunately the history of cancer drug development is littered with examples of drugs that looked good based on shrinking tumors in small numbers of patients early on, but then you take it to phase three and it doesn't improve survival and that's, you know, not a great outcome for a variety of reasons.
So what we need are.
Things that really correlate well to survival ultimately. And if you can get to the point where you have dose expand you know, so as Ryan said, you know, you.
Pick a few doses, two or three doses that you.
Think might be the best dose, the right dose, and if you could truly compare those, you know, thirty forty patient dose expansion cohorts on os impact, that would sort of be the the you know, to really establish an
optimal dose. And there are approaches around that based on tumor kinetic modeling where you incorporate a time component much more robustly than just looking at you know, time to progression or things like that, where you know, you decompose growth into an exponential decay in an exponential growth component, and you you know, based on either total tumor burden or per lesion back out of the of the curve fit the growth kinetic parameter that that you know, in
a bunch of academic work has been shown to be.
Robustly correlated with survival.
And that's seems like a very promising method for once you've got your two or three dope potential doses, getting to what should you really take forward? Because you know what what you know really gives you the best os impact, so those so that distinction so you can have the kind of the the more radiomics you know, and and by the way, it doesn't have to be just radiomex it can be other changes.
That look at that get to biology you.
Know in aggregate like ct DNA or other you know, blood biomarkers or imaging markers on one hand, on the on the pharmacodynamics and then the getting as close as you can to the biology that makes tumors lethal which is growth. Right, what kills you is the growing tumor, So that combination could let us optimally estimate benefit.
Great.
Yeah, thanks, So Andrea, a question for you, you know, you deal?
Can I comment on the last one first of the.
Yeah, yeah, please go ahead.
I'm just wondering, like listening to the conversations, you know, I think that it's I like the last speaker in terms of the modeling aspect of it. But I worry with the first speaker, especially is you know, don't we can't let perfect be the enemy of the good, And like how many patients do we want to put on all these various doses and taking forward and so forth?
And is it really necessary that the other thing? I just would question or caution and I would love to hear the perspective of, especially as we are moving into earlier stage disease and earlier on is LS really the endpoint we want to be looking at there, especially for these types of dose escalation studies?
Yeah, so I think to clarify.
That, you know, I think Greg and Wim are trying to develop imaging endpoints that can help predict.
Overall survival perfect.
Yeah, not necessarily obviously, you know, you wouldn't have the statistical power, or the randomization, or the numbers to follow, or the time to follow patients on phase one dose escalation out to survival.
Thank you to answer the the the how many patients do you need?
I mean, just to refer to.
There's one interesting publication from I think it was twenty twenty the Maitland paper, where this was an effort of the foundation of the NH analyzing data from a colorectal cancer trial. I think I think it was a I think it was a TKI and so interestingly so in this trial, the hazard ratio on OS was zero point eight two, so, you know, not a huge impact, right, you know, a.
Modest survival effect.
And what they did is they did kind of a resampling method where they would draw you know, n patients whatever, five hundred or one thousand times with replacement and see on the basis of that sample, if they used these various kind of kinetic parameter assessments, how how easily could they detect the OS impact. And what they found is that they had if they used tumor diameters, they were able to detect the OS impact with eighty percent power with like forty five patients, and if they used whole
tumor volume it was something like twenty five patients. So we're talking about dose expansion cohorts of you know, thirty or forty patients. If this is done appropriately, should be able to differences. I mean, again, that's a modest different os difference that was predictable with a relatively small sample. So, Andrew, to answer your question of how many and how long, those are roughly the numbers I think that we're talking about here.
Yeah, there's one thing I want to add to that, because it's in the end for the patient's survival is the most important thing. However, if we're developing a drug that would kill in all patients, like let's say half of the tumors very well, very rapid, but it might not have a survival benefit because the other tumors will continue growing and we kill that drug, we might throw away a valuable asset that, in the reality of combination therapies could add to the therapies.
And I think there is a true.
Danger there that if we don't have the right approach, that we're killing a lot of assets, or we're putting a lot of assets on the bench that could really add a benefit for the patients, and I think Project
Altmost is a good step in that direction. That we're in this phase where we have to put the concepts together because on one hand, the FDA wants and the patient need additional survival, but on the other hand, we also need to focus in and zoom in onto what are the assets that have the potential to add to combinations that could really dramatically up the level of survival.
I think, as as Excel showing earlier today, the next generation that will bring the survival up another twenty to thirty percent.
Yeah, great point.
So so back to you, Andrew and bringing us back to the here and now. You know, we've had some very future oriented thoughts which are really promising and exciting.
But you deal a lot with the FDA, you deal a lot with.
Investigators, and so what are you hearing about, you know, from a drug development standpoint, what are some of the current pain points around Project Optimist is it's sort of the inconsistency that you're of advice and kind of randomization of small cohorts for randomization's sake, just to check a box, or lack of clarity. I mean, what are you hearing and what are you trying to convey to the regulators.
Yes to all of the above, especially from the investigators, many of them that I've spoken with, you know, they they keep coming back to that medicine is more of an art than a science, and obviously we need to have the science to approve drugs and to demonstrate efficacy and so forth. But almost all of them have said, you know, the irrespective of what the label says, they're going to be playing around with it anyway. And with to do a randomized trial with such small number of people,
what are we really learning and gaining from it? Is sort of the perspective that I've heard from, and they're mostly academic researchers, you know, at the at the NCI aser centers that have been talking about this and complaining about it. From the drug developer side again, you know, I think, I don't know that anybody is quibbling with
the concept of it, but it is. It has been difficult to figure out what to do and how to do it, just because it's still evolving and trying to get to those right surrogate points and what can be done and what can't be done, and just the concern about adding a lot of additional time and cost into the whole paradigm that we're talking about with it, So at least that's what we're hearing.
Yep, great, thank you so so Ryan. I'm wondering what are your thoughts. We've heard a little bit about combination strategies. If you're combining two drugs together, do you need to re backfill and re randomize each combination? And then also, what are your thoughts if you've established it a mon therapy does say in one tumor type, do you need to repeat the exercise in a different tumor type?
Or how do you think about that?
It's a really good and complicated question with no easy answer.
I think it really does depend.
I think if you're giving a monoclonal antibody like pemberleism app, the dose optimization happened in a remarkably large phase one trial. You know what the toxicity is, you know what the benefit is, you know what the benefit is at large numbers across a lot of different treatments for a lot
of different diseases and indications. So I'm not sure you need to optimize the dose of pemberleism AP plus drug X. I think it becomes very difficult when you're using two agents that aren't monoclonal antibodies that are perfectly well characterized over fifteen years or whenever it started, And so that's really challenging in particularly trying to dose escalate two small
molecule inhibitors. Now that's usually done in the setting of trying to target signaling pathways, but you know, I meane cells have signaling pathways too, and more and more of our therapies can be small molecules in combination with other things, and I think if you're trying to escalate two small molecule inhibitors, or certainly anything that can augment the toxicity of the other agent, you probably have to be really careful in your escalation. You can escalate one with a
low dose of another and escalate the other. You can try these tandem escalation. It looks like a box with lots of different weird things happening, but ultimately it's complicated, and those often are really they don't.
End up being all that successful.
And so in many ways, I think when we're going to be combining immunotherapy agents, one probably has to be easy and stable and fixed and then you can kind of escalate the other one up or you've already done escalation, and then you just when you start escalating your combo with PD one, which is happening all the time, then you're just sort of starting at a level or two below where you think you know you're going to be, and then you just kind of escalate the other agent.
That's what's been done.
And I think on some level you may still need to do the opt I mean, you have to do the optimist piece, but I guess the question is I don't think you need to. You're going into it with a much narrower idea of where you're going to end up, whereas if you have two drugs you have no idea how they're going to mix together, you're probably screwed, and it's going to cost a lot of money and a
lot of patients, and you're probably going to hurt. Like it's just going to be hard, because those trials are remarkably difficult to do and and very infrequently have they led to approve combinations.
So presumed overlapping toxicity versus not different approaches.
For sure, and and more importantly, some.
We may be augmenting toxicities that we have no idea that that would happen. I've seen PD one inhibitors augment TKA toxics, TKI toxicities. You're like, well, how the heck does that happen?
Right?
Right?
Like, it's probably not t cell driven. Why does anti PD one x cause drug wise whatever neuropathy to get worse or anything, you know, And it's it's just so I think I think we have to go in with our eyes wide open, because there are surprises from a
toxicity signal that we sometimes see. But by and large, I think it's fair to you know, I think we do have to have an idea of where we're going with the two drugs, what the doses are individually before we start getting too creative, because it's it's complex, it's expensive, in it often doesn't go anywhere good.
Yep, and Greg from operational you know radiology standpoint, I mean, what do you think would need to change for these more forward thinking futuristic endpoints to be embedded.
Do we need like more.
Frequent imaging in early stage development to really capture tumor kinetics or how does it feel need to adapt?
So that's a great question. Let me go back to one of the problems that I'm sure many of you all recognize just intuitively from RESIST and then put this in contrast with these other methods. So a problem with RESIST rate is one of the things you do is you select quote unquote target lesions, meaning the things you're gonna follow quantitatively, and then most of the assessment of efficacy. You know, response rate is based on what proportion of
patients had these sample of tumors shrink. Okay, whenever you sample a population, which is the population here is the tumor burden overall, and the sample is the lesions you pick to measure. Whenever you sample, there is a built in assumption that's not voiced, and that is that the outcome of interest. And I'm gonna say that that's survival is driven by the average. And that's not true.
Right. It's not the average lesion that kills the patient.
It's the fast it's the most aggressive, most treatment resistant lesion. And if you didn't happen to choose it when you were picking out lesions to measure, then whatever methods you apply, you're blind to it.
Right.
So one of the biggest things is I think you know to apply any of these methods including radiomics for pharmacodynamics, tumor growth kinetics, for os impact whatever.
You need to basically be able to measure everything.
And of course one of the big problems with that is that radiologist time is expensive. So when you're going through and you say, okay, no, mark every single lesion in the scan in a widely descemined you know, we've got a melanoma patient with dozens and dozens of lesions.
That's hard, that's expensive.
Now, fortunately we've got you know, kind of exponentially improving power of AI. AI right now is great at doing things like if you can just point to a tumor, it'll segment it in three D, it'll propagate it across time points. Actually finding the tumors still really requires people in most cases.
That's that's the hard task.
So there are a lot of efforts in progress now to develop that part of it, the identifying and outlining all of the tumors so that you can look at the biological impact across the entire tumor burden collectively, or a zim was saying, you may really need to be able to look at, you know, lesion by lesion, but
you got to find all allegians. So that's the effort of a that's the focus of a lot of development work that's happening right now by pharma, by tech companies, by consortia and across the board.
And is there a financial model that's viable because this is just software that sits on top of a scanner, or how do you how do you make sure that's scalable?
So well?
In a clinical trial context, a lot of work is done by central by core laboratories. So you know, as long as you have you know, as long as all the scans are going through a central facility, they can
take care of that. If you want to do this at a site, then what you will eventually need is software that sits either on the process on the workstation next to the scanner, or that is done somewhere in the cloud between the scanner and the pack system, which allows that sort of more robust detection and segmentation of all allegians.
Okay, I'll open this up to anyone. What else? What else are we missing here?
We have novel cool imaging techniques, what else blood based biomarkers? Should we look at patient reported outcomes? Do we need you know, predictive markers for toxicity.
What else I'll mention one and that is you know we heard earlier the molecular imaging, right, so molecular imaging gets you know where I said is you need to get closer to biology. Well, that's definitely a way to get closer biology. So you know, for io C D eight pet tracers for example, that you look at the distribution and movement of T cells, and so that's important.
The challenges with all of those things is that, of course you know novel pet tracers are you know, you can easily do that at mg H, but you you know you can do that at at Gustav Russ, but you can't do that at you know, East Lands and Community Hospital or whatever, right, so so you know you need to be able to That is a challenge of the molecular imaging side of it, but it is very powerful and in early development that is.
A tool that you could also be applied.
But I think one of the big operational challenges there is cost increase because if you want to add all these different biomarkers into your phase one study, well you will see some kind of cost increase in your phase one The only thing you can hope for is that a better understanding early on in your drug will lower the cost later on. But it changes the risk model, and I think that's something that can be seen with investors today that Project Optimists might make some investors more
hesitant to invest into oncology drugs. So that is also one of the consequences I think that is coming from Project optomists. This changed risk profile where we don't know exactly where it's going.
Right, true, although an argument against that would be that if you can optimize the dose in phase one and early phase two, that it'll.
You're more likely to succeed in phase three.
But until that happens a few times and we have those examples where we can point to to our investors, then it becomes much more difficult to sort of make that point, you know, in the in the ether without those tangible examples.
Yeah. Great, So I'll turn it to Andrea.
So we have is it a chilling effect that I'll dissuade investors and developers to go intown collegey and go into obesity and diabetes or is it de risking programs and and you know, getting a better return on investment.
What do you think.
I think it's somewhere in the middle, quite honestly. Well, firstly, your first thought about going abandoning on college and going to obesity is horrifying on so many levels. But I think de risking the dosing such that the phase three does become a little bit more you know, not nothing is one hundred percent, but it leads a little more predictable is a good thing. In incorporating modeling and some
of these new modalities into that is good. But I'm kind of also just wondering why we don't, you know, go old school and this drugs have been approved for many, many years without all of this and maybe start looking at post marketing studies and using AI with real world data. What can we learn from that to land on an optimal dose. So I think we've taken it from one extreme to the other and there's probably a nice sweet spot in the middle.
Somewhere, right.
And your thoughts on patient reported outcomes and early is that too early to do?
It's too early to do, And I don't think you would have a representative sample, and I don't know that really what you would learn from it, quite honestly, Look, I think patient report outcomes are always a good thing if you ask the right questions. You know, ask the wrong question, you get the wrong answers. So really understanding
what it is that we're looking for. But quite honestly, even with the patient reported outcomes, I think you'll get better information and better data doing something postmarket on real world information.
Right, we can open it up if anyone has questions. I'm going to throw at another question of the panel if we see I see some hands going up, but before while we get a mic, four mics.
Okay, all right, let's open it up. Go ahead. Hi, thank you.
That was very interesting and great talk, something which is close to my heart as well. So my question to Greg, so, if you're doing it in early phase, how do you convince your clinical development counterparts to invest in something? It has to be done relatively faster, closer, right, So you have to invest in getting the imaging, getting a collap to do all these assessments and bring it back to the clinical development team in time to make the decision.
So how do you.
Especially if there is not a lot of data out there to give a basis to so, how do you convince your clinical counterparts to do that?
If you have the magic answer to that. I'll I want to talk to you. I mean, you look, ultimately, it all boils down to data, right, nothing speaks like evidence. I think you have to start with potentially retrospective analysis.
You know, you can take you know, phase three trials where there's been a success actually there so there was a nice there was a nice paper a couple of years ago, I think Renee Bruno is the lead author showing that basically, so taking a roche it was a TESSO plus a chemo backbone versus the chemo backbone alone in non small cell lung and they used a kinetic modeling approach and they did a simulation where they took you know, a lot of they basically simulated Phase one
B trials either by sampling from just the just the treatment arm at the experimental arm and control arm or just control arm versus control arm as sort of like a negative control and they were able to show, you know, that you made better decisions on go no go with the modeling based approach, and you know, just of course, I think I'm sure everybody understands this, but that is such a high stakes decision. The do you go to
phase two do you go to phase three. That's a huge stakes decision and you have to make it with really small patient numbers. So I think if you can bring evidence saying that the quality of decision making is improved with these methods since starting with retrospectively and then implementing them prospectively as an experiment as a as an exploratory analysis in a trial, that would be an approach.
To do it.
Thank you, sorry, just a fallow up question and a whim on that. So I know we are not there yet based on Greg's response, So what do you think how we can do it real time?
Not real time?
Like today we are doing the trial and have results, but like within weeks or months to give it to the back to the team. Are we there yet to do all these analysis radio mixed tumor growth, kinetics and total tumor burden in real time?
I think we are at least we as Radiomics, are are ready for that at the moment. It took us a while to find the right setup to get through this phase of having sufficient capabilities to segment tumors rapidly, but today we are ready to have data delivered within the typical clinical trial analysis timelines, so have your last patient rat within two weeks after the last visit, feeding
it into the statistical program. So it is something that is there, but the big challenge tastes convincing people to use it. And as great that data is the key to the solution, however, you need some early convinced people to help guide and create that data.
So maybe I can add on another question here. You know, this is an interesting topic where you basically are between the two extremes of being fast and having to get to the next stage of clinical development in a pretty competitive environment for many assets, and trying to make the best possible decision with that as much data as possible. And yes, modeling can help more patients help, randomization helps,
but it does take time. So the way I have understood the Project Optimus design that FDA has asked for was somewhat are hybrid here in an attempt to not ask too much from sponsors and not add on too much extra time, but still get more data than three patients per cohored before you make a decision.
So you know, I.
Call that a practicality.
So the question is, you know, how can we keep as much practicality here and let sponsors move forward with their choices without adding too much extra complexity that has merit in some ways. So you know, it's a dance between those two components.
I mean, I think the backfill concept actually fixes a lot of that. As a phase one doc, if you have three sites, you always have patients each cohort, and you get experience with that cohort and you feel like you know the drug and you know the dose, and you go up. If there's seven or eight or nine sites in a phase one trial and it's for some reason a three plus three design, then you're not You
may maybe three months before you see a patient. If you have backfail, you can always have patients, and so if you have it, it actually works much better for a larger number of investigators, whereas a smaller group you always have patients and you're always sort of so I think actually encourages recruitment. If you say, oh, I have a patient with X and I want to get them on that trial, do I have a slot? If I know I always have a slot, it's way easier to
do that. And so whereas if my turn comes up three months later, it's like, hey, do you have a patient now? I's like, oh, I forgot I had that trial open. Yeah, let me give me, give me a couple of days and I'll have somebody for you.
I think that that those.
Types of designs that encourage active participation by the investigators and the sites lead to.
More efficient a cruel and it sort of solves itself on some level.
But you have to get a few dose levels in before you feel like there's a biological signal of some activity, before you feel comfortable exposing the patient to that to that dose level. But I think I think that's probably the most efficient way to go through phase one and dose escalation.
I love that answer. Backfields are quite useful tool.
Yes, yeah, absolutely, I think we have you have quick quick question.
We have twenty seconds, Okay, a quick one.
I'll take out the first part, just keep the second part. A lot of the talk today focused in on you know, volume change, looking at growth rates, to k rates, these kind of things all focused around the lesion. You know, do you think in an early phase trial where you're trying to make these go no goes or choose the right dose, there's value in looking at things outside of the lesion, like, for example, in brain looking at edema
changes in the lungs, looking at like pleural fusion. Do you see a place for quantifying these things in early stage or should we be doing that in the later stages once we've decided on dose.
Thank you, good question him.
Yeah, I think there's definitely space for those kinds of things. They add value into your understanding. Of course, they won't give you the definitive answer. I think at your early stage it's all about hypothesis generation that you do then validate later on. But yeah, looking at edema, for example, in the brain is very valuable.
So totality of evidence is key. As with everything in OPTI, miss Well, I really want to thank the virtual and in person participants. Has been a great discussion.
M