Adaptive Cancer Therapy: Using Evolution to Fight Drug Resistance
Featuring: Bob Gatenby, MD
In short
Dr. Bob Gatenby from Moffitt Cancer Center explains why treating cancer at the maximum tolerable dose until resistance often backfires — and how a strategy called adaptive therapy, which uses planned treatment breaks and lower doses, helped prostate cancer patients live roughly twice as long as those on standard care in a small study. He also walks through principles like combining treatments in sequence rather than all at once, and using mathematical models to guide decisions.
- •Ask your oncologist about adaptive therapy: in one study of men on abiraterone, cycling the drug on and off based on PSA levels extended median survival from about 30 months to about 58 months compared to continuous dosing.
- •Patients on adaptive therapy in that study were off treatment roughly half the time, which cut drug costs by an average of $70,000 per person per year — a real quality-of-life and financial consideration worth discussing.
- •If you have metastatic cancer, ask whether treatments might be given in sequence rather than all at once — hitting the cancer in steps may reduce the chance that resistant cells survive and take over.
- •Bring up the idea that keeping some drug-sensitive cancer cells alive (rather than wiping them out) may actually help control drug-resistant cells — this counterintuitive point is central to adaptive therapy and worth exploring with your care team.
Watch on Cancer Patient Lab YouTube
Ask anything about this — free, no signup
Instant answers grounded in real guidelines, not the internet.
May 18, 2022 Brad Power
Meeting Summary
Bob Gatenby, MD, Moffitt Cancer Center, Co-Director, Center of Excellence for Evolutionary Therapy, and Department Chair, Diagnostic Imaging, shared his novel strategies for advanced cancer treatment based on evolutionary and game theory and experimental models.
The strategies include combining therapies, low doses, sequencing treatments, and using mathematical simulation models, in contrast with the predominant treatment strategy of maximum tolerable dose until resistance.
An incidental theme is his concern about limitations in the clinical gold standard of evidence – double-blind randomized trials, which he argues have some benefits, but also some drawbacks, which are more apparent as we move away from maximum tolerated dose treatments to personalized cancer care with molecular targets.
Bob advocates using “adaptive therapy”, rather than continuously applying the maximum tolerable dose until resistance: you treat enough to knock the tumor back a little bit, and then pull the treatment away, allowing the tumor to grow. But since the sensitive cells do not have the burden of the resistance mechanisms that the resistant cells have, the sensitive cells have a fitness advantage, and outcompete the resistant cells.
The strategy is to use the sensitive cells that you can control to control the resistant cells that you cannot control. Bob shared a study using an androgen suppressing drug (abiraterone) in this adaptive therapy on-off mode, monitoring disease progression through PSA tests.
17 patients completed the trial and were compared to 16 demographically similar patients who had the same 50% PSA decline with the initial abiraterone dose, but who then got standard of care dosing. 5 months on adaptive therapy. 4 months vs. 5 months in the adaptive therapy group. 4 of the 17 adaptive therapy patients are still alive and on treatment, over six years since their treatment began.
The patients on the adaptive therapy did not get treatment about half of the time, which resulted in a cost reduction of an average of $70,000 per patient per year.
•Low dose: Hit the tumors with enough treatment to perturb their system, but not so much to kill the sensitive cells and leave the resistant cells to proliferate. The cancer group is heterogeneous and the resistant cells can be controlled by keeping the sensitive cells around.
•Combinations (first strike second strike): Lacking magic bullets, metastatic cancers can be cured through a strategic combination of pretty good bullets. None of these bullets could by themselves cure the cancer, but the combination could.
•Sequencing (not a combina combination cocktail): If you have a combination cocktail, especially as a first strike, you're applying the therapy to the largest possible population. The heterogeneity is such that almost certainly you will find tumor cells that can be resistant to the combination. It is better to hit the cancer with therapies in sequence, as each knocks the population down and can drive it to an extinction.
•Mathematical models : Having a hypothesis and a simulation of what should happen based on evolutionary theory helps in understanding why and enables insights from much smaller trial cohorts. To apply this strategy in real life we asked Bob to give his off-the-cuff thoughts on the case of advanced cancer patient Brian McCloskey, who is currently on abiraterone, the drug that Bob used in his study. Bob felt that because Brian currently has a low tumor volume that he is a candidate for an attempt at extinction, to hit his cancer with a ladder of 3 different drugs to successively knock down the population, with the hope that it could be nudged down the vortex. Requests
•Do you have any feedback on Bob’s principles for treatment strategy?
•This adaptive strategy seems intuitive, yet it’s not widely practiced. What are the barriers or objections to it? The information and opinions expressed on this website or platform, or during discussions and presentations (both verbal and written) are not intended as health care recommendations or medical advice by Cancer Patient Lab/Prostate Cancer Lab, its principals, presenters, participants, or representatives for any medical treatment, product, or course of action. You should always consult a doctor about your specific situation before pursuing any health care program, treatment, product or other course of action that might affect your health. — Meeting Notes Brian McCloskey: I'm super excited to have Dr. Bob Gatenby with us today. Bob is the co- director of the center of excellence for evolutionary therapy and the department chair of diagnostic imaging at the Moffitt Center. He's also a radiologist who specializes in exploring theoretical and experimental models of evolutionary dynamics in cancer and cancer drug resistance. He's developed an adaptive therapy approach for treating cancer, which has shown promise in improving survival times with less cumulative drug use. He's also led the formation of a program called integrative mathematical oncology, which brings together a group of applied mathematicians to collaborate with tumor biologists and clinical oncologists. And the goal of tha that is to develop non-linear dynamic systems, to examine the physiology of tumors, incorporating factors, such as phenotype evolution, intercellular communication pathways, and the interaction of microenvironmental factors, including therapies for prostate cancer patients. We are particularly fortunate to have Bob join us because his adaptive therapy has been applied in prostate cancer specifically, a study, including the dosing strategy for abiraterone, a novel hormone therapy, which I happen to be on currently. And I know some of the other patients that are on here have been on that drug as well. So to summarize, Bob is really a breath of fresh air. We're so fortunate to have him here. He has very novel ideas around cancer treatment that challenge conventional practice, particularly around the notion of maximum tolerable dose. Bob Gatenby: Apologies for this formal presentation. Uniquely in cancer centers, we have a department of mathematics. We now have nine applied mathematicians and two evolutionary biologists. Our goal is to use a physics-related paradigm, where we have theoreticians and experiments working together in complex dynamic systems like cancer. The physicists have taught us for centuries that this is really the only way you could address these kinds of problems. That's what we're trying to do. I'm going to talk about two topics:
•An evolution-based game-theoretical approach to treating metastatic cancers for control or cure.
•An incidental theme is my concerns about limitations in the clinical investigation gold standard of double-blind randomized trials, which I think have some benefits, but also some drawbacks. I think we're seeing those drawbacks now as we come into this era in which we're moving away from maximum tolerated dose treatments. Currently, personalized cancer care is all about molecular targets. Consider this example of a woman with metastatic lung cancer. She has EGFR mutations identified, and there are targeted therapies for that. You give it, and she gets a great response. This is a beautiful example of the problem that if you keep giving it, almost inevitably, the tumor will come back. It will evolve resistance and return, leading to treatment failure, and ultimately to the death of the patient. The basic idea here is that this evolution of resistance is patient-specific, drug-specific, and tumor-specific. This should be part of the personalized medicine initiative. The general principle is that by applying drugs in this way, the cancer cells will inevitably access the selection for resistance. The human genome will find ways to get around it, but it's the proliferation of that population that is meaningful. A small population of resistant cells is meaningless. It becomes meaningful only when it proliferates sufficiently to become a clinically evident cancer. One cc of tumor usually is about a billion cancer cells. So this requires a considerable amount of proliferation, and those are dynamic and subject to Darwinian forces, and that's where we want to focus. One of the approaches you can use is a game theoretic model – treat cancer treatment as a game. The oncologist is playing against the tumor. The oncologist plays the game by applying a therapy, and the tumor plays the game by applying an adaptive strategy. When you look at this as a game theoretic, the oncologist actually has two enormous advantages. One is that he or she plays first. The cancer cannot begin to evolve a resistance until the oncologist has placed a therapy down. This is called a Stackelberg game, and it's the equivalent of playing white pieces in chess – the oncologist always leads the game. The second, and probably more important advantage, is that the oncologist is sentient and can play dynamically – she or he can anticipate the future and can change therapies on the fly. Depending on the evolutionary dynamics that are going on in the tumor cancers, any evolving population can never anticipate the future. It can never adapt to an environment that it's not seen before. The problem is that the dogma in cancer for the last 50 years has been that treatment is applied continuously and at the maximum tolerated dose until progression. That's the standard approach. But in doing that, you'll notice that the oncologist loses both advantages because the oncologist simply plays the same move over and over again, applies the same therapy again and again, and the cancer cells do not have to change their adaptive strategies. They can just keep adapting to that. The other thing is that because the therapy has changed only when the cancer is observed to grow, you're ceding control of the game to the cancer. The cancer is leading the game, and the oncologist is following. Part of what we're trying to do here is to use these advantages in a dynamic way, by understanding the underlying evolution. If we restate the game from an evolutionary point of view, suppose we start with a mixed population of cells, most of which are sensitive: if you give a maximum tolerated dose, you get a great response: the tumor shrinks away. But what is left behind – almost inevitably – is a small population of cells that are resistant. If you keep giving the same treatment over and over again, you're treating resistant cells, and you're having no effect on them. Even though the tumor may be not visible, you may not be seeing this growth because it's below the resolution. It is in fact growing, and the conviction of the oncologist that the tumor is under control with this therapy is really an illusion. Eventually you get proliferation that you can see. This is a very well known evolution dynamic. It's called competitive release. And as the name suggests, what's happening is that by using this approach, you're maximally killing off the cells that are sensitive, leaving behind the cells that are resistant, and eliminating all of their potential competitors. To them now they're only competing against each other, and that's bad for the patient because that means that these guys are slowly growing, getting better and better at what they do. An alternative approach, one of many, is what's called adaptive therapy, which is to deliberately pull your punches. You could apply treatment, but you specifically want to keep intact a sufficient population of sensitive cells. You treat only enough to knock the tumor back a little bit, and then you stop treatment. You pull it away and the tumor will grow. But since you're not applying a selection for resistance, and because in general, the sensitive cells, which do not have the burden of the resistance mechanisms that the resistant cells have, the sensitive cells have a fitness advantage. So when the tumor grows back, you recapitulate the initial population, meaning that you still have a large number of sensitive cells. You treat again, cycling the minimum necessary therapy to keep the cells in check. The goal here is to use the sensitive cells that you can control to control the resistant cells that you cannot control. This is the basic idea. Mathematically, this is using the sensitive cells as a forcing function. This is not a passive approach. You're driving them forward and backward. The idea is to use those oscillating dynamics to create a nearly steady state. We did preclinical experiments, but the first group that we applied this adaptive therapy approach on was abiraterone therapy for men with metastatic, castrate-resistant prostate cancer. Most of the cells in a typical prostate cancer at the beginning require exogenous androgen produced by the body. When you start to knock those down with the initial androgen deprivation therapy, some of those tumor cells begin to make their own testosterone, so they can proliferate and become resistant. Abiraterone blocks that enzyme, so that they are no longer making testosterone. This is typically a second line therapy. The time to radiographic progression typically is in the range of 8 to 60 months, depending on pretreatment. We established a mathematical model for this. The details are not important, but just to give you some idea, although you could look at this in a variety of ways, we just used a very simple model. We have a population of cells:
•T+ cells that require exogenous testosterone.
•TP cells that make testosterone. These are the ones sensitive to abiraterone.
•T- cells that proliferate independent of testosterone. These are the bad guys that we want to suppress. We have a mathematical model. There's an interesting dynamic here because the TP cells not only make testosterone, they spill it into the environment. The T plus cells can use that. That's called cheating. They have the testosterone as a common good, so that can complicate some of these dynamics. We have two mathematicians and an evolutionary biologist here as part of the group working on this. We developed a payoff matrix. Jingsong Zhang is the very brave oncologist who brought this into clinical practice. It was a very simple protocol. We said that you would give abiraterone, and when the PSA got to 50% of pretreatment value, then you stop, and you let the tumor go. And when the PSA comes back to what it was pre-treatment, you start again. You cycle this over and over again. Our mathematical models predicted that we could control the tumor for 2 to 20 cycles, and this actually turned out to be incorrect. I'll talk about why it was incorrect shortly. We completed the trial. There was no funding for this. We scraped money together to do this. We had 17 patients that completed the trial. We compared those to 16 patients who had the same 50% PSA decline with the initial abiraterone dose, but who then got standard of care dosing. These groups were quite similar demographically. mographically. This is where we are now with this. The red line is the adaptive therapy group, and the blue line is the standard group. The difference in median time to progression was 14.3 months on the standard of care, compared to 33.5 months on adaptive therapy. You can see how the different cohorts cycled on and off abiraterone. It's interesting that four of these patients are still alive and still on treatment, over six years since their treatment began. Overall survival in the standard of care cohort was 30.4 months vs. 58.5 months in the adaptive therapy group. Because of this cycle on and off, the patients in the trial did not get treatment from abiraterone about 46% or half of the time. Some economists examined this, and they found a cost reduction of an average of $70,000 per patient per year, largely because of drug costs. Four patients that are still stably on cycle, still going strong. Now they're more than six years out, and this was not predicted by the model simulations, and this becomes an important part of the whole analysis. Because we use the mathematical model to guide the trial, we can use the mathematical model to analyze the trial. We first used the longitudinal data from the trial to update our parameter estimates, to see if we were correct. We went through the models and predicted what the pretreatment fraction of resistant cells would be. When we looked at that, we could see that the higher the predicted fraction, the shorter the time to progression, which makes sense. The circles are the patients that were on the standard of care and the triangles are the patients that were on adaptive therapy. For every fraction of resistant cells, the adaptive therapy group did better than the standard of care. This is an example of two patients that had roughly the same fraction of resistant cells estimated by the models. And you can see that one person progressed after 220 days. The other person went through multiple cycles before progressing at 1400 days. So this is an example of how this was different between the two. This was a critical estimation error that we made. If you're using the sensitive cells and the fitness advantage to suppress the proliferation of the resistant cells, the critical factor is the difference in their relative fitness in the absence of treatment. The greater that difference is, the greater the suppression of the resistant cells by the sensitive cells. Our pretreatment estimate was that this would be 2. What we found in looking at the data though, is that the number is closer to seven, about a threefold increase. “An Evolutionary Treatment Strategy” (Bob Gatenby, I’m going to go all nerd on you here for a minute, because this turns out to have clinical consequences. If you look mostly at the lower left corner, this is assuming a ratio of three, so close to what we assumed originally. This is what we thought would happen. Whenever you see a decline, you treat the tumor, and you see a decline in the sensitive population. When that happens, you'll see an increase in the resistant population when you then stop therapy and let the sensitive cells go back. We predicted that this would plateau. It would suppress it so that it stayed stable. What we thought would happen with each cycle was that it would go up, plateau, up, plateau. And because of that steady increase, that was why we had an absolute ceiling on how many cycles you could go, no more than 20, according to the model. Here's the difference. If this ratio is higher, then the fitness benefit of the sensitive cells is greater than expected. And so instead of this up and plateau is that whenever the sensitive cells would go up, the resistance cells would go down. And we found that if we could hit this right, if we stuck to the protocol exactly, we would actually reduce the resistant population somewhere near extinction, somewhere near zero. Did we see that? We went back and looked at the four patients who were still cycling and we found that now that we have an updated model, we can apply it to every patient on the trial and the model can tell us what happened leading to the outcome that's observed. Instead of just looking at cohorts, which is the standard way clinical trials are run – statistical differences between A and B – what we want to understand is why A was better than B. What was it that was successful? And more importantly, how could we have done it better? What did we do right? And what did we do wrong in this case? The math models show that for the patient who's still being treated, in the very beginning we hit these cycles perfectly, and knocked the population down to something close to zero. In some of the cycles, which are not as perfect as some others, we didn't go up as high as we would have expected. We didn't get a blip, nothing happened, suggesting that this population of resistant cells is really very small and possibly zero. I should point out that we have looked at circulating DNA for these four patients and have not been able to detect any markers of resistance. Why did we fail? This is as important as understanding what we did right. Why did 13 of the 17 patients that we treated with adaptive therapy progress? It turned out we made an error. What the protocol said was that at a 50% drop of the PSA from the pretreatment value, stop therapy. The protocol had to be confirmed by radiographic study. We could get PSA results every month, but we could only get radiographic studies every three or four months. Instead of this ideal cycle that we wanted, the PSA would go to zero in this patient and stay there for a prolonged period of time, months, before it was allowed to come back up. We fell below the ideal cycle. We were overtreating the patient. Instead of allowing a substantial population of sensitive cells to remain, we were eliminating them. By making this population very small, the resistant population kept going up. We almost got back control, but lost it. What this is telling us is that we could take these patients and say what would have happened if we had done it exactly right. What if we had hit this perfectly? What it showed was that if every patient in both protocols had done it correctly, we could have had very long term control of the tumor. We've learned that by using the models, we can now say, how do we design our next protocol and what are we going to learn from this? And how can we do it better? The lesson here is that the current approach to evolution of resistance is quite fatalistic among oncologists. It can be changed. We can interfere in these evolutions and do things that will at least slow it down, if not totally stop it. The value of the classic randomized, double blind trials is inversely proportional to our understanding. If you treat intratumoral evolution dynamics as a black box, and if you don't try to investigate it, it's always going to be unknown. The problem is that you get results from these trials, but you don't know what to do. We keep adding new drugs to the mix without really understanding the dynamics that led to the results in the first place. By integrating these evolution-based mathematical models into the trial design, you can get more information from smaller trials. You can critique your trial and go to an improved trial and follow up studies. Please bear with me for a descent into pure speculation. Everybody that is associated with cancer treatment has probably had the experience at a family gathering, or at some kind of party, where you get cornered by an acquaintance or a relative who says, “you guys know the cure for cancer, but it's not being used because you want to make more money.” This is one of the conspiracy theories that's been out there for decades. This concept of the cure for cancer comes from the work of Ehrlich, who popularized this concept of magic bullets, the idea that you could identify an agent that was selectively toxic to cancer cells, but left normal cells alone. The antibiotic era made it look like that should be possible in cancer. I'd argue that all of drug development is in some way related to this idea of finding a magic bullet. This idea of a magic bullet seems highly improbable to me because of the size and the phenotypic diversity and spatial dispersion of cancers. These seemed like insurmountable barriers. But I was persuaded otherwise, that at least we should be looking at other things, by Damon Reed, who is a pediatric oncologist who treats teenagers with metastatic rhabdomyosarcoma and other sarcomas. In the metastatic setting, these kids usually die. We talked about adaptive therapy. They would be very good candidates for adaptive therapy because they typically respond very well to initial therapy, but in talking to other pediatric oncologists, they said, “That's nice, but not good enough. We're about curing things. You’ve got to go back and find a cure.” This is a group of people that have been involved in this multidisciplinary idea. Going back to the drawing board, let's start with the idea that a cancer cure is like an extinction. It's an extinction of a large, diverse, spatially dispersed group in a reproducing clade (a group of organisms believed to have evolved from a common ancestor). Whenever we think about extinction, we always go to the idea’s source. This idea in the popular imagination is that extinction requires a “KT impact” (the sudden mass extinction of three- quarters of the plant and animal species on earth about 66 million years ago) – the application of a massive evolutionary force that wipes out this population in a single cause, single event. The problem with that approach – which is what we try to do with the maximum tolerated dose continuously applied – is that it's indiscriminate. The KT impact also destroys 60% of the other land animal species. Brute force as an extinction strategy is always limited by its indiscriminate effects. And in the case of cancer, you're always going to be limited by the effects on normal cells. Is that the only way we can make extinctions? We live in what's called the anthropocene era, meaning that our species is killing off other species. That's terrible, of course, but the silver lining is that these have actually been very well studied. There is new data that's being generated about the organic system of extinction. Examples include the passenger pigeon, which was the most populous bird on earth at the time, probably 4 billion birds in the population. The humans eradicated it within a hundred years. Another example is an intentional anthropocene extinction, and that's the Galapagos goat. These goats were introduced by sailors into the Galapagos Islands. They were probably left there as potential food sources for whenever they returned. Over the years, this population grew, until in the 1980s and 1990s there were about 500,000 goats. They were both destroying the native plants and crowding out some of the native animal species. An international consortium decided that it was time to kill off the Galapagos goats. This is an intentional anthropocene extinction. Well, how do they do it? They got New Zealand sharpshooters, gave them automatic weapons, put them in helicopters or on trucks, and they drove around or flew around the Galapagos Islands, came across a herd of goats, and gunned them down. This was just mass slaughter. The population plummeted, but it did not go away. The reason is that some of the goats became sensitive to the sound of helicopters and trucks. As soon as they heard them, they would run into the woods, where they would hide. They could not be found. This population not only survived, but began to proliferate. They had to come up with something else. The strategy that they came up with is called “the Judas goat”. This is a female goat that was neutered. They put a radio collar on her, spread hormones all over her, and sent her out. She ended up joining these small herds, and they could track them down due to the radio collar, and kill them. In the end that eradicated the population. Now look at the analysis of this extinction. This is often divided conceptually into a first strike and a second strike. The first strike involved a mass reduction in the population, but it also selected for certain resistance populations. You could keep applying the first strike all you want, rike required a different strategy than the first strike. It required a strategy that would never have been effective as a first strike because this is really very targeted and small. The point is that anthropocene extinctions are not a single cause, single event dinosaur eradication. It's a multi- cause, multi-event process. You will notice that none of the treatments that were applied here by itself could cause extinction. It's the strategic sequence. Can we think about this in terms of a cure? This is an example of a breast cancer pre- and post-neoadjuvant therapy. On the left is the biopsy before treatment, and on the right is the surgical specimen. You will notice that we start with this large continuous population, and what's left behind are the survivors – these small islands of tumor cells. It’s very analogous to the Galapagos goats. How can we deal with these small populations? Large populations and small populations of the same species are entirely different. They respond differently to perturbations. The reason is that small populations are sensitive to stochastic events (random or involving chance or probability). Small changes in birth rate or death rate can cause a population to go extinct, whereas that would never happen with a large population. Remember that the extinction here is an “absorbing boundary” (takes in things without reflections, i.e., things go in, but they don’t come out) – once this population goes to some threshold, it's gone, and it cannot come back. Another factor is “allee effects” (the benefits of social aggregations to individual fitness). Large populations typically have a fitness advantage over small populations for a number of reasons. Allee effects have been well recognized in cancers. What's seen in these small populations is now characterized by evolutionary biologists as the “extinction vortex”. That means that every time you knock the population down a little bit, it becomes more vulnerable to extinction because the smaller the population is, the more stochastic events and allee effects have an impact. This idea is that, although we talked about a second strike in the case of the Galapagos Islands goats, that here the second strike is really a series of small perturbations being applied to a small population, with the goal of continuously sending it down this extinction vortex. These become self-reinforcing, and therefore can be used strategically to essentially eliminate the population. We've modeled this, and you can see on the lower left that this is a typical U-shaped response. You give treatment, get a great response, stay stable for a while, it comes back up. Suppose you add another treatment to that. So now we do two treatments. How do we do it? Typically we have them in front. They are given at the beginning (as a combination). The problem with that is that you're applying the second therapy to th the largest possible population. The heterogeneity is such that almost certainly you will find tumor cells that can be resistant to both. The model shows, yes, that works very well in the sense that the response is more prolonged, but as is typically observed in cancer populations, it eventually comes back. With 3, 4, or 5 drugs you typically see the same result. Eventually you get to a point where you're not getting any better results, and you're just increasing toxicity. But now take the same drug, and instead of giving it at the beginning, give it when you've got the tumor on the ropes, when the tumor population has declined, when it is badly damaged by the initial therapy. These guys are now vulnerable to extinction. When you give it during this period of decline and toward the lowest level of the population, you can see that the second perturbation can cause extinction. Not always, and it's obviously more complicated than that, but the idea is that we need to be thinking about giving cancer therapy in ways that are strategic and are exploiting the vulnerability of small populations to extinction. There have been a couple of papers that we've published on this. Pediatric oncologists, by the way, embrace this almost immediately because they know it. This is basically the strategy that's been empirically derived to cure kids with leukemia. They start with an induction therapy (chemotherapy used at the beginning of cancer treatment before starting radiation therapy). They've learned historically that although at the end of induction therapy you frequently cannot find cancer cells in the blood or in the bone marrow, historically they know it always comes back. So rather than waiting for the tumor to come back, which is what adult oncologists typically do, they've been treating immediately. After the induction treatment, they then do “consolidation”. This is the second strike. They're applying a sequence of different drugs subsequently, and this is commonly a curative outcome. Lacking magic bullets, maybe metastatic cancers can be cured through a strategic combination of pretty good bullets. One of my colleagues likes to say that we use the boxing rules now. You knock a tumor down, and what do you do? You just stand back and wait for it to come back. This is what we do. This is prostate cancer therapy. You start with androgen deprivation therapy, you get this kind of U- shaped response. You can normalize PSA in the vast majority of patients, and frequently it becomes undetectable. But historically it comes back. When do we start therapy again? Or we wait for it to come back. You know, it gets up off the mat, shakes it head, and is ready to go. We said, “This is a knife fight.” When you've got your opponent on the mat, you close and kill him. It can be curative if you give the therapy when the PSA goes down, when it's at its nadir. This is where you should pull out the stops and hit it with everything you've got. This is the Moffitt evolutionary tumor board. I think it's unique. It's run by oncologists, but it includes mathematicians, evolutionary biologists, and a wide range of people. I also want to thank our funding. Brad Power: I would like to hear the thoughts from Saed Sayad, a fellow mathematician with simulation models and an MD. Saed Sayad: I mentioned this before: every simulation we have done using -omics data on psoriasis, cancer, or other diseases support less concentration and combination therapies. A single therapy and increasing the concentration just increases the reaction from other genes or systems. I just want to say I completely support this hypothesis. I'm not sure about the sequencing of the treatment. Should it be at one time or should it be in sequence is a different story. The one question I have here is about the biological system. They are a massively parallel system. How do you deal with this issue using a simple mathematical equation? Bob Gatenby: We started simple. We go by the old adage that the model should be as simple as possible, but no simpler. We started the prostate trial, for example, using three populations because we know they made sense biologically. We felt certain that there were other populations present, subpopulations and certainly more complex dynamics. The question was that sufficient to at least make some predictive models. What we'll find is that as time goes on, our models are going to have to change. It should be similar to the way we've learned about how to predict weather. We need to get models, and the models have to be sufficiently complex. For the weather they have to be really quite complex, but we also have to then get data to parametrize the models. You can't just get any data. My own kind of impression of cancer therapy right now is that we just keep gathering data without any clear idea of what the fundamental principles are and how these data can be used to parameterize models. The idea is that t the models can help you understand what data is most important. We need to get models that can work some. They're clearly not going to work all the time, and all models are wrong to some extent. We have to build the model into complexity as needed to get the outcome that we want. But given the very simple starting point, that's where we started. Saed Sayad: It's a strong start, but the point is this should be a general model. There are other variables it should include, such as the phenotype and genotype, and other data, but this is a great idea. Bob Gatenby: It's a very good point, and by no means do we say this is the way to go. This is simply the starting point. It'll be fascinating to see this happen. It's hard to get mathematicians and oncologists and biologists to talk with each other; it's not traditionally done. That's why when they recruited me to head their radiology department here, and I had them over a barrel, I said that's what I wanted, because I really felt that unless you get people in the same spot, mathematicians and experimentalists working together, that it's really hard to overcome all of the barriers. This program has been going on for 15 years. It took some time to overcome these barriers, and we are still overcoming them. Clearly we have much more interaction than we ever did before. Brad Power: I'd like to bring the conversation from theory into practice and take a moment to ask you, Bob, to comment on Brian’s specific case and the specific decisions he is facing right now to see how your theory would apply in practice in his case. Brian McCloskey: The reader's digest version of my case is that I've been battling this for six years. I've had surgery, radiation, and Lupron. I did a combo of Lupron and apalutamide, which was very effective, but did not use adaptive therapy. I beat it to death. Then I decided to take a complete holiday from everything. This is point number five here. As soon as I did that, I saw my PSA jump up significantly. I went from no evidence of disease to six lesions in my peritoneum. In August of 2020 I had surgery. They didn't get it all because of margins, et cetera. Then I started a combination of chemo plus pembro, which was probably questionable. The chemo was probably fine, but in the event, I did see a decline in my PSA from about 1.2, all the way down here to about 0.2, 0.3 or so. Then I saw it begin to rise, and I had three lesions in 2020. I've been on abiraterone since November of 2021. I have seen a decline. So if I look at your model, I started at 0.91, got all the way down to about 0.45. Now I just jumped a little bit to 0.49. Is it too late to even consider using adaptive therapy? Could you work with my oncologist to enable that? And if not, there's going to be another therapy for me around the corner. It could be one, or it could be several. It could be a combination. How would we attack that problem? Bob Gatenby: Are you on Lupron too? Or just abiraterone? Brian McCloskey: Yes. Bob Gatenby: Looking at your graph, I would not take abiraterone away right now because the problem is that if you kill off too many of the sensitive cells, it will not be as effective, or effective at all. So the fact that you've plateaued makes me think that it's probably not going to be an effective approach. On the other hand, you have very low volume disease. We probably have to talk. For example, this is where the Lutetium PSMA (Pluvicto) might be a treatment option with having low volume disease. This is where the oncologist is really critical. What does he or she think? Could you, for example, go back to apalutamide or some of the others? Brian McCloskey: Darolutamide. I don't think there's any benefit to cross over to enzalutamide once I'm on abiraterone. Bob Gatenby: What’s the new therapy that blocks the androgen receptor? Brian McCloskey: Darolutamide, the Bayer drug? Bob Gatenby: I don't remember, but the question in this mix of hormone-related therapies is what are the options to continue therapy right now? You've got such a mix that in the past I would first ask the oncologist to take their best shot at what they think would be the best approach. I would think about PSMA Lutetium as an addition. Given your small volume, you could try extinction treatments, a sequence of three hits with a goal to keep pushing it down. An extinction ladder. As an aside, one of the problems is that we don't know whether treatments currently are approved based on their monotherapy effects in cancer. But these are big cancers. These may be entirely different. They may be far more effective in smaller tumors. ably at the very edge of their metabolic existence. In that setting, the antiangiogenic effects may be very useful. Some of the immunotherapies, in this setting of small tumor volume, would be very effective, and some chemotherapies that are not typically given, could in this setting be highly effective. Again, you have to just nudge it down the vortex. You don't have to hit home runs with each of them. You just have to hit bunt singles. The way this usually works is that I work with the oncologist who provides the expertise in terms of the most likely treatments and so forth. One of our extinction patients is a scientist at the Huntsman Institute who has extensively investigated non-traditional medical things, like the role of exercise and some dietary things. I like to get him involved in that, because he has reasonable evidence that some of these things are good. That would be my approach. The way we typically do this is that in patients who are not Moffitt patients, we have an oncologist, an evolutionary biologist, a mathematician, and me, and the oncologist would typically be your oncologist. We meet with you present and talk about the options because what we try to tell the oncologist is that you are a modeler, you have models in your head about what's going on. Our goal here is to take those models, frame them mathematically, and then analyze them mathematically. We're not trying to displace you. We're simply trying to allow your model to be simulated. We can do thousands of simulations, and that's typically how we meet with the oncologist. We do the model, and we run thousands of simulations. And then about 10 days later, the full tumor board meets and the computer simulations are then presented to the board. It's part of their deliberations. There's clinical intuition. There are all sorts of clinical issues that have to be dealt with. Some things are just not practical, and out of that comes a decision. The decision is made by the oncologist and the patient. Our goal is to be informative, and to do the best we can to simulate what we think will happen given various treatment options. There are other factors. For example, sometimes we think this is the therapy you should try, but the insurance companies won't pay for it. Then we have to make some kind of compromise. For example, when we started our trial, we wanted them to get their PSA every two weeks. With all apologies to people with Y chromosomes here, what I learned is that men are terrible patients, and absolutely would not get their PSA every four weeks. Women will get it every week if you want, but men will not. Brian McCloskey: You can take mine daily. Bob Gatenby: There are also other approaches. Michelle Lockley, who is an oncologist in the UK, has just started a trial in ovarian cancer using an approach that we developed in animals, but did not feel we could get into the clinic. It is an approach where you try to keep the tumor stable in size, and you continuously adjust the dose of the drugs to keep it there. In animal models that was the best approach, that actually got the best results. Over time it turns out you need less and less drugs to keep the tumor under control. They observed that in their preclinical experiment. And so they brought that into the clinic. On a practical level, that's very difficult, but that could be done. They obviously are doing it. Something like that would be a potentially good approach. And in fact, in our preclinical models, it was the best approach. What we did with the abiraterone trial was kind of the next best approach, because that fit into the practicality of clinical medicine more than the other one. Questions from the Chat (For Future Discussion) Rick Stanton: How does the ratio of resistant to sensitive cells vary from patient to patient? Saed Sayad: Biological systems are highly parallel. How do you deal with this massively parallel system? Rick Stanton: How does the evolutionary concept apply to chemo? Might it be assumed that chemo hits cell division - which might be difficult to gain resistance against? Brian McCloskey: This seems so intuitive. Why is adoption of the adaptive therapy approach slow? In other words, what are the arguments against it? Mike Yancey: In the first example using abiraterone, was ADT continued or stopped when starting abiraterone? Rick Stanton: My oncologist - Dr. John Shen - UCLA commented: the evolutionary concepts are intriguing and hopeful, but so far, we don’t have enough evidence to switch - comments? Emma Shtivelman: Any comments on bipolar androgen deprivation? Chandra Kota:We fractionate radiation treatments, but we hope to make the resistant cells more sensitive for subsequent fractions. How should radiation treatments figure in the sequence of treatments? Kerri Gogolin:Chandra, and that reminds me of XRT followed by HDR brachy boost. First strike one type; second strike another? Rick Stanton: I’m nervous about anti angiogenesis approaches - when you hold something underwater - it will do its best to move to get air = metastases? Saul Priceman: Would be good to incorporate the spectrum of sensitive and resistant cells and move away from a binary model. Also, the Galapagos goats example reminds me of stem-like cells as resistant populations, where growth and dormancy impact the second punch.
Want to learn more about your specific case?
Upload your medical records and ask Navis questions tailored to your diagnosis.