Why Game Theory Belongs in Cancer Clinics—But Isn’t Yet
Evolution never sleeps. But sure, let’s hit it with the same drug until it stops returning your calls.
In my professional life, I leveraged game theory and scenario-building approaches to model a client’s risk profile and then design mitigating strategies. The primary difference between the two is that Game Theory is a mathematical framework for analyzing strategic interactions between rational “players” (individuals, groups, cells, companies, etc.) that affect each other's outcomes. It assumes the players are actively trying to optimize their outcome. It focuses on interdependent decisions: my outcome depends on what you do, and vice versa.
Scenario Building is a narrative or planning tool used to explore multiple possible futures in the face of uncertainty. It doesn't assume strategy or rationality. It's about imagining “what ifs”, not predicting a winner.
In its simplest form, game theory asks: “If I make this move, and the other player makes that move, what happens next?” “And how can I maximize my chances, knowing they’re trying to do the same?”
It’s not guessing. It’s modeling interdependence, anticipation, and adaptation. Think chess, poker, diplomacy, even parenting. In all of these, you’re constantly adjusting based on what the other side might do next.
Now apply that to cancer.
Cancer isn’t just an out-of-control growth. It’s an evolving population of cells that compete, adapt, and respond to treatment. Give it a drug? It evolves resistance. Block one pathway? It reroutes. Kill the easy cells? The hard ones take over. From a game theory perspective, cancer acts like a strategic adversary, not intelligent, but evolutionary. It plays to win. So why isn’t game theory more widely used in cancer treatment?
The reasons are many, and none of them particularly satisfying:
1. Medicine is conservative by design. Doctors are trained to do what works, based on large-scale evidence. Game theory requires individualized strategy, often using real-time data, and that's harder to standardize, trial, or bill for.
2. It's mathematically complex. Game theory models aren’t plug-and-play. They need specific inputs, like tumor growth rates, resistance patterns, and patient tolerances. Clinicians don’t always have time or tools to run simulations between appointments.
3. Clinical trials aren’t designed for it. Traditional trials test fixed treatments, not dynamic, adaptive strategies. But game theory thrives on flexibility; pausing therapy, changing doses, adjusting timing. That’s hard to fit into the rigid framework of Phase I, II, III.
4. It feels risky, even when it isn’t. Holding back treatment to manage resistance? Stopping therapy when the patient seems to be doing well? These go against medical instincts. Game theory often suggests counterintuitive moves, and that makes people nervous.
5. It doesn’t promise a cure. Just control. Game theory is about managing cancer as a chronic condition, not eliminating it. That’s a shift in narrative—from heroism to realism, and not everyone’s comfortable with that.
But here’s the truth: in cancers like prostate, breast, and melanoma, game theory is already showing results. Not as a miracle, but as a method. A strategy to outsmart cancer’s adaptations, delay resistance, and preserve quality of life. It’s not that the science isn’t ready. It’s that the system isn’t. Yet. And that’s changing. As we understand cancer less as a single enemy and more as a dynamic, evolving ecosystem, the appeal of game theory becomes clear. In a world where cancer plays dirty, thinking like a strategist might just be our best move.
It starts, as most unsettling things do...
...with someone wearing a white coat and a quiet voice saying something like,
“Well, it’s not behaving the way we expected.” You’ve heard the words. Maybe “prostate,” maybe “recurrence,” maybe “resistant.” The details differ, but the theme is always the same: you’re in a game you didn’t sign up for, and the opponent is one that doesn’t follow the rules, mostly because it *is* the rules. Cancer, after all, is not a virus or a bacterium. It’s you. Just a version of you that decided to stop paying taxes and started building its own country.
So now what? Do you nuke it? Carpet bomb it with chemo? Starve it with hormones? The old approach was: **Kill it all, quickly.** But the new thinking is, well, smarter.
Maybe a little sneakier. It’s called **game theory**.
So What Is Cancer Treatment Game Theory, and Why Should You Care?
Game theory is what happens when smart people try to explain why other smart people do dumb things under pressure.
It started with mathematicians trying to model war and economics, and somewhere along the way, someone looked at a tumor and said: “You know, this thing behaves a lot like a bad actor in a prison yard.” It adapts, it lies low, it teams up, and when you look away, it punches.
Game theory is the study of **strategic behavior**. It’s not about IQ, it’s about tactics:
- “If I do this, how will they respond?”
- “If I wait, will they blink first?”
- “Do I go all in now, or hold my hand and wait for a better chance?”
In cancer, the other player isn’t a human. It’s a bunch of rogue cells playing evolutionary poker. And they cheat. Constantly. But if you understand their *strategy*, you just might beat them at their own game.
The Problem with "Kill It All"
Let’s say you have prostate cancer. The doctor prescribes hormone therapy because those cancer cells really, really like testosterone. Take away their fuel, and they die off. Perfect, right?
Except no. Because deep inside that tumor are a few **mutants**. Little survivalists in camouflage. They don’t need testosterone, or they figure out how to make their own.
When the rest of the cells die off, these guys **spread like weeds**.
Now you’ve gone from “mostly manageable” to “biochemically aggressive.” Why? Because you won the first battle too hard. You created a power vacuum, and evolution filled it.
This is where game theory enters and says, politely: “Maybe don’t try to win so fast.”
Fighting Smarter: Adaptive Therapy
Instead of going for the knockout, adaptive therapy says: **fight like a chess player, not a boxer. ** Give just enough treatment to keep the bad cells in check, but **not so much** that the resistant ones take over. It’s the biological version of “divide and conquer.”
Think of it like this:
- You have sensitive cancer cells (easily killed with treatment).
- You have resistant cancer cells (hard to kill, but slower-growing).
- If you kill all the sensitive ones, the resistant ones throw a party.
- But if you leave a few sensitive ones around, they **outcompete** the resistant ones, because even cancer has turf wars.
So instead of going full throttle, your doctor may back off the treatment at times. Let the weaklings keep the tough guys in check. Then, when it starts to tip too far, hit it again.
It sounds insane, I know. Like raising a few wolves to keep the coyotes from taking over. But it works. At least better than doing the same thing over and over while hoping this time the cancer won’t adapt.
What This Means for Prostate Cancer
In prostate cancer, this means **intermittent hormone therapy**. You start treatment. Your PSA drops. Great. Then, instead of staying the course until your body forgets what testosterone ever was, your doctor says, “Let’s stop for now.”
And you go, “Stop?! Isn’t that dangerous?” And they say, “No more dangerous than building a super-cancer we can’t control.”
So you pause. The sensitive cells come back (because they love testosterone). The resistant cells start to lose their edge. Then, you hit them again. Rinse and repeat.
It’s not a cure. But it’s control. It’s buying time. Quality time. Sometimes a lot of it.
And in this business, **time is the only real currency**.
What’s the Difference Between Game Theory and AI?
You may have heard about artificial intelligence helping doctors now. It’s true, AI is everywhere. It reads scans, predicts recurrence, and finds patterns in blood tests your doctor hasn’t even thought about.
But here’s the difference:
- **Game theory** is the *rules of the game*, the logic, the strategy, the models.
- **AI** is the *overcaffeinated intern* who runs every simulation, every outcome, and says: “Here’s the best move, statistically.”
You still need a doctor to look at the AI’s report and say, “Yes, but this patient isn’t a spreadsheet.” And you still need game theory to make sense of why the tumor is playing the long game, and how you can outplay it.
Where Does This Leave Us?
Cancer isn’t fair. It’s not a story, or a test, or a cosmic lesson in humility. It’s just biology with the brakes off. But there’s comfort in knowing we’re not just **fighting**, we’re **thinking**. That doctors aren’t just armed with syringes and scans, but with **strategy**. That your treatment plan isn’t one-size-fits-all; it’s a move in a game, designed to outwit a very stubborn opponent. And that sometimes, the smartest move isn’t to fight harder, but to **fight smarter**.
Because this isn’t checkers. It’s chess. And as long as the board is still in play,
you’ve got a chance.
Absolutely true—and it's a core principle of how I approach things as well. My medical oncologist trained at Moffitt, so she’s not just ahead of the curve—she’s light years ahead.
I simulated predicted results (assuming I was a moderate responder which from my genomic makeup appears likely) and the output floored me. Showed a distinct rise in PSA as populations jockey for position, then a leveling off, then a plunge. I watched the rise, the level, and now I've been going through the plunge for almost two years. The simulation predicted a PSA of 0.13 at year ten. It's only been four and I'm at 0.17. At the rate of decrease, I hope to be less than 0.13 in a month or two.
Per SOC I should have been CRPC years ago. I'm still HSPC with a zero tumor burden, a dropping PSA, an undefined PSADT, and a negative PSAV. And I have to add, the side effects? So good I'll do this even if my MO declares me cured.
"The poorly differentiated, low-PSA cancer cells are like hardened criminals armed with deadly weapons, whereas the high-PSA–producing cells resemble civilians wielding only sticks and knives. Over time, the criminals band together and even recruit the civilians, arming them with the same lethal tools.
Our goal is a population composed largely of those “civilians,” so eradicating them all is counterproductive. Equally unwise is weakening the civilians to the point that the criminals can easily conscript them. Instead, we aim to regulate the civilians’ modest armaments—allowing only pocket-knife level force—while neutralizing the criminals’ arsenals.
"
Four-Group Classification System
Understanding tumor heterogeneity is essential for aBAT optimization:
Group A: External DHT–Dependent Cells
• Androgen Sensitivity: Rely on circulating DHT for survival; they undergo rapid apoptosis under standard ADT while secreting the majority of PSA, making them reliable biomarkers of therapeutic response.
• Competitive Suppression: Maintaining a residual population preserves competitive pressure that restrains expansion of resistant clones.
• aBAT Targeting: Supraphysiologic testosterone pulses exploit high AR expression to induce TOP2B-mediated DNA double-strand breaks and preferential cell death in this group.
Group B: AR-Upregulated DHT–Dependent Cells
• AR Overexpression: Elevated AR levels sensitize these cells to SPA, situating them between hormone-sensitive and fully resistant phenotypes.
• Transitional Phenotype: They retain partial sensitivity to ADT or ARSIs but are vulnerable to SPA-induced cytotoxicity.
• Dual-Phase Targeting: aBAT’s alternating high- and low-testosterone phases combine SPA toxicity with subsequent AR blockade (typically using darolutamide) to prevent full resistance emergence.
Group C: Internal DHT–Producing Cells
• Intratumoral Steroidogenesis: Convert adrenal precursors (DHEA, DHEAS, androstenedione) into DHT, sustaining growth despite systemic androgen deprivation.
• ARSI Integration: Rather than abiraterone, aBAT protocols reinstate an ARSI (e.g., darolutamide) during the low-testosterone phase to suppress both circulating and intratumoral androgen signaling.
• aBAT Strategy: Pulsed SPA peaks followed by low-T ARSI phases suppress de novo androgen production while reintroducing competitive, hormone-sensitive cells to maintain tumor control.
Group D: Adaptive/Plastic Cells
• Phenotypic Plasticity: Can switch dynamically between androgen-dependent and independent states under sustained therapy pressure, representing a reservoir for adaptive resistance.
• Resistance Reservoir: Intermediate molecular features confer robust resistance, allowing survival through both hormonal extremes.
• Non-Hormonal Interventions: Often require local therapies (radiation, metastasis-directed SBRT) or immunotherapy for eradication.
• aBAT Strategy: Impose rapid cycles of supraphysiologic androgen exposure followed by deep suppression, thereby destabilizing transcriptional programs that sustain plasticity and forcing cells toward committed states that are more therapy-sensitive
• Experimental Concept: A single ultra-high androgen pulse administered concurrently with an ARSI (e.g., darolutamide during SPA) might destabilize plasticity programs. This hypothesis is novel and remains untested in clinical settings.
The poorly differentiated, low-PSA cancer cells are like hardened criminals armed with deadly weapons, whereas the high-PSA–producing cells resemble civilians wielding only sticks and knives. Over time, the criminals band together and even recruit the civilians, arming them with the same lethal tools.
Our goal is a population composed largely of those “civilians,” so eradicating them all is counterproductive. Equally unwise is weakening the civilians to the point that the criminals can easily conscript them. Instead, we aim to regulate the civilians’ modest armaments—allowing only pocket-knife level force—while neutralizing the criminals’ arsenals.
Chart of theoretical population changes.
Overall, Population C (androgen-independent) undergoes a steep decline from 120 million to under 1 million cells by year 10, reflecting effective long-term suppression of resistant clones. Population A (androgen-dependent) and Population B (testosterone-producing) initially recover during the adaptation phase and then decline as competitive dynamics stabilize. Population D (adaptive/plastic) remains at minimal levels throughout the decade, indicating limited phenotypic switching under optimized cycling. PSA levels, plotted on the secondary axis, mirror these population shifts by rising modestly during cell recovery phases and then declining steadily, reaching 0.13 ng/mL at year 10.