Don’t Shoot the Messenger

On ketones, cancer, and the confusions we keep mistaking for evidence

A tumor learning to make its own ketones isn’t proof that ketones are evil. It’s proof that cancer is clever. So are we.

A quick note before we begin. I’d been sitting on this piece for a while, letting my biochemist husband pressure-test every line of the science, when my colleague Amanda King, ND landed her own in my inbox — Why the Keto Backlash is Mostly Bad Science. Two of us, different doors, same room, same week, neither of us knowing the other was writing. With all the noise and buzz in this space right now, nothing makes me happier than watching good colleagues pop the same myth bubbles from their own angles. So read hers right alongside mine. This is a conversation, not a competition.

Photo by Olga Petnyunene on Unsplash

For the better part of a decade, I was “the keto gal.”

It walked into rooms before I did. It got printed under my name, said from stages, typed into comment sections by people who never read a word past a subtitle. When our book came out in 2017, the publisher chose a subtitle built for search engines — the ketogenic diet sat right there on the cover, doing its SEO job — and a whole world decided that was the sum of me. A diet lady. The keto doctor. Never mind that the book was a hell of a lot more than a diet. Never mind the deep nutrition, the bio-individual testing, the whole terrain underneath it. People read the label on the jar and never opened it, and the label stuck to my skin for ten years.

I’m telling you this because I know exactly what it feels like to be headlined into something you are not. To watch the nuance boil off until only the scary, sticky, clickable word is left. And that — precisely that — is what is happening right now to the science of ketones and cancer.

Here is a distinction almost no one making noise about this bothers to draw. Keto is a diet. A noun, a marketing word, a set of macros somebody can sell you. Ketosis is a physiologic state — ancient, ordinary, the metabolic gear your body was built to shift into every time it fasted, slept, moved, or went without. One is a brand. The other is biology. When we blur the two, we make the same mistake the world made with me: we confuse the label for the living thing underneath it.

I am not the keto gal. I never was. I’m the metabolic flexibility gal, the terrain gal, and I’ve been quietly working my way out from under that word for years. So when I read the latest wave telling frightened people that ketones feed cancer, I don’t only see shaky science. I see a label being slapped on a jar nobody opened. Let me open it.

Every few months the current shifts and a new fear comes floating downstream. Lately, again, it’s ketones. The posts and threads take a familiar shape — the tumor makes its own ketones, ketones feed cancer, the ketogenic diet doesn’t starve the disease, it arms it. The mechanism gets drawn in confident diagrams. The language is certain. And the people already frightened — patients, parents, the caregivers holding it together at two in the morning — read it and feel the floor tilt underneath them again. I know, because the fallout lands in my inbox: hundreds of messages a week flooding my DMs, my texts, my email, all some version of the same question asked with a shaking hand — do I have to be afraid of this too?

I’ve spent over three decades in this field, on both sides of the diagnosis, and I’ve learned to notice what my body does when a claim like this lands. If it makes me flinch, that’s not a reason to argue. It’s a reason to get curious. So let me do here what I keep asking all of you to do: slow down, and read the thing underneath the thing.

Curiosity is the whole posture I’m inviting you into. Not credulity, not cynicism — curiosity. The willingness to hold a frightening claim up to the light and turn it over in your hands before you let it move into your body and take up residence. That is a different nervous system than fear. It happens to be the one that healing lives in.

Because I don’t think the data is wrong. I think it’s being misread — headlined, flattened, stripped of the context that is the entire story. And the misreading is happening in a few specific, nameable places.

The tumor built the kitchen itself

Start with the biology everyone is citing — the elegant work showing that certain cancers, leukemic stem cells and pancreatic adenocarcinoma among them, switch on their own ketone-making machinery. They turn up the enzymes of ketogenesis, they run their own fatty-acid oxidation, they build a private little β-hydroxybutyrate factory inside themselves, and they use what they make to shield against the very cell-death pathways we’d otherwise like to provoke.

That is real — in the lab. Real in cell lines, real in mouse models, real in careful mechanistic work I have no wish to wave away. But read the sentence slowly, because everything hinges on it: the cell makes its own. This is endogenous. It is the cell’s own machinery, running on the cell’s own fat, sealed inside the cell’s own membrane. It has almost nothing to do with what’s on your fork.

Photo by Michael Lock on Unsplash

Picture a house that’s decided to distill its own moonshine in the basement. It bought the copper still, it runs the mash, it drinks what it makes. Now someone reads a report about that house and concludes the danger is the corner liquor store down the block. That is the leap being made. To take the leukemic stem cell manufactures its own ketones to defend itself and turn it into therefore you must fear ketosis, fear fat, fear the diet is a category error dressed up as caution. The tumor was not waiting on your breakfast. It built the still itself, in the basement, with the doors locked.

And now let me add something from the exam room, and let me name it for what it is — conjecture — because that is what a careful clinician does with a mechanism she cannot yet prove. In my decades of walking alongside people with these two exact cancers, the leukemias and the pancreatic cancers, I have watched them respond, often remarkably, to ketosis. That sits oddly against the alarm, until you remember what these cells actually are. Leukemic blasts express, and frequently overexpress, insulin and IGF-1 receptors on their surface; they are exquisitely tuned to insulin and the glucose it ushers in. And pancreatic cancer’s whole reason for being seems lashed to insulin and glucose signaling — that is close to its soul job. A metabolic state that lowers insulin and quiets that signaling is not the obvious villain here. This is mechanism, not proof, and a living body is not a single cell line in a dish. But that is exactly the point. What happens in a whole terrain, over time, in a real person, keeps refusing to match what the isolated-cell story predicts. The clinic and the petri dish are telling different stories, and I have learned which one to trust with a life.

If the ketone is the enemy, so is fasting

Here is the part that should give us pause, and it’s the part almost no one following this logic to its end wants to sit with.

If endogenously produced ketones are the danger — if the mere presence of β-hydroxybutyrate is what “arms” these cancers — then everything that raises your own ketones is implicated. Fasting raises ketones. Time-restricted eating raises ketones. Carbohydrate restriction raises ketones. Exercise raises them. A long night’s sleep raises them. The nursing newborn, living on its mother’s high-fat milk, is in ketosis by design.

You cannot warn people away from ketones and prescribe them a fast in the same breath. It’s one or the other. Either the body’s own production of this molecule is dangerous — in which case intermittent fasting, the darling of nearly every metabolic protocol, including the protocols of the very people raising the alarm, is dangerous too — or the ketone was never the problem. The machinery the cell hijacked to make and use it was the problem. Those are not the same thing, and the whole confusion lives in the gap between them.

There’s a cousin to this fear worth naming, because it comes up in my messages almost as often: the worry that fasting switches on autophagy — the cell’s recycling program — and that a clever cancer can hijack that recycling to feed itself. It’s a real mechanism, and autophagy genuinely is double-edged; the biology can protect a healthy cell and, in some late, stressed contexts, protect a malignant one too. So I hold the concern with respect. But then I go looking for where it actually plays out in humans, and I keep finding the opposite. The weight of the fasting and fasting-mimicking research runs toward sensitizing cancer to treatment while shielding healthy tissue, boosting the immune response, and improving how people tolerate chemotherapy — not toward feeding tumors. In three decades and dozens of carefully held extended water fasts, my own among them, I have watched measurable tumor debulking, not acceleration. Not cure — I don’t use that word, and I won’t start now — but real support, always best when it is layered with other interventions at the right time, the right dose, the right duration, and the right combination. And if I genuinely worried autophagy had gone rogue in someone, I wouldn’t guess at it from a headline; I’d watch the same real signals I always watch — the markers, the imaging, the inflammation, the person in front of me — and let the body tell me. Nothing here is absolute. Timing and terrain are everything. A tool that heals in one window can do nothing, or harm, in another. That is not a reason to fear the tool. It’s a reason to learn to read the window.

Photo by Daniele Levis Pelusi on Unsplash

We don’t blame the smoke for the fire

Let me borrow a logic we already half-accept, and then let me sharpen it, because the usual version gets it backward.

We block hormones all the time in the cancer world. Tamoxifen, the aromatase inhibitors, androgen deprivation — an entire architecture built around shutting a signal down. And the story we tell ourselves is that the hormone went bad, so we silenced it. But I don’t read estrogen as the villain, and I am no fan of simply flipping the switch off. I see a disordered estrogen picture — the elevated level, the skewed metabolites — as the expression of a problem, not the problem itself. It is smoke. The real questions live underneath it: what lit the fire? What drove the metabolism to stumble in the first place, and how do we correct that? To attack the smoke while ignoring the fire is to mistake the smoke detector for the arsonist.

Nuance is the whole point here, so let me hold both hands open. Nothing is absolute. There are people, with the right genetics and the right situation, for whom shutting a signal down is a precise and useful scalpel for a season. I’ve watched that work. But it is a scalpel, not a worldview, and it should never be the reflex. The reflex — the one I keep fighting — is to name the messenger as the menace and stop looking. We do it with estrogen. We do it with the calcium we watch pile up in an artery. We do it with the iron a body sequesters when it’s inflamed. And we are doing it right now with ketones. Smoke, smoke, smoke, and everyone busy blaming the smoke while the fire burns on, untended.

So read ketones the same way. β-hydroxybutyrate is not a poison a tumor cleverly weaponizes. It is one of the oldest signaling molecules we carry. It quiets the NLRP3 inflammasome, one of the central engines of the chronic inflammation that feeds nearly every hallmark of cancer. It behaves as a histone deacetylase inhibitor, changing which genes get read. It leaves its own epigenetic signature on our chromatin — β-hydroxybutyrylation, a mark we only discovered we carry a few years ago. It speaks to cells through its own surface receptor. And in a colorectal model published in Nature in 2022, that very same β-hydroxybutyrate suppressed tumor growth, through the Hcar2 receptor and the Hopx regulator, in mice and in human tissue alike.

Think of a ketone as rain. The same rain that greens a pasture will flood a town that got built in the wrong place. The rain didn’t turn evil somewhere between the field and the town. It simply met different ground. β-hydroxybutyrate is rain. It falls on the terrain it’s given. Ask about the ground, not the weather.

Same molecule. Opposite outcome. Read that twice. If a single metabolite can help arm a leukemic stem cell in one context and suppress a colon tumor in another, then the molecule was never the story. The context is the story. The terrain is the story. It always was.

Corn-oil mice and petri dishes

Now the question I keep putting to the alarm-ringers, and keep not getting a straight answer to: where is the human harm?

Because nearly all of the “ketones feed cancer” mechanism lives in two places — the petri dish and the mouse. And we need to talk about that mouse.

The standard rodent “ketogenic” diet is not what you’re picturing. It’s a fixed industrial formula. The most common one is built on lard, butter, and corn oil in a roughly six-to-one fat ratio, with powdered cellulose standing in for fiber and not a single phytonutrient in the bowl. Corn oil alone is a heavy dose of omega-6 linoleic acid, inflammatory on its own terms; some formulations lean on hydrogenated vegetable shortening outright. This is the “keto” that fattens rodent livers and drives the very inflammation these studies then turn around and measure. A mouse eating industrial seed-oil chow, continuously, with no fiber and no plants and no metabolic rhythm to its days, is not a human being eating wild fish, olives, avocado, nuts, pastured fat, and a wide base of low-glycemic vegetables. Calling both of them “the ketogenic diet” is like calling a gas-station hot dog and a wild-caught salmon dinner both “protein” and pretending the two will do the same thing to a body. What lights me up — what I cannot let pass — is how casually we’ll scare the shit out of a frightened family on the strength of a mouse fed corn oil.

And when we finally do look at humans, the picture doesn’t just soften. It inverts.

The strongest human evidence we have to date is a randomized Phase II trial in metastatic pancreatic cancer — one of the very ketone-utilizing cancers this alarm is built around. A medically supervised ketogenic diet, added to chemotherapy, improved progression-free survival, improved overall survival, and did it with no added toxicity. In the small leukemia pilot people point to, the diet was well tolerated and the leukemic cells showed more DNA damage, not less. I have read these papers looking for the harm. It is not there. What’s there is signal, tolerability, and — in the disease that was supposed to be the worst case — longer life.

Photo by Immo Wegmann on Unsplash

Most cancers are sugar sinks — so start there

Here is what gets lost while we argue about ketones: for most cancers, the fight was never really about ketones at all. It’s about sugar. And let me qualify that before anyone runs with it, because I refuse to trade one cartoon villain for another. Sugar does not cause cancer. Sugar is smoke too. But chronic, runaway sugar does something the terrain feels everywhere: it dims immune function, pours fuel on inflammation and oxidative stress, scrambles hormonal signaling, drives the obesity and insulin resistance that make a body hospitable to disease. It doesn’t light the match. It soaks the kindling. And it has gone off the rails for almost all of us — only about seven percent of American adults are in anything like optimal metabolic health, which means the other ninety-plus percent are walking around with a terrain that could use tending. That is not a fringe intervention. That is the majority report.

Now, the great majority of cancers are glucose-avid — sugar sinks, glycolytic engines that guzzle glucose and burn it fast and wasteful. This is the Warburg effect, and it is so reliable that we built an entire imaging technology on it. Hand the body a pinch of radioactive sugar, wait, and photograph where it piles up; that glow on a PET scan is the tumor’s sweet tooth, lit up for the camera. Estimates vary and I won’t pretend at false precision, but the glycolytic phenotype shows up in something on the order of three-quarters of cancers, which is exactly why FDG-PET is a near-universal cancer-imaging tool. A minority don’t glow the same way — some prostate, some lobular breast, some kidney and neuroendocrine tumors — and that exception is itself a clue worth reading. It doesn’t mean those tumors ignore the sugar lever; in my experience they respond to sensible carbohydrate restriction too. It means the degree differs, and degree is a thing you measure, not assume.

So here is what I actually do, with every single client, and it is not “go keto.” It is carbohydrate restriction, titrated to the person, and then I let their labs tell me how strict or how loose we can be. That one practice does two things at once. It turns down the fuel most cancers prefer — a broad, layerable support that tends to make almost every other therapy work better. And it reads the terrain in both directions: how the tumor responds tells me how rigid or flexible the cancer is, and how the person responds tells me how much metabolic flexibility the human still has to work with. The labs are the conversation. The body votes.

Notice what I did not say. I did not say strict ketogenic diet. I did not say force everyone into deep ketosis. Carbohydrate restriction, nutritional ketosis, and a branded ketogenic diet are three different things, and collapsing them into one word is how this whole conversation got dumb. To a fifth grader: most cancers have a sweet tooth and run hot on sugar, so turning the sugar down dims the furnace’s favorite fuel — that’s the baseline, and it helps almost across the board. Everything after that — how far, how deep, layered with what, for how long — is individual, and it’s decided by measurement, not by a headline and not by a brand.

What I am not saying

I am not telling you ketones are universally therapeutic, or that everyone carrying a cancer diagnosis should be in nutritional ketosis, or that a diet is ever the whole answer. There are tumors that appear truly able to burn ketones as fuel, and the mechanistic work describing them is careful and worth respecting. A ketone-utilizing tumor is a real category. And diet as a solo act is never the story in any cancer. It is one lever among many, and I have been saying that for as long as I’ve been doing this.

Let me define the thing that decides it, because this is where the real answer lives. Whether a cell can burn a ketone at all comes down to an enzyme called SCOT — its gene name is OXCT1. Picture SCOT as a can opener. A ketone is a sealed can of fuel; without the opener, the cell can’t get in, and the ketone drifts right past, useless to it. Most healthy tissues keep the opener in the drawer. Cancer is where it gets interesting. Some tumors run the opener low — glioblastoma is the classic example, and that ketolytic weakness is part of why a ketogenic state has earned a place there. Others keep the opener, or forge extra copies of it. You cannot know which kitchen you’re standing in by guessing. You have to open the drawer and look.

And carrying the tool is not the same as using it. This is the trap we fell into with MTHFR — a whole generation panicked because they carried the gene variant, as though the gene itself were the diagnosis. But a gene you carry is not a gene that’s expressing. With MTHFR we don’t convict on the genotype; we look at homocysteine, at family history, at what the body is actually doing right now, and only then decide whether it’s expressing and needs addressing. A tumor holding a copy of the ketone can opener is the same story. Possession is not proof of use. You confirm it downstream, in what the tissue actually expresses and what the person’s biology actually does, before you let a strand of DNA write anyone’s ending.

Which is where the alarmists overreach, and where I refuse to overreach back. This is not one number. When people ask me what percentage of cancers “use ketones,” I tell them what’s true: there is no clean figure, and I went looking hard for one. Anyone who hands you a tidy percentage is selling a certainty the science has not earned. And the human record runs the opposite direction from the fear — everywhere ketogenic therapy has actually been studied in these ketone-utilizing cancers in living people, the pancreatic trial and the leukemia pilot among them, it was tolerated and it helped. Not a single human study shows ketosis driving these cancers in an actual patient. The alarm is built almost entirely of cells and mice. OXCT1 runs low in glioblastoma and, variably, in colorectal cancer — the ketone-vulnerable end. It often runs high in liver, lung, prostate, and pancreatic cancers — the ketone-utilizing end. And it varies patient to patient inside a single diagnosis. That is not a weakness in the argument. That is the argument. There is no population answer. There is only this tumor, in this person, measured.

Which brings me to the thing I most want you to hear, because it’s the whole reason I can stay calm while the headlines shriek. Naming a cancer as “ketone-capable” is not a sentence. It is a variable. In my own practice, across the very tumor types I just listed as the ketone-utilizing end — the leukemias, the pancreatic cancers, the liver and lung and prostate cancers — I have used ketosis, effectively, as part of the recovery strategy. Not by ignoring the biology. By respecting it. The difference is the approach: never ketosis as a lonely weapon flung at a tumor, but ketosis reached by whatever road fits the person — sometimes clean whole-food fats, sometimes simply fasting, sometimes carbohydrate restriction alone, and in select cases exogenous ketones used to potentiate another therapy like radiation or hyperbaric oxygen — timed and dosed and calibrated by their labs, and, this is the non-negotiable, deployed inside a fully tended terrain rather than on top of an untended one. A ketone-utilizing tumor in a neglected terrain may well take the fuel. That same tumor, in a terrain where you’ve closed the other doors, meets a state it can no longer exploit. Same molecule. Different room. The tumor’s ability to use ketones tells you how carefully to build the room. It does not tell you to burn the house down.

And sit with the irony a moment, because it’s a good one. The exogenous ketones people fear most — the supplement, the ester, the exogenous BHB — are the very things that, in animal models, have lowered tumor-cell viability and, paired with hyperbaric oxygen, extended survival. The molecule cast as the villain keeps showing up in the data behaving like an ally. That should make us curious, not comfortable, and certainly not afraid.

So how would we actually know, for a given human being, which room we’re standing in?

How we’d actually know — the terrain answer

This is the part where I get to be a doctor instead of a critic.

If someone wants to claim that ketones are feeding a particular person’s cancer, that is a testable claim. Not a vibe, not a diagram — a claim you can put a metric on. Here is what I’d want to see before I’d take it seriously for the human being in front of me. Think of it as three questions: what is the tumor built to burn, what can this body safely do, and what is actually happening over time.

But before any of the three, there’s a gate almost everyone walks straight past: was the person even in ketosis to begin with? Most people who believe they are “in ketosis” are not. You cannot know it by how you feel, and urine strips and breath meters drift and lie; you confirm it in blood, by measuring β-hydroxybutyrate, and you follow it with the glucose-ketone index — the GKI, blood glucose divided by blood ketones, the single number that tells you what fuel a body is actually running. The typical modern pattern sits far up in the double digits. Meaningful nutritional ketosis lives down in the single digits, and the therapeutic window studied in metabolic cancer care runs lower still, often somewhere between one and three. That is a sharpened-pencil state, and it is nowhere near where most people casually “doing low carb” actually live. So when someone tells me, “I did keto and my cancer progressed,” I don’t hear proof that ketones are dangerous. I hear a missing chart. What did you actually eat, at what time of day, under what emotional weather, and what — if anything — did you measure? Nine times in ten there was no measurement, and no real ketosis either. In three decades I have never once watched ketosis itself fail a person. I have watched the terrain fail, and I have watched a physiologic state — not a diet, a state — never be given the conditions to land.

First, the tumor’s own machinery — that can opener. Does this cancer carry it? You read that off the tumor, not off the person, and you read it best with tools we already have. Comprehensive tumor sequencing that includes RNA — whole-transcriptome profiling, the kind Tempus runs as xR alongside its xT DNA panel, or Caris runs as whole-transcriptome — can show you the actual expression of the ketone machinery: OXCT1 and BDH1 on the using side, HMGCS2 on the making side. The same profile shows the glucose machinery running hot or quiet — GLUT1, hexokinase — and the lipid machinery, CPT1A and FASN, that some tumors lean on instead. And the DNA drivers that come standard on these panels are fuel tells in their own right: a KRAS or PIK3CA tumor is usually a sugar story. The point is that this is knowable, per tumor, from tissue — and that a plain DNA mutation panel alone won’t answer it, because expression is an RNA question. If you want to know what a tumor is actually burning, ask for the transcriptome, not just the mutation list.

And here I have to correct a hope I get asked about constantly, because getting it wrong is exactly the sort of thing that will get thrown back at us. You cannot read your tumor’s ketone habit off a spit-in-a-tube consumer DNA kit, and you can’t bolt it onto your genomic profile as if it were eye color. Think of it as two different documents. Your germline DNA is the blueprint you were handed at birth, copied into every cell of the house. The tumor’s genome is a diary it started keeping later, in secret, after it went rogue. There is a germline OXCT1 test, but it exists to catch a rare inherited enzyme deficiency that sends infants into ketoacidosis — it reads the blueprint, and it says nothing about what this particular arsonist is burning now. The tumor’s ketone use is somatic, written in the diary. Read the diary. It comes from tissue, not saliva.

That doesn’t make the blueprint useless — it just answers the second question. Not will the tumor eat ketones, but can this person safely make and clear them, and how will their body respond. This is where germline SNPs earn their keep, and where I’d wave a caution flag before putting anyone into deep ketosis. A whole family of inherited errors in fat and ketone handling are hard contraindications to a ketogenic state: the fatty-acid oxidation disorders — CPT1 and CPT2, VLCAD, MCAD, LCHAD, the carnitine-transporter defect OCTN2 — along with pyruvate carboxylase deficiency and the ketone-making and ketone-clearing defects themselves. If a body physically cannot burn fat or break down ketones, flooding it with both is not therapy, it’s a crisis. Softer flags live in the lipid genes: APOE4, APOB, and the familial-hypercholesterolemia variants can send LDL climbing on a high-fat diet — the so-called lean-mass hyper-responder. Whether that rise carries the danger conventional cardiology assumes is, right now, a genuinely open and lively argument; the work of Dave Feldman, Nick Norwitz, and Philip Ovadia is pressing hard on the old certainties, even as the newest plaque-imaging data suggests the answer is individual and worth watching closely rather than dismissing in either direction. Either way it is a cardiovascular conversation, not a cancer one — and a real reason to watch the labs and loosen the fat rather than white-knuckle a strict protocol. Nothing here is one-size, including the fat. Your birth blueprint tells you whether the house can safely run a fire at all; only the tumor’s diary tells you what this particular fire is eating. None of these inherited flags tell you the tumor will use ketones. They tell you whether this human can wear the intervention. That distinction is the whole game, and it’s the one the headlines never make.

Second, a word about what imaging can and can’t do, because I want to name a real limitation rather than paper over it. The FDG-PET we scan with lights up glucose avidity, not ketone use. There is no validated, clinical “ketone PET” a patient can go and get. So anyone showing you a standard scan as proof that ketones are feeding a tumor is showing you something the scan cannot actually say. Say that out loud, and a great deal of false certainty falls away. What I will say, because it matters and because it’s coming, is that the imaging world is moving fast in exactly this direction. Glutamine-based PET tracers are already in human trials, letting us watch a tumor’s appetite for that fuel. Deuterium metabolic imaging is teaching us to see how tissues actually burn what we feed them. Amino-acid tracers are in use for particular cancers today. We are, slowly, building the eyes to watch metabolism live instead of arguing about it in the dark. We’re not there for ketones yet. Say so, and keep watching.

Third, and to my mind the most important, the individual’s own dynamics over time — the third question, and the one no lab draws for you in a single sitting. This is where the lived data lives, and where I will defend the so-called anecdote against anyone who tells me it has no place in science. Track the person. Their β-hydroxybutyrate, yes — but alongside their tumor markers, their imaging cadence, their inflammatory markers, their glucose, their glucose-ketone index, their symptoms, their vitality, their color when they walk in the door. If a person’s disease truly tracks with rising ketones — markers climbing, imaging advancing, in a tumor you’ve shown carries the machinery to use them — that is a real signal, and I would reassess the dietary lever for that person without hesitation. But if it doesn’t — and across a great many people, it doesn’t — then the fear was theoretical, and the terrain just told you so. That is Test, Assess, Address, Don’t Guess. Not guess, and not fear, and not extrapolate from a rodent. Measure the actual human.

Because here is the thing we keep forgetting. We were built to be hybrid engines. We are designed to move fluidly between glucose and fat, between sugar and ketone, the way our ancestors moved between feast and famine, summer and winter, feeding and fasting. The disease of modernity is not ketones. It is the loss of that flexibility — terrain gone so metabolically rigid it can only run on one fuel. The question was never “ketones: good or bad?” It is “does this terrain, this tumor, this person still have the capacity to use ketones as fuel and as signal in the service of healing — and where along that spectrum do they sit today?” That is the whole art. It cannot be answered by a meme.

Photo by Ivan Kazlouskij on Unsplash

Nothing works in isolation

And even that is too small a frame, so let me widen it one more turn, because if I let you walk away thinking ketones, measured carefully is the answer, I’ve just handed you a smaller headline to hide behind.

There is no such thing as a metabolic input acting alone. The same ketone, the same fast, the same lowered carbohydrate lands in a terrain already shaped by a dozen other forces, and those forces decide what the input becomes. A person’s state of mind and the stress chemistry underneath it. Their COMT status, quietly governing how fast they clear the stress hormones and estrogens the rest of this depends on. Their smoldering inflammation. Their toxic burden — the exposome, the lifetime of exposures the liver is still negotiating. The co-infections and the immune dysregulation riding along with them. Their hormonal signaling. Whether the tissue is well-oxygenated or starved into hypoxia, which flips a tumor’s metabolism all by itself. Change any one of these and you change how the terrain answers any input you give it — including a ketone. This is why two people can eat the identical diet and get opposite results, and why the study that averages them learns almost nothing about either.

At the center of all of it sit the mitochondria, and this is the part I most want to land. We were taught to think of them as batteries — little power packs cranking out energy and nothing more. They are so much more than that. Mitochondria are the body’s listening posts. They receive information from every one of those signals — the stress, the oxygen, the toxins, the hormones, the fuel on the fork — they translate it, and they signal back out to the cell what state the world is in and how to respond. They are not the engine. They are the switchboard. When people ask me what “terrain” really means, this is the answer: it is the whole conversation a body is having with itself, with the mitochondria sitting at the switchboard translating every line. A ketone is one word spoken into that conversation. Whether it heals or harms depends on everything else being said in the room.

So this was never a one-pony show, and anyone selling you a single villain or a single hero — ketones bad, ketones good, sugar bad, fat bad — is selling you a pony. The body doesn’t work that way. It never did.

Photo by Christian Harb on Unsplash

Two can tango

Here is what I’ve come to respect about this disease, and it’s the reason single-lever thinking fails every time.

Cancer is elusive. It hides, it dims its own signals, it slips out of the frame just as you focus. Cancer is resilient. It survives the insult you were sure would end it. And cancer is resourceful — this is the trait the whole ketone panic actually stumbled onto without understanding it. A tumor that learns to build its own ketone still in the basement is not proof that ketones are evil. It is proof that cancer is inventive. Close the sugar door and the clever ones reach for fat. Close that and they reach for glutamine, or they make their own fuel, or they go quiet and wait. Pull one lever in isolation and a resourceful adversary simply routes around it — the ultimate game of whack-a-mole, where you slam one metabolic pathway down and another pops up grinning across the board. That is not a reason for despair. It’s the single most important clue we have about how to actually win.

Because here is the other half of the sentence, and it’s the half the fear-mongers always forget: so are we. We are elusive, resilient, and resourceful too. We are the most metabolically adaptive creatures on the planet, built to shift fuels, mount immune responses, detoxify, repair, and regenerate. The tumor gets to be clever, and so do we. It takes two to tango — and this is a dance, not a single punch thrown in the dark.

But you only get to lead the dance if you tend the whole terrain. This is the part I cannot say loudly enough. If you pull only the nutrient lever — just the diet, just the fast, just the ketones — and leave the inflammation smoldering, the toxins circulating, the stress chemistry roaring, the hormones scrambled, the immune system asleep, the mitochondria unsupported — you will not have good success, and the disease will route around your one clever move the way water finds the crack in the dam. You cannot out-diet an untended terrain. A resourceful cancer met with a single tactic will always, eventually, find the door you left open. The only thing I have ever seen genuinely change the dance is closing the doors together — many levers, tended at once, timed and layered and adjusted as the body answers back. Meet resourcefulness with resourcefulness. That is the whole game. That is the tango.

Reading the next scary headline

So the next time the current shifts and a new fear comes floating by — and it will, it always does — here is how I would have you read it, before it gets to tilt your floor.

Ask whether the claim is about ketones the body makes or ketones you eat, because those are two different universes. Ask whether it is talking about the molecule or the signal — whether the ketone is the villain, or just the messenger carrying a message the cell learned to forge. Ask whether the evidence is a petri dish, a mouse on industrial chow, or an actual human being, and whether it measured a mechanism or an outcome that someone lived. Ask whether the diet was tested alone or in context, in a population or in a person. And ask, always, compared to what, in whom, measured how? Run any frightening claim through those questions, and most of the panic resolves into the only real answer medicine has ever had: it depends — and here, precisely, is what it depends on.

Yes, and…

I’ve come to distrust almost every sentence about cancer that runs on the word or. Ketones good or bad. Diet or medicine. Sugar or fat. Science or the lived experience of the person in the chair. The whole field keeps trying to force a binary onto a body that has never once behaved like one.

The truer word is and. Ketones can starve one tumor and feed another. A fast can debulk a mass and be the wrong move six months later for the same person. Standard of care can be the right backbone and leave the terrain untended. The mechanism in the mouse can be real and fail to describe the human. I hold all of it — yes, and. Not because I can’t make up my mind, but because the body is holding all of it too, every second, and the moment I collapse that into a slogan I stop seeing the person.

This is why the work is never finished and never black and white. It is nuanced, it is individual, it is dynamic, it is iterative. We try something, we measure, the terrain answers, we adjust — and the answer we get is not a theory borrowed from a rodent but real-world data from a real human life unfolding in front of us. That loop is the whole method. It’s slower than a headline and less satisfying than a villain, and it happens to be the only thing I’ve ever seen actually work.

Photo by Jayne Harris on Unsplash

The messenger

The ketone is a messenger. One of the oldest we carry. A molecule of scarcity and survival that carried our species through every winter it had no business surviving, that feeds the newborn brain and the fasting heart, that quiets our fires and edits our genes and, in the right terrain, helps a body find its way back toward order.

When a cancer cell learns to forge that messenger’s handwriting to save its own skin, we do not answer by burning every letter in the language. We learn to read the forgery. We test the tumor. We measure the person. We watch what the terrain actually does, in this body, over this stretch of time. We hold the mechanism and the human being in the same hand, and we refuse to let either one bully the other.

I know a little about being read by your label instead of your contents. It taught me to distrust the headline and open the jar. Do that here. Ketones are not the enemy of the body any more than I was ever only a diet. Read past the word. The truth was always in the terrain underneath it.

And here is the quiet thing under all of it, the thing I most want to leave in you. We know far less than the confident voices pretend — about ketones, about cancer, about the staggering intelligence of a body that has kept itself alive through every fuel shortage and long winter our species ever met. I’ve made my peace with that not-knowing. It is not a threat. It is a doorway. It is the reason to stay curious, and humble, and awake to the person in front of me rather than the headline behind me. Wonder and rigor are not opposites. They are how you keep from getting fooled — by the fear, and by yourself.

We don’t shoot the messenger. We get curious about the message.

That is the work. It always was.

None of this is medical advice, and none of it is a protocol. It is an invitation to think more carefully — to trade certainty for curiosity, and fear for measurement. If you are walking a cancer path, walk it with a clinician who will test and assess before they tell you what to guess.

For the reader who wants the receipts

The human pancreatic trial: Jameson GS, Roe DJ, Borazanci E, et al. A randomized phase II trial of gemcitabine, nab-paclitaxel, cisplatin with or without a medically supervised ketogenic diet for patients with metastatic pancreatic cancer. Cancer, 2026. doi:10.1002/cncr.70343.

β-hydroxybutyrate suppressing tumor growth: Dmitrieva-Posocco O, et al. β-Hydroxybutyrate suppresses colorectal cancer. Nature, 2022;605:160–165. doi:10.1038/s41586-022-04649-6.

Ketones as signals: Newman JC, Verdin E. β-Hydroxybutyrate: A Signaling Metabolite. Annual Review of Nutrition, 2017. And Youm YH, et al. The ketone metabolite β-hydroxybutyrate blocks NLRP3 inflammasome-mediated inflammatory disease. Nature Medicine, 2015;21:263–269. And Shimazu T, et al. Suppression of oxidative stress by β-hydroxybutyrate, an endogenous histone deacetylase inhibitor. Science, 2013;339:211–214.

The endogenous, tumor-made ketone mechanism (the real work, fairly described): Chen et al. A ketogenesis–ferroptosis axis maintains leukemic stem cell survival. Cell Stem Cell, 2026. And the pancreatic OXCT1 / ketone-utilizer biology: OXCT1 high expression in PDAC and gemcitabine resistance, Frontiers in Oncology, 2021 (PMC8479148).

Why OXCT1 (SCOT) is not one number: The Function and Mechanism of OXCT1 in Tumor Progression as a Critical Ketone Body Metabolic Enzyme, Biomolecules, 2026 (OXCT1 upregulated in hepatocellular, lung, bladder, prostate, pancreatic, ovarian cancers; downregulated in glioblastoma and variably in colorectal). Expression data also viewable on the Human Protein Atlas (proteinatlas.org, OXCT1/ENSG00000083720). Ketolysis as a stress-survival switch: Huang D, et al. Hepatocellular carcinoma redirects to ketolysis for progression under nutrition deprivation stress. Cell Research, 2016;26:1112–1130. The germline test that does NOT read your tumor: OXCT1 (SCOT deficiency) single-gene testing, NIH Genetic Testing Registry — a test for a rare inherited ketolysis defect, not a tumor ketone-use assay.

The insulin / IGF wiring in leukemia and pancreatic cancer: Expression of Insulin Receptor Isoform A and Insulin-Like Growth Factor-1 Receptor in Human Acute Myeloid Leukemia, 2009 (PMC2762752); and autocrine IGF-I signaling promoting AML growth and survival via PI3K/Akt, Leukemia, 2007. Pancreatic adenocarcinoma’s dependence on insulin/IGF and glucose signaling is well documented across the metabolic-oncology literature.

Fasting, autophagy, and cancer therapy: Intermittent fasting enhances cancer therapy via autophagy-dependent and independent mechanisms, World Journal of Clinical Oncology, 2026 (fasting and fasting-mimicking approaches sensitize tumors and protect normal tissue rather than promoting tumor growth); and the broader fasting-mimicking-diet literature on differential stress sensitization and chemotherapy tolerance.

Metabolic health of the population: O’Hearn M, et al. Trends and Disparities in Cardiometabolic Health Among US Adults, Journal of the American College of Cardiology, 2022 (only ~6.8% of American adults in optimal cardiometabolic health); and Araújo J, et al. Prevalence of Optimal Metabolic Health in American Adults: NHANES 2009–2016, Metabolic Syndrome and Related Disorders, 2019 (~12% metabolically healthy).

Emerging metabolic imaging beyond FDG-PET: in-human glutamine PET with 18F-(2S,4R)-4-fluoroglutamine, Radiology, 2017; glutamine metabolic imaging review, Journal of Nuclear Medicine, 2017;58:533; and deuterium metabolic imaging as an emerging tool for reading tissue fuel use in vivo.

Most cancers are glucose-avid (the Warburg effect), which is why FDG-PET works as a near-universal cancer-imaging tool; a minority of tumor types (some prostate, lobular breast, renal, neuroendocrine) are less FDG-avid. See FDG-PET avidity and glycolytic metabolism in NSCLC (PMC7370816) and Warburg-effect reviews in the metabolic-oncology literature.

Reading the tumor’s fuel machinery by tumor sequencing: whole-transcriptome RNA profiling — e.g., Tempus xR alongside the xT DNA panel, and Caris whole-transcriptome sequencing — reports expression of ketone (OXCT1, BDH1, HMGCS2), glucose (SLC2A1/GLUT1, hexokinase), and lipid (CPT1A, FASN) machinery; a DNA-only mutation panel does not capture expression.

Germline contraindications and cautions for a ketogenic state (host, not tumor): The ketogenic diet is not for everyone: contraindications, side effects, and drug interactions, Annals of Medicine, 2025 (fatty-acid oxidation disorders — CPT1/CPT2, VLCAD, MCAD, LCHAD, OCTN2; pyruvate carboxylase deficiency; ketogenesis/ketolysis defects — as absolute contraindications). Lipid-response genetics: Ketogenic Diets Exacerbating Hypercholesterolemia in APOE Variants and APOB Mutations, Journal of Clinical Lipidology, 2022; and the lean-mass hyper-responder LDL phenomenon, Journal of Clinical Lipidology, 2022.

Measuring real ketosis (the GKI): Meidenbauer JJ, Mukherjee P, Seyfried TN. The glucose ketone index calculator: a simple tool to monitor therapeutic efficacy for metabolic management of brain cancer. Nutrition & Metabolism, 2015;12:12 (therapeutic GKI window commonly cited around 1–3); and The glucose ketone index: a proposed quantitative biomarker, Frontiers in Science, 2026. Blood β-hydroxybutyrate is the reliable measure; urine and breath methods are unreliable proxies.

Exogenous ketones behaving as allies in preclinical models: Poff AM, Ari C, Seyfried TN, D’Agostino DP. Ketone supplementation decreases tumor cell viability and prolongs survival of mice with metastatic cancer. International Journal of Cancer, 2014. And Poff AM, et al. The ketogenic diet and hyperbaric oxygen therapy prolong survival in mice with systemic metastatic cancer, PLoS One, 2013; and the combined ketogenic diet + ketone supplementation + hyperbaric oxygen protocol in VM mice, 2015.

The lean-mass hyper-responder / LDL debate (host cardiovascular question, not cancer): Norwitz NG, Feldman D, Budoff M, et al. Carbohydrate Restriction-Induced Elevations in LDL-Cholesterol and Atherosclerosis: The KETO Trial, JACC: Advances, 2024, and its ongoing coronary-plaque-progression follow-up — an actively contested, evolving area; see also the Lipid Energy Model work of Feldman, Norwitz, and Ovadia.

Human ketogenic-diet cancer evidence, in aggregate (feasibility, tolerability, no convincing harm signal): Klement RJ, et al. and Ketogenic diets in medical oncology: a systematic review with focus on clinical outcomes, Medical Oncology, 2020; alongside the pancreatic randomized trial and AML pilot cited above.

The mouse-chow problem: Bio-Serv F3666 ketogenic rodent diet — ingredient composition (lard, butter, corn oil, casein, cellulose; ~6:1 fat ratio).

The human safety picture, in aggregate: systematic reviews and meta-analyses of ketogenic diets in cancer patients consistently report feasibility and tolerability without a convincing signal of harm — e.g., Effects of ketogenic diets on cancer-related variables: a systematic review and meta-analysis of randomised controlled trials, Nutrition Bulletin, 2024.