In oncology we return, again and again, to first principles. The cell is our unit of life and of medicine. When a normal cell becomes malignant, it does not merely divide faster; it eats differently. It hoards glucose, reroutes amino acids, siphons lipids, and improvises when a pathway is blocked. We have learned to poison its DNA, to derail its signaling, to enlist T cells as sentinels.

We have been slower to ask a simpler question that sits at the cell’s kitchen table: What if we change what a tumor can eat?

For a century, metabolism was oncology’s prologue. In the 1920s, Otto Warburg observed that many cancer cells consume glucose voraciously and convert much of that glucose to lactate even when oxygen is plentiful, a seemingly wasteful choice that became a metabolic signature of malignancy. That insight eventually receded into a footnote while genetics took the stage. But tumors are not static genotypes; they are shape-shifters that adapt to therapy by rewiring their fuel lines.

Therefore, if we want longer and deeper responses, the clinic has to treat metabolism as a first-class target. That means moving from one-size-fits-all “cancer diets” that tell nearly every patient to cut “sugar,” avoid white bread and pasta, drink green juices, or adopt alkaline regimens, regardless of tumor type, treatment, or physiology. Instead, oncologists should move to tumor-informed metabolism: interventions matched to the biology of a patient’s tumor, to the drug it is receiving, and to the body in which both reside.

Consider a woman in her 50s with hormone receptor-positive, HER2-negative breast cancer whose tumor carries a PIK3CA mutation. She receives a PI3K inhibitor alongside endocrine therapy. At first, the drug seems to hold; the scans steady, the markers fall. Then, over months, the cancer advances again. Blood work shows the clue: glucose and insulin levels, driven high by the drug’s effect on insulin signaling, have opened a back door for the tumor. The escape is not written in her genome; it is metabolically improvised.

So alongside the drug, she is given a dietary plan that trades sweetened drinks, desserts, and refined starches for slow-digesting carbohydrates, lipids, and proteins, crafted to blunt those insulin spikes, a protocol timed around dosing, calibrated to her physiology, monitored with real-time metabolic biomarkers. The tumor’s escape hatch narrows; the response deepens; the remission lasts longer. What made the difference was not drug alone or diet alone, but the braid of the two into a single therapy.

Metabolism is where ecology meets oncology. A tumor shares and competes for nutrients with its microenvironment. It burns different fuels in the liver than in the lungs. It shifts when a PI3K inhibitor raises insulin, and it shifts again if hyperglycemia follows steroids. Asparagine is dispensable for some cells, essential for acute lymphoblastic leukemia, which is why depleting it works. Serine and glycine can become growth linchpins in certain breast and colorectal cancers. Methionine restriction alters one-carbon flux in ways that make some tumors more susceptible to therapy. None of this is folklore. It is cell biology in the clinic.

Tumor-informed metabolism treats food as information. For a patient receiving a PI3K-pathway inhibitor, this might mean flattening post-meal glucose and insulin peaks. For a colorectal tumor that depends on particular amino acids, it might mean restricting those substrates during a course of chemotherapy or radiation. For a patient wasting away on treatment, it might mean adding calories and protein precisely because loss of weight and muscle would blunt the very therapy we hope will work. It begins with what the tumor uses, where it lives, which drug is acting upon it, and how the host responds. It is time-bound and measurable. It is designed to make a pharmacologic mechanism work better, not to replace it. It is delivered as precisely as a medication, with safeguards for weight, strength, and metabolic health.

In my work as a co-founder of Faeth Therapeutics, my colleagues and I build such regimens prospectively, pairing diets with PI3K/AKT/mTOR inhibitors in endometrial cancer, amino acid-restricted diets in rectal cancer, and scripting each plan around a specific mechanism and treatment window. Our intention, however, is less to advance a single company than to suggest a new template for how oncology might use food as a co-therapeutic instrument. I would welcome more companies designing similarly rigorous nutrition–drug regimens and submitting them to randomized trials, and I believe sponsors and regulators will eventually have to treat diet as a prespecified element of the protocol rather than an unmeasured backdrop.

The field is still in its infancy. There are encouraging signals across model systems in which pathway-directed drugs, such as PI3K inhibitors or chemotherapy/radiotherapy, have been combined with insulin-lowering or amino acid-modifying diets, but there are also failures when diets are generic, prolonged, or divorced from drug mechanism. Many patients lose weight during chemotherapy; some have diabetes; others fast zealously and end up weaker. Tumor-informed metabolism is an antidote to both nihilism and zeal. It treats nutrition as a targeted adjuvant rather than a belief system.

What will it take to make this part of standard care? The studies must be prospective, controlled, and anchored to a drug’s mechanism. Endpoints must be objective: response rates, survival gains, dose intensity preserved, and toxicity reduced. An intervention that shifts these curves deserves to be integrated into routine practice.

But evidence alone is not enough. We will need clinical structures that can hold and act on that evidence.

First, we will need multimodal regimens in pathways such as PI3K that shut down signaling without collapsing the immune response, avoiding the brittleness of single-node inhibitors and the collateral damage of indiscriminate blockade.

Second, we will need precision nutrition, not as lifestyle advice but as therapy: food scripted to complement a drug’s mechanism and to close the metabolic escapes the drug itself can provoke.

Third, we will need what might be called a metabolic operating system, a computational model of metabolism that lets us predict flux, anticipate resistance, and explore combinations in silico before they are carried into patients.

Without the unity of drug, precision nutrition, and model, tumor-informed metabolism remains a hypothesis. With it, medicine begins to see metabolism as a stratum of biology as fundamental as DNA or protein, but more immediate, revealing a cell’s state in seconds rather than years. “Feed the patient, starve the tumor” is not a slogan but a clinical directive, to be written with the same specificity as a chemotherapy order: macronutrient and micronutrient targets, timing, and contraindications.

There will be skepticism. Some will argue that metabolism is too plastic to trap, or that its contribution will be marginal. But oncology has always been built on combinations in which agents acting through different mechanisms, together, produce results greater than the sum of their parts: targeted drugs layered on chemotherapy, immunotherapies paired with radiation, and supportive drugs that preserve dose intensity and keep patients on treatment. If a tumor-informed plan buys three more cycles of a drug before resistance, that is not incidental to the person living those weeks.

Others will raise concerns about equity. They are right. If metabolism becomes a precision tool, it must be delivered as one: covered, accessible, adapted to diverse kitchens and cultures, not relegated to concierge care.

The deeper reason to do this is not tactical but philosophical. If the cell is the unit of life, then metabolism is the first verb in its sentence. We already intervene at the genome and the immunome. We should not ignore the part that feeds both. The clinic is where this becomes more than an idea. It becomes a plan the patient can taste, and a plan the tumor cannot.

Cancer medicine has always advanced by expanding what counts as therapy. We once thought cures would come only from sharper scalpels and stronger poisons. Then we learned to listen to T cells. Now we must listen to the hungers and handicaps of malignant cells, and use them. The next generation of combination therapy will not be drug plus drug alone. It will be drug plus metabolism, food braided with pharmacology, so that a tumor cannot simply sidestep us on a different substrate.

We will still sequence tumors. We will still give the best drugs we have. We will still sit with our patients on the hard days. But we can also do something elemental that does not subtract from strength or dignity. We can feed the person and starve the cancer, on purpose. That is an old idea made new by the precision of our time.

Siddhartha Mukherjee, M.D., D.Phil., is a physician, researcher, and author. A new edition of his Pulitzer Prize-winning book “The Emperor of All Maladies” is now available with four new chapters. He has co-founded several biotechnology and health care companies focused on developing novel cancer treatments, including Faeth Therapeutics and Manas AI. He also serves as associate professor of medicine at Columbia University and as an oncologist at the university’s medical center.