What this covers
Azra Raza, an oncologist at Columbia University and widower of a cancer patient, sits down to discuss the argument of her book The First Cell: that oncology has spent decades and vast resources attacking cancer once it is advanced, where outcomes have barely budged, and should instead redirect roughly half those resources toward detecting the earliest traces of malignancy before it grows dangerous. The conversation moves through the basic biology of what cancer is, why it remains so difficult to treat, and what technologies now make early detection feasible. Raza speaks from both her clinical experience running the Myelodysplastic Syndrome Center at Columbia and her personal experience watching her husband succumb to the disease, lending the discussion an uncommon honesty about the human costs of the present approach.
The conversation covers why current treatments have stalled: chemotherapy regimens like the "7+3" combination for acute myeloid leukemia remain unchanged since the 1970s, and advances in cancer mortality over recent decades have come not from new drugs but from decades-old screening methods and anti-smoking campaigns that have "been milked dry." Cancer as Evolution drives much of the problem—tumors are heterogeneous populations under constant selection pressure, making them moving targets. Raza argues that early-stage disease runs 90–100% curable while advanced disease is essentially 100% fatal for common cancers, and explores emerging detection technologies like cell-free DNA and liquid biopsy that could catch the First Cell vs Last Cell. She also addresses the structural incentives that lock in failing treatments—pharmaceutical pricing, physician liability, and financial incentives distributed everywhere except to patients—and makes the case that animal models, though biologically useful, systematically fail to predict human drug response, yet the FDA still requires them. The conversation turns philosophical toward the end, touching on Peto's Paradox and what it tells us about aging, the distinction between preventing death and denying it, and why prevention is more tractable than chasing advanced disease once it has already evolved.
Raza argues that decades of resources spent trying to kill cancer once it is advanced have largely failed, and that we should redirect roughly half of cancer resources toward detecting the very 'first cell' of cancer through new biomarker and monitoring technologies, because early disease is dramatically more curable.
- Treatments for advanced cancer (chemo, radiation, surgery) are essentially unchanged in 50-70 years and fail for the patients who present with advanced disease
- Early-stage disease is ~90-100% curable while advanced disease is ~100% fatal for common cancers
- New technologies (liquid biopsy, cell-free DNA, wearable scanning) make early detection of the first cell increasingly feasible
Early detection before cancer progresses is dramatically more curable than treating advanced disease.
- Instead of developing ever more sophisticated techniques to identify the 'last cell' (minimal residual disease left after treatment), we should develop techniques to determine 'minimal initial disease'—detecting the first cancer cell as early as possible and preventing the disease, since the 'last cell' is what we chase after treatment and the 'first cell' is what we should catch before progression.
“why not go not for minimal residual disease, but develop techniques to determine minimal initial disease and diagnose it cancer early in fact, and try to prevent it.”
- Patients diagnosed with early-stage disease can be cured 90% of the time, whereas patients with advanced common cancers will essentially 100% die—a dramatic gradient that demonstrates early detection is far more amenable to treatment.
“patients were diagnosed with early stage disease, 90% can be cured, but those who are advanced with advanced disease, 90% or more... 100% are going to die with the common kinds of cancers, really.”
- Because cancer is a moving, evolving target, we have not been able to control advanced cancers; the outcome for a patient with advanced disease today is no different than it was in the 1930s.
“And this is one of the reasons that we have not been able to control advanced cancers until now. The outcome for a patient with advanced disease is no different than it was in 1930s.”
- Many of the same drugs that fail in advanced disease can be far more effective when applied in earlier stages—as shown in chronic myeloid leukemia where the targeted drug is curative while the disease is chronic but useless once it accelerates into acute leukemia.
“as long as it is chronic myeloid leukemia, that drug is all, is curative. But if the disease starts to accelerate, becomes acute leukemia, the same drug is useless. So earlier stages, the same drugs can work.”
- Treating early disease is fundamentally different from treating late disease: killing a million cells rather than 150 billion means a targeted laser could replace radiation, and if cancer-specific biomarkers existed, immune killer cells could be armed to distinguish cancer from normal cells—unlike current CAR T therapy, which kills every cell expressing the marker.
“let's say you have a million cells to kill rather than 150 billion cells to kill, there's a big difference.”
- We should treat the human body like a machine and constantly monitor it for the appearance of disease-caused network perturbations—Raza is working with biomedical engineers on a chip implanted under the skin from birth to death that continuously monitors, replacing at-best-annual screening so intervention can occur the moment disease appears.
“I'm working with biomedical engineers at Columbia University on a chip that we want to implant under the skin from birth to death which constantly should be monitoring.”
Current cancer treatments use antiquated methods that harm most patients and provide minimal survival benefit.
- About 68% of cancers diagnosed today are cured, but they are cured with essentially the same 'slash, poison, burn' methods—surgery, chemotherapy, radiation—used 60-70 years ago, while the remaining 32% who present with advanced disease have outcomes as bad as 50 years ago.
“the good news is that 68% cancers, we are... That are diagnosed today are cured. But cured with what? The same things we were using like 60-70 years ago. "Slash, Poison, Burns", surgery, chemotherapy, radiation therapy.”
- Acute myeloid leukemia is treated in 2019 with the exact same two-chemotherapy combination, known as '7+3', that was used in 1977, with the same side effects and the same dreadful results decade after decade.
“Today in 2019, I'm using the exact same combination of two chemotherapy drugs, popularly known as 7+3 to treat acute myeloid leukemia. I was using 7+3 then, I'm using it now.”
- About 95% of the drugs brought to the cancer patient's bedside fail, and even the 5% that are FDA-approved as great successes typically prolong survival by only a few months for a small minority—so of 100 patients given a drug, 20-30% respond briefly, 70-80% never respond, and 100% suffer physical and financial toxicity.
“we are being bringing drugs to their bedside which have a failure rate of 95%. And the 5% that are FDA approved and are ostensibly great successes, they should have failed because they are only prolonging survival by a few months for a small minority.”
- Oncologists continue to administer drugs like 7+3 that they know will harm 80-100% of patients because key opinion leaders have set treatment guidelines, and deviating from those guidelines opens the physician to legal challenges—so harmful but standardized treatment is enforced by liability.
“If I deviate from doing that, then I'm opening myself to legal challenges.”
- Raza's husband Harvey did not die of his cancer but because the horrible chemotherapy given to him wiped out his immune system—an illustration that current aggressive treatment can itself be the lethal agent.
“Harvey did not die of his cancer, he died because we wiped out his immune system with the horrible chemotherapy we gave him.”
Cancer drug development relies on animal models and platforms whose results do not translate to human outcomes.
- Animal models are excellent biologic tools for testing intricate questions about biology, but they fail specifically for drug development: you cannot create a tumor in an animal, treat it with a drug, and assume the same drug will work the same way in a human tumor, because the cross-species flip-flop of toxicity and efficacy information does not hold—yet the FDA still requires mouse testing.
“animal models are excellent biologic tools where we can test a large number of very intricate questions about all sorts of biologic phenomenon... The only thing I object to is you can't create a tumor in an animal treated with some drug and think now when you use the same drug in the human tumor, it's going to have the same results. So for drug development, animal models have not been helpful.”
- If 95% of cancer drugs fail at the bedside, the root cause is the failing development platform: drugs are tested mostly in mice, petri dishes, or artificial in-vitro conditions whose results repeatedly do not extrapolate to humans, so we should take off the blinders and admit cancer is too complicated to fix by chasing intracellular signaling pathways.
“What is that failing platform where we are even developing these drugs to bring to the bedside? Mostly in animals, or in in-vitro petri dishes or in extremely artificial conditions that we keep insisting will be extrapolated to humans. Well, over and over, it's been shown that doesn't happen.”
- The best model for a human's disease is that same human's tissue; rather than testing on mice or on whole human beings, we can and should interrogate individual human cells and tissues directly—an approach not yet seriously pursued.
“somebody said the best model for a cat is another cat preferably the same one.”
Cancer emerges from accumulated cellular and systemic instability with age, not single mutations.
- Using Per Bak's sandpile model, cancer can be understood as the collapse of an unstable system: the final grain of sand that causes the pile to collapse is no different from the thousands before it, so it may not be a single mutation that causes cancer but the accumulated instability of the aging body that finally tips over.
“if you are dropping grains of sand, it forms a little pile. But the sandpile, after you keep dropping the sand, will eventually collapse. The last grain of sand, which caused the pile to collapse, was no different than thousands of grains that came before it, it was the pile that had become unstable.”
- Cancer becomes more common with age not merely because one is older but because the per-year likelihood increases through a convergence of factors captured by the acronym MIST: Mutations accumulate per cell, the Immune system grows decrepit, Senescent cells accumulate and produce inflammatory garbage that poisons the surrounding tissue/soil, and Topographic increases in distance between controlling and target cells weaken normal regulation.
“I actually call it "the MIST of aging" to remember it easily. It's an acronym, MIST. "M" stands for mutations”
- Senescent cells will never become cancerous themselves, but by accumulating with age and producing pro-inflammatory cytokines they poison the surrounding tissue (the 'soil'), weakening normal cells and giving any slightly abnormal cell a selective growth advantage—so the cancer outcome depends on the fitness of the landscape to the seed.
“by poisoning the atmosphere in which they are living, poisoning the soil in which they exist... causing normal cells to have less than robust existence, whereas supporting the proliferation and growth of something which is even slightly out of balance but has a selective growth advantage in that setting.”
- While some cancers come from specific events, most are random; at least 60% of all cancers happen because of DNA copying errors as cells divide, since each division copies about three billion base pairs and retains three to four new mutations.
“And probably I would say at least 60% of all cancers happen because of DNA copying errors as the cell is dividing.”
- Fully curing cancer is extremely hard because it is essentially equivalent to curing aging, given how many destabilizing processes converge under the umbrella of aging—so prevention rather than chasing the disease is the more tractable target.
“Of course, it's going to be very hard, like curing cancer is like, you're going to cure aging basically but one thing we can target is instead of always running after disease and treating disease preventing it.”
- Cancer evolves under the same Darwinian constraints as life itself: a founder/driver mutation in a vital gene (e.g., a tumor suppressor controlling proliferation) starts the process, but each subsequent division adds passenger mutations, so that after ~30 doublings the tumor mass may share nothing but the founding mutation with the original cell—making cancer a constantly metamorphosing, selectively evolving collection of cells.
“I think of cancer as something which tends to evolve like life itself. The same constraints and the same principles that apply to evolution apply to initiation, expansion, growth of a cancer also.”
Recent advances in cell-free DNA detection and monitoring enable early cancer identification before clinical disease.
- The company GRAIL announced it can not only reliably detect cell-free DNA 99% of the time but also identify which organ it is coming from—pancreas, liver, lungs—based on the kind of mutations, made possible because the human genome was sequenced and we now know what normal DNA should look like.
“it was announced by the company GRAIL, which is really going after early detection of cancer, they announced that not only can they detect cell-free DNA reliably, 99% of the time, they can also say which organ it's coming from.”
- The discovery that a fetus sheds cell-free DNA into maternal blood (cffDNA)—replacing amniocentesis with a few drops of the mother's blood to assess fetal health—has been adapted to cancer: when cancer starts it sheds cell-free DNA (cfDNA) into the blood, allowing detection of the mutated cancer DNA.
“When cancer starts, it also sheds off cell-free DNA, cfDNA, into the blood”
- A nascent cancer divides faster than normal cells, needs more nutrition, and attracts more blood supply, making the area literally hotter—a 'hot spot' detectable by scanning (e.g., a smart bra with 200 nano sensors detecting temperature and circadian rhythm changes, or bed sheets that scan overnight); cancers also shed cell-free DNA into the blood detectable by liquid biopsy.
“a cancer when it begins divides faster than normal cells, so it needs more nutrition, it attracts more blood supply, area becomes hot, it can be picked up by scanning techniques.”
- When an early lesion is detected, rather than immediately doing invasive surgery, one can 'yell' at it with ultrasound: hitting the small mass with ultrasound makes it wiggle and shed more protein, which a follow-up liquid biopsy can then detect to confirm a real, potentially lethal presence.
“How do we yell? Using ultrasounds. We hit it with ultrasound, when ultrasounds hit this little mass, it wiggles, it shakes, and they shed off more protein”
- Twenty years ago the first human genome sequencing took 15 years and $2.7 billion; today it costs a few thousand dollars and three weeks, and we are only now seeing the dividends because knowing normal DNA lets us pick up mutations.
“20 years ago, it took 15 years and $2.7 billion to do the first human genome sequencing. Today, it costs a few thousand dollars and three weeks of sequencing.”
Financial incentives throughout medicine favor expensive late-stage treatment over prevention and early detection.
- Instead of the current ~95%/5% split favoring treatment of advanced disease, we should divide cancer resources roughly fifty-fifty—half for present patients and half for developing better technologies to detect cancer early and prevent it from becoming an end-stage monstrosity.
“We should divide up resources into half and half. Half we should be dedicating to the present patients, half to developing better technologies to prevent and to detect early”
- Everyone benefits financially from the current cancer regime except the patient—not just pharmaceutical companies; for example a Band-Aid costing $3.99 for 100 at a drugstore costs $75 each in a hospital, and oncologists administer therapies and excess imaging knowing 80% of patients won't benefit, so blame is diffuse and individual responsibility is unclear.
“My answer is not some people but everyone is benefiting except the cancer patient.”
- In America today, 42% of individuals diagnosed with cancer become financially ruined within two-plus years, imposing a financial toxicity that harms not only patients but their next generation.
“Do you know, Sean, that today in America, 42% individuals who are diagnosed with cancer, 42% become financially ruined by two plus years.”
Cancer mortality improvements stem from prevention and decades-old screening, not from new drug breakthroughs.
- Chronic myeloid leukemia treated with Imatinib is the paradigmatic single success—one gene causing the problem and one magic bullet that targets it—but this model sadly failed to generalize because all other cancers have many genetic mutations and no single drug works.
“one is chronic myeloid leukemia being treated with the Imatinib which is one gene causing a problem and one magic bullet that targets it. That's the one cancer that has been really paradigmatic”
- Cancer mortality has dropped about 26% over the last 30 years not because of any new magic-bullet drug but because of two things: anti-smoking campaigns and four decades-old screening measures (mammography, colonoscopy, pap smear, PSA)—and these have now been milked dry, with mortality currently declining only ~1% a year from better use of existing therapies and earlier imaging.
“mortality from cancer has dropped by 26% in the last 30 years not because we invented some new magic bullet, but because... One, anti-smoking campaigns and two, screening measures.”
Dying patients demonstrate sublime grace and acceptance that physicians should honor rather than pathologize.
- Rather than being pitied for caring for dying patients, an oncologist should be envied, because being present for a human being in their final moments is an immense privilege and the most sublime form of grace—seen in the nobility of endurance these patients show.
“when a human being needs us, then it is actually an immense privilege, Sean, to be there for someone... that is the most sublime form of grace that you can see.”
- Because life is so precious and dying so scary, most patients will choose a treatment even when told there is only a one-in-a-million chance, signing consent forms after being fully informed—which is why responsibility for futile treatment is diffuse and no one can fully assume the blame.
“life is so precious and dying is so scary that most of my patients will tell me even if there's a one in a million chance, what do I have to lose? Let me sign up for it and do it.”
- We wrongly treat dying as a failure and promote only positive anecdotes; in truth dying is not a failure—denying death is—and the acceptance of mortality is what distinguishes humans from Greek gods, manifesting as a sublime grace in patients who protect their living loved ones even as they die.
“I'm getting tired of everyone only promoting the positive anecdote as if dying is a failure. Dying is not a failure, denying death is a failure.”
- The book's purpose is to bring the patient front and center into every discussion of cancer, asking at every point whether the patient benefits or is hurt, no matter how rigorously scientific the discussion.
“the purpose of this book, Sean, is to bring back the patient, front and center, into every discussion we have about cancer... We have to ask the question at every point: Where does the patient benefit from this or is hurt by it?”
- Oncologists call a patient cured only after at least five cancer-free years; before that, even absent microscopic disease, they call it remission because the cancer may recur—but after five years, if it recurs the chances are it is a new cancer rather than the original returning.
“if someone has been cancer-free for five years at least, then we call them cured... after five years, if it recurs the chances are that it's a new cancer.”
- A study found that long-term smokers who repeatedly fail to quit and then finally succeed have an inordinately high rate of subsequently developing lung cancer, with the conclusion being a selection effect: they could finally stop only because something internal—an already-developing cancer—made them able to, illustrating cancer as a silent killer operating at a subconscious level.
“When they finally manage to quit and not go back... an inordinately high number of them end up with lung cancer... the conclusion was that most likely they could stop only because something internal finally...”
- A cancer cell is fundamentally one that has overcome the Hayflick limit—the normal cap of roughly 40-45 cell divisions after which a cell commits suicide or enters senescence—becoming immortalized and capable of an infinite number of divisions.
“And cancer cells overcome this Hayflick limit and become immortalized, that they don't die. So to me, the definition of a cancer is just that, that a cell which is now capable of infinite number of divisions.”
Studying precancerous lesions and building human tissue repositories is superior to mouse model research.
- Because acute myeloid leukemia is too aggressive to cure once established, the best strategy is to stop it before it accelerates by studying its pre-leukemic precursor—myelodysplastic syndromes, a third of which progress to AML—and intervening there; this is why Raza built a longitudinal tissue repository of over 60,000 serial samples since 1984 rather than building a mouse model.
“the best way to handle acute myeloid leukemia is to stop the disease before it becomes so aggressive. Are there patients like that? Yes, it's a whole series of syndromes called myelodysplastic syndromes, which are pre-leukemic, because a third of these patients can develop acute myeloid leukemia.”
- One in five new cancers is diagnosed in a cancer survivor because whatever caused the first cancer is still present—and treatment may even worsen the conditions—making cancer survivors the people at highest risk and therefore most interested in finding the first cell.
“one in five new cancers is diagnosed in a cancer survivor because whatever was there to cause the first cancer is still there.”
- Rather than seeking a single $100 million government grant (which no agency will give to one individual), the 20 million US cancer survivors—those at highest risk and most interested in early detection—could fund the tissue repository if even one million of them gave $10 a month for a year.
“Today, in this country, there are 20 million survivors from cancer. Even if one million of them just give $10 a month for a year, that's it, that will fund the whole tissue repository.”
Individual human tissue is the best model for understanding and testing responses to that person's disease.
- Because every tumor is a heterogeneous, evolving collection of cells, treating cancer as a single disease is like treating Africa as one country; the tumor is a moving target, so drugs effective when a biopsy was taken six weeks ago may no longer work once the cancer has evolved.
“So what I say is that treating cancer as one disease is like treating Africa as one country.”
- Simply adding redundant tumor suppressor genes is not a viable anti-cancer strategy: when animal models were engineered to overexpress or carry multiple copies of p53, the animals did not get cancer but aged dramatically and died of aging.
“when this was replicated in animal models to produce, to overexpress this gene or to have insert multiple copies of these genes, the results were not so great because you know what happened. The animals, number one, didn't get cancer but aged dramatically. And died of aging.”
Cancer incidence patterns are consistent globally but best documented in Western nations with reliable data.
- Cancer incidence is broadly similar wherever it has been studied around the world; apparent differences in numbers arise mainly because some places keep good statistics and others don't, so reliable data exist only for Western countries and a few in Asia and Africa.
“The only difference in numbers that appears is because some places keep good statistics, others don't.”
- In America, 1.7 million new cancer cases are diagnosed in adults each year versus only 14,000 in children, demonstrating the large effect of age on cancer incidence even though no age is immune.
“In America, for example, 1.7 million new cancer cases are diagnosed in adults. In children, 14,000. So you see the difference that age makes.”
- Although more cells and more divisions should mean more cancer, this is not observed across species—a phenomenon called Peto's paradox; humans should have more cancer than mice but don't, whales almost never get cancer, and elephants rarely do, partly because elephants carry about 20 redundant copies of the p53 tumor suppressor gene.
“It's called Peto's paradox because the idea would be that if the number of mutations increase each time a cell divides then a larger organism should have a higher chance of having cancer, right?”