412 ‒ The evolving science of cancer: emerging approaches to prevention and early detection, & more

Episode Overview

• Oncologist and author Dr. Siddhartha Mukherjee joins Dr. Peter Attia to examine what science has learned about cancer prevention, early detection, and treatment in the 15 years since the publication of The Emperor of All Maladies.
• The conversation shifts the paradigm of oncology away from viewing cancer purely as random genetic "bad luck," reframing it as a two-part equation where dormant mutated "sleeper cells" are awakened by physiological triggers like chronic inflammation and hyperinsulinemia.
• They explore cutting-edge developments in immunology—including cryptic neoantigens, thymus involution, and lineage-specific CAR T-cell therapies—as well as the role of generative AI in scaling molecular drug design.
• The episode concludes by connecting oncology and stem cell biology (such as osteoarthritis) to human longevity, explaining why curing major pathological killers like cancer and heart disease will still leave supercentenarians facing a wall of systemic regenerative failure.

Key Concepts

• The 15-Year Oncology Landscape and Mortality Trends: Overall cancer mortality across the US population is declining, making the idea that cancer is exploding and killing everyone "nonsense." However, key exceptions exist: colorectal cancer is rising in young people (which Mukherjee hedges is a "complete mystery" as to why); breast cancer mortality gains made over the last 20 years have started to plateau; esophageal cancer is rising, probably due to alcohol and tobacco mixed together (hedged as "half a mystery"); and endometrial cancer is plateauing or rising, most likely due to increasing obesity rates. While people are drinking less today, there is always a 5-to-10-year lag in cancer rates. In early detection, continuous imaging (following patients over time) rather than discrete imaging (once-a-year mammograms of relatively poor quality) is being invented, though Mukherjee hedges, "whether they will work or not I don't know." Meanwhile, second- and third-line therapies truly expand lifespan, including Enhertu in breast cancer, 3 or 4 new medicines in myeloma, 2 or 3 in lymphoma, 2 or 3 in prostate cancer, and daraxonrasib targeting RAS to extend survival in spread pancreatic cancer by months if not years.
• "Shove" vs. "Nudge" Hereditary Mutations: In 2015 and 2017 Science papers, Bert Vogelstein and colleagues estimated that ~5% of cancer mutations were hereditary, ~30% arose from environmental triggers/carcinogens, and ~66% (two-thirds) were purely random ("bad luck"). Mukherjee explains that the 5% hereditary figure is now much larger because single-gene mutations (like BRCA1, BRCA2, and p53) only cover about a third (1/3) or a fourth (1/4) of hereditary breast cancer. Hereditary risk is divided into "shove" mutations (like BRCA1/2, which shove you off the cliff at a 9-fold to 10-fold risk above the normal population) and "nudge" mutations (hundreds of genes pushing risk by a small percentage that combine with environmental factors).
• The "Gun and Trigger" Effect and Dormant Sleeper Cells: Inflammatory states almost certainly cause cancer by mixing with mutations, though Mukherjee hedges that we are early in the inflammation story and lack a clinical biomarker (like high cholesterol or hypertension for cardiovascular disease) to tell a patient they are in a pro-carcinogenic inflammatory state. Research from Charlie Swanton’s lab in the UK shows that if you biopsied every lung cell in healthy non-smokers (like Attia and Mukherjee), you would find dormant "sleeper cells" bearing all the genetic mutation hallmarks of cancer. In a hypothetical lung biopsy of a billion cells, Person A might have 2,000 mutated sleeper cells and Person B might have 750; real "bad luck" is when a specific inflammatory response wakes one up. This is supported by a 1950s Oxford study where painting a chemical carcinogen (DMBA) on a mouse's back grew ~4 cancers, but adding croton oil (an inflammatory agent) grew 80 cancers, and by Allan Balmain’s UCSF lab, where genetically predisposed mice grew 4, 5, or 6 cancers naturally, showed nothing impressive from a small surgical slit, but grew massive cancers exactly where wounds were stapled and triggered localized inflammation.
• Hyperinsulinemia vs. Adiposity in Cancer Metabolism: About 5 years ago, Lew Cantley and Mukherjee published a Nature paper (inspired in part by Attia, who also proposed the initial idea over a sushi dinner for Faeth Therapeutics—a company co-founded by Mukherjee and Cantley that went public ~2 weeks prior to recording to develop drugs targeting cancer metabolism in endometrial and breast cancer, where Mukherjee is no longer privy to internal info). Their animal models showed that you do not need to be obese to sustain cancer; rather, chronic insulinemic tone and the insulin cascade drive cancer cell resistance to therapy, and removing that tone makes the cancer vulnerable. Mukherjee carefully hedges ("surmise," "extrapolate," "general hypothesis") that while insulin's role in treatment resistance is proven in animals, extrapolating that chronic hyperinsulinemia initiates or releases cancer cells in humans requires a "fussy," well-controlled experiment. This parallels a paper written ~15 years ago by Mitch Lazar at Penn using a 2x2 matrix of obesity vs. metabolic health, showing cardiovascular disease prevalence tracked with hyperinsulinemia, not adiposity per se. Attia and Mukherjee also discuss a hypothetical randomized clinical trial testing standard of care vs. standard of care + a GLP-1 agonist as an adjuvant cancer therapy (in endometrial, ovarian, or breast cancer), using AI to dissociate insulinemic response from weight loss (though Mukherjee does not know if a formal proposal has been submitted to Novo, Lilly, or the NIH).
• Osteoarthritis as Stem Cell Regenerative Failure: Mukherjee’s lab discovered the first osteoarthritic stem cells (published in Cell, Nature, and Nature Medicine). He rejects the idea that osteoarthritis is a weight-bearing/obesity disease as "blaming the victim," noting non-obese women get it in non-weight-bearing joints like the hands (inflamed nodes), upper neck joints, and shoulders. Joint cartilage is maintained by a balance between genesis and destruction; in obese women and very old men and women, cartilage-forming stem cells go down while destruction stays the same or rises. In young mice, injecting an inflammatory stimulus (like LPS/bacterial mimic or a small molecule) into a joint causes stem cells to stop making cartilage. When asked why an immune response shuts down host stem cell output, Mukherjee hedges, "I don't know," explaining that "why" questions are metaphysical/teleological ("Rudyard Kipling's Just So Stories"), whereas science answers "how" questions.
• Cryptic Neoantigens, Checkpoints, and Thymus Involution: While Steve Rosenberg shared data that 80% of solid organ tumors have novel neoantigens recognizable by self T-cells, Mukherjee recalls arguing with Rosenberg onstage years ago that 30, 50, or 100 mutations couldn't explain all neoantigens. Work from Tyler Jacks’ lab at MIT (and Mukherjee's lab) reveals that real neoantigens arise when cancer cells dysregulate their genetic decoding apparatus and express "cryptic antigens" (pieces of genes that shouldn't be made into proteins). T-cells don't attack them because surviving cancer cells display dozens of "checkpoints" (flags saying "I'm normal, leave me alone"), which normally prevent T-cells and B-cells from causing autoimmunity like Type 1 diabetes, lupus, or scleroderma. Furthermore, a recent talk at Harvard used AI to scan tens of thousands of patient CAT scans and found one marker correlating with wellness, longevity, and fewer autoimmune diseases and cancers: an intact thymus (the "kindergarten" where T-cells grow up, which naturally involutes with age). Mukherjee hedges that this is correlative, not causal, and dismisses a ~10-year-old open-label "kitchen sink" study using growth hormone, metformin, and DHEA to regenerate the thymus as "a mess," advocating instead for systematic science like implanting a new thymus in old animals and identifying specific thymic growth factors.
• Lineage-Specific CAR T-Cells and Pancreatic Cancer Anatomy: Mukherjee (who disclosed creating one of the first CAR T-cell companies in India, selling therapy at 1/20th the US price and treating ~150–200 patients) explains that genetically engineered CAR T-cells cure liquid cancers (with patients living 15 to 20 years post-diagnosis) because they are lineage/tissue-specific, not cancer-specific. Targeting CD19 kills all B-cells, and humans happen to be able to survive without B-cells. Thus, CAR T-cells can be applied to other expendable tissues like the prostate or ovaries/fallopian tubes (noting ovarian cancer actually originates in the fallopian tubes, which have known antigens and active CAR T trials), whereas breast and pancreas lack unique antigens not shared by vital cells. In pancreatic cancer, KRAS mutations are hidden inside the cell (making them a target for small molecules like daraxonrasib, which causes nasty facial lesion side effects as seen in a US Senator, rather than CAR T-cells). Anatomically, pancreatic tail tumors are distinct, detected incidentally, and allow some patients to live 10 years after surgery, whereas head tumors sit in a complex anatomical "traffic intersection" (intestines, blood vessels), present with painless jaundice, and are much harder to cure surgically (where Attia notes his own bias from his surgical training program regarding Whipple procedures).
• AI in Drug Discovery vs. Whole-Cell Simulation: Every drug (aspirin, Lipitor, Prozac) is a "key" binding to a wriggling protein "lock." Traditional "search and fit" screening faces a chemical space of $10^{20}$ (more than a billion billion), which exceeds current computation and will likely require quantum computers to do inferential reasoning on at least a billion keys. Instead, at his company Manas, Mukherjee uses AI for synthetic/generative molecular design—teaching AI the rules of chemistry, physics, geometry, and electrostatics of protein pockets to scale 1 synthetic chemist into 50,000. However, he hedges that he is skeptical of using AI to build simulated biological models of whole cells anytime soon because biology lacks neat rules across cells with vastly different jobs (e.g., a kidney cell filtering salts vs. a neuron vs. a skin cell).
• Chemoprevention, Geroscience, and Regenerative Failure: Prevention (specifically chemoprevention) will always save the most lives, though it won't be as simple as Lipitor (lowering LDL) or Losartan/ACE inhibitors/ARBs (acting on kidney hormonal systems for blood pressure); Attia notes clinicians now check for CHIP (clonal hematopoiesis of indeterminate potential) in cardiovascular patients and use colchicine as a crude preemptive immune therapy. However, eliminating proximal killers of 70–90-year-olds (cancer, heart disease, neurodegenerative disease) will not solve longevity. Supercentenarians (110+ years old) don't die of cancer (though Attia notes they still die of atherosclerosis and infections); they die of regenerative failure—skin falling off, bone breakage, immune failure, and bone marrow failure. This echoes Élie Metchnikoff, who coined "thanatology" in the 1890s and argued that without pathological death, humans would live to 100–102 and naturally want to die. Advancing geroscience will require a 15-to-30-year basic science "Manhattan Project" (as advocated by David Sabatini) funded by philanthropy rather than fast-return venture capital or the illness-focused NIH. (Mukherjee is currently exploring these themes in a new book on death, dying, and longevity, spending months or years in "dream time" before quickly writing 600–700 pages and editing with Nan Graham).

Quotes

• At 0:02:13 - "There is a kind of a gun and trigger effect, the gun being the inflammatory states, the trigger being the mutations that cause cancer, or a mix of some states like that." - Sid Mukherjee explaining how inflammation and genetic mutations interact to cause cancer.
• At 0:04:46 - "The idea that cancer is exploding and killing everyone in the population is nonsense. But, there's a very important but: we know that colorectal cancer is on the rise in young people." - Sid Mukherjee clarifying overall declining US cancer mortality against specific rising exceptions.
• At 0:07:03 - "So without selecting for high-risk people, without selecting for people who have had cancer before, we know that those tests are probably not going to work. They're going to lead to false positives, they're going to lead to extra biopsies..." - Sid Mukherjee on why blood-based early cancer detection tests fail in unselected general populations.
• At 0:09:37 - "Immune therapy is now known to extend the lifespan of smokers with lung cancer if given to the right population of people." - Sid Mukherjee highlighting advances in immunotherapy over the last 15 years.
• At 0:18:12 - "In hereditary cancers, there are two kinds of mutations. I call them shove mutations and nudge mutations. BRCA1 is a shove mutation; it shoves you off the cliff and puts you at a nine-fold, ten-fold risk above the normal population to get hereditary breast cancer at an early age." - Sid Mukherjee distinguishing high-penetrance single-gene mutations from polygenic risk.
• At 0:19:56 - "You and I, right now, if we were to biopsy every cell in our lung—if we magically had the capacity to biopsy every cell in our lung—we would find, almost certainly, and I don't smoke, you don't smoke, we would almost certainly find cells that genetically speaking would bear all the hallmarks of genetic mutations of cancer." - Sid Mukherjee discussing Charlie Swanton's research on dormant cancer mutations in healthy non-smokers.
• At 0:21:08 - "Our scientific publication strategy has become a form of magic storytelling. By magic storytelling, I mean you have to tell a full story about what you observe, what the mechanism is, what the explanation is, and how it's working at a molecular level." - Sid Mukherjee discussing the nature of modern scientific publishing.
• At 0:24:44 - "They stapled where the wounds were, and wherever there were staples, of course, you got an inflammatory response. And exactly, exactly the place where they put the staples, out grew massive cancers." - Sid Mukherjee describing Allan Balmain’s UCSF mouse experiments linking localized inflammation to tumor growth.
• At 0:26:40 - "You had a particular kind of inflammatory response in your body, and your 2,000—in one of your 2,000 cells, which already was carrying a mutation, turned from a sleeper cell, a dormant cell, a doing-nothing cell, into a tumor, into an active cancer cell." - Sid Mukherjee redefining "bad luck" in oncology as an inflammatory trigger waking a dormant mutated cell.
• At 0:32:23 - "It's not being obese; it is the insulin and it's the cascade of insulin and what follows the insulin cascade that is responsible, at least, for the resistance of cancer cells to therapy." - Sid Mukherjee summarizing his and Lew Cantley's Nature paper on cancer metabolism.
• At 0:40:59 - "I have never believed that osteoarthritis was an obesity disease. There are non-obese people who have osteoarthritis, there are obese people who have osteoarthritis—it's much larger—but the idea, there was something kind of like, it didn't make sense and also it was sort of like, let's blame the victim." - Sid Mukherjee reframing osteoarthritis as an inflammatory shutdown of cartilage-producing stem cells rather than mechanical wear and tear.
• At 0:49:57 - "Why is not a science question; why is a metaphysical question." - Sid Mukherjee explaining why science focuses on how immune responses shut down stem cells rather than teleological reasons why.
• At 0:51:04 - "80—eight-zero—percent of solid organ tumors had novel neoantigens that could be recognized by self." - Peter Attia citing Steve Rosenberg's data on the immunogenicity of solid tumors.
• At 0:54:33 - "A checkpoint is a gene or a flag that a normal cell floats on its surface saying, 'I'm normal, leave me alone,' to a T-cell. And that's why we don't have desperate, terrifying autoimmunity." - Sid Mukherjee explaining how immune checkpoints prevent autoimmunity and how cancer cells hijack them.
• At 0:59:16 - "People who had an intact thymus tended to have less diseases such as autoimmune diseases and cancer; people who lost their thymus as they grew older, because the thymus naturally involutes and dies, tended to have more." - Sid Mukherjee sharing correlative AI findings from Harvard on CAT scans and thymus involution.
• At 1:08:01 - "The CAR T-cells that kill B-cell leukemias and lymphomas don't just kill B-cell leukemias and lymphomas; they kill all B-cells. I'm going to repeat that fact: they just, they don't kill B-cell leukemias and lymphomas, they kill all B-cells." - Sid Mukherjee explaining that CAR T-cell therapy works by wiping out an entire expendable tissue lineage.
• At 1:18:58 - "He says that there are three zeros between 1 million and 1 billion. And that means that if you have 1 billion cancer cells and you take them out, and there's only 1 million left, that means that there's one in a thousand—you've reduced the chances of one in a thousand that you'll have a cancer cell that's resistant to your therapy." - Sid Mukherjee quoting Bert Vogelstein on the mathematical rationale for surgical debulking and adjuvant therapy.
• At 1:22:56 - "My recommendation for people who want to avoid colorectal cancer is triple or double your fiber intake, obviously reduce exposure to these forever chemicals, and, based on no evidence whatsoever, use probiotics..." - Sid Mukherjee offering practical, hedged advice for mitigating rising young colorectal cancer risk.
• At 1:38:32 - "The prevention of pathological disease—cancer, heart disease, the two biggest killers—is not the answer to the prolongation of longevity." - Sid Mukherjee introducing the orthogonal view of longevity and stem cell exhaustion in supercentenarians.
• At 1:48:25 - "The combinatorics of injecting yourself with pep—with a random group of peptides, the chances you'll find the right peptide at the right concentration are basically zero. Like zero. Not—it's like, there's no—this is a mathematical truth." - Sid Mukherjee dismissing Silicon Valley longevity peptide hacks based on the combinatorics of 20 amino acids.

Takeaways

• Reserve blood-based early cancer detection tests for high-risk populations: Avoid using liquid biopsy screening tests in unselected, general populations, where they are unlikely to decrease mortality and often lead to false positives and extra biopsies; instead, deploy them in high-risk individuals such as genetic mutation carriers or prior cancer survivors.
• Expand hereditary genetic screening beyond single genes: When evaluating hereditary risk for breast and ovarian cancers, use modern polygenic sequencing that looks at a multiplicity of "nudge" genes rather than relying solely on single "shove" genes like BRCA1 and BRCA2, which account for only 1/3 to 1/4 of hereditary breast cancers.
• Begin colonoscopy screening by age 40 for average-risk adults: While formal guidelines lowered the screening age from 50 to 45, Attia advocates getting screened at age 40 because colorectal cancer is the 3rd leading cause of cancer death and is rising in young people under 45.
• Double or triple your dietary fiber intake: Inspired by Denis Burkitt's 1950s/1960s observations in Africa—where high-fiber diets led to evacuating colorectal contents twice a day (versus once every 2 to 3 days in the US) and virtually zero young colorectal cancer—significantly increase dietary fiber to improve the colorectal milieu.
• Reduce exposure to pro-inflammatory "forever chemicals": Actively minimize exposure to forever chemicals, which have strong animal-model evidence for altering the gut milieu and driving inflammation.
• Incorporate live yogurt for gut microbiome support: To help counter gut microbiome disruption from antibiotics and modern food, Mukherjee recommends eating live yogurt at least once a week for probiotics (though he explicitly hedges this recommendation is "based on no evidence whatsoever").
• Maintain insulin sensitivity regardless of body weight: Because animal models demonstrate that chronic insulinemic tone and the insulin cascade drive cancer therapy resistance independently of obesity, prioritize metabolic health and low chronic insulin levels even if you are not overweight.
• Pair surgical tumor debulking with neoadjuvant or adjuvant therapy: Apply Bert Vogelstein's mathematical principle ("three zeros between 1 million and 1 billion") in clinical oncology by using surgery (such as a Whipple procedure) or radiation to remove bulk tumor cells before administering targeted drugs like daraxonrasib, reducing the probability of therapy-resistant cells by 1,000-fold.
• Target expendable tissue lineages when developing solid-tumor CAR T-cell therapies: Focus CAR T-cell engineering on organs humans can live without—such as the prostate, ovaries, and fallopian tubes (where ovarian cancer originates)—to bypass the danger of destroying vital normal tissues that share antigens with cancer cells.
• Avoid unvalidated Silicon Valley peptide injections: Do not self-administer random longevity peptides; because there are 20 amino acids per molecule, the mathematical combinatorics make the probability of randomly selecting the right peptide at the right concentration "basically zero."
• Replace "kitchen sink" longevity regimens with systematic basic science: Rather than combining random compounds like growth hormone, metformin, and DHEA in open-label trials to regenerate the thymus, focus research on systematic experiments like thymus transplantation in aged animals and isolating specific thymic hormones.
• Direct philanthropic capital toward a 15-to-30-year geroscience "Manhattan Project": Because studying wellness, stem cell exhaustion, and aging biomarkers falls outside the illness-focused NIH mandate and requires timelines too long for venture capital, utilize philanthropic funding to support foundational basic science in longevity.