How Cancer Works

TL;DR. Cancer is your own cells, with damaged control genes, dividing when they should not and spreading where they should not. It is not one disease; it is hundreds, united by a mechanism. A cell accumulates mutations over years, most of them harmless, until a particular combination breaks the accelerator (oncogenes) and the brakes (tumour suppressor genes) at the same time. The result grows, recruits a blood supply, evades the immune system, and eventually invades tissue and travels. That last step, metastasis, is what kills: roughly 90 percent of cancer deaths are from cancer that has spread, not from the original lump.

Key takeaways

• About 20 million new cancers and 9.7 million cancer deaths occurred in 2022. Roughly 1 in 5 people develop cancer in their lifetime.
• Cancer is a genetic disease of somatic cells: mutations acquired during life, not usually inherited. Only about 5 to 10 percent of cancers start from an inherited high-risk variant.
• It takes several independent hits to the right genes, which is why cancer is overwhelmingly a disease of ageing and why some inherited syndromes cause it decades early.
• Oncogenes are stuck accelerators; tumour suppressors are cut brakes. One faulty copy of an oncogene is enough; tumour suppressors usually need both copies disabled.
• Metastasis is the killer, and it is a wildly inefficient process: millions of cells enter the blood for every one that establishes a distant colony.
• Roughly 13 percent of cancers worldwide are caused by infections (HPV, hepatitis B and C, H. pylori, EBV), and those are preventable by vaccination and treatment.

What it is

In short: Cancer is your own cells with broken growth controls, and the tissue it starts in determines both its name and its treatment.

A normal cell obeys rules: divide only when instructed, stop when neighbours are close, repair DNA damage or self-destruct, stay where you belong, and die on schedule. A cancer cell has broken all of them.

Benign tumours grow locally, stay encapsulated, and do not invade or spread. They can still be dangerous by location (a benign brain tumour in the wrong place is fatal). Malignant tumours invade surrounding tissue and can metastasise. Only the second is cancer.

Cancers are named for the tissue they arise from, which is worth knowing because the name carries the biology:

A metastasis keeps the name of its origin. Breast cancer that has spread to bone is metastatic breast cancer, not bone cancer, and it is treated with breast cancer drugs, because the cells are still breast cells.

The history

In short: From a chimney sweep's cancer in 1775 to the discovery that cancer genes are corrupted copies of our own normal ones.

Cancer is ancient. The Edwin Smith papyrus (about 1600 BCE) describes breast tumours and notes, in the earliest recorded therapeutic pessimism, "there is no treatment." Hippocrates named it karkinos, Greek for crab, reportedly from the appearance of swollen vessels radiating from a tumour.

Key steps toward the modern picture:

What actually goes wrong

In short: A handful of driver mutations break the accelerators and cut the brakes, and the resulting cells then evolve inside the patient.

Mutations, drivers, and passengers

Every time a cell divides it copies 3 billion DNA letters, and it makes mistakes. Add damage from UV light, tobacco smoke, radiation, chemicals, chronic inflammation, and normal metabolic by-products, and a typical adult cell accumulates thousands of mutations over a lifetime.

Almost all are passengers: irrelevant, in non-coding regions or genes that do not matter to that cell. A few are drivers: mutations in the specific genes that control division, death, and repair. Sequencing studies find that most cancers carry between two and eight driver mutations, accumulated over years to decades.

The two families of control gene

Oncogenes: accelerators stuck on. Normal versions (proto-oncogenes) drive cell division when the body asks. A mutation makes them permanently active. One faulty copy is enough, so they act dominantly.

• RAS (KRAS, NRAS, HRAS): a switch relaying growth signals. Mutated in around 90 percent of pancreatic cancers, roughly 40 percent of colorectal, and a large share of lung.
• MYC: a master transcription factor driving growth programmes.
• HER2: a growth factor receptor, amplified in about 15 to 20 percent of breast cancers, and the target of trastuzumab.
• BRAF: mutated in about half of melanomas.

Tumour suppressors: brakes cut. They stop division, trigger repair, or force damaged cells to self-destruct. Because you have two copies, both must usually be lost, which is exactly what Knudson deduced from the age distribution of retinoblastoma: children with an inherited faulty copy need only one further hit, so they develop tumours early and in both eyes, while sporadic cases need two hits in the same cell and appear later in one eye.

• TP53, encoding p53, the "guardian of the genome." It halts the cell cycle after DNA damage, directs repair, and triggers apoptosis if repair fails. Disabled in roughly half of all human cancers.
• RB1, the retinoblastoma gene, the direct brake on cell cycle entry.
• APC, lost early in most colorectal cancers.
• BRCA1 and BRCA2, DNA repair genes; losing them means damage accumulates faster everywhere else.

The hallmarks

Hanahan and Weinberg's framework lists the capabilities a cell must acquire to become a clinically meaningful cancer:

1. Sustained proliferative signalling: it makes or responds to its own growth signals.
2. Evading growth suppressors: brakes disabled.
3. Resisting cell death: normally, a cell with this much damage would self-destruct via apoptosis. Cancer cells block that programme.
4. Replicative immortality: normal cells can divide only 40 to 60 times before their chromosome-end caps (telomeres) run down. Most cancers reactivate telomerase to rebuild them.
5. Inducing angiogenesis: a tumour beyond about 1 to 2 mm needs its own blood supply, so it secretes signals (VEGF) that make vessels grow toward it.
6. Invasion and metastasis: breaking through the basement membrane and travelling.
7. Genome instability (enabling): broken repair machinery accelerates everything above.
8. Tumour-promoting inflammation (enabling): immune cells recruited to the tumour supply growth factors and remodelling enzymes.
9. Reprogramming metabolism: the Warburg effect, in which cancer cells consume glucose at high rates and ferment it even with oxygen available. This is what PET scanning detects.
10. Evading immune destruction: the basis of modern immunotherapy.

Clonal evolution, and why cancer becomes resistant

A tumour is not a uniform mass. It is an evolving population of related cells with different mutations, competing for space and nutrients. Any treatment applies selection pressure: cells that happen to survive it are the ones that repopulate the tumour. That is Darwinian evolution running on a timescale of months inside one patient, and it is why single-agent therapies eventually fail, why combinations are used, and why the same cancer sampled in two places can respond differently to the same drug.

Metastasis, the part that kills

The steps: cells at the tumour edge loosen their attachments and acquire a mobile phenotype, digest their way through the basement membrane, enter a blood or lymphatic vessel, survive the shear forces and immune attack of the circulation, lodge in a small vessel elsewhere, exit into the tissue, and then, hardest of all, grow in a foreign environment.

It is extraordinarily inefficient. Millions of cells may enter the bloodstream for each one that founds a colony, and many disseminated cells lie dormant for years, which is the mechanistic basis of late relapse: a breast cancer can recur fifteen years after apparently successful treatment.

Where cancers spread is not random. Stephen Paget's 1889 "seed and soil" hypothesis, that the tumour cell needs a compatible tissue environment, has been repeatedly confirmed. Prostate and breast cancers favour bone; colorectal cancer goes to the liver first (its venous drainage arrives there); lung cancer favours brain, bone, liver, and adrenal glands.

What it does to the body

Locally, a tumour compresses, obstructs, ulcerates, and bleeds. A bowel cancer obstructs; a lung cancer collapses a lobe or presses on a nerve; a brain tumour raises pressure inside a rigid skull.

Systemically, it starves the host. Cancer cachexia, a syndrome of profound muscle and fat loss driven by inflammatory signalling rather than by appetite alone, affects a large share of patients with advanced cancer and contributes directly to death. Feeding does not reverse it.

Paraneoplastic syndromes are effects at a distance caused by substances the tumour secretes or by immune cross-reaction: high calcium, low sodium, clotting abnormalities, and neurological syndromes that can appear before the cancer is found.

Immune and marrow suppression, from marrow invasion or from treatment, brings infections, anaemia, and bleeding.

Pain arises from tissue invasion, nerve compression, and bone involvement. It is frequently undertreated worldwide, particularly where opioid access is restricted.

Is it deadly?

Cancer is the second leading cause of death globally, and the first in many high-income countries.

• 20 million new cases and 9.7 million deaths in 2022.
• Approximately 1 in 5 people develop cancer in their lifetime; about 1 in 9 men and 1 in 12 women die of it.
• Survival varies more than for any other disease category in this book, from over 95 percent five-year survival for testicular cancer and localised thyroid, melanoma, and prostate cancer, to under 15 percent for pancreatic cancer.
• Overall survival has improved substantially: in the United States, age-adjusted cancer mortality has fallen by roughly a third since its 1991 peak, driven mostly by fewer people smoking, plus screening and better treatment.

Is it contagious?

In short: No, with the crucial exception that several cancers are caused by transmissible infections, which is why an HPV vaccine prevents a cancer.

No. You cannot catch cancer from another person by contact, air, food, sex, or blood in any ordinary circumstance.

Three exceptions and one important indirect route are worth stating precisely.

Genuinely transmissible cancers exist, but not in humans. Devil facial tumour disease in Tasmanian devils and canine transmissible venereal tumour spread as living cell lines between animals. There is nothing equivalent in people.

Transplant and pregnancy. Cancer has been transmitted through organ transplantation from a donor with an undetected malignancy, which is rare and is why donors are screened. Mother-to-fetus transmission is documented in a handful of case reports.

Infections that cause cancer are transmissible, and this matters enormously. Roughly 13 percent of cancers worldwide are attributable to infection:

This is the most actionable fact in the chapter. A vaccine given to adolescents prevents a cancer that kills over 300,000 women a year, and countries with high HPV vaccine coverage are already seeing cervical precancer and cancer rates collapse in vaccinated cohorts.

Who gets it

In short: Age dominates, followed by tobacco, infection, obesity, and alcohol, and the mix of cancers differs sharply between rich and poor countries.

Age is the dominant factor. Cancer incidence rises steeply after 50, because the multi-hit process takes decades. Most cancers are diseases of ageing, which is why cancer rates rise as countries reduce deaths from infection and childbirth.

Known causes, ranked roughly by global attributable burden:

Random replication error. A 2015 analysis by Tomasetti and Vogelstein noted that the lifetime cancer risk of different tissues correlates strongly with the number of stem cell divisions those tissues undergo, and argued that a large share of the variation between tissues is attributable to unavoidable copying errors. The finding was widely misreported as "two-thirds of cancers are just bad luck," which is not what it says: it explains variation between tissues, not the proportion of cases that are preventable. Population comparisons still show that a large share of cancers would not happen in the absence of tobacco, obesity, infection, and alcohol.

Geography and income. Incidence is higher in high-income countries (partly real, partly because of detection), while mortality relative to incidence is far higher in low-income countries, because of late presentation and limited treatment. The cancer mix differs too: cervical, liver, and stomach cancer dominate in poorer regions, breast, prostate, lung, and colorectal in richer ones. This divergence is one of the sharpest illustrations of the global health inequity discussed in Chapter 61.

How it is described: staging and grading

In short: Stage is how far it has spread and grade is how abnormal the cells look, and molecular subtype now often matters more than either.

Two separate numbers, routinely confused.

Stage is how far it has spread, and it drives prognosis and treatment. The TNM system records T (size and local extent of the tumour), N (lymph nodes involved), and M (distant metastasis present or not), combined into stages I to IV.

Grade is how abnormal the cells look under the microscope, from well differentiated (resembling the tissue of origin, usually slower) to poorly differentiated (chaotic, usually faster).

Increasingly, a third axis matters most: molecular subtype. Breast cancer is no longer one disease but at least four, defined by hormone receptor and HER2 status, each with different drugs and outcomes. Lung adenocarcinoma is subdivided by EGFR, ALK, ROS1, KRAS, and other driver mutations, each with a matching targeted drug. This is why a modern diagnosis takes days to weeks: the tumour is being sequenced, not just looked at.

Don't be confused: "stage 4" and "terminal" are no longer synonyms. Stage 4 means the cancer has spread beyond its origin. For some cancers that remains rapidly fatal. For others, including several with targeted or immune therapies, metastatic disease is now managed for years, and a minority of patients with metastatic melanoma or lung cancer are alive and disease-free a decade after checkpoint immunotherapy. Prognosis depends on the specific cancer, not on the number.

Sources and notes

Global incidence and mortality: GLOBOCAN 2022 (Bray et al., CA: A Cancer Journal for Clinicians, 2024): 20 million new cases, 9.7 million deaths, approximately 1 in 5 lifetime risk, 1 in 9 men and 1 in 12 women dying. Hallmarks of cancer: Hanahan and Weinberg, Cell, 2000 and 2011; Hanahan, Cancer Discovery, 2022. Two-hit hypothesis: Knudson, PNAS, 1971. Proto-oncogenes: Stehelin, Varmus, Bishop, Vogt, Nature, 1976. Seed and soil: Paget, The Lancet, 1889. Infection-attributable cancer fraction (about 13 percent): de Martel et al., Lancet Global Health, 2020. Replication-error analysis: Tomasetti and Vogelstein, Science, 2015, and the subsequent critiques. US cancer mortality decline: American Cancer Society annual statistics. HeLa and consent: Skloot, The Immortal Life of Henrietta Lacks.

Open questions. How much of cancer risk is preventable in principle is genuinely contested, with credible estimates ranging from about 40 percent to substantially more. The mechanisms of tumour dormancy, and how to predict which disseminated cells will awaken, remain largely unknown.

Next: the specific cancers people actually get. 👉