Cancer — What It Is, Why It's Hard, How to Read It¶
flowchart LR
mut[mutations accumulate] --> clone[clonal expansion]
clone --> hall[hallmarks acquired]
hall --> tumor[primary tumor]
tumor --> inv[invasion]
inv --> met[metastasis]- eyes — same stacked-signs logic for a different organ
- food — dietary risk and protective factors
- signs & levels — one mutation is noise; stacked hallmarks are signal
- health as infrastructure — cancer fits the four-floor model
Investigation · rating: high. Hallmarks framework: Hanahan & Weinberg 2000/2011/2022. Screening data: USPSTF guidelines.
Cancer is not one disease. It is the name we give to a family of conditions that share a single failure mode: a clone of cells in your body has stopped obeying the rules that hold a multicellular organism together. Different tissues, different mutations, different timelines, but the same underlying story — a lineage of cells whose selfish replication has been selected for, inside you, at your expense.
This page is an investigation, not a treatment guide. The goal is to make the shape of the problem clear enough that screening recommendations, treatment choices, and news headlines stop being a fog of acronyms and start being readable. For the "one sign vs. several stacked signs" framework that runs through this page, see SIGNS-AND-LEVELS.md.
The two-sentence definition¶
A cancer is a clonal population of cells descended from a single ancestor that has accumulated mutations letting it grow when it shouldn't, ignore stop signals, evade death, and — eventually — leave its home tissue. Everything else — the tumor, the symptoms, the metastases, the treatment difficulty — falls out of those four properties.
Tumors are not cancer in themselves. A benign tumor is a clone that grows where it shouldn't but stays put and doesn't invade. A malignant tumor has crossed the basement membrane and can seed daughters elsewhere — that crossing is what the word cancer names.
The hallmarks — a checklist for what a cell has to acquire¶
Hanahan and Weinberg's 2000 framework (updated 2011, 2022) is still the cleanest way to enumerate what a normal cell has to break to become cancerous. Think of these as locks on a door; the cell has to pick most of them.
| Hallmark | What it means | A drug or process that exploits it |
|---|---|---|
| Sustained proliferative signaling | Cell makes its own "grow" signal or never turns off the receptor | EGFR inhibitors, BRAF inhibitors |
| Evading growth suppressors | Tumor-suppressor genes (RB, p53, APC) silenced or mutated | MDM2 inhibitors (reactivate p53) |
| Resisting cell death | Apoptosis pathways disabled (BCL-2 overexpression) | BH3-mimetics (venetoclax) |
| Enabling replicative immortality | Telomerase reactivated — no Hayflick limit | Telomerase inhibitors (experimental) |
| Inducing angiogenesis | Tumor recruits its own blood supply | Bevacizumab, VEGF inhibitors |
| Activating invasion & metastasis | Cells dissolve basement membrane, migrate, seed | Matrix metalloprotease focus |
| Reprogrammed energy metabolism | Warburg effect — aerobic glycolysis | 2-DG, metabolic inhibitors |
| Avoiding immune destruction | Cell hides from T-cells (PD-L1 up, MHC down) | Checkpoint inhibitors (pembrolizumab, nivolumab) |
| Tumor-promoting inflammation | Chronic inflammation feeds the tumor | Anti-inflammatories as adjuvants |
| Genome instability | Mutator phenotype — accelerated mutation accumulation | PARP inhibitors (in BRCA-mutant cancers) |
| (2022) Senescent cells, non-mutational epigenetic reprogramming, polymorphic microbiome, unlocking phenotypic plasticity | The newer additions — the field is still extending the list | Senolytics, epigenetic drugs |
You cannot become cancer by acquiring one of these. You typically need most of them. That is why cancer is rare per cell-division but common per lifetime — billions of divisions, dozens of near-misses, and eventually one lineage that has rolled the necessary dice.
The progression ladder — adenoma to carcinoma¶
The colorectal cancer pathway (Vogelstein 1988) is the cleanest textbook example, and the same shape — multi-step acquisition over years — applies to most carcinomas.
| Stage | What's there | Mutations typically accumulated | Detection window |
|---|---|---|---|
| Normal epithelium | well-behaved cells | 0 | invisible |
| Hyperplasia | crowded but still ordered | APC loss | sometimes visible |
| Early adenoma | small benign polyp | APC + KRAS | colonoscopy catches |
| Late adenoma | larger polyp, dysplastic | APC + KRAS + SMAD4 / DCC | still curable by polypectomy |
| Carcinoma in situ | malignant features, still local | + p53 | curable by resection |
| Invasive carcinoma | through basement membrane | + telomerase, angiogenic switch | treatable, harder |
| Metastatic | seeded distant organs | + invasion/EMT program | usually incurable |
The killing fact: the window from first malignant cell to symptoms is often 5–20 years. A 50-year-old's colon cancer started in their mid-30s. Screening is valuable because it inserts a sensor into that long, silent corridor.
Why cancer is hard — three structural reasons¶
1. Selection pressure is unavoidable¶
Any treatment that doesn't kill every cell selects for the survivors. A drug that kills 99.99% of a 10⁹-cell tumor leaves 10⁵ resistant cells — a tumor by themselves. This is why monotherapies fail and combination regimens dominate; it is also why HIV combination therapy was a model for oncology.
2. The tumor is your tissue¶
A cancer cell shares 99.9%+ of its proteome with your healthy cells. Anything that kills cancer cells specifically has to find the 0.1% that differs — and that difference often is a quantitative excess of a normal protein, not a unique target. This is the therapeutic window problem: cytotoxics like cisplatin and 5-FU kill cancer cells because cancer cells divide faster, but they also kill your gut lining, your hair follicles, your bone marrow.
3. Heterogeneity within a single tumor¶
A 1 cm tumor has ~10⁹ cells, and they are not clones — they are a tree. Subclones at the periphery may carry different driver mutations than subclones at the core. Sequencing one biopsy gives you one branch's-eye view. This is why liquid biopsy (ctDNA in blood) is a research frontier — it samples the whole tree.
Reading cancer — signs, screening, and what they actually catch¶
The same logic as weather signs applies: one sign in isolation is usually wrong; several stacked signs are usually right. Cancer screening lives or dies by specificity at population scale — a 99% specific test applied to 1,000,000 healthy people yields 10,000 false alarms.
Symptoms ladder — from suspicious to alarming¶
| Sign | Suspicion | Why |
|---|---|---|
| Unexplained weight loss (>5% in 6 months) | high | cancer rewires metabolism; cachexia is real |
| Persistent fatigue not explained by sleep | low alone, higher with others | cytokine load |
| A lump that is hard, fixed, painless, growing | high | benign lumps tend to be soft, mobile, sometimes tender |
| New persistent cough >3 weeks (esp. smoker) | moderate–high | lung Ca, especially with hemoptysis |
| Change in bowel habits >4 weeks | moderate | colorectal — also IBS, infection |
| Blood where it shouldn't be — stool, urine, sputum, between periods, post-menopause | high | breach of an epithelial surface — investigate |
| Night sweats, drenching | moderate | lymphoma classic; also TB, menopause |
| New persistent headache + neurological signs | moderate–high | brain mass, but also migraine, vascular |
| Non-healing skin lesion >4 weeks | moderate–high | basal cell, squamous cell, melanoma |
| A mole that changes (asymmetry, border irregularity, color variation, diameter >6mm, evolution — ABCDE) | high | melanoma |
| Difficulty swallowing, progressive | high | esophageal — late symptom, unfortunately |
| Bone pain that wakes you at night | moderate–high | metastasis or primary bone |
Heuristic: A single sign is usually nothing. Two unexplained signs in someone over 50 lasting more than a few weeks is the threshold for serious workup. Below 40, the prior probability is much lower and worry threshold can be higher (but melanoma, testicular, leukemias, lymphomas, brain tumors, and cervical do occur in young people — don't dismiss persistent localized signs).
Screening tests — what they actually do¶
| Test | Catches | What's good | What's bad |
|---|---|---|---|
| Colonoscopy | colorectal adenomas + cancer | removes precursors during the test — prevention, not just detection | invasive, prep is awful, ~10 yr interval |
| Mammography | breast cancer (lumps, calcifications) | reduces mortality ~20% in 50–70 women | overdiagnosis of indolent DCIS, dense breasts limit sensitivity |
| Low-dose CT (lung) | lung nodules in heavy smokers | reduces lung Ca mortality ~20% in eligible group | radiation, false positives → unnecessary biopsies |
| Pap smear / HPV test | cervical dysplasia | one of the great public health wins — cervical Ca rates collapsed | requires repeat sampling |
| PSA (prostate) | elevated prostate antigen | catches prostate Ca early | overdiagnosis huge — many caught cancers would never have hurt you |
| Skin self-exam + dermatology check | melanoma + non-melanoma | cheap, high yield in fair-skinned | requires literate observer |
| Liquid biopsy / multi-cancer early detection (Galleri etc.) | ctDNA from many cancers | early, frontier | sensitivity for stage I still modest, specificity for site uncertain |
The overdiagnosis problem: a test that finds a lesion that would never have killed you still puts you through surgery, radiation, anxiety. PSA and DCIS are the classic examples. "Catching cancer early" is good only when "early" means "before it would have killed me" — not when "early" means "an indolent thing I'd have died with, not from."
Treatment modalities — what each one targets¶
| Modality | What it does | Hallmark targeted | Strengths | Limits |
|---|---|---|---|---|
| Surgery | physical removal | the whole clone, locally | cures localized cancer outright | useless once it's metastasized |
| Radiation | DNA damage in a beam | proliferation + genome integrity | precise, local, can be curative | damages adjacent tissue; long-term second-cancer risk |
| Cytotoxic chemo | poison dividing cells | proliferation broadly | works systemically; cured testicular, ALL, Hodgkin | non-selective — gut, hair, marrow toll |
| Targeted therapy | small molecule blocks a specific driver protein | whichever hallmark the driver hits | dramatic when the target is right (imatinib for CML) | resistance evolves; tumor must have the target |
| Hormonal therapy | block growth signals (tamoxifen, anti-androgens) | proliferative signaling in hormone-sensitive cancers | very effective in ER+ breast, prostate | only works for hormone-driven cancers |
| Immunotherapy — checkpoint inhibitors | release the brakes on T-cells (PD-1/PD-L1, CTLA-4) | immune evasion | transformative — melanoma, lung, kidney, MSI-high colorectal | works in ~20–40% depending on cancer; can cause autoimmunity |
| Immunotherapy — CAR-T | engineered T-cells targeting CD19 etc. | immune evasion + targeted kill | curative-grade in pediatric ALL, some lymphomas | bespoke, expensive, cytokine storm, solid tumors still hard |
| Bispecific antibodies, ADCs | link a drug to an antibody pointing at tumor marker | proliferation + selective delivery | growing class — trastuzumab-deruxtecan a recent star | resistance, target loss |
| PARP inhibitors | block DNA repair backup pathway | genome instability — synthetic lethality in BRCA-mutant cells | elegant — kills cells that can't repair | only works in repair-deficient cancers |
| Hyperthermia / proton / FLASH | physical modalities being refined | local | precision | infrastructure-heavy |
The big shift since ~2014: immunotherapy works. Checkpoint inhibitors turned some metastatic melanomas from a 12-month median survival into long-term remissions. That category of drug is the single most important advance of the last two decades; combined with targeted therapy guided by tumor sequencing, it has redrawn the map of "incurable" cancers, especially in melanoma, NSCLC, RCC, and MSI-high colorectal.
Risk factors — what actually moves the needle¶
Risk factors are noisy at the individual level (a non-smoker can get lung cancer; a heavy smoker can die at 90 of something else), but they are very tight at population level. Roughly ordered by total attributable burden:
| Risk factor | Cancers driven | Magnitude of effect |
|---|---|---|
| Tobacco smoking | lung, throat, bladder, kidney, pancreas, cervix, stomach | ~20% of all cancers, ~80–90% of lung. Single biggest modifiable cause. |
| Age | nearly all | not modifiable; mutation count rises linearly with cell-divisions over time |
| Obesity | breast (post-menopause), endometrial, colorectal, esophageal, pancreas, liver, kidney | rising contribution as smoking falls |
| Alcohol | mouth, throat, esophagus, liver, breast, colorectal | linear with intake; no safe threshold for breast |
| Chronic infection | HPV → cervical / oropharyngeal; HBV/HCV → liver; H. pylori → stomach; EBV → some lymphomas; HIV (immune-related) | major in low/middle-income countries; HPV vaccine is changing this |
| UV radiation | skin (BCC, SCC, melanoma) | dose-dependent; sunburns in youth especially |
| Ionizing radiation | leukemias, thyroid, breast | medical imaging is small dose; occupational/accidental larger |
| Family history / germline mutations | varies — BRCA1/2 (breast, ovarian), Lynch (colorectal, endometrial), Li-Fraumeni (many) | rare but high penetrance |
| Diet patterns | colorectal (red/processed meat), gastric (salt, smoked) | modest individual effect, real at population |
| Air pollution | lung, possibly bladder | rising recognition |
| Occupational exposures | asbestos → mesothelioma; benzene → leukemia; vinyl chloride → liver | small denominators, very high relative risk |
| Hormonal/reproductive | breast, endometrial, ovarian — modified by parity, age at menarche/menopause, HRT | modest |
Heuristic for personal risk reduction (high-evidence, ordered): 1. Don't smoke. Quitting beats almost any other intervention. 2. Get the HPV vaccine if eligible. Cervical and oropharyngeal cancer. 3. Get the HBV vaccine. Liver cancer. 4. Keep alcohol low. The "moderate drinking is healthy" story has not survived recent meta-analyses for cancer specifically. 5. Maintain body composition you can carry — obesity is a real lever. 6. Sun protection, especially before adulthood. 7. Do the recommended screens for your age and sex. 8. Know your family history; ask about genetic testing if there's a pattern.
Why some tissues, some ages — the cell-division economy¶
Cancer risk per tissue correlates strongly with the lifetime number of stem-cell divisions in that tissue (Tomasetti & Vogelstein 2015). Tissues that turn over fast — gut epithelium, skin, blood, breast ducts — generate more replicative errors and host more cancers. Tissues that turn over slowly — heart muscle, neurons — rarely host primary cancers (heart sarcomas and brain gliomas exist, but the per-cell rates are tiny). This is the "bad luck" component of cancer risk: even with no environmental insult and no inherited predisposition, the act of being a long-lived multicellular organism generates cancer.
Age is the corollary. Cancer is overwhelmingly a disease of late life because mutations accumulate — incidence rises roughly with the 5th–7th power of age across many cancers. Pediatric cancers are a different beast: they typically involve developmental pathways (Wnt, Hedgehog, Notch) and embryonic-tissue origins, not decades of mutational drift.
The frontier — what's actually changing¶
| Direction | What's new | When it'll matter |
|---|---|---|
| Multi-cancer early detection (Galleri, Grail, others) | ctDNA + methylation signatures in blood, single test, many cancers | clinical evidence maturing 2026–2030 |
| Tumor-agnostic drugs | drugs approved for a mutation (NTRK fusion, MSI-high) regardless of tissue of origin | already here; expanding |
| Personalized cancer vaccines | mRNA vaccines coding for a patient's neoantigens — pancreatic, melanoma trials | early but promising as of 2026 |
| CAR-T for solid tumors | hard problem — solid tumors lack a CD19-like clean target | gradual progress |
| Antibody-drug conjugates (ADCs) | new payloads, new linkers — explosion of approvals | already changing breast, lung, bladder |
| AI-guided pathology and radiology | flag suspicious areas in slides and scans | being deployed; care/quality varies |
| Single-cell + spatial transcriptomics | map the actual heterogeneity of a tumor | research today, clinical perhaps 5–10 yr |
| Microbiome modulation | gut microbes shape immunotherapy response | mechanism clear, intervention messy |
| Senolytics + cancer | clear out senescent cells that promote tumor growth | early human trials |
| Theranostics (e.g. Lu-177 PSMA) | targeted radionuclides — image and treat with the same molecule | growing in prostate, neuroendocrine |
A small philosophy section — why cancer is a problem at all¶
A multicellular organism is a treaty: trillions of cells cooperate in exchange for shared reproductive success via the germline. Cancer is defection — a somatic lineage that re-discovers, via mutation, the ancestral free-running replicator strategy of single cells. Every multicellular organism faces this; whales and elephants have so many cells that they would have impossible cancer rates if not for evolved defenses (Peto's paradox: large long-lived animals have less cancer per cell than predicted — elephants carry 20 copies of TP53). Our defenses are decent for ~60 years and then increasingly leaky after, which is exactly the lifespan our recent ancestors needed.
This is also why cancer can never be "cured" once and for all in the way smallpox was. As long as we are multicellular and our cells divide, there will be defectors. The realistic goal is detection early enough to remove, and treatments precise enough that defection doesn't pay. The last 20 years have made enormous progress on both.
Cross-references¶
- SIGNS-AND-LEVELS.md — the framework behind "no single sign is reliable; stacked signs are"
- WEATHER.md — same reading-the-world logic, different domain
Further reading (canonical entry points)¶
- Hanahan & Weinberg, Hallmarks of Cancer (2000), Hallmarks: The Next Generation (2011), Hallmarks: New Dimensions (2022) — the field's spine
- Robert Weinberg, The Biology of Cancer — textbook of choice
- Siddhartha Mukherjee, The Emperor of All Maladies — history and human texture
- NEJM, Nature Cancer, Cancer Cell — current journals
- USPSTF, NCCN, ESMO guidelines — what to actually do for screening and treatment in 2026
References¶
- Hanahan, D. & Weinberg, R. A. (2000). The hallmarks of cancer. Cell 100(1). Original six-hallmark framework; the field's primary organizational spine.
- Hanahan, D. & Weinberg, R. A. (2011). Hallmarks of cancer: the next generation. Cell 144(5). Expanded to eight hallmarks plus two enabling characteristics; adds tumor microenvironment and metabolic reprogramming.
- Hanahan, D. (2022). Hallmarks of cancer: new dimensions. Cancer Discovery 12(1). Four additional hallmarks including senescence, non-mutational epigenetics, polymorphic microbiome; current state of the framework.
- Weinberg, R. A., The Biology of Cancer (2nd ed., 2013). Garland. Standard graduate-level textbook; mechanistic depth behind each hallmark.
- Mukherjee, S., The Emperor of All Maladies (2010). Scribner. Oncology history and clinical texture; contextualizes the discovery arc of the hallmarks model.
- Vogelstein, B. et al. (2013). Cancer genome landscapes. Science 339(6127). Driver vs. passenger mutation framework; the mutational basis for hallmark acquisition.