Paths to cures: interception, eradication, control
Three different ways a cancer stops killing someone: stop it before it starts, remove every last cell, or hold it in check for life. Each needs different technology.
Cure is not one goal. Interception prevents a cancer from forming or removes it while it is still precancerous. Eradication removes every malignant cell, which is what surgery plus adjuvant therapy already achieves in early disease and what cell therapy achieves in some leukaemias. Control converts advanced cancer into a managed chronic condition, which is what endocrine therapy in breast cancer and TKIs in CML already do for many patients.
Separating them matters because they have different endpoints, different evidence requirements, and different technologies. An interception vaccine needs decades of follow-up in healthy people; an eradication strategy needs a sensitive measure of residual disease; a control strategy needs sequencing, tolerability, and resistance management rather than depth of response.
- Interception: establishedhistoric
What already prevents cancer
HPV and hepatitis B vaccination, tobacco control, screening with removal of precancerous lesions, risk-reducing surgery in carriers, and tamoxifen or aspirin chemoprevention are the interventions that have measurably reduced incidence. Almost every future interception technology is competing against, or adding to, this list, and none of it is glamorous.
- Interception: nextemerging
Vaccinating and monitoring people who do not yet have cancer
Shared-neoantigen vaccines in Lynch syndrome and BRCA carriers, multi-cancer blood tests, and breath analysis all aim to act before or at the very beginning of disease. The bar is high: a healthy person accepts risk today for a probabilistic benefit later, so safety must be near-perfect and trials must run for years with surrogate endpoints. A positive interception vaccine trial would be the single largest change on this roadmap.
- Eradication: establishedcurrent
Removing every cell, and knowing that you did
Surgery with adjuvant systemic therapy cures a large fraction of early cancers; CAR-T and transplant cure a minority of advanced haematologic malignancies. What has changed is measurement: ctDNA-based residual disease testing turns 'we think it is gone' into a testable claim, and the first ctDNA-guided approval arrived in bladder cancer in 2026.
- Eradication: nextemerging
Killing the last cell, wherever it is hiding
Micrometastatic disease is where alpha and Auger emitters, in situ vaccination, and logic-gated cell therapy should have their advantage, because each kills single cells rather than bulk. The pairing to watch is a sensitive residual-disease test that says who still has disease, and a single-cell-selective therapy that can clear it, which is the explicit design of several ongoing trials.
- Control: establishedcurrent
Living with cancer as a chronic disease
Endocrine therapy in hormone-driven breast cancer, androgen-pathway therapy in prostate cancer, BTK and BCL-2 inhibitors in CLL, and TKIs in CML already keep many people alive for years or decades. Control depends less on depth of response than on tolerability, sequencing, and managing resistance, which is why supportive care and cardio-oncology belong in this row.
- Control: nextemerging
Steering resistance instead of waiting for it
If a cancer cannot be eradicated, the goal becomes keeping the sensitive clone dominant. Adaptive dosing, timing treatment to the body clock, payload switching guided by resistance biomarkers, and treating the host environment all aim at that. Cachexia therapy belongs here too: patients who keep weight and function tolerate more lines of treatment.
- The honest caveatspeculative
What would have to be true
Every path above assumes measurement improves faster than the cancer adapts: interception needs a test that finds disease while it is still local, eradication needs one that proves nothing is left, control needs one that spots the resistant clone before it takes over. Technology that measures is therefore upstream of technology that treats, which is an unfashionable conclusion but the one the evidence supports.
Probability ranges are named estimates that the claim is borne out on roughly a five-year horizon. They are meant to be argued with: propose a revision with your name and reasoning via a pull request to src/data/confidence.ts.
Story
topWhat already prevents cancer
HPV and hepatitis B vaccination, tobacco control, screening with removal of precancerous lesions, risk-reducing surgery in carriers, and tamoxifen or aspirin chemoprevention are the interventions that have measurably reduced incidence. Almost every future interception technology is competing against, or adding to, this list, and none of it is glamorous.
Vaccines that prevent the viral infections behind cervical, throat, anal, and liver cancers. The most effective anti-cancer intervention ever created.
Drugs or surgery for people at high inherited risk, before any cancer appears.
Finding and removing polyps before they become cancer. Colonoscopy prevents cancer; stool and blood tests catch it early and get more people screened.
A test of the DNA you were born with, to find inherited risk genes such as BRCA or Lynch syndrome.
Removing the fallopian tubes, where most ovarian cancer starts, during other pelvic surgery or in women at inherited risk.
Vaccinating and monitoring people who do not yet have cancer
Shared-neoantigen vaccines in Lynch syndrome and BRCA carriers, multi-cancer blood tests, and breath analysis all aim to act before or at the very beginning of disease. The bar is high: a healthy person accepts risk today for a probabilistic benefit later, so safety must be near-perfect and trials must run for years with surrogate endpoints. A positive interception vaccine trial would be the single largest change on this roadmap.
Vaccinating people who do not have cancer yet but are very likely to get it, against the antigens their future tumour will carry.
Vaccines against mutations or proteins shared by many patients, so they can be made in advance.
A single blood test intended to screen for dozens of cancers at once, including ones with no screening today.
Fragmentomics reads the sizes and positions of DNA fragments in blood, not the mutations. Cancer cells die messily and leave a recognisable fragmentation pattern.
Smelling cancer: measuring the trace chemicals a tumour puts into exhaled breath.
Giving the innate immune system a memory, so monocytes and NK cells respond harder the next time they meet a tumour.
Removing every cell, and knowing that you did
Surgery with adjuvant systemic therapy cures a large fraction of early cancers; CAR-T and transplant cure a minority of advanced haematologic malignancies. What has changed is measurement: ctDNA-based residual disease testing turns 'we think it is gone' into a testable claim, and the first ctDNA-guided approval arrived in bladder cancer in 2026.
An ultra-sensitive blood test after surgery that detects leftover cancer months before a scan would.
A patient's T cells are removed, given a synthetic receptor that recognises the cancer, multiplied, and put back as a living drug.
Replacing a patient's blood system with a donor's, so the donor's immune cells hunt down any leukaemia left behind. Still the only cure for many high-risk leukaemias.
Surgeons operate through small incisions using robotic arms with tremor-free precision and 3D vision.
Very high, very precise radiation doses in 1-5 sessions that can ablate a tumour like surgery.
Killing the last cell, wherever it is hiding
Micrometastatic disease is where alpha and Auger emitters, in situ vaccination, and logic-gated cell therapy should have their advantage, because each kills single cells rather than bulk. The pairing to watch is a sensitive residual-disease test that says who still has disease, and a single-cell-selective therapy that can clear it, which is the explicit design of several ongoing trials.
Like radioligand therapy but with alpha particles: far more destructive over a much shorter range, so single cells can be killed with less collateral damage.
Radioactive atoms that spray very short-range electrons, lethal only if the atom sits on or inside the cell's DNA.
Treating one tumour so aggressively that the immune system learns to attack every other one, using the tumour itself as the vaccine.
Cells or drugs that fire only when two conditions are true at once, so healthy tissue expressing just one of them is spared.
Instead of testing blood every three months, sampling constantly, so a relapse is caught the week it starts.
Adding a PARP or ATR inhibitor to a radioactive drug so the tumour cannot repair the damage the radiation causes.
Living with cancer as a chronic disease
Endocrine therapy in hormone-driven breast cancer, androgen-pathway therapy in prostate cancer, BTK and BCL-2 inhibitors in CLL, and TKIs in CML already keep many people alive for years or decades. Control depends less on depth of response than on tolerability, sequencing, and managing resistance, which is why supportive care and cardio-oncology belong in this row.
Pills that block or remove oestrogen signalling, the mainstay of treatment for hormone-driven breast cancer for 50 years.
Androgen deprivation lowers testosterone or blocks its receptor, and has been the foundation of prostate cancer treatment since 1941 (Nobel Prize 1966).
Pills that block the specific enzyme a cancer relies on. Imatinib in 2001 proved a cancer could be switched off by design.
Protecting the heart from cancer treatments, which is increasingly important as patients live longer.
Structured exercise during and after treatment, which the CHALLENGE trial showed improves survival in colon cancer.
Geriatric assessment is a structured check of an older patient's fitness, memory, and support that predicts treatment tolerance better than age.
Steering resistance instead of waiting for it
If a cancer cannot be eradicated, the goal becomes keeping the sensitive clone dominant. Adaptive dosing, timing treatment to the body clock, payload switching guided by resistance biomarkers, and treating the host environment all aim at that. Cachexia therapy belongs here too: patients who keep weight and function tolerate more lines of treatment.
Giving the same drug at a different time of day, because the body clock changes how much damage it does and how well the immune system responds.
Treating the wasting that kills many cancer patients, by blocking the hormone that suppresses appetite.
Cancer neuroscience is the study of how tumours talk to nerves. Nerves grow into tumours and feed them signals; brain tumours even wire themselves into neural circuits. Cutting the conversation with common drugs such as beta-blockers is now being tested.
Chemotherapy leaves behind zombie cells that will not divide but poison their neighbours. Senolytics aim to clear them.
Choosing treatment from a map of where each cell type sits in the tumour, not just from a list of its mutations.
Organoid-guided therapy means routinely growing a piece of each patient's tumour and testing drugs on it before choosing, rather than relying on genetics alone.
Testing cheap old drugs, aspirin, metformin, statins, beta-blockers, as cancer treatments, because they are safe, available and sometimes work.
What would have to be true
Every path above assumes measurement improves faster than the cancer adapts: interception needs a test that finds disease while it is still local, eradication needs one that proves nothing is left, control needs one that spots the resistant clone before it takes over. Technology that measures is therefore upstream of technology that treats, which is an unfashionable conclusion but the one the evidence supports.
A single blood test intended to screen for dozens of cancers at once, including ones with no screening today.
An ultra-sensitive blood test after surgery that detects leftover cancer months before a scan would.
Choosing treatment from a map of where each cell type sits in the tumour, not just from a list of its mutations.
Instead of testing blood every three months, sampling constantly, so a relapse is caught the week it starts.
Scanners sensitive enough to image the whole body in seconds at a fraction of the radiation dose, which raises the question of whether healthy people should be scanned at all.