OnCo
bottlenecksBottleneck

Most lethal cancers are found late

Screening exists for only a few cancers. Pancreatic, ovarian, liver, oesophageal and most lung cancers are found when cure is unlikely.

Stage at diagnosis is the largest single determinant of survival: localised pancreatic cancer has a five-year relative survival around 44% in SEER data, distant disease about 3%. Organised population screening exists only for breast, cervical and colorectal cancer, low-dose CT for heavy smokers, and prostate in some systems, and uptake is incomplete even where it exists. The cancers that kill most people at a late stage (pancreas, ovary, liver, oesophagus, stomach, most lung cancers in never-smokers) have no screening test, and the largest randomised trial of ovarian screening (UKCTOCS) found no mortality benefit despite a stage shift. Multi-cancer early detection blood tests (Galleri, Shield, fragmentomics) can detect signals from dozens of cancers but have modest sensitivity for stage I disease and have not yet shown reduced mortality; the NHS-Galleri trial is the first randomised test of the concept. Risk-stratified surveillance (new-onset diabetes for pancreas, cirrhosis for liver, Barrett's for oesophagus) is the intermediate strategy.

criticalprevention detection48 ideas to fix it
How big the problem is
~44% vs ~3%
Five-year relative survival for pancreatic cancer diagnosed at localised vs distant stage (SEER, US)
20%
Reduction in lung cancer mortality with low-dose CT screening vs chest X-ray in heavy smokers (NLST)
No significant reduction despite stage shift
Ovarian cancer deaths prevented by annual multimodal screening over 16 years of follow-up (UKCTOCS)
~20 million cases, 9.7 million deaths
New cancer cases and cancer deaths worldwide, 2022
Root causes
  • Most lethal cancers arise in deep organs without symptoms until they are locally advanced or metastatic.
  • Screening tests for low-prevalence cancers need very high specificity to avoid harm, which few biomarkers achieve.
  • A stage shift does not guarantee a mortality reduction, so each test needs a long randomised trial that few sponsors fund.
  • Screening uptake is incomplete and unequal even for tests that work, because of access, cost and awareness.
  • Health systems lack the diagnostic capacity (imaging, endoscopy, pathology) to work up the positives a population test would generate.
What is already being tried
  • NHS-Galleri randomised about 140,000 people to GRAIL's multi-cancer blood test, the first trial of MCED on stage shift and mortality.
  • The NCI Cancer Screening Research Network's Vanguard study is piloting MCED evaluation in the US.
  • Guardant's Shield blood test was FDA-approved in 2024 for colorectal screening, and Exact Sciences' Cologuard has expanded stool DNA testing.
  • The USPSTF broadened lung screening eligibility in 2021 and the NHS is rolling out targeted lung health checks nationally.
  • Cirrhosis-based HCC surveillance and new-onset-diabetes pancreatic surveillance cohorts (NCI/NIDDK New-Onset Diabetes cohort) target high-risk groups.
  • Delfi, Freenome and Harbinger are developing fragmentomic and methylation-based tests aimed at earlier-stage sensitivity.
What breaking it looks like
A randomised trial shows that a multi-cancer or organ-specific test reduces late-stage incidence and cancer-specific mortality for at least one cancer without a current screening programme, and the share of pancreatic, ovarian, liver and oesophageal cancers diagnosed at stage I-II at least doubles.

Ideas to fix it

48top
early clinicalclinicmedium cost
A 28-day national pathway for people with a positive multi-cancer blood test

A positive blood test with no known tumour is frightening and hard to manage. A standard imaging cascade with a time limit, and a registry of what was found, would make these tests usable.

preclinical evidenceresearchmedium cost
A breath test to rule out cancer in people with vague symptoms

Volatile compounds in breath differ in cancer. A breath test validated in truly symptomatic patients, not lab volunteers, could tell GPs who needs urgent scans and who can safely wait.

speculativeclinicmedium cost
A cancer blood test for older people arriving at A&E with unexplained symptoms

One in five cancers in the UK is first found in an emergency, usually late. Adding a cancer test to the blood already taken in A&E for over-60s with vague symptoms could catch some earlier.

being tested at scalepolicymedium cost
A legislated, publicly reported 28-day standard from urgent referral to diagnosis

Set a legal limit: anyone referred with suspected cancer should be told within 28 days whether they have it. Publish how every hospital performs each month.

early clinicaldatasmall cost
A live national dashboard of stage at diagnosis as the scorecard for early detection

You cannot manage what you do not measure quickly. Publishing stage at diagnosis by cancer and region every quarter, not years later, would show whether detection efforts are working.

speculativepolicymedium cost
A single 'cancer check at 60' appointment bundling all screening tests

People are invited separately for bowel, breast, cervical and lung screening and many miss some. One appointment offering all eligible tests, plus a risk assessment, would raise uptake.

being tested at scaleclinicmedium cost
A swallowable sponge test for reflux patients, offered in pharmacies

A pill on a string collects cells from the food pipe and finds Barrett's oesophagus, a precursor of cancer. Offering it in pharmacies to people on long-term heartburn drugs would find it early.

speculativephilanthropymedium cost
A ten-dollar blood test for the five cancers that kill most people in poorer countries

Most cancer deaths are in low and middle income countries, where scans and endoscopies are scarce. A cheap methylation blood test tuned to liver, stomach, oesophageal, cervical and breast cancer could fill the gap.

early clinicalclinicmedium cost
A urine DNA test to decide who with blood in the urine needs a camera test

Most people referred for blood in the urine do not have bladder cancer, yet all get cystoscopy. A urine DNA or methylation test could safely spare most of them.

speculativepolicylarge cost
A whole-population cancer interception programme: risk-stratify every adult, detect and intercept early

Instead of separate screening programmes for a few cancers, assess every adult's overall cancer risk and offer blood tests, imaging and preventive treatment tuned to that risk, all inside one system that learns.

early clinicalclinicmedium cost
AI clears the normal lung screening scans so radiologists read only the suspicious ones

Most screening CT scans are normal. Letting a validated AI clear them, and sending only flagged scans to a radiologist, would let screening scale without more radiologists.

preclinical evidenceresearchmedium cost
AI that spots pancreatic cancer on scans taken a year before diagnosis

Pancreatic cancer is often visible in hindsight on earlier scans. Software trained on those pre-diagnostic scans could flag subtle changes while surgery is still possible.

speculativedatamedium cost
Bank yearly blood from cancer survivors so future tests can be validated

To prove a leftover-cancer test works you need blood taken years before relapse. Collecting and freezing yearly samples now makes every future test testable.

being tested at scalepolicymedium cost
Blood EBV DNA screening for nasopharyngeal cancer across southern China and Southeast Asia

Nasopharyngeal cancer is common in southern China and is caused by a virus. A blood test for viral DNA finds it early, and a large study showed better survival. Scale it up.

early clinicalclinicmedium cost
Cancer risk scores running automatically in GP records to prompt urgent referral

Computers can combine minor symptoms, blood tests and age into a cancer risk score in the background. Showing that score to the GP could get more people referred earlier.

early clinicalindustrymedium cost
Cheap DNA fragment-pattern blood test as a first sieve before expensive cancer tests

How DNA fragments in blood are chopped up differs in cancer and can be read with cheap, shallow sequencing. Used as a first sieve, it could cut the cost of population screening.

early clinicalpolicymedium cost
Community health workers with smartphone AI screen for mouth cancer in South Asia

Mouth cancer is common where tobacco is chewed and is visible to the naked eye. Health workers with a phone camera and AI could find it early in villages.

early clinicalengineeringmedium cost
Every routine CT scan checked by AI for early cancer signs, with a tracked follow-up pathway

Hundreds of millions of CT scans are done each year for other reasons. Software could check each one for early lung, kidney, liver and pancreas changes, but only if a follow-up system exists.

speculativedatasmall cost
Glucose monitor data as an early pancreatic cancer signal

Millions now wear continuous glucose monitors. A sudden, unexplained worsening of glucose control in a middle-aged wearer could be flagged as a possible early sign of pancreatic cancer.

speculativeregulatorsmall cost
Let MCED trials read out on late-stage incidence, with mortality follow-up mandated

Blood tests that look for many cancers at once take a decade to prove they save lives. Regulators could accept fewer late-stage cancers as the first answer, if trials keep counting deaths afterwards.

being tested at scalepolicymedium cost
Lung screening eligibility by risk score, not pack-years, including high-risk never-smokers

Pack-year rules miss many people who get lung cancer, including East Asian women who never smoked. A risk score with family history and ancestry would find more cancers per scan.

speculativepolicysmall cost
Make a two-minute mouth cancer check part of every dental visit

Dentists see the mouth more than any doctor. A standard, recorded oral cancer examination with a referral route would catch cancers earlier at almost no cost.

speculativeregulatorsmall cost
Mandatory interval-cancer audit for every blood-based screening test

When a screening test misses a cancer, we should know. Linking every negative result to the cancer registry and publishing what was missed, by stage, should be a condition of use.

early clinicalclinicmedium cost
Mobile diagnostic units that biopsy, scan and treat on the same visit

Vans equipped with ultrasound, biopsy kits, cervical screening and a link to a distant pathologist could bring a cancer diagnosis, and for cervical pre-cancer immediate treatment, to villages far from any hospital.

preclinical evidenceresearchmedium cost
Molecular indolence classifiers bundled with every screening programme

Screening finds cancers that would never have caused harm alongside dangerous ones. Pair every screening test with a test that says which is which, so people with harmless findings can safely watch and wait.

being tested at scalepolicylarge cost
MRI-first prostate screening with genetic pre-selection

PSA screening finds too many harmless cancers. Using PSA plus a genetic risk score to select men, and MRI before any biopsy, finds the dangerous ones and skips the rest.

early clinicalclinicmedium cost
Multi-cancer blood test as the first step for patients with non-specific symptoms

People with vague symptoms like fatigue and weight loss get many scans. A blood test could point to which organ to look at first, or safely reassure.

early clinicalclinicmedium cost
New diabetes after 50 plus weight loss triggers a pancreatic cancer check

About one in a hundred people who develop diabetes after 50, more if they are losing weight, have pancreatic cancer. A simple score could send them for a scan or blood test.

early clinicalpayermedium cost
Offer the blood test for bowel cancer only to people who refuse stool tests or colonoscopy

Blood tests for bowel cancer miss most precancerous polyps, so they should not replace stool tests. But for the third of people who never do any screening, a blood test may beat nothing.

being tested at scaleclinicmedium cost
One-stop breast clinics: imaging, biopsy and a preliminary answer in a single visit

A woman with a breast lump should be examined, scanned and biopsied on the same day, and hear the result within a few days, rather than visiting four times over two months.

speculativeresearchsmall cost
Open-source models that translate stage shift into lives saved, for every cancer

Whether finding cancer earlier saves lives depends on the cancer. Public models, one per cancer, would let trials and payers predict the mortality benefit from a stage shift honestly.

early clinicalpolicysmall cost
Personalised stool-test cut-offs by age, sex and prior results

Bowel screening uses one blood-in-stool threshold for everyone. Setting it by age, sex and the person's previous results would find more cancers with the same number of colonoscopies.

speculativeregulatorsmall cost
Pre-agreed update rules so an MCED test is not obsolete when its trial reads out

A cancer blood test improves every year, but a ten-year trial tests the old version. Regulators and sponsors could agree in advance how updates are validated and carried into the result.

early clinicaldatamedium cost
Push residual disease detection a hundredfold deeper with whole-genome methods

Current blood tests miss leftover cancer in many patients. Reading thousands of mutations at once, rather than a few dozen, can detect far smaller amounts.

being tested at scaleclinicmedium cost
Rapid diagnostic centres for people with vague but worrying symptoms

Weight loss, fatigue and unexplained pain do not point to one organ, so patients bounce between specialists. A single clinic that investigates such symptoms quickly finds cancers that would otherwise be found late.

speculativedatasmall cost
Record and publish the symptom-to-diagnosis interval for every cancer, by hospital

How long people wait between first noticing something wrong and being diagnosed is barely measured. Recording it routinely and publishing it by hospital would expose where the system loses time.

speculativeresearchlarge cost
Registry-based randomised MCED trial: a million people, no study visits

Randomise people through the national health system, post the blood kit, and read cancer deaths off the registry. That is ten times cheaper per participant than a classic trial.

early clinicalclinicmedium cost
Replace six-monthly ultrasound with a blood test for liver cancer in cirrhosis

People with cirrhosis are meant to have ultrasound scans twice a year, but most do not. A blood test done with their routine liver bloods could catch liver cancer earlier.

early clinicalregulatormedium cost
Require stage-shift or interval-cancer endpoints for AI in cancer screening

AI for screening should be judged on whether it finds dangerous cancers earlier and misses fewer, not just on whether it agrees with radiologists on old images.

being tested at scalepolicymedium cost
Roll out AI-supported mammography nationally as a stepped-wedge trial

Sweden's MASAI trial showed AI can safely replace one of two radiologists. Rolling it out region by region in a randomised order would prove it works at national scale and that interval cancers do not rise.

early clinicalclinicmedium cost
Set each woman's mammogram interval from her last mammogram, using AI risk

Instead of every woman every two or three years, an AI reading of the current mammogram would set who comes back in one year and who can safely wait four.

speculativeclinicsmall cost
Smart scales and health records flag unexplained weight loss for a cancer check

Losing weight without trying is one of the strongest signs of hidden cancer, but it is rarely measured. Automatic alerts from recorded weights could prompt a check-up.

speculativeclinicsmall cost
Stomach cancer serology screening for high-risk migrant communities in low-incidence countries

People who grew up in East Asia, Eastern Europe or Latin America keep a high stomach cancer risk after migrating. Cheap blood tests could select who needs an endoscopy.

early clinicalclinicmedium cost
Tailored screening for second cancers in survivors with known high-risk exposures

Survivors who had chest radiotherapy as young women, or certain chemotherapies, have much higher risks of specific second cancers. They should be screened like people with inherited risk, and today most are not.

being tested at scalepolicymedium cost
Take lung screening scanners to supermarket car parks in the poorest areas

People most at risk of lung cancer are least likely to attend hospital screening. Manchester showed mobile scanners in car parks reach them. Make this the default model.

early clinicalpolicylarge cost
Use a polygenic risk score to set when screening starts

Common gene variants shift a person's cancer risk several-fold. A one-time genetic score could tell each person when to start breast, bowel or prostate screening.

speculativedatasmall cost
Use the trend in routine blood counts, not a single threshold, to spot cancer

A platelet count that is normal but rising year on year, or haemoglobin drifting down, can signal cancer. Records already hold these trends; software could use them.

early clinicalresearchmedium cost
Whole-body MRI plus blood DNA surveillance for people with Li-Fraumeni syndrome

People with an inherited TP53 mutation face a near-certain lifetime cancer risk. Yearly whole-body MRI catches cancers early; adding blood DNA tests may catch them earlier still.

Key papers

12top
rctLancet Oncology 2023
MASAI: AI-supported mammography screening finds more cancers with half the radiologist workload

AI can take over one reader's work in double-reading screening programmes while finding more cancers. Whether the extra cancers found are ones that would have harmed women, and whether interval cancers fall, is the question the trial's primary endpoint will answer.

observationalThe Lancet 2023
PATHFINDER: the first prospective test of a multi-cancer blood test in people without symptoms

A multi-cancer blood test can be run in ordinary clinics, and most positive results can be resolved with imaging. But six in ten positives are false alarms that take months to resolve, and the test misses most cancers. Whether it reduces late-stage cancer or deaths is unknown; that requires randomised trials.

methodsCancers 2022
NHS-Galleri: design of the largest randomised trial of a multi-cancer blood test

NHS-Galleri is the trial that will decide whether a blood test for many cancers at once should be offered by a health system. Its endpoint is a reduction in late-stage cancer rather than deaths, so even a positive result leaves the mortality question to be settled by longer follow-up.

rctNew England Journal of Medicine 2022
NordICC: inviting people to a screening colonoscopy reduced bowel cancer, but less than expected

A colonoscopy probably does reduce bowel cancer risk for the person who has it, but a programme that offers colonoscopy achieves much less if most people decline. Programmes based on stool tests with high uptake may deliver as much population benefit at lower cost and risk.

observationalScience 2020
DETECT-A: a blood test plus PET-CT found treatable cancers in 10,000 women with no symptoms

A blood test can find early, treatable cancers in people who feel well, including cancers for which no screening exists. It is not a replacement for mammography or colonoscopy but a possible addition. Larger randomised trials are needed to show benefit outweighs harm.

rctNew England Journal of Medicine 2020changed practice
NELSON: volume-based CT screening reduces lung cancer deaths with fewer false alarms

Lung screening works when it uses volumetric nodule management, and it works against a no-screening control. The protocol underpins the UK Targeted Lung Health Check programme and European recommendations. Benefit in women remains less precisely estimated.

meta analysisBMJ 2018changed practice
Self-collected HPV samples are as accurate as clinician samples and reach women who never attend screening

Women can collect their own screening sample at home with no loss of accuracy if the laboratory uses a PCR test. Sending kits directly is the most effective way to reach women who do not attend, which matters because most cervical cancers occur in under-screened women.

basicScience 2015
Martincorena: normal sun-exposed skin is a patchwork of cancer-mutation clones

Carrying a cancer mutation is normal; most mutant clones never become cancer. This means blood or tissue tests that look for driver mutations alone will produce false positives, and that the question of what tips a mutant clone into cancer (tissue environment, further hits, immune surveillance) is as important as the mutation itself.

observationalNew England Journal of Medicine 2014
Jaiswal: clonal haematopoiesis, the pre-leukaemic clones in most people over 70

Many older people carry blood clones one or two steps from leukaemia, and those clones also drive heart disease through inflammation. CHIP is why blood-based cancer tests must filter out mutations from blood cells, and it opens a route to preventing both leukaemia and cardiovascular events in carriers.

reviewScience 2013
Cancer genome landscapes: about 140 driver genes, and each tumour needs only a handful

There are not thousands of cancer genes, and any one patient's tumour is driven by only a few of them. That makes targeted sequencing panels sensible, but because most drivers are lost tumour suppressors, drugs exist for only a minority, which is why the same group turned to early detection.

rctNew England Journal of Medicine 2011changed practice
NLST: yearly low-dose CT scans cut lung cancer deaths in heavy smokers

For people with a heavy smoking history, an annual low-dose CT scan is one of the few screening tests proven to reduce cancer deaths. Most abnormal scans are not cancer, so screening must be paired with careful nodule management. It does not apply to never-smokers or light smokers.

rctNew England Journal of Medicine 1993changed practice
Minnesota trial: a yearly stool blood test cuts bowel cancer deaths by a third

A cheap home stool test, repeated yearly or every two years and followed by colonoscopy when positive, prevents bowel cancer deaths. This is what national bowel screening programmes do today, with FIT replacing the older guaiac test.

Connected

120top

cancers

7

fronts

1

technologies

9

drugs

2

companies

6

institutions

9

terms

6

trials

4

roadmaps

1

ideas

52
A 28-day national pathway for people with a positive multi-cancer blood testA breath test to rule out cancer in people with vague symptomsA cancer blood test for older people arriving at A&E with unexplained symptomsA legislated, publicly reported 28-day standard from urgent referral to diagnosisA live national dashboard of stage at diagnosis as the scorecard for early detectionA single 'cancer check at 60' appointment bundling all screening testsA swallowable sponge test for reflux patients, offered in pharmaciesA ten-dollar blood test for the five cancers that kill most people in poorer countriesA urine DNA test to decide who with blood in the urine needs a camera testA whole-population cancer interception programme: risk-stratify every adult, detect and intercept earlyAI clears the normal lung screening scans so radiologists read only the suspicious onesAI that spots pancreatic cancer on scans taken a year before diagnosisBank yearly blood from cancer survivors so future tests can be validatedBlood EBV DNA screening for nasopharyngeal cancer across southern China and Southeast AsiaBlood-based HCC surveillance to replace six-monthly ultrasoundBlood-based pancreatic cancer detection in new-onset diabetesCancer risk scores running automatically in GP records to prompt urgent referralCheap DNA fragment-pattern blood test as a first sieve before expensive cancer testsCommunity health workers with smartphone AI screen for mouth cancer in South AsiaEvery routine CT scan checked by AI for early cancer signs, with a tracked follow-up pathwayFAPI PET as the workup for MCED positivesGlucose monitor data as an early pancreatic cancer signalLet MCED trials read out on late-stage incidence, with mortality follow-up mandatedLung screening eligibility by risk score, not pack-years, including high-risk never-smokersMake a two-minute mouth cancer check part of every dental visitMandatory interval-cancer audit for every blood-based screening testMobile diagnostic units that biopsy, scan and treat on the same visitMolecular indolence classifiers bundled with every screening programmeMRI-first prostate screening with genetic pre-selectionMulti-cancer blood test as the first step for patients with non-specific symptomsNew diabetes after 50 plus weight loss triggers a pancreatic cancer checkNon-endoscopic screening for Barrett's oesophagus and early adenocarcinomaOffer the blood test for bowel cancer only to people who refuse stool tests or colonoscopyOne-stop breast clinics: imaging, biopsy and a preliminary answer in a single visitOpen-source models that translate stage shift into lives saved, for every cancerPersonalised stool-test cut-offs by age, sex and prior resultsPre-agreed update rules so an MCED test is not obsolete when its trial reads outPush residual disease detection a hundredfold deeper with whole-genome methodsRapid diagnostic centres for people with vague but worrying symptomsRecord and publish the symptom-to-diagnosis interval for every cancer, by hospitalRegistry-based randomised MCED trial: a million people, no study visitsReplace six-monthly ultrasound with a blood test for liver cancer in cirrhosisRequire stage-shift or interval-cancer endpoints for AI in cancer screeningRoll out AI-supported mammography nationally as a stepped-wedge trialSet each woman's mammogram interval from her last mammogram, using AI riskSmart scales and health records flag unexplained weight loss for a cancer checkStomach cancer serology screening for high-risk migrant communities in low-incidence countriesTailored screening for second cancers in survivors with known high-risk exposuresTake lung screening scanners to supermarket car parks in the poorest areasUse a polygenic risk score to set when screening startsUse the trend in routine blood counts, not a single threshold, to spot cancerWhole-body MRI plus blood DNA surveillance for people with Li-Fraumeni syndrome

collections

2

people

9

key papers

12