The hardest cancers are found late

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.

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.

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.

Pancreatic cancer grows and spreads quickly, and delays between the scan that finds it, the specialist meeting, the biopsy, the bile duct stent and the first chemotherapy or operation are measured in weeks. The proposal is a dedicated fast-track pathway with a national standard for the interval from first imaging to first treatment, reported by the audit.

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.

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.

People are invited separately for bowel, breast, cervical and lung screening, and partial participation is common. One appointment at 50 and 60, modelled on the NHS Health Check, offering every eligible test plus family history and risk assessment with navigation support, would raise uptake.

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.

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.

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.

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.

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.

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.

No randomised trial has shown that screening survivors for a second cancer reduces death from it. A registry-based randomised trial, which invites rather than enrols, is the only design that could answer this at an affordable cost.

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.

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.

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.

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.

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.

The rules that decide who gets a lung scan count cigarettes. A risk model that also uses age, sex, family history, deprivation and lung disease would find more cancers in the same number of scans, and would stop excluding people who smoke less but are more likely to get the disease.

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.

Bowel cancer in people under 50 is rising by 2 to 8 percent a year on both sides of the Atlantic and nobody knows why. The strongest lead is a toxin made by some gut bacteria whose damage signature is three times more common in young patients and is stamped on the colon early in life. If that is the cause, the fix is in childhood, not in a screening programme.

Bowel cancer that has seeded the lining of the abdomen responds badly to drugs but well to a complete operation, and the heated chemotherapy that everyone added turned out to do nothing. What decides the outcome is whether the disease is found while it is still limited and whether the patient reaches a centre that can remove all of it.

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.

Screening trials are judged on deaths, which take fifteen years to count, and on cancers found, which is the wrong direction. Preventing a man from turning up with cancer already in his bones happens three times as often as preventing a death, arrives years earlier, and is the outcome he cares about.

Multi-cancer early detection blood tests take a decade to prove they save lives. Regulators could accept a fall in late-stage cancers as the first answer, validated against pooled data from completed screening trials, provided approval is conditional on continued mortality follow-up.

Pack-year rules miss people who get lung cancer without heavy smoking, including East Asian women who never smoked, as Taiwan's TALENT study showed. Eligibility by a validated risk model with a set threshold, plus a never-smoker arm where family history matters, would find more cancers per scan.

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.

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.

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.

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.

PSA screening finds harmless prostate cancers that would never cause trouble. Selecting men by PSA plus a polygenic risk score, then MRI before any biopsy and targeted biopsy only for PI-RADS 4 to 5 lesions, finds the dangerous cancers and skips the rest; GOTEBORG-2 showed MRI-targeted biopsy halves overdiagnosis.

People with vague symptoms such as fatigue and weight loss are worked up with repeated scans. A multi-cancer blood test as the first step in Rapid Diagnostic Centres could point to which organ to examine first, or safely reassure; SYMPLIFY showed the Galleri test has high specificity in symptomatic patients.

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.

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.

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.

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.

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.

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.

Every targeted result on this roadmap depends on a test happening fast enough to act on. In England nobody publishes what share of lung cancers are tested, how long the test takes, or, in Wales and Northern Ireland, how long the lung pathway takes at all.

Current blood tests for leftover cancer track a few dozen mutations and miss low-level disease. Whole-genome and error-corrected methods integrate signal across thousands of tumour-specific sites plus methylation and fragment features, reaching detection near one part per million in research settings; cost, turnaround and reproducibility are the barriers.

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.

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.

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.

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.

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.

In parts of Chile and northern India gallbladder cancer is common enough that a cheap ultrasound programme aimed at the highest-risk people might catch it while surgery can still cure it. Nobody has run the 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.

About 1 in 100 people who develop diabetes after 50 has a pancreatic cancer behind it. A score built from weight change, blood sugar change and age, calculable from records already in primary care, picks out a group where the rate is nearer 1 in 30; the proposal is to scan that group rather than wait for symptoms.

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.

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.

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.

One patient in twenty with pancreatic cancer carries an inherited gene fault, and most have no family history. Guidelines now say test every patient, which finds relatives who carry it too, but the yearly scans that catch cancer at stage I in carriers are still offered only in research programmes. The proposal is to make surveillance follow the test result automatically.

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.

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.

About one lung cancer in five happens to someone who never smoked, and they are outside every screening programme in the world. The genomes show it is a different disease that grows more slowly, which is exactly the kind of cancer a screening test could catch.

The only randomised trial of screening colonoscopy cut bowel cancer by 18 percent because only 42 percent of the people invited turned up. A test that is 20 percent more sensitive but is taken by the same people buys far less than an invitation that 20 percent more people accept.

Every positive stool test in England, from the screening programme and from the symptomatic pathway, ends in a colonoscopy. The thresholds have been lowered faster than the capacity to act on them, and who performs the test decides whether it prevents anything.

The fastest rise in bowel cancer is in people two decades younger than any screening programme, who reach diagnosis through symptoms, often after several visits. Screening cannot help them; the referral pathway can.

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.

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.

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.