Overdiagnosis and false alarms

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.

People are told a blood test can find fifty cancers, but not how many false alarms or how much is unknown. A short, tested decision aid before the test would make consent real.

Scans find unexpected lumps in the adrenal, thyroid, pancreas and lung. Nobody knows how many matter. A national cohort following them for ten years would tell us whom to watch.

Precancer is treated by cutting away part of the cervix, which raises pregnancy risks. A vaccine that makes the immune system clear the infected cells would avoid surgery.

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.

Blood tests increasingly find precursor conditions like MGUS and smouldering myeloma, but most never progress. A registry with clear rules would stop early treatment outside trials.

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.

Papillary thyroid cancers under 1 cm almost never cause harm. Japanese hospitals have watched thousands safely. Make watching, not surgery, the default everywhere, with a registry.

Low-risk DCIS is a breast change that may never become cancer. Trials now show watching it is safe in the short term. The next step is a proper pathway and a name that does not say cancer.

Most lung nodules on CT are harmless but trigger years of follow-up scans. A validated AI score could discharge low-risk nodules immediately.

Whether a lesion is called precancer or cancer varies between pathologists, and over time the bar has drifted lower. AI reference reads could hold the line.

Some things called cancer, such as low-grade prostate lesions, almost never spread. An expert body could reclassify them, as cervical precancer was, so fewer people are overtreated.

For choices where the right answer depends on what the patient values (watching a slow prostate cancer, adjuvant chemo at 80, breast reconstruction), a tested decision aid becomes part of the consultation.

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.

Screening finds cancers that would never have caused harm, but programmes only report cancers found. Publishing the estimated overdiagnosis rate alongside would make the trade-off visible.

Frail women over 80 with small hormone-sensitive breast cancers may do as well with a daily tablet as with surgery. A trial would define who can safely avoid the operation.

Melanoma diagnoses have soared while deaths barely changed, a sign of overdiagnosis. AI skin apps should be judged on whether dangerous thick melanomas fall, not how many spots they flag.

Trials show older women with the lowest-risk breast cancers gain almost nothing from radiotherapy after lumpectomy. Yet most still get it. Track and reward omission.

Old autopsy studies found hidden prostate, thyroid and breast cancers in many people who died of other causes. Repeating them with modern methods would show how large the reservoir of harmless cancer is.

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 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.

Many low-risk patients get operations and radiotherapy they may not need. Health systems would fund the trials that find out who can safely skip them, and keep the savings.

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.

Men on surveillance for prostate cancer have repeat biopsies every year or two, which puts many off. Using MRI, PSA density and new markers to trigger biopsy only when needed could be just as safe.

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.

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.

South Korea's thyroid cancer rate rose 15-fold after ultrasound screening, with no fall in deaths. A firm rule not to biopsy nodules under 1 cm, and not to screen, would prevent this harm elsewhere.

People with Barrett's oesophagus without dysplasia have endoscopies every few years, but the BOSS trial found no survival benefit. Redirecting effort to those with dysplasia would save harm and cost.

Screening someone unlikely to live ten years causes harm without benefit. A life-expectancy estimate in the medical record could stop invitations and prompt a conversation.

Small pancreatic cysts are found on many scans and followed for life. Evidence suggests those unchanged for five years rarely turn to cancer. A stopping rule would spare years of scans.

Gleason 6 prostate lesions do not metastasise. A trial could show whether describing them without the word cancer leads more men to choose monitoring.

Most kidney tumours under 3 cm found by chance grow slowly and a fifth are benign. Watching them, with surgery only if they grow, could spare many operations.