The screening argument

What the argument is actually about

Mammography can find cancers that a woman or her physician would never have noticed. That is not in dispute. The images are real, the technique has improved steadily since Robert Egan standardised exposure and positioning in the late 1950s, and the pathology is genuine tissue with genuine malignant cells. The argument is not about whether the machine detects things. It is about what happens next, and to whom, and at what cost.

The central concept is overdiagnosis: a cancer detected at screening that would never, in the patient's lifetime, have caused symptoms or death. Because breast cancers vary enormously in their biology — some doubling in days, some dormant for decades — a proportion of screen-detected tumours belong to the slow or arrested end of that spectrum. The problem is that, once detected, they cannot be distinguished from the fast ones, so they are treated. The woman undergoes surgery, often radiotherapy, sometimes systemic therapy, and is entered into a long follow-up programme — for a cancer that would otherwise have remained silent. She is counted, in institutional records, as a life saved. Whether she is is unknowable at the individual level, and that is precisely the problem.

Overdiagnosis is not the same as a false positive, though the two are often conflated. A false positive is a recalled scan that turns out, after biopsy, to be normal tissue. Overdiagnosis is a true positive: cells that genuinely meet the pathological definition of malignancy, in a tumour that genuinely exists, but that would have gone undiscovered and harmless without the screening programme. Estimating how large that proportion is requires modelling or long follow-up of randomised cohorts, which is why the numbers remain contested.

Where the evidence comes from, and what it says

The randomised trials that launched population mammography were run mainly in the 1960s through 1980s, before modern digital imaging and before the molecular subtype era. The Health Insurance Plan of New York trial, begun in 1963, was the first to allocate women by chance to screening or no screening. Swedish trials followed through the 1970s and 1980s — the Two-County trial in Kopparberg and Östergötland being the most cited. These trials found reductions in breast cancer mortality in the screened groups, and those mortality reductions became the foundation for national programmes.

The Early Breast Cancer Trialists' Collaborative Group, working from Oxford, is best known for pooling individual patient data from treatment trials, and pooled analyses of the screening trials by other groups consistently found that screening reduced breast cancer mortality. That finding is not disputed. What is disputed is how large the reduction is when the trials are scrutinised for allocation quality, and how that benefit compares with the harm of overdiagnosis in contemporary conditions, where background treatment is far better than it was in 1963 and where some screen-detected cancers might resolve without treatment even if left.

The Cochrane Collaboration's systematic review, first produced by Peter Gøtzsche and Ole Olsen and updated repeatedly since, took the position that when only the better-randomised trials are included, the mortality benefit shrinks, and that for every life saved, a considerable number of women are overdiagnosed and overtreated. The review's estimates of overdiagnosis — at various revisions suggesting somewhere between one in three and one in two screen-detected cancers — provoked fierce rebuttal from epidemiologists working within screening programmes, who argued the modelling assumptions were too pessimistic and the trial quality exclusions too aggressive.

The disagreement fractured along institutional lines. The World Health Organization's International Agency for Research on Cancer, in its 2015 breast screening handbook, concluded that there was sufficient evidence of benefit to justify population programmes while acknowledging overdiagnosis as a real cost. The US Preventive Services Task Force has revised its recommendations multiple times, navigating explicitly between the mortality reduction evidence and the harms of overdiagnosis and false positives, and has at different revisions set different starting ages for routine screening. The Independent UK Panel on Breast Cancer Screening, reporting in 2012, estimated roughly one death prevented for every three overdiagnoses — a ratio its authors described as a reasonable basis for offering screening while being transparent about the trade-off.

What the disagreement reveals

The screening argument is partly statistical and partly philosophical. The statistical disputes concern model assumptions: what proportion of cancers detected at screening would have surfaced clinically in the absence of screening, how long follow-up must be before lead-time bias — finding a cancer earlier without actually extending survival — can be discounted, and how to separate the effect of screening from the effect of improved treatment over the same decades. These are not trivial disputes, and they have not been resolved by any single analysis. Randomisation as evidence can answer the question of whether screened populations do better — it cannot easily separate the mechanism when background treatment is improving at the same time.

The philosophical dispute is about whose values count and how trade-offs are expressed. A mortality reduction expressed as a relative risk — screening reduces breast cancer deaths by roughly 20 percent in trial populations, according to IARC's assessment of the evidence — sounds unambiguously good. The same reduction expressed in absolute terms, over ten years of screening a population of a thousand women, produces a smaller number of deaths prevented and a larger number of overdiagnoses. Neither framing is dishonest; both are genuine, and the choice between them shapes public communication in ways that advocacy and institutional positioning have not always made neutral.

There is also the question of which women and at what age. Screening programmes in most high-income countries were designed around population averages, but the biology of breast cancer is not uniform. Women who carry a pathogenic variant in BRCA1 — the gene that Mary-Claire King localised to chromosome 17 in 1990 — face a different risk distribution from the general population, at younger ages, and the screening interval and modality appropriate for them is a distinct question from what suits a 55-year-old at average risk. Molecular stratification has made the single population answer less adequate, not more.

What is clear is that the argument is real, conducted between researchers of good faith working from overlapping but differently weighted bodies of evidence, and that it has practical consequences. Countries with active screening programmes continue to see stage shifts — more cancers detected early, fewer presenting at advanced stage — but the relationship between that shift and actual mortality reduction depends on how much of the early detection is genuine rescue and how much is overdiagnosis. The answer differs by tumour biology, screening interval, age group, and the treatment landscape surrounding it. It is not a question that admits of a single number, which is why the conversation continues.