Reanalysis of Mammography Trials Pushes Overdiagnosis Estimates Lower
A fresh reanalysis of the major randomized mammography trials is forcing a reassessment of one of breast screening’s most disputed harms. Breast cancer overdiagnosis has long been discussed as a meaningful downside of population screening, but this new synthesis suggests the problem may be far smaller than the estimates that have circulated for years.
Instead of figures reaching 30% to 50%, the combined evidence now aligns with an overdiagnosis rate below 5%. That is a striking shift, and it changes how the balance of benefit and harm should be explained to women invited for mammography.
The biological reason is straightforward. Screening pulls diagnosis forward in time, finding some cancers earlier than they would have appeared clinically. If investigators stop watching too soon, the early surge in detected cases can look like extra disease rather than earlier diagnosis of cancers that would later have surfaced anyway.
Why the Early Spike in Diagnoses Can Mislead
The core argument in this analysis is temporal. When mammography begins in a population, incidence rises because more cancers are found before symptoms develop. Some of those cancers are then removed from the pool of future diagnoses, which should create a later decline if follow-up is long enough and if screening exposure is balanced between trial groups.
In several older studies, those conditions were not fully met. Women in the control groups later underwent screening after trials ended or once routine screening expanded, and the number of screening rounds differed across studies. Follow-up also varied. Each of those factors can distort the apparent excess of screen-detected cancers and make overdiagnosis look larger than it really is.
This is not a minor statistical detail. It changes the clinical interpretation of the entire evidence base.
Eight Trials, One Reframed Signal
The reanalysis pooled all eight randomized mammography trials: the New York Health Insurance Plan, Malmö, Two-County, Edinburgh, the Canadian National Breast Screening Study, Stockholm, Gothenburg, and UK Age. Investigators then compared incidence patterns from those trials with two Danish regional screening programs that offered a real-world reference point.
Denmark provided a useful comparator because organized screening started 17 years earlier in some regions than in others. That staggered rollout made it possible to examine how incidence changed immediately after screening began and how those changes evolved over time. Matching the trial data against those regional patterns showed a similar structure: an early rise in diagnoses, followed by a later shift that reduced the apparent excess.
The analysis also included both invasive breast cancer and ductal carcinoma in situ (DCIS), which matters because DCIS is one of the main lesions detected by mammography and one of the diagnoses most often tied to concerns about overtreatment. When the timing of diagnosis and the duration of follow-up were handled more carefully, the signal pointed away from large-scale overdiagnosis and toward a much smaller effect.
Denmark Offered the Real-World Counterweight
The Danish data mattered because they reflect how screening behaves outside a trial setting. In organized programs, the first wave of mammograms will inevitably uncover disease earlier, and later years should reveal whether those diagnoses were merely advanced in time or truly added to the total burden of clinically meaningful cancer.
Elsebeth Lynge described the central interpretive problem clearly: if researchers do not account for the fact that some women in both arms continue screening after a trial ends, the initial rise in cases can be mistaken for overdiagnosis. The reanalysis treated that issue directly by comparing incidence at matching points in time rather than freezing the evidence at an early, misleading moment.
A small statistical error in time can become a large public-health error in policy.
What a Below-5% Estimate Means at the Bedside
Sisse Helle Njor said the goal was to bring together all randomized controlled trial evidence to clarify the extent of overdiagnosis in breast cancer screening. The result is a more restrained estimate than the one many clinicians and patients have heard for years.
For women deciding whether to enter a screening program, this matters because the discussion has often been framed as a tradeoff between the chance of early detection and the risk of substantial unnecessary cancer labeling. This reanalysis narrows that harm. The study does not erase the possibility of unnecessary treatment, but it suggests that the proportion of cancers detected through screening that would never have become life-threatening is much smaller than some earlier estimates implied.
Matejka Rebolj said older high estimates were based on evidence before the trial data had fully matured. Interpreted in full temporal context, the randomized trial data are consistent with overdiagnosis of less than 5%, not figures approaching 50%.
For clinicians, the practical consequence is in counseling. Mammography still finds cancers earlier, and earlier detection can prevent premature death in some women. Yet the language used to explain screening should not lean on inflated estimates of harm if those estimates came from incomplete follow-up or unequal screening exposure.
The Remaining Tension in Screening Communication
Breast cancer screening will continue to generate debate because it sits at the intersection of population benefit, individual preference, and uncertainty about downstream treatment. Overdiagnosis remains real, and even a low rate can matter when millions of women are screened. DCIS and invasive cancers also do not carry identical clinical weight, so aggregate numbers never tell the full story for a given patient.
Still, the shape of this evidence matters. A screening program built around a false assumption of 30% to 50% overdiagnosis invites a very different policy response than one anchored in a rate below 5%. The first implies widespread unnecessary diagnosis; the second suggests a smaller but still important harm that must be communicated honestly.
The study used a meta-analytic reanalysis of the randomized trial literature, not a new screening experiment. Its conclusions depend on how completely the old trials are reconstructed in time and how carefully later screening is handled. Those methodological choices are central to the result and should remain part of any future guideline discussion.
For now, the strongest message is narrow and concrete: mammography appears to cause substantially less overdiagnosis than many long-standing estimates suggested, and the timing of diagnosis is the key reason those earlier numbers may have been too high.