Mammogram overdiagnosis runs below 5%, not the 30% to 50% claimed for decades

Overdiagnosis from mammography screening runs below 5%, according to a reanalysis of eight randomized trials published online in the Journal of the National Cancer Institute on September 14, 2026, and not the 30% to 50% that estimates drawn from the same trials have suggested for decades. The team, led by Sisse Helle Njor of the University of Southern Denmark and Lillebælt Hospital, argues the older figures came from reading immature data at the wrong moment.

Overdiagnosis means finding a breast cancer that would never have caused symptoms or death in the woman’s lifetime. It cannot be seen in any individual patient, only inferred from populations, and that inference has driven an argument for decades over whether the harm of unnecessary surgery, radiation and anxiety should weigh more heavily against the lives saved. A figure of 30% to 50% would make screening look like a trade in which a large share of treated cancers were never a threat; a figure below 5% makes it look like a trade with a small downside.

Eight trials read against a Danish yardstick

The analysis covers every major randomized mammography trial: the New York Health Insurance Plan trial, Malmö, Two-County, Edinburgh, the Canadian National Breast Screening Study, Stockholm, Gothenburg and UK Age. Njor’s co-authors are Casper Urth Pedersen, Elsebeth Lynge of the University of Copenhagen, Robert A. Smith of the American Cancer Society and Matejka Rebolj of Queen Mary University of London. According to the journal abstract, the team examined 52 timepoints across the trials and compared them with a reference pattern built from Denmark’s Funen screening program.

Denmark supplied the yardstick because it introduced organized screening region by region over a 17-year span, which created something close to a natural experiment. Diagnoses can be followed from before screening began, through the initial surge, to the years afterward, in a country where overdiagnosis is thought to be below 5%. The trials were then checked against that pattern: if a trial’s excess of cancers in the screened arm matched what the Danish curve predicted for its screening rounds and follow-up length, the trial was compatible with very low overdiagnosis.

Three things earlier estimates missed

Screening pulls cancers forward in time. When it starts, diagnoses jump because tumors that would have surfaced years later are found now. If the comparison stops too early, that temporary bulge reads as extra cancers that never would have appeared. Lynge explained the pattern to Healthcare in Europe: cases should fall back later as the advanced cancers of the unscreened group catch up, and researchers must allow for that to avoid overestimating overdiagnosis.

The team identified three factors that the earlier calculations handled unevenly: whether women in the control groups were offered screening after the trial ended, how many screening rounds each group received, and how long follow-up lasted. The University of Southern Denmark release, distributed through EurekAlert, says the older high estimates were “based on evidence before trial data had fully matured.”

Once those factors were accounted for, the abstract says, the mammography trials “yield consistent excess-incidence patterns compatible with very low overdiagnosis (<5%).” The abstract also describes earlier estimates as ranging from about 0% to 50%, attributing the spread to methodological inconsistencies and not to real variation between studies.

Questions the 5% figure leaves open

Rebolj put the conclusion this way in a statement carried by News-Medical: “When interpreted in their full temporal context, randomized trial data are consistent with overdiagnosis of less than five percent, rather than with estimates nearing 50%.” Njor added that the benefits of early detection and prevented premature death “will outweigh the small risk of unnecessary treatment.”

The claim is narrower than it sounds. It concerns what the randomized trials show. The earliest of them began in the 1960s, they ran with older imaging and treatment, and are not the same thing as modern screening programs. Models built for the U.S. Preventive Services Task Force reached a different number for contemporary practice: its 2024 recommendation for biennial screening at ages 40 to 74 cites roughly 14 overdiagnosed cancers per 1,000 women screened over a lifetime, with individual models spanning 4 to 37. The task force also flags ductal carcinoma in situ as a particular uncertainty.

Model estimates and the new trial-based figure answer related but different questions, and the reanalysis leaves the modeling standing. What it removes is the argument that the randomized trials themselves prove overdiagnosis of 30% to 50%.

The ScienceDaily version of the release reached general readers on September 24. What has not appeared in the sources reviewed is a published response from the researchers behind the higher estimates, and whether the Funen reference pattern holds up under their scrutiny is the next test of the below-5% claim. Njor’s stated aim is practical: accurate numbers so that women invited to screening can weigh the benefit of early detection against a small risk of unnecessary treatment, rather than a share of screen-detected cancers approaching half.

This article was produced with the assistance of AI and reviewed by Morning Overview editors prior to publication.