Harald Weedon-Fekjær sits on the advisory board of BreastScreen Norway, the national program that has invited women aged 50 to 69 for a mammogram every two years since 1995. On October 15 the Oslo University Hospital biostatistician put his name to a paper estimating that the program overdiagnoses about one in every 86 women it screens, finding breast tumors that would never have troubled them in their lifetimes. He still supports screening, because he believes it saves lives. Asked by The Epoch Times how a woman ought to weigh one against the other, he answered in four words: “People have different values.”
The paper, written with Torunn Heggland, Lars Vatten and Signe Opdahl and published in the Journal of Medical Screening, exploits an accident of administration. Norway didn’t switch screening on everywhere at once; it began in four counties and spread over the following decade, so for years some women were being invited while their neighbors were not. The researchers fed national screening and cancer registry data from 1987 through 2019 into an extended age-period-cohort model and put a simple question to it about what happens once the invitations stop.
The question comes straight from the theory behind screening. A cancer caught early on a mammogram is supposed to be a cancer pulled forward in time, one that would have surfaced later anyway, which means that once women age out of the program at 70 their diagnoses should fall below the expected rate by enough to repay the surplus screening created. In Norway the diagnoses did fall. They did not fall far enough, and the gap was still open when the researchers followed women out to age 84.
What remains is overdiagnosis. The study estimates that invited women carried a lifetime risk of a breast cancer diagnosis 6.6 percent higher than women who had not been invited, which the authors translate into roughly 18 percent of screen-detected cancers, counting ductal carcinoma in situ (DCIS), or about one overdiagnosed woman for every 86 screened. Set against earlier studies’ estimates of the deaths screening prevents, that comes to roughly three women overdiagnosed for every breast cancer death averted. The abstract calls overdiagnosis “a considerable problem” and allows that it may be smaller than some earlier work suggested, a concession that is fair and also revealing, since 18 percent is the modest figure. Most of the overdiagnosed tumors, the authors think, grow very slowly. None of this can be read off an individual woman. Her pathology report doesn’t say whether she is the one in 86, and neither can her oncologist.
Three to one, again
Norway’s ratio has a familiar ring. In 2012 the British government asked Sir Michael Marmot to chair an independent review of the National Health Service screening program, and the panel concluded that screening cut breast cancer mortality by 20 percent in relative terms. For every 10,000 women invited from age 50 for 20 years, it estimated 43 breast cancer deaths prevented and 129 cancers overdiagnosed, which is three to one almost to the decimal. The panel then added a line about its own arithmetic that screening brochures have not tended to reprint, warning that “the figures quoted give a spurious impression of accuracy.”
The Cochrane review by Peter Gøtzsche and Karsten Jørgensen in 2013 was blunter. For every 2,000 women invited to screening over ten years, it estimated, one avoids dying of breast cancer, about ten healthy women are treated unnecessarily, and more than 200 go through significant psychological distress from false alarms. Its handling of trial quality deserved more attention than it got. Pooling the three trials it judged adequately randomized produced a relative risk of 0.90, an estimate not statistically distinguishable from no reduction at all, while the four with suboptimal randomization pooled to 0.75. The better-run group found the smaller benefit.
Twice the early cancers, barely fewer late ones
The American version of the ledger came from Archie Bleyer and H. Gilbert Welch, who worked through three decades of national registry data in the New England Journal of Medicine in 2012. As coverage of the paper laid out, early-stage breast cancer among American women 40 and older more than doubled between 1976 and 2008, from 112 to 234 cases per 100,000. If those extra early cancers were dangerous tumors caught in time, late-stage disease should have dropped by something like the same amount; it went from 102 to 94. Bleyer and Welch estimated that 1.3 million American women had been overdiagnosed over those 30 years, more than 70,000 of them in 2008 alone, or 31 percent of that year’s breast cancer diagnoses, and they credited most of the decline in breast cancer deaths to better treatment rather than to screening.
The American College of Radiology, the professional society for the physicians who read mammograms, called the paper “deeply flawed and misleading.” Daniel Kopans of Massachusetts General Hospital called overdiagnosis estimates “smoke and mirrors.”
The college reached for nearly identical language two years later, when the Canadian National Breast Screening Study published its 25-year follow-up in the BMJ. That trial randomized more than 89,000 women aged 40 to 59, and by the end 500 women in the mammography arm and 505 in the control arm had died of breast cancer. The authors put overdiagnosis at 22 percent of screen-detected cancers, about one for every 424 women screened. The college and the Society of Breast Imaging called the study “incredibly flawed and misleading” and said it should not be used to set screening policy. Kopans’s sharpest objection went beyond the “secondhand” machines he complained about: clinical breast exams performed before randomization, he argued, revealed which women had lumps, opening the door to sicker women being steered into the mammography arm and weighting the trial against screening. That is a serious charge about how the trial was run. It is not evidence that the missing benefit exists, and lead author Anthony Miller’s reply was that “mere detection of a cancer does not equate with benefit.”
Zero days, give or take
The most basic question a woman can ask is whether screening will help her live longer, and in 2023 Michael Bretthauer and colleagues tried to answer it for six common cancer tests in JAMA Internal Medicine. Mammography had two usable randomized trials, 73,634 women and 13 years of follow-up. The estimated lifetime gained was zero days, with a confidence interval running from 190 days lost to 237 days gained, and deaths from all causes were identical in both arms at 0.65 per 100 person-years.
Screening’s defenders answer, correctly, that those trials were too small to detect a modest effect on overall survival. The interval allows a modest gain. It equally allows a modest loss. What it does not allow anyone to say is that a lifespan benefit has been shown.
A yardstick that doesn’t predict deaths
The second paper in this month’s Journal of Medical Screening tests the measure screening programs most like to quote. Julia Geppert of the University of Warwick and 13 coauthors, among them Adam Brentnall of Queen Mary University of London, gathered 58 randomized screening trials covering ten cancer types and asked whether the trials that turned up more early-stage cancers were also the ones that cut cancer deaths. Across 43 trials, the correlation was negative 0.25, with a confidence interval spanning from negative 0.49 to 0.07, and the authors concluded that none of the three early-stage measures they examined showed a reliable correlation with cancer mortality. A trial that finds more early cancers, in other words, cannot be assumed on that basis to be saving lives. Brentnall argues that a fall in advanced cancers is the better early signal, and his group’s earlier work found that trials with fewer advanced cancers did tend to show fewer deaths.
That stricter yardstick is the one Grail’s Galleri blood test, which looks for more than 50 cancer types, was measured against in a British trial of more than 142,000 people. The test found additional cancers but did not significantly reduce combined stage III and IV diagnoses across the 12 cancers it targeted, although stage IV diagnoses alone were lower. In September an FDA advisory committee voted 7 to 2, with one abstention, that its benefits outweigh its risks. The agency has not made a final decision.
More scans, on the strength of models
None of this has slowed the push for more mammography. The US Preventive Services Task Force now recommends a scan every two years from 40 to 74, and the American College of Radiology wants one every year from 40 for as long as a woman stays healthy. Much of the quantitative case comes from simulation, including six National Cancer Institute-funded CISNET models published in JAMA alongside the Task Force’s 2024 recommendation. Under those models, biennial 3D mammography from 40 to 74 averts a median 8.2 breast cancer deaths per 1,000 women while generating 1,376 false-positive recalls, 201 benign biopsies and 14 overdiagnosed cancers, with the models’ overdiagnosis estimates ranging from 4 to 37. Going annual lifts averted deaths to 10.3 and the costs with them, to 2,096 false-positive recalls, 288 benign biopsies and 21 overdiagnoses. Stamatia Destounis, who chairs the radiology college’s breast imaging commission, wants patients to understand “the risks of biennial screening.” Karla Kerlikowske of the University of California, San Francisco, points out that “no randomized trial has ever shown that yearly mammograms prevent more deaths than screening every two years.”
The models count breast cancer deaths averted. The trials that tried to measure lifespan itself could not find the gain. Those are different measures, and the guidelines are built on the first.
What to do with a lesion once it is found is a separate question, and it is starting to get tested. The COMET trial randomized women with low-risk DCIS to active monitoring or standard treatment, and after two years invasive cancer had been diagnosed in the affected breast in 4.2 percent of the monitoring group and 5.9 percent of the standard-treatment group. Destounis cautioned that the result “does not establish that every patient with DCIS can safely forgo treatment indefinitely.” Weedon-Fekjær’s view was more human: “It would probably be very hard to leave a cancer untreated if you know about it.”
He offered one more sentence, about the women his program may have helped: “Mammography screening might have saved your life, but we can never know.” It holds just as well for the one woman in 86 on the other side of his ledger, diagnosed with a cancer that, by his own estimate, would never have found her, and with no way of learning that the estimate was about her.
Sources
- Heggland et al., breast cancer overdiagnosis in BreastScreen Norway, Journal of Medical Screening (2026)
- Geppert et al., early-stage cancers and cancer mortality across 58 screening trials, Journal of Medical Screening (2026)
- The Epoch Times, Mammograms Find Cancer Early, but How Many Lives Are Saved? (2026)
- JAMA Internal Medicine, Bretthauer et al., estimated lifetime gained with cancer screening tests (2023)
- JAMA, Trentham-Dietz et al., CISNET collaborative modeling of breast cancer screening strategies (2024)
- British Journal of Cancer, Independent UK Panel on Breast Cancer Screening, full report (Marmot, 2013)
- Cochrane, Gøtzsche and Jørgensen, screening for breast cancer with mammography (2013)
- AuntMinnie, Bleyer and Welch NEJM study on three decades of mammography and overdiagnosis (2012)
- AuntMinnie, Canadian National Breast Screening Study 25-year follow-up (2014)