Past estimates of how often mammograms find harmless breast cancers have ranged from nearly 0% to about 50%, a spread that has shaped screening debates for years. A reanalysis of randomized mammography trials, published in the Journal of the National Cancer Institute, found their results consistent with overdiagnosis below 5%. The authors are at the University of Southern Denmark, Queen Mary University of London, the University of Copenhagen and the American Cancer Society.
If you have put off mammograms, or spaced them out, because you heard screening often finds cancers that would never hurt you, this paper is worth bringing to your doctor when you decide whether, when and how often to screen. It does not pick a schedule for you, and it says nothing new about how many lives screening saves.
The finding lands in a live fight. In April, the American College of Physicians (ACP) issued new screening guidance. ACP listed overdiagnosis among the harms.
Key takeaways
- The randomized mammogram trials fit a pattern of overdiagnosis below 5%, not the estimates of up to about 50% from past analyses.
- Past estimates swung because trials were counted in different ways.
- This is a new look at old trials, not a new trial, and it did not measure deaths prevented.
What the study found
The eight trials were the New York Health Insurance Plan, Malmö, Two-County, Edinburgh, the Canadian National Breast Screening Study, Stockholm, Gothenburg and UK Age.
As a yardstick, the team used screening data from Funen, Denmark. There, researchers could compare women from birth years invited to screening with women from birth years that were not, and watch diagnoses jump when screening started, as cancers were found early, and then dip later.
The researchers compared the trials with that Danish pattern at 52 points in follow-up. The trials’ 73 measurements of extra cancers closely matched the Danish prediction, both for invasive breast cancer alone and with ductal carcinoma in situ (DCIS), an early form confined to the milk ducts. For most measurements, the range where the true value very likely falls included the prediction. The authors attribute the mismatches to incomplete reporting in the trials’ early years.
Trial data “are consistent with overdiagnosis of less than five percent, rather than with estimates nearing 50%,” said Matejka Rebolj, a senior epidemiologist at Queen Mary University of London.
Dr. Kumar’s take
Overdiagnosis is not a false alarm. It is a real cancer, found by screening, that would never have caused symptoms or threatened a woman’s life. It also covers women who die of something else soon after diagnosis. No test can tell which cancers fall in that group, so they are treated like any other.
The estimates swing because of how you count. Screening pulls diagnoses forward, so the screened group always shows extra cancers at first. That head start can be mislabeled as overdiagnosis depending on how long women were followed, how many screening rounds each group got, and whether the comparison group got mammograms after the trial ended, which was common.
This paper does not calculate a new overdiagnosis rate. It shows the trials fit the Danish pattern, then relies on Denmark’s under-5% estimate. If that estimate is off, so is this conclusion. One coauthor is an American Cancer Society scientist.
The study says nothing new about deaths prevented, false alarms or biopsies. And the balance still shifts with age. ACP notes that overdiagnosis becomes more likely as women get older, since fewer years remain for a slow cancer to matter.
Overdiagnosis is a large problem in some screening, such as the PSA blood test for prostate cancer. For mammograms, this reanalysis points to levels below 5%. It weakens one argument for starting later or screening less often, but it does not decide the right start age or schedule.
What it means for you
- If fear of treatment for a harmless cancer is the main reason you delayed mammograms, bring this paper (Njor and colleagues, JNCI) to your doctor.
- False positives and extra testing still belong in the discussion.