Breast Screening Finds Fewer Harmless Cancers Than Past Estimates Suggested

Researchers in Denmark, Britain and the United States have re-read the randomized trials of mammography screening within a single framework and report that the trials are compatible with overdiagnosis below 5%, against published estimates that range from about zero to about 50%. The paper was published Sept. 14 in JNCI: Journal of the National Cancer Institute.

Overdiagnosis is not a false alarm and not a mistaken diagnosis. It is a cancer that screening correctly finds, and that is correctly called cancer, but that would never have caused symptoms or harm in that person's lifetime. The authors of the paper write that a coherent way of interpreting trial results is needed to aid clinical counseling and guideline development.

Sisse Helle Njor of the University Hospital of Southern Denmark and colleagues used one long-running screening program in Funen, Denmark, as a yardstick. They measured how the excess number of cancers among women invited to screening rises and falls over time there, then checked whether the trials followed the same pattern. The trial figures came from published reports found through PubMed and earlier reviews, so no one was newly screened for the work. Control groups that had been offered a final screen were re-counted as screened, one of the differences in method that the authors say produced the wide spread in earlier estimates.

Across 52 points in follow-up, 73 measurements closely matched what the Funen pattern predicted, the authors report, and the match held whether or not non-invasive growths in the milk ducts were counted as cancers. Where the numbers diverged, they write, incomplete reporting in the trials' early phases was a plausible explanation.

The estimate rests on trial results already published and on one national program as the reference for how excess diagnoses should rise and fall over time. The authors conclude that unadjusted counts of extra diagnoses in screened groups should not be used on their own as overdiagnosis estimates, and that a consistent framework supports balanced clinical and policy discussion.