The Background Review for the USPSTF Recommendation on Screening for Breast Cancer FREE

TO THE EDITOR:

Nelson and colleagues calculate the reduction in breast cancer mortality on the basis of randomized trials (1) but present no calculations for the most important harm: overdiagnosis. They quote a mixture of observational studies and results from statistical models with unverifiable and doubtful assumptions. Only for the Malmö trial do they quote real data, but these data were obtained 15 years after the trial ended, when many additional cases of cancer had occurred in both groups. When we corrected for this dilution and for the 24% of the women in the control group who were also screened during the trial, we found a 25% overdiagnosis rate (2), rather than the 10% reported by the authors.

It is indefensible that Nelson and colleagues base their estimate of overdiagnosis on flawed studies when data from about 600 000 randomly assigned women are available, which we pooled in our Cochrane systematic review in 2009 (3) and before that in 2006 and 2001. We found 31% more lumpectomies and mastectomies. In July, we reported an overdiagnosis rate of 52% in a systematic review of publicly organized mammography screening programs, without using assumptions or statistical modeling (4). It is also curious that Nelson and colleagues do not quote our Cochrane review, as they searched the Cochrane Library.

Nelson and colleagues reported that most studies found an overdiagnosis rate between 1% and 10%. However, systematic reviewing is not a consensus conference—it is a scientific discipline—and Nelson and colleagues overlook that the small estimates of overdiagnosis are based on poor science, mostly produced by researchers with vested interests in screening. If one opens one's eyes, unaided by any statistical tricks, one cannot escape seeing a huge amount of overdiagnosis, such as in the United Kingdom (4 - 5).

The Task Force now recommends against breast screening in women aged 40 to 49 years, but the harms may outweigh the benefits in all age groups. An effect of 15% and an overdiagnosis rate of 30% mean that for every 2000 women invited for screening throughout 10 years, 1 woman will have her life prolonged and 10 healthy women, who would not have breast cancer diagnosed if there had not been screening, will be treated unnecessarily (3). Furthermore, about 1000 women in the United States will have had a false-positive diagnosis (3). The psychological strain until one knows whether it was cancer can be severe. The harms caused by overdiagnosis are lifelong.

Karsten J. Jørgensen, MD

Peter C. Gøtzsche, DrMedSci

The Nordic Cochrane Centre

2100 Copenhagen, Denmark