Mammography screening for breast cancer is now almost universal in developed countries. However, the net benefit of this intervention is not completely clear. Lives have undoubtedly been saved, but it is known that some screened women undergo un-necessary treatment for tumours that would not have caused problems for decades, and many others endure the distress of false positive results. In fact, the time interval between screenings leads to a bias in favour of detecting slow-growing and relatively benign tumours.

Peter Gøtzsche and Margrethe Nielsen from the Nordic Cochrane Centre, Copenhagen, Denmark have conducted a detailed assessment and meta-analysis of clinical trials of mammography screening, and this has now been published as a review in the acclaimed Cochrane Collaboration. They identified eleven such trials and subsequently excluded two small ones and one with inadequate randomisation; the eight included randomised controlled trials involved a total of 600,000 women from many countries. The cohorts recruited varied in age group and the randomisation conditions differed between the trials. Since the trials were conducted over a very long period – the oldest trial considered, the New York trial, recruited women between 1963 and 1966 – the reporting standards changed throughout, generally improving over time. The outcome measures reported were: mortality from breast cancer, from any cancer and from all causes; frequency of surgery and adjuvant chemotherapy; and harm caused by the mammography itself. Four included trials had sub-optimal randomisation, and all-cause mortality could not be analysed in these because of insufficient data.

The meta-analysis showed that results from all trials taken together did show a statistically significant decrease in mortality from breast cancer, although quite a modest one. However, the greatest effect was seen in those trials without adequate randomisation. The relative risk (RR) of death from breast cancer in the screened group was reported as 0.81 when all trials were aggregated, but as 0.75 for the trials classed as having sub-optimal randomisation. The effect of screening on the probability of death from all cancers or from all causes (including cancer) was not seen to be statistically significant. All the trials that considered one or more specific cancer treatments – radiotherapy, chemotherapy, mastectomy or lumpectomy – as outcomes found that the frequency of these was significantly increased in the screening group. There may be particular concerns here, as the treated group includes women with indolent tumours and perhaps some false positive cases, and no cancer treatment is without some risk.

Gøtzsche and Nielsen concluded that for every 2,000 women enrolled in a mammography programme over a period of ten years one case of life-threatening cancer would be detected. This difference is tiny because of the relative rarity of death from breast cancer in the age groups generally screened: it represents a 15% decrease in breast cancer mortality. However, ten women in the same group would be over-treated for cancers that would not necessarily cause problems, and a further 2,000 would undergo the distress of a false positive diagnosis. There can therefore be no consensus yet about the balance between the benefits and harms of breast cancer screening.

The authors advised further that women offered screening will need to be fully informed of this risk-benefit analysis as they decide whether to take it up. A leaflet explaining the issues to a lay audience is available on the authors’ website http://www.cochrane.dk/ with copies in thirteen mainly European languages.

Reference

Gøtzsche, P.C. and Nielsen, M. Screening for breast cancer with mammography. Cochrane Database of Systematic Reviews 2011, Issue 1. Art. No.: CD001877. DOI: 10.1002/14651858.CD001877.pub4.