Commentary: screening mammography: a decision analysis.

over the credibility of the randomized screening mammography trials are red herrings. The overall results of the trials suggest that screening very likely reduces breast cancer mortality. Despite some peccadillos of the trials, I agree with the overall conclusion. I always have. So why have I been painted as being anti-screening? It is because I am less enthusiastic about the consequences of screening than are most medical researchers. And I believe that it is essential for women to understand and take into account the risks as well as the potential benefits of screening. 1 Regular screening may be a rational choice for some women and not for others. Estimated reductions in breast cancer mortality due to screening are modest. For example, the most positive results have been from the five randomized trials that were conducted in Sweden. Because of the evaluation bias, 1 the most appropriate assessment is via a so-called 'follow-up analysis'. Recent results of these trials 2 indicate that regular screening reduces breast cancer mortality by an estimated 11% for women aged 40–60 years. This is a relative risk reduction, which is meaningful statistically but has little relevance for women deciding whether screening is right for them. A better measure for use in decision making is absolute risk. An example measure of absolute risk reduction is the expected increment in life for a woman who gets regular screening. In the Swedish trials the average additional life per woman in her 40s or 50s (out to 18 years after randomization) was about 3 days. Comparable average increments in expected additional life could be achieved by, for example, losing an ounce of body weight (and keeping it off) or wearing a helmet for 10 hours of bicycling. 3 There are good reasons to think that this 11% reduction is an underestimate. One is that the design of the Swedish trials diluted any true benefit for screening. Consider a woman assigned to screening in the Gothenburg trial. She was scheduled to have five mammograms at 18-month intervals, over a period of 6 years. A woman in the control group was scheduled to have a mammogram as well, one timed to coincide with the screened woman's fifth mammogram at 6 years after randomization. After the sixth year, both women got mammograms or not as they wished. So the comparison was a prescription of five mammograms versus a prescription of one mammogram rather …