A recent news article in Science (2/19/2010, vol. 237, pp. 936-938) reviews the issues concerning breast-cancer screening for women between 40 and 50. What seems clear is that NOBODY involved in this controversy has done even the most rudimentary decision analysis, yet the question is obviously a good one for such analysis.
According to the article, the major issues are these:
1. Many cancers detected by mammograms in younger women will be detected anyway, even without mammograms. Of the cancers detected without mammograms, most will be cured. The upshot is that, while mammograms do detect cancers (and many women who have had cancers cured after detection with mammograms think that the mammograms saved them), they do not reduce deaths from cancer very much.
2. Still, mammograms do prevent cancer deaths. The effect is small, and its size varies from study to study. The lowest figure seems to be that cancer deaths (which are rare anyway) are reduced by 15%. Or, conservatively (it seems), one death is prevented for every 1900 women screened routinely between ages 40 and 50. (The article is a little unclear about this.) Screening at age 50 prevents one death for roughly every 1300 women screened.
3. So why not? The article puts aside the cost. Let's say the cost is $100. This comes to $100 times 1900, which comes to $190,000 to prevent one death. A very good deal if you ask me. (I suspect that standard insurance policies cover treatments that cost $100,000 to save one year of life. Surely prevention of cancer death in a 45-year-old will save many years of life.) So cost does not seem to be the problem.
4. The article also puts aside the increased risk from x-rays from repeated screening. I'll take their word for it that this is a small factor. (Wikipedia agrees, and cites a reference. Someone correct me if I'm wrong.)
5. The big trade-off according to the article is the risk of false positives. And the big problem with false positives is anxiety. Note that a biopsy is typically not immediate after a positive mammogram, so the anxiety can go on for some time, possibly weeks. The article gives a ratio of 5 biopsies for every case of cancer detected, which means that 4 are false positives. If 10% of these cancers would be fatal if undetected, then this means that the risk of a false positive is about 40 times the risk of death from undetected cancer. (I'm guessing at the fatality rate. It may be much lower.)
It is conceivable that some women would rather take the risk of undetected cancer than the much higher risk of anxiety from a false positive. I would think that research is needed that compares the relative utilities of death from cancer and anxiety from false positives. It may turn out that women differ enough so that they differ in the option that would maximize utility for each patient.
It seems likely to me that very few women would have a sufficiently high disutility for anxiety as to tilt the decision toward no screening after age 40. But the research remains to be done.
Again, what strikes me here is the total absence, so far as I can tell, of analysis of this question in terms of expected utility. Why is this idea so distant from even scholarly discourse?
I welcome comments by email, and I may post a follow-up: email@example.com.