More on mammography: just because you don't like the results doesn't make research junk science

A recent study published in the New England Journal of Medicine, “Effect of three decades of screening mammography on breast-cancer incidence”, by Archie Bleyer and H. Gilbert Welch[1], has generated enormous controversy. This has been caused by a combination of the study’s findings, the interpretation of them by the popular press, and the reactions of those who have a vested interest in the status quo – a combination that regularly occurs any time anyone publishes any research questioning the current conduct of screening or treatment for breast cancer (see, for example, my blog post Breast cancer screening: conflicting evidence? what are the important questions for health?, October 30, 2010). It happens in other areas, also, but breast cancer is the most common and in some sense most personal of cancers for women, and has a huge advocacy community, as well as powerful groups who profit from both treating it and screening for it.

What did the study show, what does it mean to people, what is the implication for cancer screening and most important, for the health of people (overwhelmingly women) who might get breast cancer? Before addressing these questions, I think it might be helpful to review a little about screening tests, cancer, and people’s hopes and beliefs. People want to not get sick, and especially don’t want to get cancer. If they do get it, they want to be treated and get all better. Of course, despite the use of “cancer” as if it were a single disease, and the existence of organizations such as the American Cancer Society, the federal National Cancer Institute, and the many Cancer Institutes, hospitals, and specialists, it is in fact a variety of diseases that all share certain characteristics but differ in many others. These include commonness, severity, cause, and likelihood of progression or death with or without treatment. Known causes for some cancers include smoking, radiation and viruses, and for many (including most breast cancer) the cause is unknown. It is even more complicated, because just naming the organ affected (breast, lung colon) is not all there is to it, as there are different kinds of cancer that affect the same organ. Whew. This is why the idea of “a cure for cancer” is unlikely; there are cures for some, and may be cures for others in the future, but there is unlikely to be “a” cure.

Some cancers, like breast cancer, are common enough, and well-publicized enough, that women realize that there is a real risk. In that case, the hope is that there exists a screening test that can identify it early enough to intervene and make a positive difference in the outcome. The first thing is that screening tests, by definition, are only for people who have no symptoms of a disease; once they do, a test, even if it is the same test, is no longer “screening” because the probability of the disease is greater in people with a symptom. For example, if one has a lump in the breast, a mammogram may be a good diagnostic test, but it is no longer a screening test. In looking at the criteria for a good screening test, there must be:

1)      A disease a test can screen for (while this seems obvious, doctors still do tests in asymptomatic people that do not effectively screen for any disease),
2)      A reasonable sensitivity and specificity to the test (meaning people with disease are more likely to have a positive test and those without the disease to have a negative test),
3)      A test that is reasonably cheap and acceptable to patients (tests like mammography and colonoscopy, for example, are both more expensive and more uncomfortable than, say, a blood test),
4)      A more definitive test available to say more definitely whether people who screen positive actually have the disease (for most screening, although those who screen positive are more likely to have the disease than those who do not, the majority of those who screen positive still may not have the disease),
5)      An intervention that can be done in the asymptomatic stage that will prevent the disease from progressing (or else, why not wait until it is symptomatic?)

So how does mammography stack up? This is a big part of what is addressed by the Bleyer and Welch study. They have looked at 30 years of screening in the US and found that screening mammograms have uncovered a large number of early-stage breast cancers; in fact, over that time, the number of early-stage breast cancers identified has doubled (from 112 to 234 cases per 100,000 women per year). This is a good – particularly if criterion #5, above, is met – and they can be treated and prevent women from dying or suffering serious morbidity. If this is happening, then (assuming the actual rate of cancer stays the same) the number of cancers diagnosed in later stages, where intervention is less successful, should go down. That is, those cancers detected early and treated would not progress and should mean that many fewer women present with later stage cancer.  Unfortunately, this study demonstrates, that has not occurred. The decrease in late-stage cancer diagnosis has been about 8%, or 8 per 100,000 women per year. So, for every 100,000 women, we are diagnosing an additional 122 early stage cancers, but only decreasing the number of late stage cancers by 8. This means that most of the additional women found by mammography to have early stage breast cancer would not have progressed to late-stage cancer. This, then, leads to their assertion that cancer was over-diagnosed – in 70,000 women in 2008 alone. Any estimate of the number of lives saved by screening and early intervention is inflated if it includes large numbers of women whose cancers would not have progressed. In other words, many of these women diagnosed with cancer, many of whom had non-trivial interventions (surgery, radiation chemotherapy) had cancers that would, basically, have not required any treatment.

Some radiologists who do mammograms have said that this is “junk science” (“Study links mammograms to overtreatment”, Boston Globe, November 21, 2012), but it is clearly not; the findings are the findings. The implications, however, are harder to assess. Does this mean women should not get mammograms? No, certainly that would be a premature conclusion. Some of the women diagnosed with early stage breast cancer would have gone on to develop late stage cancer; if you are one of the 8, you are lucky to have been found; if one of the 114, maybe not, especially if you had to endure the potential harms of chemotherapy or radiation or both, not to mention mastectomy. It may suggest that aggressive interpretation of mammography findings are not warranted. What would be useful would be to identify mammographic findings and subsequent pathology findings on biopsy that required aggressive intervention and those that could be safely followed. One type of breast cancer that is likely to be the subject of future studies is that called ductal carcinoma in situ, or DCIS, which may be more likely than some other types to resolve.

It may well be too soon to know the answer on mammographic screening, but it is clear that it is far from the perfect screening test that everyone would like it to be. We need more studies, and more information, and mostly we need a willingness to accept the accumulated findings of research. Certainly, what we do not need is for those who have a financial stake in screening and treatment to call good research “junk science” because it comes to conclusions that they do not like.

---

[1] Bleyer A, Welch HG, “Effect of three decades of screening mammography on breast-cancer incidence”, NEJM 22Nov2012;367(21):1998-2005.