The background for this debate:
Last week The New England Journal of Medicine published a study with the potential to change both medical practice and public consciousness about mammograms.
Published on Thanksgiving Day, the research examined more than 30 years of United States health statistics to determine, through observation, if screening mammography has reduced breast cancer deaths. The researchers found that, as expected, the introduction of mammogram screening led to an increase in the number of breast cancers detected at an early stage.
But importantly, the number of cancers diagnosed at the advanced stage was essentially unchanged. If mammograms were really finding deadly cancers sooner (as suggested by the rise in early detection), then cases of advanced cancer should have been reduced in kind. But that didn’t happen. In other words, the researchers concluded, mammograms didn’t work.
This is a bold claim for an observational study. There are countless reasons why conclusions from such studies are commonly fraught with error. What if, for instance, the lion’s share of advanced cancers occurred among women without access to screening mammograms—a fact often not available in health statistics? Or what if mammography successfully prevented a major increase in advanced cancers, leaving the health statistics unchanged?
Hippocrates, the father of medicine, called experience “delusive.” He recognized that uncontrolled observations may lead to faulty conclusions. For centuries the flawed logic of observational data seemed to validate bloodletting, an unhelpful and often harmful therapy. But most who were bled eventually improved—no thanks to the bloodletting—an observation that led medical authorities to believe in the practice.
Fortunately, we have learned something about bad logic. Today we seek studies designed to neutralize illusions. By enrolling people in a study and assigning them randomly to treatments, for instance, groups tend to be evenly balanced in every way except one: the treatment. Controlled studies led to the discovery that bloodletting is harmful rather than helpful, and randomized trials of screening mammography would therefore be a worthy gold standard to answer once and for all the question of whether the test saves lives.
It may be surprising, therefore, to learn that numerous trials of mammography have indeed randomly assigned nearly 600,000 women to undergo either regular mammography screening or no screening. The results of more than a decade of follow-up on such studies, published more than 10 years ago, show that women in the mammogram group were just as likely to die as women in the no-mammogram group. The women having mammograms were, however, more likely to be treated for cancer and have surgeries like a mastectomy. (Some of the studies include trials from Norway, the Netherlands, Sweden, and this major review of the data.)
In other words, mammograms increased diagnoses and surgeries, but didn’t save lives—exactly what the researchers behind last week’s observational study concluded.
Those European trials and the recent observational study might not be enough to conclusively determine that mammograms don't do very much to cut back breast cancer mortality. But what if they are correct? What then?
These questions obviously matter. They bring up agonizing arguments about women losing the one tool of secondary prevention (i.e. screening) they thought they had against breast cancer. IF mammograms are of limited value in prolonging survival, what can women do instead? Just wait for a lump in the breast tissue?
But they also suggest that mammograms as a secondary prevention tool might never have done very much prevention, that the cancer survivors who thank an early mammogram and consequent early treatment for their survival may, in fact, have had one of the cancers which doesn't advance very rapidly or at all.
This is tricky ground, mostly because our understanding of the possible types of breast cancer is so limited. Perhaps some types of tumors remain small and never spread much? Perhaps some types of tumors might even disappear on their own? Perhaps the most aggressive (and often lethal) cancers develop so very rapidly that a mammogram is of value only if it happens to be performed at the exactly right moment?
There is something distasteful about those arguments. But on the opposite side, consider the potential distastefulness of spending vast amounts of money providing routine mammograms to all women above a certain age if mammograms, indeed, turn out not to have much value in reducing mortality rates. Those funds could be used for developing alternative screening methods or better treatments for breast cancer.
Then consider the radiation risks from mammograms every year or two and the financial costs and medical risks from potentially unnecessary treatments those mammograms may cause.
Clearly, something better than hope or entrenched industry interests is necessary to justify routine mammograms. What is needed are proper prospective randomized trials where women are randomly assigned to a group which will have regular mammograms and a group which is somehow kept from having them. Ideally, the two groups would otherwise be as similar to each other as possible. A follow-up of these groups for a few decades might give us decisive information about the value of mammograms.
But studies of that sort are an impossibility in most places. Mammograms are widely available and the control group could not be ethically kept away from having them, too. And I'm sure that many would find such a study unethical, because it would keep one group of women away from what is regarded as an established part of breast cancer prevention. (That's the snag, of course. If we truly are not sure about the mortality-reducing effects of mammograms, then they probably shouldn't be part of the established prevention pattern.)
All this may be premature. It could be that new, and better, studies find mammograms of greater value in reducing breast cancer mortality or it could be that better understanding of the etiology of the disease allows us to use mammograms more efficiently. Some alternative way of diagnosing breast cancer early might be developed and make mammograms but a part of medical history.
Still, when I was writing those musings I felt that odd kind of discomfort. Even wondering about the validity of mammograms makes me feel like a traitor, like someone who would abandon an orphan in a life-boat in the middle of the Atlantic Ocean, like a callous and mean-minded Echidne. But then looking at the other side of the arguments makes me feel pretty miserable, too.
Yet what ultimately matters is whether mammograms work or not. That should be distinguished from the general efficacy of various forms of medical screening. Some, such as the PAP-smears for cervical cancer, are very effective and fairly low-cost. Recommendations about others, such as prostate cancer screening, are now more qualified. We may simply not have sufficient information on breast cancer to make the best recommendations about who should get mammograms and when.
Where does that leave us, as individuals? Trying to muddle through the best we can, I guess, with the help of those health care professionals who base their advice on the most recent state of relevant knowledge.