Can We Unmuddle Mammography?
A new study of mammography, showing lack of survival benefit, has once again muddied these waters and muddled the relevant messaging. The study, generating considerable controversy, as has much prior research on the topic, looked at breast cancer mortality over a 25 year period in nearly 90,000 Canadian women assigned to mammography or usual medical care without mammography during the initial five years of the study period. There was no appreciable difference between groups.
Perhaps you see a major problem already. To study the effects of mammography, or any cancer screening, on mortality over time requires… time. Time goes by at its customary pace no matter the research goals. So, if it takes 25 years to get the desired data, the intervention needed to take place 25 years ago. And so, inescapably, this study is entirely blind to any advances in mammography technique, technology, or interpretation over the last 20 years at least. In medicine, two decades is just about forever.
Perhaps the value of mammography is perennially muddled — if just a bit less so than prostate cancer screening — for the most obvious of reasons. The truth is in the middle, between slam-dunk and fuhggeddaboudit. With the apparent exception of titillating (if not salacious) novels, we don’t tend to like shades of gray. But that’s where mammography falls. It’s pretty close to a toss-up.
There is a long history of research on the topic, and conclusions have been anything but consistent. Some studies suggest clear potential benefit for women who would not otherwise be screened. But, of course, women who would not be screened are apt to differ in a variety of ways from those who would — including, perhaps, their access to, and the quality of, primary medical care. Unbundling such influences is nearly impossible.
But, if, instead, you attempt to study women who would be screened anyway, how do you randomize them to a control group? What woman, inclined to get mammograms, would go without for 20 years for the sake of a clinical trial? Not very many I know.
Enrollment in a trial itself can exert an influence. Regardless of assignment to mammogram or control, there may be more attention to breast health and a greater likelihood of finding breast cancer early among all women participating in a study. This effect obscures any real world, and potentially important differences between intervention and control arms.
We are, as well, dependent on an imperfect technology. Even if finding breast cancer early through imaging is decisively beneficial, studies will produce variable results based on flawed imaging, variable performance of the same technology in different women, and variation in the quality of interpretation of mammograms. That much more so when today’s data are the product of mammography done 20 to 25 years ago. There have been improvements in scans, scanners, and the training of radiologists during that span.
And complicating things further, mammography is a source of radiation, and may be doing some direct harm as well as good.
The false positive error rate of mammography is notoriously high, and unavoidably so if we want to avoid false negatives. False positives occur when we think we’ve found cancer that isn’t there. It can be avoided by raising the bar, but then there is a risk of missing cancers that are there. We tend to favor the former error over the latter, and in the absence of perfect tests, are forced to choose.
We may have failed to translate good evidence into practice. Pre-menopausal mammography would likely be more useful if performed more than once a year. Breast cancer tends to be more aggressive and progress faster in younger women. Post-menopausal mammography might be just as useful done every other year. A one-size-fits-all approach may attenuate benefit and raise the rates of harm to both groups.
And then, perhaps most important: not all the trouble we find through screening deserves the attention it gets. Some tiny breast cancers, like the majority of prostate cancers, are destined to do nothing if just left alone. These are cases where cure is very likely to be worse than disease — but we are not good yet at differentiating. Doing so requires analysis at the level of histopathology (i.e., tissue and cell analysis), and molecular genetics. This can be done, but it’s not routine and our abilities in this space remain limited.
One very important issue routinely ignored when parsing the benefits of any cancer screening modality, mammography included, is that screening does not prevent cancer. The goal of screening is to find cancer early — which is generally much better than finding it late. But it’s not nearly as good as not getting it in the first place. The evidence is strong that optimal lifestyle practices can slash risk for all major chronic diseases, cancer included. Related evidence shows that lifestyle as medicine can modify gene expression in a manner projected to protect against cancer development and progression. DNA is not destiny. Dinner may be! There is interesting literature on the relevant timing as well. It may be the best way to improve breast cancer in women is to focus on healthy living in childhood. That we could dramatically lower rates of cancer overall by living well across the life span is all but undisputed.
There are many reasons why decisive evidence that mammography confers net survival benefit at the population level, or that it lacks benefit and should be abandoned — is elusive. The result is something of a muddle for epidemiology. Until technology, interpretation, application and histopathological confirmation all rise to consistently high standards, we can’t unmuddle mammography for populations.
But by combining what we know about the test with what you and your doctor know about you, a basis for a good decision should be at hand. Inquire about the technology, making sure it is state of the art. Ask about the training of the radiologist reading the film. Ask as well about plans for immediate next steps if the mammogram is abnormal. Good breast care centers follow up right away with additional testing to differentiate false from true positives. Ask whether screening has been personalized — taking into account your age, breast density, family history and risk profile.
Evidence-based recommendations about mammography for the population at large where one size must fit all are, for now, ineluctably muddled. By personalizing the decision, as good clinical medicine always should, we can, I believe, unmuddle things for you.
Dr. Katz was recently named one of the most influential people in Health and Fitness (#13) by Greatist.com. He has co-authored textbooks on both Preventive Medicine, and Clinical Epidemiology — the science of good clinical decisions. His new book,DISEASE PROOF, is available in bookstores nationwide and at:
Follow David Katz, M.D. on Twitter: www.twitter.com/DrDavidKatz