Does Screening Mammography Lead to Breast Cancer Overdiagnosis?

Laurie R. Margolies, MD; Saurabh Jha, MBBS, MRCS; Daniel B. Kopans, MD


January 20, 2017

This feature requires the newest version of Flash. You can download it here.

Laurie R. Margolies, MD: Hello. I am Laurie Margolies, associate professor of radiology at Mount Sinai Hospital. I am here today at the Medscape studio in New York City with two prominent radiologists who find themselves at odds on the controversies in screening mammography and whether overdiagnosis of breast cancer is real.

Joining me today is Saurabh Jha, MBBS, MRCS, associate professor of radiology at the University of Pennsylvania, where he specializes in cardiovascular CT and MRI. Dr Jha recently wrote a commentary for Medscape arguing that overdiagnosis is an inevitable result of screening. I am also joined by Daniel B. Kopans, MD, professor of radiology at Harvard Medical School and a pioneer in the field of breast imaging. He founded the breast imaging division within Massachusetts General Hospital in 1978 and is also the inventor of the Kopans wire, which was introduced in 1980 to help guide surgeons to lesions detected by screening mammography.

We are going to start by discussing the evidence for overdiagnosis. Dr Jha, will you explain why you believe that breast cancer is being overdiagnosed?

Dr Jha: Thank you, Dr Margolies. First, I just want to say that the term "overdiagnosis" is a very dreadful term; it implies failure in diagnosing. What we really have is failure in prognosticating. To start from the beginning, you have to realize that cancer is a spectrum, not a monolith, and you have slow-growing and fast-growing cancers. Once you accept that, then it is inevitable that any process to detect cancer will disproportionately detect slow-growing cancer. Of course, some of them will progress and cause the patient harm in their lifetime, but some might not. What is the evidence of overdiagnosis? The problem with overdiagnosis is that once a patient has been treated, it is impossible to tell whether (in foresight or in hindsight) they have been overdiagnosed. Overdiagnosis can only be found in populations as opposed to individuals.

It has been found not just in breast cancer but in other cancers that patients who have autopsies for reasons other than malignancy have an underlying reservoir of disease. Once you have established that as a biologic plausibility, then you are conscripted to using data obtained indirectly in order to show overdiagnosis.

The ideal way to show overdiagnosis would be with a randomized controlled trial (RCT) where you have patients in a screening arm and nonscreened arm and follow them for 35-40 years to completely tease out lead-time bias. The problem is that such an RCT is not feasible, so now you are down to secondary databases, which Dr Kopans will explain. I will then explain why I believe in their results to an extent.

Dr Margolies: Thank you, Dr Jha. Dr Kopans, in a commentary that you wrote for Medscape, you said that there are no data to show that mammography overdiagnoses invasive cancers. Will you expand on that?

Dr Kopans: Yes. First of all, I appreciate the opportunity to speak with the Medscape subscribers. This is a very important topic. I am a little surprised because Dr Jha has redefined overdiagnosis to a certain extent. The common understanding of overdiagnosis is that you would find cancers that would not become clinically evident during a patient's lifetime. In fact, folks have argued in support of overdiagnosis, claiming that cancers would disappear if left undetected by mammography.

I would point out that no one has ever seen a mammographically detected breast cancer disappear, and yet it has been argued in a paper[1] that tens of thousands of cancers would have disappeared. I think it is important to separate invasive breast cancers from ductal carcinoma in situ. There are legitimate issues to discuss about ductal carcinoma in situ, but I think we have agreed that we are just going to talk about invasive cancers today.

The argument that there are invasive cancers that would not become clinically evident is based on studies that I would argue should have never passed peer review because they are scientifically unsupported. The problem has been that people have tried to extrapolate what the incidence of breast cancer would have been in the absence of screening. I think we all agree that screening began in the mid-1980s. You can see that on this slide from the Surveillance, Epidemiology, and End Results (SEER) database (Figure). The Bleyer and Welch paper[1] suggested that breast cancer would have increased 0.25%, when, in fact, it actually increased at a much faster rate of 1% per year based on the background data from the Connecticut Tumor Registry.[2] If you used the correct extrapolation, there is no overdiagnosis from invasive cancers.

Figure. Age-adjusted SEER incidence rates by cancer site. SEER = Surveillance, Epidemiology, and End Results. Courtesy of Daniel B. Kopans, MD

Dr Margolies: How good are we at predicting the risk for overdiagnosis for a specific woman? You talked about populations, but if I am trying to tell my patient whether or not she should have a mammogram, how good are our prediction models at determining who is at high risk for an invasive breast cancer and who is not?

Dr Jha: That is an excellent question. I want to go back to Dr Kopans' point about redefining overdiagnosis. I think it needed to be done. Yes, overdiagnosis is the detection of cancer that would not cause a problem during a person's lifetime. That is essentially a failure to prognosticate, a failure to say which is the more aggressive and which is the less aggressive cancer.

It is a genuine problem that you cannot tell who is going to be overdiagnosed and who is not, even if you accept that there is overdiagnosis. How is knowing that there is overdiagnosis going to change anything? What it means is letting women know that mammograms can extend your longevity and save lives, but it can also lead to overdiagnosis and overtreatment.

Dr Margolies: Dr Kopans, would you like to respond?

Dr Kopans: The argument has been made by people who are strong advocates of overdiagnosis that mammographically detected invasive breast cancers would disappear on their own. I think we have no direct evidence of that. The arguments are all coming from indirect evidence. This time is too short to go into it in detail. If you use extrapolations that are based on guesses, which the paper that I referred to and subsequent papers have used, what they guessed would have been the incidence of breast cancer had there been no screening mammography.

Had they used 40 years of information, which we have from the Connecticut Tumor Registry, they would have seen that there is no overdiagnosis of invasive cancer. Instead, they used the most uncertain period in the SEER database, right after Happy Rockefeller and Betty Ford (the wives of the Vice-President and President of the United States) had breast cancer, when there was a flurry of screening. They used that to guesstimate what the incidence would have been, and they came out with overdiagnosis, which is simply not based on the scientific evidence.

Dr Margolies: Let us turn now to a related topic. There is considerable disagreement about the age at which women should begin screening. The American College of Radiology says that screening should start at age 40 years. The American Cancer Society says 45 years. The US Preventive Services Task Force (USPSTF) says 50 years.

Dr Jha, would beginning screening at a later age perhaps reduce the cases of overdiagnosis that you believe are occurring?

Dr Jha: Age cut-offs are always going to be arbitrary. There are two ways to look at this. The first is to give a diagnosis of cancer earliest in women who are younger in order to give them maximum life expectancy, maximum years of life added. The way that we are measuring success of screening is wrong. We should measure success of screening in terms of quality life years gained. In that regard, I would support an earlier phase of screening rather than a later phase. I want to make two other points. The USPSTF is a national treasure because they are so politically naive that they came up with their recommendations for screening in the midst of healthcare reform. Machiavelli would probably be turning in his grave. Having said that, there should be a specialist on the committee.

There are two types of biases that people have. One is that specialists are always biased, and the other is that specialists know everything. The pendulum in the middle is that you need to have primary care physicians, epidemiologists are also important (although Dr Kopans does not like them as much as I do), and you need to have specialists. You also need to have a uniform application of how you arrived at the figures, what the number of maximum false positives should be, and what that trade-off should be. And every organization—the American College of Radiology, the USPSTF, the American Cancer Society—should be explicit about that and should apply it uniformly to all forms of screening.

Dr Margolies: Thank you, Dr Jha. Dr Kopans, what is your take on this?

Dr Kopans: Dr Jha has raised a huge number of issues, but one thing I am happy to hear is that he agrees that we should start annual screening at the age of 40 years because that saves the most lives. The age at which to start is based on science. The science is based on RCTs, which for whatever reason only included women who were 40-74 years of age. One trial[3] included 39-year-old women, but 40 is not totally unreasonable.

Those data all show significant mortality reduction for screening starting at the age of 40 years. Absolutely no data show that anything changes at the age of 50 years. There is no scientific support for using the age of 50 years as a starting point. I agree with Dr Jha that that is completely arbitrary.

I think we can agree that women should be provided with accurate information. The information is that you save the most lives by starting annual screening at the age of 40 years. Let women decide whether the so-called harms of screening will affect their decision to participate in screening, not some panel using their own values and biases to make that decision. That is what I object to.

Dr Jha: The reason that the USPSTF's guidelines are important is that when they give a grade of A or B, they have to offer screening at 0 copayment. I don't like big government, but in a sense, we are bound to them because it's not really free-market healthcare. The arbitrary decisions that they make would be less arbitrary if they had more specialists and if they were more explicit with their calculations.

Dr Kopans: The problem is that the people who are trying to reduce access to screening don't want to say that they use numbers needed to screen, for example, which is a surrogate for cost of screening. If you tell women that it saves lives, but we don't want to pay for it, you would have an argument, and the people who are trying to reduce access to screening know that they would lose that argument.

Furthermore, the recall rate, which epidemiologists have called "false positive" (epidemiologists do that because they have to have those terms), has been misinterpreted by suggesting that women are being told that they have breast cancer. I saw in a blog that someone said that she was going to have her breasts removed as a result.

That is not true. They are called back for an ultrasound or a few extra mammographic images, and everything is fine in the vast majority. We should tell women the truth—that the harms are from primary recalls from screening. I would make two other points. Overdiagnosis is not because of screening, it is because pathologists cannot tell an indolent cancer from a less indolent caner. As for overtreatment, I have never treated a breast cancer. Overtreatment is done by oncologists.

Dr Margolies: One thing that strikes me is that we talk about the harm of overdiagnosis being related to recalls. We know that there is a significant recall rate, which is anywhere between 5% and 12% in many practices. Yet, all of these studies that have been analyzed by the American Cancer Society and the USPSTF were done with analog mammography. In the United States, there is almost no analog mammography being done. In fact, 25% of our mammograms are now digital breast tomosynthesis, which has been shown to significantly decrease recalls. Do you think that will have an impact on perhaps a revised USPSTF recommendation?

Dr Jha: Technology changes and treatment changes, which is why any analysis of the net benefits of screening that are derived from studies that were done 25-30 years ago may not be applicable today. There is a wide bandwidth of imprecision.

People do not like dealing with cost, but this is where this type of discussion will bring about greater honesty. The USPSTF, the American College of Radiology, or any organization should state up front what those maximum numbers should be and how they have arrived at those maximum numbers. As technology improves, the USPSTF should take into consideration the most recent data from the emerging technology and revise their calculations accordingly.

Dr Kopans: The American College of Radiology has said that recall rates should be 10% or less. We have tried for years to reduce the recall rate. We developed digital breast tomosynthesis in large part to find more cancers but also to reduce the recall rate, and it is very satisfying that that has happened.

We did a study at two of the largest Harvard teaching hospitals looking at women who died from breast cancer. It is called failure analysis—what went wrong. We looked at women who died from breast cancer, and more than 70% of the women who died from breast cancer in the Harvard hospitals (presumably getting the best treatment) were among the 20% of women who were not being screened.

That is not proof that it is screening, but it is more frosting on the cake. All of the scientifically derived data support the fact that screening is a major component in saving lives.

Dr Jha: Yes. One of the take-home messages from the Bleyer and Welch paper is that screening has had a net effect. That is probably the most incontrovertible proof that screening has had a net effect.

Where it gets messy is where treatment comes into it as well as deciphering the contributions of the two. I don't think it is unreasonable to study that, but the methods to study that are inherently imperfect, and it doesn't necessarily change the management other than the parsimonious information that we give to women who are invited to be screened.

Dr Kopans: You did not hear or did not acknowledge what I said. The observational studies in which women have access to treatment—all the ones that have looked at individual patient data, not just the SEER database, which has no information on mammography screening—show that women who participate in screening have a better outcome and longer survivals than women who do not.

Dr Jha: Dr Kopans' point is something that I have been writing about a lot. The measuring instruments—I don't mean the mammogram, but the studies used to show overdiagnosis—are, unfortunately, imperfect. They are imperfect because they rely on assumptions, and they are very sensitive for assumptions. You can see from the 2012 paper[1] that if you go from an APC (age-period cohort) of 0.25, then change to 0.5, the overdiagnosis rate changes from 31% to maybe 28% or 22%.

Dr Kopans: If you go up to 1%, there are actually fewer invasive cancers.

Dr Jha: The problem is that we are at an impasse because while I am sure that Dr Kopans could very well be correct, I am not that sure. I am only talking about myself, but the problem here is that empirically, this is going to be very, very difficult to resolve. The important thing to do at this point is not to let the data from overdiagnosis alter the screening paradigm.

Dr Margolies: Do you mean do not stop screening yearly beginning at age 40 years?

Dr Jha: No, do not stop offering screening as a national program.

Dr Kopans: If I am understanding you correctly, and based on Dr Margolies' question, we agree that annual screening starting at the age of 40 years saves the most lives. You should tell women that fact. Tell women to be aware that at least for your first screening, there is probably a 10% chance that you will be called back for a few extra pictures and an ultrasound. There is a 1%-2% chance that we will recommend an imaging-guided breast biopsy under needle under local anesthesia, which is a safe biopsy, and then let women choose. Everyone is saying that they agree about informed decision making, and yet, we are misinforming women.

Nothing happens at age 50 years for the task force to say that if you are in your 40s, you can talk it over with your doctor and make an informed decision, but at 50 years or older, you should be screened every 2 years. Everyone should talk it over with their doctor and know the upsides and downsides of screening. If women are presented with accurate information, they will make the decision that is best for them.

The problem is that these panels are making the decisions for women. Had Congress not intervened, the USPSTF would have ended payment for screening for women 40-49 years of age because they gave it a C rating. These are not just trivial pronouncements of inexpert panels. These have actual, real-life impact. You need to provide women with accurate information, and physicians need to critically examine the papers they read. The articles on breast cancer screening in the New England Journal of Medicine are embarrassing. These extrapolations guessing at what the use of mammography was, they have no data in these papers, and yet they are claiming that mammography is responsible for overdiagnosis. That needs to stop.

Dr Jha: I am a little uncomfortable that I am agreeing with Dr Kopans on almost everything. Access to healthcare these days means access to information as well. Screening is always going to be an individual decision, and there is nothing irrational about not wanting to be screened.

Where do we disagree? We disagree with the overdiagnosis. I am going to say that the overdiagnosis rate is somewhere between 15% and 25%. Dr Kopans will disagree with me, and the reason we will disagree is the way we think about the imperfection of the Bleyer and Welch study. The other reason is that other studies have confluence around the 20% increase. I grant that there is another factor in this, which is the lead time. Without actually knowing the lead time, we will never be able to precisely quantify the exact rate of overdiagnosis. My point is that we cannot perfectly quantify it. However, there is sufficient evidence that it is a problem that ought to be communicated to the patient. I don't think we ought to be precise about it, but it is still a possibility that should be mentioned.

Dr Margolies: How do we tell the patients that it is a population issue and that I can't tell if it relates to them? Dr Kopans, let us start with you.

Dr Kopans: You provide the patient with accurate information. The problem is that physicians have been derelict in critically reviewing publications that have become accepted when the science is not behind them. In medical school, we are taught to critically read papers. Yet, when we get out of medical school, most of us read the abstracts, and then we assume that because it passed peer review, the abstract must be correct.

You would probably be the first to agree with me that this is simply not true. Doctors have to read papers. It's very difficult for primary care doctors to take the time to go through all of the breast cancer screening papers, which is why they rely on supposed experts to advise them. I am one of those supposed experts, and I am perfectly happy if people read papers themselves, understand the papers, and make decisions based on them.

Most doctors are not. They are using the American College of Physicians, for example, to provide the primary care doctors with their guidelines. Radiologists rely on the American College of Radiology to decide about breast imaging. You have to know that there are probably biases in those situations. The physicians need to be informed, and if they don't understand it, they should admit that they don't understand it and send the patient to someone who does understand the data.

Dr Margolies: This has been an absolutely fascinating discussion. Dr Kopans, do you have any concluding remarks to share with our audience?

Dr Kopans: I think all three of us agree that you save the most lives by annual screening starting at the age of 40 years. I think women should be given that information as well as information on recall rates. I think overdiagnosis is manufactured, but you can suggest that there may be cancers that can get treated that maybe would not bother them during their lifetime, as long as you provide them with accurate, scientifically supported information.

Dr Margolies: Dr Jha, may I ask you to make a concluding statement as well?

Dr Jha: I would say that mammography saves lives and, in some cases, leads to overdiagnosis. You just tell women about these possibilities and encourage them to make their own decision. Once you do that, screening is very much a personal choice. It depends on one's risk preference and risk tolerance, and it is not something that can be mandated from government or demanded by the medical profession.

Dr Margolies: Thank you both for joining me for this fascinating debate. I think that our patients and viewers will have learned something. We can all learn that screening mammography saves lives. There are considerable debates that will persist for many, many years. I want to thank you for watching. Please be sure to tell us how you feel about this topic in the comments section of this video. We look forward to hearing from you.


Comments on Medscape are moderated and should be professional in tone and on topic. You must declare any conflicts of interest related to your comments and responses. Please see our Commenting Guide for further information. We reserve the right to remove posts at our sole discretion.