Overdiagnosis of Breast Cancer Is Clinically Insignificant

An Interview With Dr Stephen Feig

Tricia Ward; Stephen A. Feig, MD


August 21, 2015

Editor's Note:
Dr Saurabh Jha guest-edited the August 2015 print issue of Academic Radiology, devoted to overdiagnosis in radiology. He invited two authors to contribute opposing viewpoints on breast cancer screening. Dr Stephen Feig, interviewed below, does not believe that there is a significant overdiagnosis problem. For the interview with Dr Archie Bleyer, who believes that there is overdiagnosis, click here.

Medscape: You have been a vocal critic of the New England Journal of Medicine paper[1] by Bleyer and Welch that brought the issue of overdiagnosis and breast cancer screening to prominence. Why do you think their findings are incorrect?

Dr Feig: The best, most accurate, most scientific way to estimate overdiagnosis is through a randomized clinical trial, where you have two groups of women who are identical except that one group is offered screening and the other is not. Then, you measure the number of cancers detected in the study group that was offered screening (and that includes women who didn't get screened after being offered it) vs the number of cancers in the control group.

You have to have long-term follow up to do this because when you screen, you initially find more cancers. If you measure the number of cancers after year 1, 2, 3, 4, or even 5, you'll have more in the study group than in a control group. But if you follow these women for a long period—as was done in the Swedish Two-County Trial,[2,3] which followed women for over 30 years—then, when you look at the number of cancers, they're absolutely identical in the two groups, so there's no overdiagnosis. You initially have more in the screening group, but by the end of follow-up, they come back to normal.

Any other way of measuring is very hypothetical and inaccurate. The degree of inaccuracy with other methodologies, including the one used by Bleyer and Welch, can lead to totally misleading conclusions.

Yen and colleagues[4] looked at the Swedish Two-County Trial results by age. For women screened in their 40s, 50s, and 60s, they found absolutely no excess of cancers in the study group vs the control group. This refers to cancers, not cancer deaths—no more cancers were actually detected.

There was a slight excess with screening in the 70- to 74-year age group, but it was not statistically significant. If overdiagnosis exists, it's very small and is confined to older women.

Medscape: You think there is some overdiagnosis, just not at the levels cited in the New England Journal of Medicine article?

Dr Feig: If overdiagnosis occurs, it is at an extremely small rate—not 31%, as Bleyer and Welch suggested, but perhaps 1%, 2% at very most.

"Overdiagnosis" means that the cancer looks like a cancer under the microscope to the pathologist, but it might not kill a patient. Maybe the patient is old and dies of something else, such as cardiac disease; meanwhile, she was treated for cancer and she really didn't have to be. That's the concern of overdiagnosis—that women are being unnecessarily treated with radiation therapy and chemotherapy.

Medscape: Do you accept any data on overdiagnosis that are not from randomized clinical trials?

Dr Feig: You can also estimate overdiagnosis from trials that are not randomized, such as service screening studies (and there have been plenty of those). But you have to follow the women for long enough; when you screen women, you're going to find more cancers that you would normally see in an unscreened population for a couple of years because we're pulling cancers forward in time.

We're lowering the threshold of detection because we're seeing cancers before they can be palpated. When you do mammography, you find more cancers initially, but these cancers eventually, if unscreened, would be real cancers. They'd become palpable. When you use nonrandomized trial data, such as service screening studies, you have to be sure to have a long enough follow-up period.

Second, you have to account for lead time. Lead time is the amount of time that screening advances cancer detection— that is, how much earlier it was detected compared with when it would arise without screening. You have to make an adjustment for this. This is done by reputable statisticians and epidemiologists around the world.

You have to take into account differences in risk factors between a screened group and a nonscreened group and adjust for that. When you look at studies that did take lead time into account, had long enough follow-up, and took account of the differences in risk factors—and this has been done in a study by Puliti and colleagues[5] from Italy—the rate of overdiagnosis in the well-conducted studies may be about 2% or 3%. It's that low. The Bleyer and Welch article is a total outlier.

Medscape: One criticism of Bleyer and Welch's analysis by you and others is in reference to the estimated background annual increase in breast cancer incidence of 0.25%.

Dr Feig: They made one fatal flaw: They assumed that breast cancer incidence in the United States was increasing by only 0.25% per year, from the beginning of the screening year to the current day, and that's not true. If you look at the incidence of breast cancer in the United States, it fluctuates. It's higher some years vs others. It is gradually increasing. The reasons for this increase are unclear, but it may be on the basis of environment or diet—things like that. But it really is not on the basis of screening.

If you look at the 40-year trends using the Connecticut registry, which is the most accurate registry in the United States, the incidence increased by 1% per year.[6] Now if Bleyer and Welch had used that 1% increase per year in their article instead of 0.25%, that would have made all the difference. They would have shown no overdiagnosis. That was a major mistake, and you have to be careful with these studies because they have consequences. They can persuade women not to be screened.

Women can die because they don't get screened. It's not an academic exercise, which I think is not appreciated by a lot of the people who are raising an uproar about overdiagnosis rates of 30%. Women are going to read this. They won't have their cancers detected on screening and, frankly, they're going to die.

Medscape: Dr Bleyer addressed the background incidence rate in his article in Academic Radiology.[7] He doesn't agree with the 1% figure, but even when they used a background increase of 1% per year, they estimated that 31,000 women in the United States were overdiagnosed during 2011 (a rate of 12%).

Dr Feig: This is an oversimplistic and naive viewpoint, because epidemiology studies are very complex. Breast cancer rates are influenced by a lot of confounding factors. It's influenced by risk levels of the population and by diet. It's so complicated that the only way to accurately study it is with a randomized clinical trial. You can't say, I'm going to take country X, and this is a screening rate in that country, and this is a breast cancer death rate. It's too complicated for that.

You need a more controlled scientific trial to show that, and it's been done in study after study. We have trials from Europe and overall, among the women who are offered screening, there's a 30% reduction in breast cancer death rates, and among the women who are actually screened, it could be as high as 50%.[8] These are reputable studies done under controlled scientific conditions in numerous European countries. The way in which Bleyer and Welch did their study using this country and this screening rate is nonscientific; it's not even worthy of the name "science."

Medscape: Does the difference between a 30% vs 50% reduction in death rates with screening relate to women offered screening who do not undergo mammography?

Dr Feig: That's exactly it. When you take the women who were actually screened, it could be as high as 50% in some of these studies. Women who agree to be screened are in better health, and there are other factors that might lead to their more positive outcomes.

That's the benefit of randomized trials: They completely circumvent the hypothetical arguments that we're not really reducing deaths, but we're just finding the cancers earlier or finding slower-growing cancers because of screening. Randomized trials underestimate the benefit from screening because you're ignoring the women who are offered screening but don't get it.

Medscape: What do you say to the argument that overdiagnosis isn't a problem, it's overtreatment—you're discovering cancers that are very small and maybe shouldn't be treated?

Dr Feig: Cancers that are small will eventually become large cancers, because cancers grow. Ductal carcinoma in situ (DCIS), which might be the poster child for what they term "pseudocancer," represents a spectrum of disease. The best way to assess the aggressiveness of DCIS is to look at calcification, because calcification represents cell death and that's the more aggressive type.

My point is that when a radiologist calls something "possibly DCIS," this is at one end of the aggressive spectrum. We're not detecting that many cases of low-grade DCIS. High-grade DCIS is more apparent on mammography, and by and large, our detection of DCIS is skewed toward the more aggressive kind. This has been found in a lot of studies. Even in the United Kingdom (where the quality of mammography is not great), only about one third of the DCIS was low-grade and about two thirds was intermediate- or high-grade.[9]

That's similar to the data we have in the United States. In my paper in Academic Radiology,[10] I cited Mel Silverstein (a surgeon in California) and colleagues[11]; in his practice, at least 80% of the DCIS is intermediate-grade or high-grade.

Medscape: And intermediate- or high-grade DCIS will progress to breast cancer?

Dr Feig: Yes. Even low-grade cases are going to progress to medium-grade. I have done some studies showing that the proportions of high-grade cases increase over time.[12] In other words, low-grade turns into intermediate-grade because you have cell turnover. Cancers are a mixture of different cells; some are more aggressive than others. And what happens? The higher-grade cells will overpopulate the rest of the cells.

It's like in Darwinian evolution. The more aggressive cells grow faster and become the predominant cells over time. That's why even low-grade DCIS has the potential to become cancer, although not as fast as intermediate-grade and certainly not as fast as high-grade.

Medscape: If the lead time and follow-up time are not long enough, then you may not show this in a study?

Dr Feig: My point regarding lead time is that screening finds cancers. If you screen women in 2015, you'll find cancers that ordinarily wouldn't be apparent, clinically, until 2016, 2017, 2018, and later.

Let's say that without screening, the incidence is 100. Then we screen women, and we may get 200 cases. Some may say we doubled the rate, so the overdiagnosis rate is twofold. That's not true, because we're finding cancers that you wouldn't find for another few years.

Even without screening, the incidence would be higher next year and higher the year after that. These are the reasons we're finding more cancers than you would expect.

But overdiagnosis is very minor, negligible, and that's why I used the term "clinically insignificant" in the title of my paper. They can make all the fuss they want about overdiagnosis, but we have the proof that in randomized trials, screening has reduced breast cancer mortality. There's absolutely no way to negate that scientific fact.

Medscape: In terms of the cutoff age for breast cancer screening, do you support 40 years of age rather than 50 years?

Dr Feig: Every woman should begin by at least 40 years of age, because about 20% of all breast cancers in our country are found in women in their 40s. If you say that we're not going to screen women in their 40s, we are basically ignoring 20% of breast cancers.

Women in their 40s have normal life expectancies. If you find an early breast cancer at age 45 years, it has more significance than finding it in a woman aged 85 years, because that 85-year-old woman (not to diminish her) doesn't have that many more years of life expectancy.

But if you find cancer in a woman in her 40s, you're going to give so many more years of life to that woman. Not only is she younger, but she has the younger family.

Breast cancer incidence increases with age, but there is a precipitous increase in incidence between 35 and 40 years of age, so 40 years seems like a reasonable age at which to start screening women.

Medscape: What about the countries that use an age cutoff of 50 years (including the United Kingdom and Canada)[13,14] without an increase in breast cancer deaths?

Dr Feig: There are a lot of different factors between the countries, but we have reduced our breast cancer death rate more than these European countries. From studies in Sweden, when you look at women in their 40s in counties that had screening vs counties that didn't screen, the breast cancer death rate is lower with screening.[4]

Medscape: Could you comment on the paper in JAMA Internal Medicine showing an increased incidence of small breast cancers (< 2 cm), but no decreased incidence of large breast cancers, that attributes this difference to overdiagnosis?[15]

Dr Feig: That, again, is very conjectural, because those larger cancers could have been in women who weren't screened. The women who don't get screened get larger cancers. You don't see a decrease in the late-stage cancers right away. But there are a lot of factors. That is why what they're doing is basically amateur science.

Medscape: Do you think there's a role for patient discussions around the uncertainty about the age to start screening?

Dr Feig: That's tricky because it's a complicated subject, and if you say, "Well, we'll give patients a brochure or have them speak to a counselor," it could depend on the spin put on it. Is the average woman in a position to read every article in the scientific literature?

Medscape: What about being less aggressive in treating smaller cancers revealed on screening?

Dr Feig: In terms of treatment, it could be addressed with treatment that's tailored to the individual cancer. That really is the job of the pathologist, to look at the pathology and the molecular factors. That's an individual decision the patient should make.

Then, you have a limited number of patients who get a diagnosis of cancer. Do they get a mastectomy? Do they get a lumpectomy? Do they get chemotherapy? Do they get a contralateral mastectomy, or bilateral prophylactic mastectomy?

That's an individual decision. We are learning more about individual cancers by looking at molecular markers and assessing how to best measure the aggressiveness of a cancer. That's something that can be done.

People such as Bleyer and Welch say, "This is overtreatment. Women didn't need chemotherapy." But I would say, look at the other side of the coin: Look at the women in whom we're finding early cancers, and we're sparing them from aggressive chemotherapy. If a woman has a cancer found at a later stage because she's not screened, then certainly she'll need much more radiation therapy, much more chemotherapy. She'll need different chemotherapeutic drugs. Basically, we're sparing women from all of these treatments because we're finding cancers earlier.

Medscape: What message do you have for primary care physicians and the ob/gyn audience who are talking to these patients before they come to the radiologist?

Dr Feig: As primary care physicians, they have an opportunity to help save patients' lives by recommending screening mammography according to American Cancer Society guidelines. Patients listen to their physicians. These are complex issues, and they regard their primary care physician or their gynecologist as an authority.

I'd urge patients to make sure that they get a good doctor who's knowledgeable. Speak to them. Get a second opinion. For any patients diagnosed with DCIS, I'd recommend having their slides reviewed by several pathologists. Make sure it is DCIS, because there are a lot of borderline lesions. Get concurrence from several pathologists that you really do have DCIS and an assessment of how aggressive it is. And get opinions from several surgeons and oncologists on which chemotherapy or surgery you should get, if any.


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.