This article is more than 5 years old. Information may no longer be current.
Study results renew debate over value of mammography
Mammography has only marginally reduced the rate at which women present with advanced breast cancer and has had minimal effect on the rate of death from the disease, study findings published in The New England Journal of Medicine in November suggested.
The paper — along with an accompanying article in which three clinicians discuss if or when they would recommend mammograms for a woman aged 40 years deemed at average risk for breast cancer — renewed the debate over the value of mammography, long considered the gold standard for breast cancer detection.
“No one is denying that screening mammography doesn’t have some benefit,” Archie Bleyer, MD, clinical research professor at the Knight Cancer Institute at Oregon Health & Science University and co-author of the NEJM paper, told HemOnc Today. “The problem is that screening mammography has also created this situation of the overdiagnosis of breast cancer.”
Archie Bleyer
Ductal carcinoma in situ (DCIS) “hardly existed” before mammography, but it has since become something of an epidemic, Bleyer said.
“A significant proportion of DCIS is not cancer and does not need to be treated, but because it is called ‘carcinoma,’ we had to treat it. Millions of women have gotten therapy who did not need it,” said Bleyer, who also is professor of pediatrics at the University of Texas Medical School at Houston and chair of the Institutional Review Board for the St. Charles Health System in central Oregon.
Critics contend mammography has saved millions of lives, and they question the data and methods used by Bleyer and his co-author, H. Gilbert Welch, MD, MPH, professor at The Dartmouth Institute and The Geisel School of Medicine at Dartmouth.
William Farrar
“Screening is the only way to detect an abnormality in the breast,” William Farrar, MD, professor of surgery and chief of surgical oncology at the Ohio State University Comprehensive Cancer Center — James Cancer Hospital and Solove Research Institute, said in an interview. “Because of this, I am bothered by this talk of overdiagnosis and overtreatment. I don’t know how you can overdiagnose something without diagnosing it.”
HemOnc Today spoke with several clinicians about the benefits of mammography, its imperfections, and the future of breast cancer diagnosis and treatment.
Study findings
Bleyer and Welch used the SEER database to identify trends in rates of early-stage breast cancer — defined as DCIS and localized disease — and late-stage cancer — defined as regional and distant disease — among women aged at least 40 years from 1976 to 2008.
Their findings showed that, since the introduction of mammography in the United States, the number of early-stage breast cancers detected each year has doubled, from 112 to 234 cases per 100,000 women. This translates to an absolute increase of 122 cases per 100,000 women.
Bleyer and Welch observed a concomitant decrease of 8% for late-stage breast cancer presentations, from 102 to 94 cases per 100,000 women, for an absolute decrease of eight cases per 100,000 women.
“For a screening test to help patients, it must not only cause more early-stage cancers to be found, it must also cause fewer late-stage cancers to be found, demonstrating its ability to advance their time of diagnosis to an earlier time and earlier stage,” Bleyer said.
Based on the assumption of constant underlying disease burden, eight of the 122 additional early-stage cancers diagnosed could be expected to progress to advanced disease.
In their analysis, Bleyer and Welch excluded the transient excess incidence linked to hormone therapy and adjusted for trends in breast cancer incidence among women aged younger than 40 years. Based on results of this analysis, they estimated that breast cancer was overdiagnosed in 1.3 million women in the United States in the past 30 years.
They subsequently estimated that, in 2008, 70,000 women were overdiagnosed with breast cancer. That finding suggests overdiagnosis — defined as tumors that were detected on screening that never would have led to clinical symptoms — accounted for 31% of all breast cancers diagnosed that year.
Bleyer and Welch concluded that screening mammography has only marginally reduced diagnoses of advanced disease, and that screening has had little effect on breast cancer mortality.
“Moreover, compared to other diseases like prostate cancer, breast cancer requires more therapy,” Bleyer said. “Everybody gets surgery and radiation or a full mastectomy, like radical prostatectomy, and 5 or more years of hormone therapy. Some also receive chemotherapy and/or IV molecular therapy. The total treatment impact is more significant.”
Because proponents advocate for mammography screening in younger women, the potential harms of overtreatment of breast cancer are greater than for other diseases, Bleyer said.
“We have to acknowledge the difference between treating a 40-year-old woman for breast cancer and treating a 75-year-old man for prostate cancer,” he said. “Unnecessary treatment in that woman could negatively impact 25 or 30 years of her life.”
‘Harm of screening’
Bleyer said he knew the findings — specifically the idea of overdiagnosis — would cause controversy.
“Once I began to look at the numbers, I couldn’t walk away from it, and I couldn’t worry about the possible conflict,” he said. “Somebody needed to bring this to the attention to women, their spouses and their families.”
However, this was not the first paper to explore the idea of overdiagnosis.
In a study published in Archives of Internal Medicine in 2011, Welch and Brittney A. Frankel of the Dartmouth Institute for Health Policy and Clinical Practice concluded that “most women with screen-detected breast cancer have not had their life saved by screening. They are instead either diagnosed early, with no effect on their mortality, or overdiagnosed.”
Autier and colleagues published a study in the British Medical Journal in 2009 that arrived at a similar conclusion.
“The contrast between the time differences in implementation of mammography screening and the similarity in reductions in mortality between the country pairs suggest that screening did not play a direct part in the reductions in breast cancer mortality,” they wrote.
A 2009 NEJM paper by Kalager and colleagues concluded that “the availability of screening mammography was associated with a reduction in the rate of death from breast cancer, but the screening itself accounted for only about a third of the total reduction.”
Robert A. Smith, PhD, senior director of cancer screening in the cancer control science department at the ACS, acknowledged the growing body of evidence that suggests breast cancer may be overdiagnosed.
“It is important to appreciate that overdiagnosis is a statistical concept measured by comparing observed vs. expected incidence,” Smith told HemOnc Today. “There is no ability to distinguish an overdiagnosed cancer from a progressive cancer based on pathological criteria. It is a concern because an overdiagnosed tumor results in overtreatment. Thus, to the extent that it exists, it represents a harm of screening.”
However, estimates of how frequently breast cancer is overdiagnosed vary greatly, Smith said.
He cited data from Puliti and colleagues, who conducted a literature review of observational studies to provide estimates of breast cancer overdiagnosis in population-based mammography screening programs in Europe.
Results indicated 13 primary studies that reported 16 estimates of overdiagnosis that occurred in seven European countries. Unadjusted estimates of overdiagnosis ranged from 0% to 54%.
“Reported estimates adjusted for breast cancer risk and lead time were 2.8% in the Netherlands, 4.6% and 1% in Italy, 7% in Denmark, and 10% and 3.3% in England and Wales,” they wrote.
Puliti and colleagues concluded that the most plausible overdiagnosis estimates likely range from 1% to 10%.
“Substantially higher estimates of overdiagnosis reported in the literature are due to the lack of adjustment for breast cancer risk and/or lead time,” they wrote.
For Smith, differences often are attributed to whether researchers adjust for contemporaneous trends in incidence and lead time, allow adequate follow-up, and include DCIS in their estimates.
Bleyer and Welch did not adequately adjust for these factors, Smith said.
“The question to ask is: If there were no screening at all, would the incidence rate of breast cancer have stayed the same over time, risen or fallen?” Smith said. “If you underestimate the actual increase in screening independent of the screening effect, then you will incorrectly expect to observe lower incidence over time and attribute the excess to overdiagnosis.”
Smith also questioned whether there were cohort effects in the screened group.
“If incidence rates are rising in one age group and are not changing or falling in another age group, that will affect your estimate of the underlying trend in incidence over time,” he said. “We know screening increases incidence rates because it brings forward the detection of cases that would not be expected to be diagnosed for one or several years.”
Independent factors
Overdiagnosis is a concern with screening mammography, Diana L. Miglioretti, PhD, dean’s professor in biostatistics in the department of public health sciences at the University of California, Davis, School of Medicine, told HemOnc Today.
Diana L. Miglioretti
“However, I believe the estimates of overdiagnosis present in [the Bleyer and Welch] study are likely too high,” she said. “I do not think the authors adequately accounted for the background increase in breast cancer incidence that is unrelated to screening.”
Breast cancer incidence may be increasing due to greater prevalence of risk factors, such as obesity and older age at the birth of their first child, Miglioretti said. Advances in treatment also may play a role.
“As breast cancer treatment improves, the benefits of screening decrease,” Miglioretti said. “We are now able to successfully treat cancers at a later stage than before. The benefits of diagnosing breast cancer at a very early stage likely aren’t as great as before, when most of the clinical trials were conducted.”
Although Smith acknowledged that intuitively one could expect to see a larger decline in advanced breast cancer over time, he noted limitations in the methods Bleyer and Welch used to conduct their analysis.
“Trend analysis is not a good tool for measuring this kind of influence,” he said. “It does not allow you to measure the influence of screening on women who actually attend screening.”
Bleyer and Welch did not acknowledge that only about half of US women receive regular mammograms, Smith said.
“Not all women diagnosed with breast cancer have access to mammography, and not all women diagnosed with breast cancer are in an age group for which screening is recommended,” he added. “Trends in staging have also influenced the advanced cancer rate, and a sizable fraction of the increased incidence is being diagnosed as localized breast cancer. Bleyer and Welch’s expectations are appropriate if you have a closed system, where there is no change in incidence, everyone gets screened and screening has nearly perfect accuracy. The actual world is a lot more complicated.”
Technological advances
Sandhya Pruthi, MD, associate professor of medicine and past director of the Breast Diagnostic Clinic at Mayo Clinic, said clinicians must make sure patients understand there are limitations and potential risks associated with screening mammography.
Source: Photo courtesy of Sandhya Pruthi, MD, reprinted with permission.
The quality of mammography has improved with the advent of digital technology, Sandhya Pruthi, MD, associate professor of medicine and past director of the Breast Diagnostic Clinic at Mayo Clinic, said in an interview.
“A lot of [the data] reviewed by Drs. Bleyer and Welch were based on film-detected mammograms,” she said. “We have come a long way since then.”
Breast MRI is becoming more widely available, and this screening method may become the most attractive option, especially for women with dense breast tissue and who are at high risk, Pruthi said.
“It is the best test for very high-risk patients, with good sensitivity,” she said. “However, at about $4,000, it is expensive, and you can’t offer it to every woman. Also, there are no studies showing a decrease in mortality with breast MRI screening.”
Whole breast ultrasound also is expensive, but it could prove viable in the future, Pruthi said.
“The other new option is tomosynthesis, which involves a slightly higher radiation dose, but which has proven effective and sensitive,” she said.
Another tool in the clinical armamentarium is looking at tumors at the molecular level.
The Oncotype Dx test is now being evaluated in patients who have DCIS and may be able to identify which patients need further treatment outside of excision.
“We need to further evaluate this test, as well as other tests that can predict tumor behavior,” Farrar said. “This would eliminate unnecessary treatment but would also help define patients who would benefit. This would increase the success of screening.”
New adjunctive methods hold promise for increasing accuracy, especially in women with dense breasts, Smith said.
“This should be an ongoing goal,” he said. “We should be improving both the sensitivity and specificity of mammography, which would increase benefits and decrease harms.”
However, although improving accuracy is important, the development of more sensitive screening tools may not solve the problem of overdiagnosis, Miglioretti said.
“In contrast, it could make the problem worse by detecting even more unimportant lesions,” she said. “We need to develop tools that can determine which breast cancers, once detected, are likely to be harmful.”
There also should be a push to make sure a thorough and accurate family history is obtained when assessing breast cancer risk, Pruthi said.
“About half of women diagnosed with breast cancer do not have family history of breast cancer as an identifiable risk factor,” she said. “For these women, their screening mammogram was what led to the early detection of breast cancer, resulting in better prognosis and less need for aggressive therapy.”
Cost and clinical perspective
The mammography debate is similar to most others in the health care arena in that both individual costs and public health costs must be considered.
“From a public health perspective, we look at this in terms of overall benefits vs. overall harms when giving a screening test to many millions of asymptomatic women,” Miglioretti said. “We try to quantify how many women have to undergo harms to benefit others. Based on this, we try to decide what is best for the population as a whole.”
At the population level, most agree the benefits of screening mammography outweigh the harms in women aged 50 to 74 years, she said.
“It is not as clear for women aged 40 to 49 years, which is why I believe women in this age group should be encouraged to make a personal choice about what is right for them,” Miglioretti said.
Pruthi went a step further.
“Breast cancer is the leading cause of death for women in their 40s,” she said. “We know that these women are being diagnosed with cancers because of mammography. We are going to save lives by detecting cancers early, regardless of false positives, unnecessary mammograms or unnecessary biopsies.”
Because of this, Pruthi emphasized the value of keeping patients informed and employing a shared-decision approach.
“We need to make sure our patients understand that there are risks and limitations to having a mammogram,” Pruthi said. “They need to know that they may be called back for additional mammograms and/or need a biopsy. My experience has been that women are comfortable having an extra mammogram or biopsy to ensure that it was nothing.”
Smith put the cost debate in a larger perspective.
“In many ways, this is a unique tension in the United States because we make population-based recommendations, but, ultimately, women will make personal decisions about the balance of benefits and harms based on values and personal preferences,” he said. “Ideally, they will have clear, unambiguous information to help them make that decision. How to reconcile an academic judgment and opinion that the balance of benefits to harms is unfavorable with the possibility that a significant fraction of the target population may see things differently is something that is largely unexplored by any organization that issues guidelines.”
Lack of long-term data
Bleyer acknowledged many of the arguments against his position and welcomed the debate. He said he and Welch re-analyzed the data dozens of times before publication and did not put forth the study lightly.
“In the long run, if it is helpful — and it may take 10 to 15 years for the controversy to be debated and the truth to prevail — it will be worth it,” he said. “At the moment, though, there are still a lot of unresolved issues to be explored.”
Although Miglioretti disagrees with some of the findings, she said it is important that studies such as the one by Bleyer and Welch are conducted.
“We need to try to understand how frequently overdiagnosis occurs,” she said. “Unfortunately, it is not something we can directly measure.”
Pruthi agreed.
“We don’t have sufficient long-term data on outcomes on this term ‘overdiagnosis,’” she said. “We don’t have prospective data to be able to reliably counsel patients that, ‘This is the kind of cancer we are going to treat’ and ‘This is the kind of cancer we are not going to treat.’”
Smith offered suggestions for how future analyses should be conducted.
“These arguments about benefits and harms would be better served if there was greater emphasis on the importance of reducing the rate of harms by improving quality,” he said. “Thus far, the approach to reducing harms has been to reduce exposure to screening.”
Bleyer agreed that the accuracy of the screening process must be improved to determine whether a carcinoma must be treated and sufficiently reduce the number of women who have to be screened.
“At the advent of chemotherapy, we saw how much harm it could do,” he said. “We remembered the pledge we took as physicians, and we were afraid to use it. We have since learned that it is sometimes necessary to do some harm to do some good.
“It is a question of balance,” he added. “How much harm are we willing to accept? In the case of women who are treated who do not need to be treated, it’s pure harm. That is what we have to avoid.” – by Rob Volansky
References:
Autier P. BMJ. 2011;343:d4411.
Bleyer A. N Engl J Med. 2012;367:1998-2005.
Kalager M. N Engl J Med. 2010;363:1203-1210.
Puliti D. J Med Screen. 2012;19(Suppl 1):42-56.
Welch HG. Arch Intern Med. 2011;171:2043-2046.
For more information:
Archie Bleyer, MD, can be reached at St. Charles Medical Center, 2884 N.E. Neff Road, Bend, OR 97701; email: ableyer@stcharleshealthcare.org.
William Farrar, MD, can be reached at Surgical Oncology, 410 W. 10th Ave., N924 Doan Hall, Columbus, OH 43210; email: william.farrar@osumc.edu.
Diana L. Miglioretti, PhD, can be reached at 1 Shields Ave., Med Sci 1C, Room 144, Davis, CA 95616; email: dmiglioretti@ucdavis.edu.
Sandhya Pruthi, MD, can be reached at 200 1st St. SW, c/o Gonda Building, Breast Diagnostic Clinic, Mayo Clinic, Rochester, MN 55905; email: pruthi.sandhya@mayo.edu.
Robert A. Smith, PhD, can be reached at Cancer Control Science and Trends, American Cancer Society Inc., 250 Williams St., Atlanta, GA 30303; email: robert.smith@cancer.org.
Disclosure: Bleyer reports a consulting relationship with Sigma-Tau Pharmaceuticals. Farrar, Miglioretti, Pruthi and Smith report no relevant financial disclosures.
Perspective
Back to Top
This is an incredibly complicated and emotional issue for many in the breast cancer community. Two independent currents in American medicine over the past 3 decades have brought us to where we are today: (1) improvements in screening technology for breast cancer, and (2) better treatments for those diagnosed with breast cancer. Adjuvant chemotherapy, adjuvant hormonal therapy and adjuvant trastuzumab have changed the natural history of early-stage breast cancer dramatically for the better. Therefore, it is clear that women diagnosed with breast cancer face a far better prognosis at all stages of disease. However, we still do not know for sure which women diagnosed with early-stage breast cancer require treatment and which do not, especially in an era when screening modalities such as MRI find earlier- and earlier-stage cancers. Are our better results with adjuvant therapy for more advanced disease a reason to abandon widespread screening altogether? I think not. This is not a repeat of the prostate cancer story, where many men are diagnosed in their 70s and 80s with perhaps a decade of normal lifespan left. Many women are diagnosed with breast cancer in their 50s and 60s, with decades of potential life lost if their breast cancer in many cases is not detected and treated appropriately. We should continue to screen everyone with the emerging understanding that not everyone will require treatment, and step up our research into genomic and other types of assays that can guide us to determine which women require treatment and which do not. We are on our way there with the widespread adoption of Oncotype Dx, and the successors to this type of test will likely be even better at helping us determine who needs treatment. However, we should not abandon finding these women through screening simply because we lack complete discriminatory power in our assays at this point in time.
Perspective
Back to Top
While provocative, the analysis by Bleyer and Welch is flawed and misleading. They conclude that screening has had little impact on the decrease in breast cancer mortality and that nearly a third of all women diagnosed with cancer are treated unnecessarily. Their analysis is misleading for the following reasons:
• Their main data source is not specific to breast cancer screening. Their "best guess estimate," as they call it, is extrapolation from generalized health survey data that contain self-reported items on mammography use, but not actual screening mammography utilization data linked to tumor registries.
• Their critical assumption is that underlying incidence of breast cancer has remained constant since the 1970s, which we know is false. Breast cancer incidence has increased about 1% annually over this time period. They argue that because the annual percent change in breast cancer incidence for women aged younger than 40 years — those who do not undergo screening — was 0.25% per year, then incidence for all other age groups should be similar. As almost all breast cancer specialists know, breast cancer incidence increases with age, and that’s why screening is recommended for those over age 40.
• They assume that a later stage at diagnosis is required for breast cancer mortality. However, women with all stages of breast cancer may eventually die of the disease, even those with the earliest stages of cancer at the time of diagnosis. Catching the disease at earlier stages gives women the best chance to save their lives.
• While they report "absolute" numbers, none of the reported figures are actual calculations. They are estimates based on multiple assumptions that are then scaled to the entire population. They concede that all of these extrapolated numbers "warrant a more critical evaluation." Suffice it to say, with a few different assumptions, completely different "absolute" numbers and conclusions could have been reported.
Mammography still remains the most effective and cheapest form of breast cancer screening. It can detect about 85% of the breast cancers among the screening population and is the only screening test shown to save women’s lives through randomized clinical trials. However, it is not a perfect technology.
Supplemental screening with MRI has been shown to be cost-effective — with benefits outweighing the risks — in women with high personal lifetime risk of breast cancer, such as those with BRCA gene mutations. There is now a great amount of research effort being put forth to personalize screening for individual women based on risk factors, such as dense breasts. New technologies, including tomosynthesis or automated whole breast ultrasound, may be useful adjunct studies for women at increased cancer risk down the road, but the evidence is just not there yet to inform clinical practice. Mammography remains our best bet for detecting cancers early, when women can be more effectively treated and cured.
Perspective
Back to Top
The article by Bleyer and colleagues sheds more heat than light on the subject of screening mammography. It is already well known that we overdiagnose breast cancer, particularly ductal carcinoma in situ, which rarely is associated with breast cancer mortality. We also overdiagnose and overtreat hypertension and hypercholesterolemia. We also already know that for patients with a lower incidence of breast cancer, such as those aged 40 to 50 years, the ratio of additional tests and biopsies to cancer diagnoses is higher. We already know that some patients with screen-detected breast cancer who undergo breast surgery, as well as adjuvant hormonal and cytotoxic therapy, may never had died if their breast cancer were undiagnosed. What we don’t know is how to effectively differentiate the dangerous cancers from those with no threat. We clearly need better screening tests with higher positive and negative predictive values, and we need better prognostic factors for those cancers we do diagnose. We also need to know the death rate from diagnostic biopsies — which is presumably exceedingly low, although those estimates seem to be missing in many of the editorials on this subject. Might it be lower than fatalities due to hypertensive and lipid-lowering drugs? There has been a call for the implementation of risk-adjusted screening mammograms, especially for younger individuals. This makes sense, and this concept should be a part of the discussion physicians have with patients and should be available to the public as educational/decision-making tools. However, the majority of cancer diagnoses and deaths occur in patients with no or few of the classical risk factors, so from a public health standpoint, there is no scientifically proven set of criteria upon which to base such decision-making or to maximize the number of lives saved by screening. The Bleyer article does point out the disparity between the increase in cancers diagnosed with screening compared with the impact on the fraction diagnosed with earlier stages, but this negates the well-documented improvement in outcome in mammographically detected cancers adjusted for stage. In summary, estimates of the percent of overdiagnosed breast cancer, just as with overdiagnosed hypertension and hyperlipidemia, do little to help the clinician and patient decide about health screening if they are not accompanied by evidence-based stratification guidelines that are informed by quantitative benefits and harms of screening and treatment. Until such estimates are provided to the public, they will rightfully be confused and anxious. They deserve some light to accompany the heat.