Healthydebate.ca has run a series of stories on breast cancer screening mammography, stimulated by the recent guidelines from the Canadian Task Force on Preventive Health Care.
This last story focuses on how women and policy makers must balance the benefits and harms of screening mammography.
The issues raised in this series are relevant to screening for other diseases, such as PSA for prostate cancer.
The Canadian Task Force on Preventive Health Care recently issued recommendations about screening mammography for women at average risk of breast cancer. For women between 40 and 49 years of age, the Task Force recommended against screening. For women older than 50 years, they recommended screening every 2 to 3 years. The Task Force characterised both of these recommendations as “weak.”
The reason the Task Force gave for the difference in recommendations is that the risks of harm from regular screening mammography were felt to outweigh the benefits in most women under 50, but the benefits were felt to outweigh the harms in most women over 50.
The harms of screening mammography estimated by the Task Force are shown in the infographic below.
The importance of individual choice
Much has been written against and in support of the Task Force recommendations. Some have criticized the Task Force of under-estimating the benefits of screening mammography because they claim newer mammography technology will detect more treatable cancers. Others have suggested that the benefits of screening mammography were exaggerated because most of the improvement in deaths from breast cancer has been due to better treatment, not earlier detection.
The Task Force concluded that screening mammography in average risk women likely reduces the risk of dying from breast cancer. However, they note that, depending on the women’s age, between about 450 and 2100 women need to be screened for 11 years to prevent one death from breast cancer.
As well, some women who undergo screening mammography will have an unclear test result which will require further tests, some women will undergo biopsies that turn out not to find cancer, and others will have some cancers treated unnecessarily with surgery, radiation and chemotherapy.
Are the harms of an indeterminate or false positive mammogram worth it, to prevent a death from breast cancer?
Weighing the evidence
The relevant emphasis placed upon the benefits and harmsof screening mammography will vary from woman to woman. As Shelagh McRae, a family doctor on Manitoulin Island has pointed out, the impact of repeated testing might be quite different for someone in rural Ontario who has to drive many hours back and forth to appointments than for someone who lives in a big city.
Many argue that primary care physicians must provide women with accurate information about screening and support them in informed decision making about whether they wish to undergo screening mammography. However, the ability for primary care providers and patients to have accurate and detailed conversations about screening can be limited by time and other factors.
A dedicated screening mammography program is a population-level choice
In contrast to the individual-based decision making described above, organized screening programs generate general guidance for broad age categories, and provide simple messages encouraging screeening. These screening programs use approaches like educating health care providers about the benefits of screening for their patients, providing financial incentives to practitioners to achieve screening targets, and public education campaigns aimed at women and advising them to have a regular mammogram.
Dedicated screening programs consume limited public resources which could be used on other health care priorities. Therefore, it is important to ensure that these programs provide good value. In many ways this is a more difficult decision than an individual woman’s choice about mammography because policy makers are forced to make a decision that applies to a broad population.
It is much more difficult politically to stop a screening program once it is in place, than to decide not to institute one in the first place. Steven Lewis, a health policy expert in Saskatchewan noted that screening mammography programs were first developed decades ago when the evidence was different than it is now. He argues that based on current evidence, screening programs would likely not be instituted. However, Lewis also suggests that most governments wouldn’t want to face the backlash of ending an existing program.
Ontario has a population-wide screening mammography program focused on women aged 50 to 69. Given what is known about the benefits and harms of screening mammogrpahy in women of average risk, should Ontario stop the screening program altogether, provide additional information to women about the benefits and harms of screening mammography, more aggressively advertise screening or maintain the status quo?
The comments section is closed.
Andreas Laupacis and Karen Born have provided a nice summary of the logic for relying on randomized controlled trial evidence in the evaluation of screening, but I disagree with some aspects of their discussion. The comparison of randomized trials with non-randomized studies, specifically their example of the broad acceptance of the alleged benefits of post menopausal hormone therapy (HRT) based on inferential evidence, is instructive. As it turns out, when we eventually had evidence from a prospective randomized controlled trial, the long held assumptions about the benefits of HRT were stood on their heads.
However, the criticism of the Canadian Task Force on Preventive Health Care is based not only on their decision to consider only RCT evidence, but the fact that they relied on meta-analysis to estimate the effectiveness of mammography. Furthermore, they included data from studies that used mammography that is entirely inadequate by today’s standards. While it is quite reasonable to do meta-analysis to calculate a summary estimate of the breast cancer mortality reduction from all of the RCTs, it is another thing altogether to defend that it is the best, most unbiased estimate of the true benefit of screening. This makes about as much sense as concluding that the very best meal you can make from last week’s dinners is to put every remaining item in your refrigerator into one pot, especially when some of the older items show signs of wilting and mould.
The RCTs served us well by providing experimental evidence of the efficacy of mammography, free from known biases. With respect to age-specific benefits, meta-analysis was particularly important to determine the efficacy of screening women in their forties, since most trials were not designed with large enough sample sizes to measure this benefit. However, the trials had variable outcomes, and it is nonsense to assert that the differences in the RCTs are best addressed by averaging their outcomes. Further, it makes no sense to extrapolate from the trials that they are the best measure of real world outcomes since we would expect to see variable rates of compliance with screening recommendations in the population. This argument blurs the fundamental point of criticism of both the U.S. and Canadian Task Force conclusions. At issue is that they concluded that the individual benefit of screening was marginal in women under age 50 based on the results of a meta-analysis of intention to treat results, and reinforced that conclusion with improper estimations of the number needed to screen to save one life.
RCTs are generally analyzed according to the “intention to treat” principle. This term originates from the fact that the early RCTs were all treatment trials and participants were randomized according to the intention to treat them with the new therapy being tested or the conventional therapy (which could be no therapy). Randomization is done to insure that the two groups are at least nominally equivalent on all known and unknown factors that could, in addition to the protocol being tested, influence outcomes. Thus, the analysis plan is based on the original intent for the patient at the time of randomization, i.e., to be treated with the experimental protocol, or not. Ideally, randomization is effective, and all study participants are adherent with the protocol. Measuring the effectiveness of the intervention being tested depends on two important factors: 1) delivering the intervention properly, and 2) the fraction of the intervention group that actually receives the intervention. The converse, of course, is true for the control group. The first factor describes the efficacy of the therapy – does the intervention work? I will refer to the second factor as the “compliance” with the intervention.
It can be difficult to obtain high compliance in an RCT of a screening intervention. The reason for this is that the only situation in which it is ethical to perform an RCT is when the relative benefit provided by the interventions that are included in the study and control arms of the trial are not known. If this is the case, a participant is not being placed at a disadvantage by being assigned to a particular arm. But, in an RCT, precisely because the investigator does not know which arm provides greater benefit, it is not ethical to promote the intervention being tested to attempt to optimize compliance.
By the way, it is because we now know that mammography screening of women over the age of 40 is efficacious, that we almost certainly won’t have any more RCTs with “no screening” as an arm. Simply put, it would no longer be ethical to do this.
I contend that while it may be possible simultaneously to evaluate the efficacy and effectiveness of a new therapy in a RCT, this usually is not the case in a screening trial of an imaging intervention. In a treatment trial, participants with yet untreated disease are randomized to be treated with a new therapy vs. standard therapy. It is fair to say that they enter a RCT with the hope of receiving a treatment that they may believe is more effective than the standard therapy. The individual is, therefore, highly motivated to accept the intervention, and in the absence of serious side effects, to stick with it. A breast cancer screening trial is an effort to evaluate the efficacy of a screening intervention. A healthy woman agreeing to participate in a trial of an unproven screening modality may be doing so more as a gesture of generosity toward the advancement of medical science than as an act of self preservation. Thus, it is more likely for a participant to not comply with the assigned screening regimen in a RCT entirely, or for one or more of the screening rounds (inconvenient to attend the screening appointment, discomfort of compression, fears of radiation, etc.). Similarly, an individual assigned to the conventional therapy in a treatment trial likely will not have access to the new drug being tested, while a woman in the no-screening arm of a screening trial who wishes to receive mammography (perhaps a woman at higher risk) may be able to do this outside the study. This phenomenon is often referred to as “crossover”.
Thus, to those unacquainted with these methodological issues, it may seem strange to read that the outcome of RCTs are expressed as a mortality reduction associated with an “invitation to screening.” The implication may be that an invitation to screening and being screened are one and the same, but in fact in some RCTs significant proportions of the intervention group never received a mammogram, or did not receive one at each round of screening. Yet, according to the rules of an intention to treat analysis, the death rate is compared in the invited group with the control group, independent of the screening history of the individual participants. Although this may appear counter intuitive, the advantage is that the comparison holds known and unknown biases in check. The disadvantage is that we do not usually achieve a good measure of effectiveness, that is, the benefit that accrues to women who actually get a mammogram.
In a RCT, if benefits of an imaging screening intervention related to earlier detection of a cancer exist, they will only accrue to those who receive the intervention. If a trial suffers from a significant rate of noncompliance with the randomization assignment (ie, if women who are supposed to get mammograms don’t, and women who are not supposed to get mammograms do so), then the benefits of screening will be underestimated. Therefore, an attempt to evaluate effectiveness within a RCT of screening by analysis according the intention to treat principle is likely to be biased toward showing reduced benefit, unless compliance is perfect, which has not occurred in any trial. For example, in the UK Age Trial, nearly one-third of women randomized to an invitation to mammography did not accept the first invitation. Overall, only 81% of women in the intervention group received at least one mammogram, and the average attendance was only to approximately half of the invitations to mammography. While simply measuring the outcomes among those who were strictly adherent with the randomization assignment will be influenced by selection bias, methods have been developed that allow for a truer measure of the effectiveness of the intervention by adjusting for rate of acceptance of the invitation to screening, on the assumption that it would have been similar in the control group. In the recent publication of the UK Age Trial, the intention to treat analysis yielded a 17% mortality reduction, while the estimated mortality reduction among those who actually attended screening adjusted for non-compliance was 24% fewer breast cancer deaths. Even this estimate is an underestimate of the benefit that likely would accrue to women who attended screening regularly, because in the Age Trial. Only single-view mammography was performed after the initial screening round. This does not represent the standard of practice.
I contend that proper evaluation of the effectiveness (i.e. to include the compliance factor) of a screening intervention (one whose efficacy has already been established), requires that the prospective participants be made fully aware of the proven efficacy of that intervention, so that they can make an informed choice as to whether or not they want to be screened. This could not be done in an RCT.
Similarly, for women to be able to make informed decisions regarding participation in an organized screening program, they require accurate information regarding its strengths and limitations. By providing inaccurate data on cancer detection, false positives, negative biopsy rates and the number of women who must be screened to save a life the Canadian Task Force Recommendations do a disservice to women, by undervaluing the benefits and overstating the limitations of screening. By replicating these incorrect data in the figure in your article, Healthy Debate adds to the harm. Lets put the correct numbers on the table.
Martin J. Yaffe, PhD
Senior Scientist, Sunnybrook Research Institute
Professor, Medical Biophysics and Medical Imaging
University of Toronto
Director, Smarter Imaging Program
Ontario Institute for Cancer Research
I celebrated new years last night with two very life long friends. there were 8 women at the party and 3 of us have had breast cancer. friend one was diagnosed on her 40th birthday, friend number 2 at 43 and myself at 44 just this fall. clearly, i am in support on mammograms for women in their 40’s. it should be noted however, that 2 of us detected our cancers on our own and then sought out mammograms to investigate. the other friend only found out through mammogram.
We have all eaten well, exercised, not smoked, lived a healthy life. thank goodness for mammograms!
happy new year! laura