| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Editorials |
1 Department of Radiology, Massachusetts General Hospital, Ambulatory Care Bldg, Suite 219, 15 Parkman St, Boston, MA 02114.
Index terms: Breast neoplasms • Breast radiography, utilization • Editorials
Editor's Note: A copy of this editorial was sent to Richard D. Klausner, MD, Director, National Cancer Institute, and Leon Gordis, MD, Conference and Panel Chairperson, offering each an opportunity to respond. We received no responses. Anthony V. Proto, MD
Introduction and Background
With the stroke of a press conference, the National Cancer Institute (NCI), in the spring of 1997, appeared to end one of the most contentious debates in modern medicine: whether or not to recommend mammography screening for women in their 40s. Since few details have been provided explaining the basis for the NCI reversal of its 1993 position, and its decision to support the screening recommendation of most other medical organizations, the decision itself has been labeled a political concession (1) because Congress became involved in the controversy. Supporters of screening have been labeled as greedy, or as riding "roughshod over scientific evidence" driven only by "emotion" and "commercial. . .compulsions" (2). Others have trivialized the benefit from screening younger women, characterizing the benefit as being so insignificant that a woman can decide whether to be screened by the flip of a coin (3). More recently, supporters of screening have been called "wishful thinkers" who have ignored the scientific evidence (4), and others have stated that those who recommend screening to women aged 40–49 years somehow wish to deprive them of "shared decision making" (5,6). Some analysts now suggest that screening women aged 40–49 is not worth the cost (7). Much of this response is related to the statement, and the information it contained, issued by a Consensus Development Conference (CDC) Panel on January 23, 1997 (8). The CDC had been convened to review the latest data on breast cancer screening and concluded that it could find no reason to support a recommendation in favor of screening women before age 50 years. The Panel statement formed the basis of many media reports. Its comments have been expanded to form continued arguments against screening women aged 40–49 (1–7).
The following has been written to provide additional perspective about the CDC and the Panel statement and to explain the response that occurred, and was prominently reported by the media, when the Panel released its statement on January 23, 1997 (8). The reader should be aware that the Panel statement has been rewritten several times since its initial release. It is the information that was provided to the media and transmitted to the public on January 23, 1997, that is the subject of my comments since it had the most impact. The publication of the final, revised Panel statement (9) in the fall of 1997—more than 6 months later—was barely noted by the media.
It is well known that I have been an advocate for screening women aged 40–49 years (10). This is based on the scientific evidence and what I believe to be the misinterpretation of that evidence by opponents of screening women in their 40s.
For the reader to understand my observations, a historical perspective is needed. I first became aware of the plan to convene the CDC in the spring of 1996. I had been trying since September 1995 to convince the new director of the NCI, Dr Richard Klausner, to revisit the NCI's 1993 position against screening women in their 40s. I had redoubled my effort when the latest data that were emerging from the screening trials in 1996 showed that the NCI requirement had been met and that there was statistically significant benefit for these women (11). Although initially not invited, I ultimately gained an appointment to the Planning Committee that proposed an agenda and topics for the CDC and was subsequently invited as a speaker at the conference itself to provide information in support of screening women aged 40–49 years. I attended the entire CDC.
In my opinion, the Panel statement has never been critically reviewed either in the scientific literature or by the media. The subsequent pronouncements have all been based on the Panel's conclusions, with little or no reference to the data that the Panel had been asked to incorporate into their analysis. Compounding the confusion has been the fact that since 1993 the preponderance of information on the screening controversy, published in several of the closely monitored medical journals that are reported on by the media, has been arguments against screening women aged 40–49 (1–7,12–27). Limited material has been published in those journals to provide arguments in support of recommending screening (28–30). The New England Journal of Medicine has even published an editorial defending the fact that they did not feel an obligation to publish balanced information on controversial topics (31).
The Controversy Has No Biological Basis
The scientific evidence clearly shows that the original controversy, based on an analysis of women aged 40–49 years as a separate group, was in fact artificial. The age of 50 years has no biological importance and at best was chosen as a surrogate for menopause. None of the parameters of screening change abruptly at the age of 50. This fact was presented to the CDC and has been reaffirmed most recently in a review of the largest modern screening program in the United States (32). The perpetuation of this age as a threshold was an artifact of repeated data analyses that group women aged 40–49 as if they were a uniform population and compare them to women aged 50 and over as if they too were a uniform group. This has the effect of making factors that actually change steadily with increasing age appear to change abruptly at the age of 50 (33).
Furthermore, the randomized controlled trials of screening were not designed to be broken into age subgroups. They lacked the statistical power to permit accurate, unplanned, subgroup analysis of women aged 40–49 separately (34). The use of unplanned subgroup analysis of data lacking statistical power to advise women has never been justified. Despite warnings concerning the pitfalls of using this approach (35,36), and the inappropriateness of using these types of analyses for making medical recommendations, this type of evaluation was used by the NCI in 1993 as the basis for its withdrawal of support for screening women aged 40–49.
Why a CDC?
With longer follow-up of the women in the screening trials, it became evident that with an increase in the numbers of breast cancer deaths in the trials, the benefit had become statistically significant (37). The CDC was convened at the request of the director of the NCI specifically to review the newest data. In 1994, the NCI had been criticized for not having followed an objective procedure when it withdrew support for screening these younger women (38). The NCI had based its 1993 decision on the summary report from the International Workshop on Screening for Breast Cancer that had been held in February of that year. That Workshop, and the report that summarized it (17), was, in my opinion, grossly imbalanced and biased in its review against screening women in their 40s (39). This view was also expressed by Dr Charles Smart, the former head of the Detection Branch at the NCI, when he testified before a Congressional Committee that reviewed the 1993 decision (38). That Congressional review labeled the NCI's 1993 decision to drop screening guidelines for women aged 40–49 as "misused science" (38). For the first time in the history of the NCI, the advice of the National Cancer Advisory Board (NCAB), which had voted 14–1 to not change the guidelines, was ignored (38). Not only did the NCI drop support for screening women aged 40–49 against the nearly unanimous advice of the NCAB, but, contrary to the advice of the Board of Scientific Counselors (40) as well as the NCAB, the interval for women aged 50 and over was extended to every 1–2 years. On the basis of its use of unplanned subgroup analysis of the data from trials that were not designed to permit such analysis, the NCI had concluded at the time that, despite the fact that five of the eight trials were showing a mortality reduction for women in their 40s, since the benefit was not statistically significant, they could not support screening these younger women.
Unfortunately, the NCI had neglected to inform women and their physicians that it was mathematically impossible for the trials to show a statistically significant mortality reduction of an expected approximately 25% for the subgroup of women aged 40–49 in the early years of follow-up. The NCI had been alerted to this fact in 1993 (41), and it was subsequently confirmed by the NCI (42) and the International Union Against Cancer (43), but this fact was never conveyed to women and their physicians by the NCI. The NCI apparently ignored the fact that if the trials were analyzed as they were designed and not broken into subgroups by age, there was statistically significant support for screening beginning by the age of 40 (44).
Dr Richard Klausner, who became NCI director in 1995, sought an unbiased review of the new data and the NCI policy. He turned to the National Institutes of Health (NIH), of which NCI is one part, to conduct the CDC. CDCs are organized by the Office of Medical Applications of Research (OMAR) of the NIH. The CDC on Breast Cancer Screening for Women Ages 40–49 was convened with the intended goal of conducting an objective review of the new data. As an office in the NIH, OMAR is separate from the NCI.
By OMAR's own description, a CDC is organized in a similar fashion to that of a jury trial (45). "Testimony" is given by both advocates and opponents as presentations to a panel that acts as a "jury" to review the available data and reach a consensus on the issue. The panel members are to be "carefully screened to exclude anyone with scientific or commercial conflicts of interest" and are to be chosen so as to "not be professionally identified with advocacy or promotional positions with respect to the conference topic" (46).
For 1
days, the 13-member Panel was presented the results from the longer follow-up of the randomized controlled trials of screening as well as data from other "observational" studies. At the conference, the Panel was informed that, despite the fact that the trials were not designed for age stratification, the benefit for women aged 40–49 had become statistically significant (47). Despite all of the turmoil that resulted from the Panel statement, this fact was never directly reported to women or to their physicians. It was not even mentioned in the CDC Panel statement presented to the media on January 23, 1997. The response to the Panel statement at the end of the CDC was one of "outrage" as reported by the media (48). There was very little, however, that was published to explain that response except to suggest that it was purely emotional as has been suggested by subsequent authors (2,4).
The consternation was due to the fact that the Panel had in their conclusions not only appeared to have ignored virtually all of the recent data that they had been charged to review, but also they did not even mention the new data. They appeared to have based their decision on old data. The Panel statement made no mention of the fact that investigators from the major screening programs in Sweden had reported to the Panel mortality reductions of 44% in Gothenburg (more recently increased to 45% [49]) and 35% in Malmo (50), each statistically significant by itself. There are only two trials that by themselves have statistically significant results for women aged 50 and over. Nevertheless, paradoxically, the Panel acknowledged a benefit and urged screening for women aged 50 and over, but not for younger women. They never mentioned that the new data revealed mortality reductions for younger women that are actually greater than those found in these two trials that show statistically significant benefit for women aged 50 and over.
When the data from the trials were combined, the overview of the Swedish trials revealed a 29% mortality reduction for women aged 40–49 that is also statistically significant. Adding the Edinburgh and Health Insurance Plan of New York trials to combine all of the population-based trials of invitation to screening provided a 26% reduction for women in their 40s, which is also statistically significant (47).
The Trials Underestimate the Benefit
Not only were the results all the more compelling in that these randomized controlled trials were not designed to be split into subgroups, and they were not optimized for detection of early cancers in younger women, but also the results actually underestimate the benefit of screening due to noncompliance and contamination. Legitimate trial analysis requires that women who refused the invitation to be screened and died of breast cancer are still counted as having been screened (noncompliance), and women in the control groups who may have been saved by undergoing mammography on their own outside the trials are still counted as unscreened controls (contamination). This must be done to avoid selection bias, but both effects mean that the trials likely underestimate the actual benefit from screening.
Feig (51) has estimated, for example, that if all of the women had attended the Gothenburg screening program rather than the 75%–85% who actually participated, the mortality reduction among the women under age 50 in that trial might have been as high as 55%.
The Results Are Significant Even if the Canadian Trail Is Included
The data remained statistically significant even when the results from the National Breast Screening Study of Canada (NBSS) were combined with population-based trials. There is good reason to exclude the NBSS from the analysis. Unlike the other trials, the NBSS was a trial that screened volunteers, so that the population was self-selected. There have been major concerns raised over the design and execution of the NBSS; these concerns included the quality of the mammography, unblinded randomization, as well as the paradoxical excess of cancer deaths among the screened women, and the remarkable, and unexplained, survival of the women in the control group (52–54). The excess of cancer deaths in the NBSS, that was reported to the CDC, with 10 years of follow-up, remains unexplained (55) and differs from all of the other trials. Nevertheless, despite all of the problems with the NBSS, even including its poor results, the overall benefit for all of the trials combined remains statistically significant.
The excess of cancer deaths in the NBSS dilutes the benefit of the other trials. It is the diluted benefit of an 18% mortality reduction that was reported in a preponderance of the media coverage and alluded to by Pauker and Kassirer (3) as suggesting little more than a coin toss decision for women. Even if the benefit is "only" an 18% mortality reduction, this means that almost 1,000 lives can be saved of the approximately 5,700 women who die each year from breast cancer while in their 40s. This does not include a percentage of the 6,500 women who die in their 50s who might be saved if their cancer were diagnosed with breast cancer screening while in their 40s.
In their effort to suggest that screening women aged 40–49 was not cost-effective, Salzmann et al (7) estimated the overall benefit to be even lower at only 16%, the figure that is included in the final Panel statement (9). Instead of giving a range of estimates based on the range of benefit seen in the trials, the authors used only the lowest number to decide that the cost-benefit ratio was too high to make screening worthwhile for younger women and, once again, using inappropriate age grouping, ignored the fact that none of the parameters of screening change abruptly at age 50.
Without ever being asked for their reasons for the harsh criticism of the Panel statement, critics have been castigated and dismissed as having vested interests, since the Panel is supposed to consist of unbiased experts in the field. To the contrary, in an effort to avoid having anyone on the Panel who might appear to have a preconceived bias, based on the guidelines for selecting panel members for CDCs, the members were chosen specifically because they lacked any prior involvement with the subject. Thus, I believe that some of the Panel members also lacked the expertise to evaluate the topic. In addition, I had raised major concerns in writing many months prior to the conference concerning the imbalance of its formulation and the selection of the Panel.
Failures of the CDC Panel
The Panel stated that it wished women to make up their own minds based on the facts as to whether or not to participate in screening, implying that somehow those who thought women should be advised to participate in screening were trying to deprive them of the ability to decide for themselves. This theme has been repeated in more recent reviews (4,5). This was just one of the elements that fueled the reaction of the participants to the Panel statement. On the one hand, the Panel argued appropriately for informed decision making, while on the other hand the CDC Panel never explained its decision not to provide women with the recent data that had been provided to the Panel.
The Panel never explained its decision to require that a benefit from the trials must appear with only 7 years of follow-up. The 7-year data were little more than the same data that were analyzed in 1993. The Panel never explained reliance on early data that lacked statistical power when greater power was achieved with longer follow-up. The purpose of the CDC was to analyze the new data. The Panel seemed to trivialize the new data without explanation.
The statement by the Panel demonstrated, in my opinion, its major failure to follow its charge and its own stated guidelines. Many of the Panel statements appeared to have no basis in fact. Despite the fact that the charge to the Panel was to determine how risks and benefits "change with age" (9), the Panel statement repeatedly gave estimates out of context and only for women in their 40s. Any discussion of age either compared women aged 40–49 years to all women aged 50 years and over (grossly skewing the comparison), or selectively compared them to women aged 50–59.
Other Panel statements were simply incorrect, in my opinion. The original Panel statement suggested that the studies may show a benefit that is at most "as great as 30%" (8) when, in fact, the data provided from Gothenburg showed a statistically significant 44% decrease in breast cancer deaths and Malmo a 35% decrease in deaths for these women.
The Panel stated that none of the trials showed a benefit until 7 or more years, yet the benefit from the HIP, Malmo, and Gothenburg trials began to appear by years 5–7.
The Panel suggested that any benefit might be due to women reaching the age of 50 during the trials and having a benefit suddenly appear. Not only did the Panel fail to explain the biological importance of age 50 (there is none), but they ignored the repeated warning that analyzing trials by the age at diagnosis introduces major bias against detection of cancers in women under the age of 50 (56,57). Even overlooking this fact, three of the trials provided analyses by age at diagnosis and the Panel was provided with data that showed that the majority of the benefit was from cancers detected before the women reached the age of 50. In the HIP trial, the benefit was actually greatest for women aged 40–44 (a 30% mortality reduction) who never reached the age of 50 during the 4 years of screening (58). The Kopparberg trial showed that the majority of the benefit for women in their 40s was for women diagnosed while in their 40s (59), and the Gothenburg trial presented the data showing the same result (49). Despite these facts, women were not informed of the same.
The Panel statement suggested that the benefit may have been due to clinical breast examination among the screened women. This was one of the more astonishing pieces of misinformation because the Swedish trials, which showed a 29% statistically significant benefit, did not provide clinical breast examination (60).
The statement repeatedly provided analysis out of context. The Panel suggested that only two women in 1,000 would have their lives extended if screened in their 40s (8). They neglected to point out that, using their same analysis, "only" three to four women in 1,000 would benefit from screening at ages 50–59 and "only" four to six women in 1,000 will benefit from screening women in their 60s. It is well known that screening programs, such as those for cervical cancer, benefit only a small number of the screened women (61). Without any frame of reference, the benefit for younger women was made to appear trivial.
Although the Panel stated in its introduction that they examined "observational studies and randomized trials," they appear to have selectively used only the former. The Panel pronouncement that mammography fails to enable detection of 25% of cancers among women in their 40s while it fails in only 10% of women aged 50 and over has been repeated in the media. This was information derived from the screening program at the University of California at San Francisco (62). Not only does this ignore the fact that a detection rate of 75%, which translates into reduced deaths, is not inconsequential, but also the Panel failed to point out that the same data show an only 15% false-negative rate for women screened in their 40s if they were screened annually. Furthermore, the Panel neglected to point out that data from another modern screening program, presented to them by the X-ray Associates of New Mexico (63), reported virtually the same sensitivity (85%) for women under the age of 50 as for those aged 50 and over. Similar detection rates, regardless of age, for early-stage cancers were reported in these screening programs. The Panel analysis also ignored the fact that inappropriate data grouping makes the age of 50 appear as a true break point when it is not.
The Panel dwelt on the "harms" of screening (anxiety from undergoing mammography, anxiety from being recalled to evaluate an abnormal screening mammogram, and the anxiety and trauma associated with a breast biopsy), yet they ignored the data presented from multiple modern screening programs (63), including our own (64), which showed that the recall rates and "biopsy recommended" rates are virtually the same regardless of age, as is the sensitivity of mammography. The only thing that varies with age is the positive predictive value, and this changes steadily with increasing age (reflecting the prior probability of breast cancer in the population) with no abrupt change at the age of 50 or any other age (65,66). Are these "harms" somehow appropriate for women aged 50 and over but unacceptable for women aged 40–49?
Despite the fact that the most recent radiation risk assessments show that if there is any risk from radiation it is extremely low, and greatly outweighed by the benefit (67,68), the Panel statement, in my opinion, provided a pessimistic view of the radiation risk.
Finally, the Panel suggested that women should decide for themselves whether to be screened. This suggestion does not provide any guidance and seems to imply that anyone who provides guidance is somehow trying to force women to be screened. There has never been any suggestion that an individual should not decide for herself. Guidelines are not requirements. Women and their physicians, however, seek guidance in making these decisions. Informed and "shared" decision making requires access to information. Ironically, at the least, the Panel did not report to women and their physicians the most recent information and did not provide a balanced presentation of the facts so they could make informed decisions. Defenders of the Panel have pointed out that some of the data had not yet appeared in peer-reviewed journals; they do now (47,49,50,60,63). Given the magnitude of the issue, and the fact that the trials have been rigorously monitored (11,49,50,60) with major governmental oversight, the Panel should have reported the data with the caveat that they were awaiting publication. Had they relied only on results published in peer-reviewed journals, the Panel would have had to ignore most of the information that was presented to them at the CDC. To not even mention the new data is, in my opinion, totally inappropriate.
Unfortunately, because of the haste to issue a statement, as required by the CDC guidelines, the time for discussion over the Panel statement was limited and was terminated before all opinions could be expressed.
Events Following the CDC
Few are aware that, although the initial statement was declared unanimous, subsequently one of the Panelists resigned after the meeting and two others issued a "minority" report that concluded that the data, in fact, support a recommendation in favor of screening (69).
Those who believe that the director of the NCI was influenced by the Congressional review ignore the statement that he made immediately after the CDC made its conclusions public and certainly before any hearings took place. He had heard the same presentations and data as the Panel and stated that he disagreed with the Panel's conclusions and that the data did support a recommendation for screening women in their 40s (70). The subsequent action of the NCAB and the director of the NCI to once again support screening women aged 40–49 was not based on politics but on the available scientific evidence. Theirs was an effort to correct the failure of the CDC Panel while trying to avoid embarrassing the Panel members. To suggest that the NCI decision was a political decision is to ignore the facts.
I fully agree with Dr Suzanne Fletcher, who wrote that Congress should have no role in influencing scientific analysis (1). The government should not be involved in scientific debates. Scientific controversy should be resolved by objective debate based on the facts. If it is not resolved, then the process may become open to outside intervention.
Subsequent to the CDC, the American Cancer Society held a similar review of the new data. A conference of experts in the field concluded that the data did support the recommendation for screening women aged 40–49 and that women in their 40s should be advised to be screened on an annual basis (71). An expert panel of the American College of Radiology also reviewed the data and agreed with the American Cancer Society. Both groups were influenced by data that suggested that cancers among younger women grow more rapidly and that the time between screening mammographic examinations should be annual (72).
Medicine is under great pressure to reduce the cost of health care. If health planners do not believe it is worth paying for an intervention, such as screening women in their 40s, they should state this openly for public debate. By the most rigorous scientific analysis, there is clear proof that screening, with mammography, beginning by the age of 40 can save lives and that this should be done on an annual basis. Women should also know that the primary reason to go for a longer time between screenings is to save money.
There is not one shred of data that shows any parameter of screening that changes abruptly at the age of 50. Therefore, in the future, analysts should not be permitted to group women under the age of 50 as if they are a uniform group and compare them to those aged 50 and over as if they are a uniform group. To do so would artificially make any factors that change with age appear to change abruptly at the age of 50. If age grouping is needed for statistical analysis, the analyst should alert the reader as to how it can influence the interpretation of the results.
Despite all of the misinformation concerning the CDC, I believe the facts show that the Panel statement issued on January 23, 1997, failed to provide critical information to the public and provided women and their physicians with misinformation. The decision by the NCI to once again advise women in their 40s that there is a potential benefit from screening was not a political decision but one based on facts and science.
Footnotes
Address reprint requests to the author.
Received October 28, 1997; revision requested December 19, 1997; revision received September 16, 1998; accepted September 28, 1998. References
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| RADIOLOGY | RADIOGRAPHICS | RSNA JOURNALS ONLINE |
