Twenty five year follow-up for breast cancer incidence and mortality of the Canadian National Breast Screening Study: randomised screening trial
BMJ 2014; 348 doi: https://doi.org/10.1136/bmj.g366 (Published 11 February 2014) Cite this as: BMJ 2014;348:g366
All rapid responses
Rapid responses are electronic comments to the editor. They enable our users to debate issues raised in articles published on bmj.com. A rapid response is first posted online. If you need the URL (web address) of an individual response, simply click on the response headline and copy the URL from the browser window. A proportion of responses will, after editing, be published online and in the print journal as letters, which are indexed in PubMed. Rapid responses are not indexed in PubMed and they are not journal articles. The BMJ reserves the right to remove responses which are being wilfully misrepresented as published articles or when it is brought to our attention that a response spreads misinformation.
From March 2022, the word limit for rapid responses will be 600 words not including references and author details. We will no longer post responses that exceed this limit.
The word limit for letters selected from posted responses remains 300 words.
Screening with mammography is itself a serious methological flaw. How radiation should be accepted as a tool for cancer early detection as it perverts the test result? Just because radiation induces cancer, no matter how low may be the radiation dose used, it is not ethical to propose radiation as an agent to seek for cancer on healthy women. Nobody would accept a blood sugar level test done with a sugar reactive, neither an exudate done with non sterile tools. Because it's like setting an incriminating piece of evidence on an innocent victim in order to accuse him.
Screening with mammography is a destructive trial, which only proves radiation induces cancer. Radiologists are not able to distinguish between a radioinduced cancer from a spontaneous one. Never a woman should accept screening with mammmography early detected breast cancer on her, but it radioinduced it.
I've had 3 normal mammograms done and when I realized that radiation over and over until a positive result, will only produce cancer, I decided to stop it. Wilson and Jungner principles don't detail maybe because it is
common sense: the test must detect without producing the disease it seeks.
Competing interests: Screening with mammography is compulsory in Uruguay for every female worker age 40-59. I'm age 53 and I'm litigating against the uruguayan government in order to stop violating my human rights.
Is it possible to find the above mentioned article in german language?
Many thanks for your answer and greetings from Austria
Competing interests: No competing interests
A failure from the beginning
It is a well-known fact that the Canadian trials were not population-based. Instead, they used motivated volunteers who represented only 26% to 27% of the invited population. It is particularly important to note that the Canadian trials could not evaluate the independent impact of mammography because of the confounding effect of physical examination.
For these reasons, the Canadian trials were excluded from the meta-analysis of the impact of early detection using mammography on mortality from breast cancer by the World Health Organization's International Agency for Research on Cancer (IARC) workshop in 2002, even though Miller chaired the section. It is quite surprising that three authors of the current BMJarticle are on record as having agreed with the following conclusions of the WHO-IARC publication (IARC Handbook of Cancer Prevention, Vol. 7: Breast Cancer Screening. Lyon: IARC; 2002):
• "There is sufficient evidence for the efficacy of screening women aged 50 to 69 years by mammography as the sole screening modality in reducing mortality from breast cancer."
• "There is inadequate evidence for the efficacy of screening women by clinical breast examination in reducing mortality from breast cancer."
• "There is inadequate evidence for the efficacy of screening women by breast self-examination in reducing mortality from breast cancer."
The scientific community, with good reason and with the assent of Miller and colleagues, excluded the Canadian trials from the assessment of mammography screening. Nevertheless, 12 years later, Miller and colleagues revert to their earlier opinion that mammography does not have an effect on mortality from breast cancer.
A total failure for many reasons
The Canadian trials stand apart as the only ones among the published screening trials that failed to show any effect on mortality from breast cancer. The flagrant professional error in figures 2 to 4 of the current BMJ paper, claiming that survival curves are actually "breast cancer specific mortality" curves, does little credit to the peer-review process carried out by BMJ as well as to the authors. The article does not provide news for anyone, since these trials have been plagued with serious quality assurance problems that have been well-known and carefully documented over the past few decades.
After Drs. Wende Logan and Stephen Feig resigned as outside reviewers of the Canadian trials, I (LT) was invited by Miller to review the trials' image quality. Dr. Cornelia Baines was assigned to record my evaluation of the relevant mammography quality assurance parameters for 50 selected cases from the Canadian trials.
At case 15, she terminated the review because of the overwhelmingly negative ratings given to every case. The next day at the committee meeting I should have recommended immediate termination of the Canadian trial, and I regret that I did not do so. My protests would most likely have been to no avail, as were the previous protests of Logan and Feig, followed by the vehement protests of Dr. Dan Kopans from Harvard University and Drs. Martin Yaffe and Norman Boyd from Toronto, among others.
It is worthwhile to mention the publications of Boyd and colleagues in Radiology (1993, Vol. 189:3, pp. 661-663) and in the Canadian Medical Association Journal (1997, Vol. 156:2, pp. 207-209). He and his co-authors wrote:
Questions about randomization in the [CNBSS,] however, have subsequently been raised because of the study's findings that showed an imbalance in the number of women with advanced breast cancer. In particular, the mammography arm had an excess of patients with involved axillary lymph nodes whose breast cancer had been detected by physical examination at baseline (i.e., before randomization).
Boyd's early warning about the imbalance in the number of advanced cancers in the two arms of the Canadian trials was right on target, because it is a well-known fact that the most sensitive predictor of the forthcoming mortality results in a screening trial is whether or not the screening in question decreases the advanced cancer rate. The Canadian trials are the only ones among all the trials that did not report any decrease in the advanced cancer rate. This is not surprising because of the above mentioned imbalances, but also because the radiologists were not trained for the demanding task of finding early breast cancers.
Boyd and co-authors also wrote:
We now know that names were altered and that there were more alterations in the mammography arm. Although 78% of these changes could be accounted for in some way (e.g., clerical errors), the remaining 22% (representing 101 names) could not. We also now know that a coordinator at one of the [CNBSS] centers was suspected of assigning her friends to the mammography arm and that the suspicion was strong enough to remove her from her position. We do not know the method of subversion thought to have been used by this coordinator, nor do we know whether an examination of name alterations revealed what she was alleged to have done.
Comparison with other trials
There is no comparison, since the Canadian trials are uniquely poorly designed and carried out and, therefore, have been much criticized for decades. Miller and his colleagues make an attempt to compare the Canadian trials only with the Swedish Two-County Trial, although their subtitle is "Comparison with other studies."
Comparing the Canadian trials with the Swedish Two-County Trial is inappropriate for many reasons, and the statements published in the BMJ article are contrary to fact:
• The randomization did not take place on a "county level" as the authors claim, since a Swedish county corresponds to a province in Canada or a state in the U.S.
• The average tumor size in the control group was not 28 mm but 22 mm.
• Their false claim that the significant reduction in mortality observed in women invited to screening in the Swedish trial is due to an initial imbalance has already been shown to be unfounded, and Miller and his colleagues should be aware of this publication (Duffy et al, Annals of Oncology, 2003, Vol. 14:8, pp. 1196-1198).
• The fact that mortality from breast cancer was compared among women invited versus those not invited implies that the true impact of screening among those who actually attended was considerably higher than 31%.
• The Swedish Two-County Trial was population-based while the Canadian trials were not.
The following data can inform the reader about the true difference between the CNBSS and the Swedish Two-County Trial: The Swedish trial did decrease the advanced cancer rate in women invited to screening, whereas the Canadian studies had an excess of advanced cancers in the mammography arm. The Swedish Trial has accomplished a significant decrease in mortality from breast cancer, the Canadian trials did not!
According to the original report of CNBSS, the relative rate of node-positive cases in the mammography arm versus the control arm in the 40 to 49 age group and the 50 to 59 age group were 1.53 (95% confidence interval [CI]: 1.11-2.09) and 1.07 (95% CI: 0.80-1.44), respectively, and 1.27 (95% CI: 1.02-1.57) for the two age groups combined. The failure to reduce advanced breast cancer therefore resulted in the lack of benefit in reducing breast cancer mortality in CNBSS.
In contrast, the corresponding figure of reducing advanced cancer for women ages 40 to 74 was 0.75 (95% CI: 0.65-0.87) as reported in the first article of the Swedish Two-County Trial (Tabár et al, 1985). The reduction in advanced breast cancer also led to a 31% reduction in mortality from breast cancer in women invited to screening.
Miller and colleagues' current article in BMJ confirms Boyd's conclusion published 21 years ago in Radiology (1993, Vol. 189:3, pp. 661-663): “Taken at face value, the results of the [CNBSS] argue for abandoning mammographic screening as a population-based means of controlling death rates from breast cancer. We believe such a conclusion to be unjustified and unsupported by the findings of the [CNBSS]” ... [and] “the results of these trials should not be used to change the prevailing scientific view of the potential benefits of screening with mammography”.
Competing interests: No competing interests
I am somewhat surprised that many of your respondents have failed to understand our study (1). Indeed, some of the comments they make are very similar to previous comments on our study, to which we responded at the time (2,3).
The most serious, and also previously answered criticisms, are those of Daniel Kopans, echoed by Eva Rubin, Paul Tartter and Jacob Levman.
Our cancer detection rate and other measures of screening quality was similar to that in the Two-County Study and superior to many others, disproving the claims that our mammography quality was poor (4). The centres in the our study did not use second hand mammography machines, many were newly purchased for the trial, all were carefully assessed by our reference physicist, who ensured the radiation dose was low and the study films of high quality. Grids were used as necessary, the study radiologists were advised by our reference radiologist and they met periodically. The technologists were properly trained, and positioning met North American standards of the time (3).
The proportion of cancers detected by mammography alone is a reflection of the high quality of the breast examinations performed. There is no comparison possible with other trials, all with the exception of the early New York HIP study did not use breast examinations. The similarity of average sizes in our two study arms only reflects the high quality of all our screening examinations.
The comments on randomization are a repetition of previously raised concerns. After Kopans accused me of scientific misconduct because of his misunderstanding of the process of randomization in the study, the National Cancer Institute of Canada arranged for randomization in the trial to be assessed by two internationally recognized epidemiologists. After full evaluation, they determined the randomization complied fully with accepted standards (5). In brief the nurses had no role in the randomization. They were trained to recognize abnormalities, not make a diagnosis, and could have no opinion on whether the woman had a cancer or not and certainly could not judge the curability of any cancers found. The numbers of women referred to the study surgeon after the first screen breast examinations was equal in the two arms. It is strange that Kopans should fail to recognize that good mammography can bring forward the diagnosis of both cancers later found to be node positive (not advanced as he stated) as well as node negative cancers.
To turn to the other comments, it is astonishing that Keith Ericson regards a hazard ratio of 0.99 as indicating a benefit of mammography screening. All his subsequent computations are based on this so-called 0.01 benefit. No one could seriously claim such a non-significant finding could be the basis of policy decisions determined by evidence. I also find it astonishing that Jacob Levman should ignore the well-recognized biases associated with survival, which as we state in the paper cannot be used to evaluate the efficacy of screening. I also find it surprising that Ruth Etzioni should largely dismiss our mortality analyses, especially that relating to the cases ascertained in the 5 year screening period, on the grounds of the apparent discrepancy between the numbers of cases detected and breast cancer mortality, The excess cases in the mammography arm are due to a combination of lead time and over-diagnosis, and therefore, the correct evaluation must be based on breast cancer mortality.
References
1. Miller AB, Wall C, Baines CJ, Sun P, To T, Narod SA. Twenty five year follow-up for
breast cancer incidence and mortality of the Canadian National Breast Screening Study: randomised screening trial. BMJ 2014;348:g366 doi: 10.1136/bmj.g366
2. Miller AB, Baines CJ, To T. Re: "Randomization in the Canadian National Breast Screening Study: a review for evidence of subversion" In response to: D.B. Kopans. CMAJ 1997;157:247-8.
3. Miller, A.B., Baines, C.J., Sickles, E.A. Canadian National Breast Screening Study. AJR 1990; 155:1133-1134.
4. Fletcher SW, Black W, Harris R, Rimer BK, Shapiro S. Screening for Breast Cancer. J Natl Cancer Inst. 1994 Apr 6;86(7):558-61.
5. Bailar JC, MacMahon B. Randomization in the Canadian National Breast Screening Study. Report of a review team appointed by the National Cancer Institute of Canada. CMAJ 1997; 156: 213-5.
Competing interests: No competing interests
Dear Editor,
We read with interest the work by Miller et al1, a large screening trial attempting to answer a longstanding debatable issue: does screening mammography reduce breast cancer-related mortality in women from 40 to 59 years old ? According to the authors, the answer is clearly no; consequently, this paper unavoidable controverts the Swedish study by Tabar et al2 who concluded in a 37 percent reduction in breast cancer mortality, attributable to screening mammography.
Can a screening tool be evaluated, by means of mortality only? It is incontestable that screening mammography can reveal non-palpable lesions at an earlier stage; furthermore, one should expect that a substantial proportion of mammography-detected tumors tend to be smaller in size, or small enough to escape clinical examination whatsoever. Conversely, mean size in the two arms of the study is comparable. Early detection of breast cancer has given rise to substantial evolution of therapies towards the disease. There is much controversy to outstrip in order to eventually debate the role of screening mammography as an adjunct to clinical examination.
Most importantly, a woman who does not die from breast cancer does not mean she does not strive with it; breast cancer remains a very hard and consuming personal, psychological, familial and social adventure for millions of women worldwide. Early detection of a non-palpable breast cancer promptly leads to appropriate management in order to fight the disease at an early stage, and offers optimal care and quality of life.
Lastly, the way that the study results penetrated the media could lead to perilous interpretations of a largely utilized screening tool, which, according to other, important studies, saves lives.
1 Miller, A. B. et al. Twenty five year follow-up for breast cancer incidence and mortality of the Canadian National Breast Screening Study: randomised screening trial. Bmj 348, g366, doi:10.1136/bmj.g366 (2014).
2 Tabar, L. et al. Swedish two-county trial: impact of mammographic screening on breast cancer mortality during 3 decades. Radiology 260, 658-663, doi:10.1148/radiol.11110469 (2011).
Competing interests: No competing interests
The study is silent on treatment for these discovered conditions.
If treatment differed across the arms, we have no way of knowing.
For that reason, we must assumed randomized equivalent intervention, yet assumptions of randomness are precisely the methodological weaknesses addressed by readers.
In short, we don't know.
We might well be left with: two forms of diagnosis with unknown treatment followup, but with somewhat significant earlier and smaller-size-growth detection effects by mammogram, produce equivalent outcomes, if untreated.
Competing interests: No competing interests
Dear Sir:
United States (U.S.) radiologist critics of the Canadian National Breast Screening Study promote a “talking point” that the quality of the Canadian screening mammography was poor or substandard. Therefore, it would be useful to compare the Canadian study results with 21st century screening mammography in the U.S.
Table 2 of the study shows that the 212 non-palpable invasive cancers in the screening arm (mammography screen-detected) had a mean size of 1.4 cm. The 454 palpable cancers in the screening arm (physical exam-detected) had a mean size of 2.1 cm. The overall average size was 1.9 cm, versus 2.1 cm in the control group.1
The Breast Cancer Surveillance Consortium reports that the mean size for invasive cancers detected after 2.1 million screening mammography exams (asymptomatic women) in the U.S. from 2004 to 2008 was 1.8 cm. The median size was 1.4 cm. For 360,000 diagnostic exams in symptomatic women, the mean and median sizes were 2.6 and 2.1 cm.2
How is it possible that mammography screen-detected invasive cancers in the “substandard” Canadian study (1.4 cm) were smaller than is typical today in the U.S. (1.8 cm)?
Overdiagnosis would also be lower with “substandard” mammography, especially without recently adopted digital technology or computer-aided detection. The Canadian study invasive overdiagnosis estimate of 22% equals excess cancers divided by any cancer detected in the screening arm (106/484). As briefly noted but quite alarming, if the denominator is replaced with only mammography screen-detected cancer, overdiagnosis increases to 50% (106/212). This is double the estimate from the Malmo trial.3
Will the authors provide overdiagnosis estimates including in situ lesions? These lesions represent about one third of U.S. screen-detected cancers. 4 The recent cumulative overdiagnosis estimate for screen-detected cancer including in situ lesions in the U.S. is 50%, assuming that 60% of all breast cancer is detected by screening mammography. 5
Yours truly,
John D. Keen, MD
References:
1. Miller AB, Wall C, Baines CJ, Sun P, To T, Narod SA. Twenty five year follow-up for breast cancer incidence and mortality of the Canadian National Breast Screening Study: randomised screening trial. BMJ;348:g366.
2. Breast Cancer Surveillance Consortium: National Cancer Institute:BCSC Data & Statistics. Accessed February 14, 2014, at http://breastscreening.cancer.gov/data/benchmarks/.
3. Welch HG, Schwartz LM, Woloshin S. Ramifications of screening for breast cancer: 1 in 4 cancers detected by mammography are pseudocancers. BMJ 2006;332(7543):727.
4. Keen JD. Promoting screening mammography: insight or uptake? J Am Board Fam Med 2010;23(6):775-82.
5. Bleyer A, Welch HG. Effect of three decades of screening mammography on breast-cancer incidence. N Engl J Med 2012;367(21):1998-2005.
Competing interests: I receive income from reading mammograms.
The results of the 25 year follow-up of the Canadian Breast Screening Trials should not be a surprise to anyone who has followed their earlier publications. With no reduction in node positive tumours in the mammography arm during the intervention period (204 v 170) and no reduction in mortality seen earlier there was no reason to anticipate any reduction of mortality with 25 years follow-up. The publication has been accompanied by calls for further reviews of mammography (“that the rationale for screening by mammography be urgently reassessed by policy makers”). The rational approach however is to try to understand why the results of this trial are different from that of all other randomised controlled trials of mammography. There are four obvious possible explanations:
1. It is a chance finding consistent with the 20-25% reduction in mortality seen in the other seven randomised controlled trials of mammographic breast screening considered by the Marmot review[1];
2. The breast examination in the control arm is as good as mammographic screening at detecting progressive breast cancers early;
3. The quality of mammography in this trial was substantially worse than in all other trials;
4. The randomisation in this trial did not result in balance between the two arms.
The Marmot review determined that “There was some heterogeneity in the RRs from different trials, but this was not statistically significant”. Even if one were to only compare the Canadian trial(s) to the other seven randomised controlled trials (without making an allowance for the implicit multiple comparisons), the P-value is not very small (P=0.015) and the difference could easily be due to chance.
It is informative to consider the data presented in the current paper. First it should be noted that the excess in breast cancer in the mammography arm is not statistically significant (despite what is written in the accompanying editorial). Over 25 years, the proportions of participants diagnosed with breast cancer in the two arms were 7.23% and 6.98%: an excess of 0.25% with a 95% confidence interval ranging from a reduction of 0.08% (i.e. fewer breast cancers in the mammography arm) to an increase of 0.59%. One might argue that even though the excess was not statistically consistent that it is consistent with other trials. That is true, but the data presented strongly suggest that the cancers screen-detected in year 1 were not over-diagnosed. There were 253 breast cancers diagnosed in year 1 in the mammography arm compared to 170 in the control arm. Those that were screen-detected resulted in 52 and 26 deaths respectively. Additionally there were 46 and 44 deaths from interval cancers in years 1-5 (approximately 11 per year in each arm). Thus the proportion of year 1 cancers resulting in death was greater in the mammography arm ((52+11)/253=25%) than in the control arm ((26+11)/170=22%). From this perspective it is difficult to argue that the extra cancers in the mammography arm were over-diagnosed, because (by definition) an over-diagnosed breast cancer cannot result in a breast cancer death.
Table 3 of the current paper suggests that the incidence rate of fatal breast cancer “in year 5” in the control arms was exceptionally high compared to earlier years. If this is true it would suggest that there was a deficit of undiagnosed breast cancers (at the time of randomisation) in the control arm. A more likely explanation is that there is an typographical error in the Table and that the number presented is of fatal breast cancers incident “in years 2-5”.
It is not possible to determine the reason for the results of the Canadian trial being different from those of all other trials based on the published summary of results, but the Conclusions of the current paper completely ignore the rest of the literature and even the detailed results of their own trial.
[1] Marmot MG, Altman DG, Cameron DA, Dewar JA, Thompson SG, Wilcox M. The benefits and harms of breast cancer screening: an independent review. Br J Cancer. 2013 Jun 11;108(11):2205-40.
Competing interests: No competing interests
The evaluation of mammography impact on breast cancer mortality in this study poses major challenges that the primary analyses of disease-specific deaths do not overcome. Consequently, the primary mortality analyses which show no benefit of mammography during the screening interval are not as informative about mammography benefit as one might hope.
There are two mortality analyses.
The first looks at the breast cancer death rate restricted to the cases detected during the first five years (the screening period). This is the cumulative death rate in the population but only allows deaths from the cases diagnosed in the first five years. The analysis finds similar death rates on the two arms.
This seems strange because there were there were 666 cases on the mammography arm and 524 cases on the control arm. Thus, if, as the investigators conclude, mammography has no effect, we would actually expect a higher observed cumulative death rate on the mammography arm, unless all of the (142) excess cases are overdiagnosed which is unlikely. Thus, this analysis may bias results against mammography benefit.
The second mortality analysis looks at all breast cancer deaths. But this includes all the cases diagnosed after the screening period. If, after this point, screening behavior evolves similarly in the two arms then the accumulation of a comparable case population in each arm will lead to a dilution of any mortality effect.
In conclusion, both mortality analyses are likely to yield attenuated disease-specific mortality risk ratios even if there is a benefit of mammography.
Competing interests: No competing interests
Re: Twenty five year follow-up for breast cancer incidence and mortality of the Canadian National Breast Screening Study: randomised screening trial
Those criticising this study are missing the point.
Whether this trial does or does not show that breast screening doesn’t reduce breast cancer mortality is neither here nor there.
What anyone eligible for screening needs to know is what chance screening offers them of postponing death at what price.
Critics of this study haven’t got an honest answer to that question.
They claim that screening does postpone deaths at a price that makes it worth offering. They base that claim on other trials which are not better than this one at substantiating that claim.
Those studies were not less flawed than they claim this one is. Those studies did not follow up for long enough to yield a conclusion on whether anyone lives longer as a result of earlier detection. Those studies, scrutinized inside out, upside down and back to front, as they have been, do not yield a definitive number of overdiagnosed and overtreated. Those studies are out of date. Those flawed, old, unsatisfactory studies base a wide range of unreliable estimates.
All the evidence there is does not yield enough to warrant offering screening.
The critics have chosen to take what they can get out of those studies and spin it as a story favourable to screening. The evidence they themselves adduce does not support their claim. In making that claim they have gone beyond the evidence.
Quibbling over this study, when there never has been enough evidence to support screening, is bluster.
Competing interests: Diagnosed through screening