Breast cancer mortality in 500 000 women with early invasive breast cancer diagnosed in England, 1993-2015: population based observational cohort study
BMJ 2023; 381 doi: https://doi.org/10.1136/bmj-2022-074684 (Published 13 June 2023) Cite this as: BMJ 2023;381:e074684Linked Opinion
Risk of breast cancer death after a diagnosis of early invasive breast cancer
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.
Dear Editor
While we agree with the substantive conclusions of Taylor et al [1] – the prognosis for women with breast cancer has indeed improved – we are troubled by their use of the term “mortality”. Mortality is conventionally measured in a population at risk to develop disease, not among those diagnosed with the disease. Case-fatality is the conventional term for what was measured in this paper [2].
The problem is not simply one of semantics; improvements in case-based measures – rising survival and declining case-fatality exaggerate the progress being made [3]. The problem is in the denominator, which is sensitive to who gets diagnosed with early-stage breast cancer and when. That changes over time. More widespread mammography screening and increasingly sensitive diagnostic practices (such as the addition of breast tomosynthesis, ultrasound and MRI) will reliably shift the spectrum of breast cancer towards more slowly-growing, less aggressive forms – including breast cancers not destined to progress (overdiagnosis)[4]. Even if there is no change in population-based mortality, survival will reliably rise and case-fatality rates will reliably fall.
Comparisons of case-based measures across time are thus biased by diagnostic practice. Falling case-fatality from breast cancer in England is some combination of the real effect of improved medical care (largely more effective breast cancer treatment) and the spurious effect of women being diagnosed earlier with no change in their time of death (lead time bias) and women being diagnosed with a disease not
destined to cause death (overdiagnosis bias).
It is unfortunate that the authors – and the BMJ editors – conflated case-fatality with mortality. It is challenging enough for physicians, policymakers and reporters to understand the misleading feedback from screening without the addition of misleading language. Declining population-based mortality is the best measure of progress against cancer, as it is the least likely to be biased by changing diagnostic practices [5]. If the author’s interest was only to provide data to help clinicians estimate the risk of breast cancer death for newly diagnosed women, they should have simply presented the most recent case-fatality data.
[1]. Taylor C, McGale P, Probert J, Broggio J, Charman J, Darby S C et al. Breast cancer mortality in 500 000 women with early invasive breast cancer in England, 1993-2015: population based observational cohort study BMJ 2023; 381 :e074684 doi:10.1136/bmj-2022-074684
[2] Principles of Epidemiology | Lesson 3 - Section 3 (cdc.gov) (https://www.cdc.gov/csels/dsepd/ss1978/lesson3/section3.html)
[3] Oke, J.L., Brown, S.J., Senger, C., Welch, H.G. et al. Deceptive measures of progress in the NHS long-term plan for cancer: case-based vs. population-based measures. Br J Cancer 129, 3–7 (2023). https://doi.org/10.1038/s41416-023-02308-9
[4] Hayse B, Hooley RJ, Killelea BK, Horowitz NR, Chagpar AB, Lannin DR. Breast cancer biology varies by method of detection and may contribute to overdiagnosis. Surgery. 2016;160(2):454-462. doi:10.1016/j.surg.2016.03.031
[5] Extramural Committee To Assess Measures of Progress Against Cancer, Measurement of Progress Against Cancer, JNCI: Journal of the National Cancer Institute, Volume 82, Issue 10, 16 May 1990, Pages 825–835, https://doi.org/10.1093/jnci/82.10.825
Competing interests: No competing interests
Dear Editor
Our recent study (1) included all 512,447 women who were registered in England with early invasive breast cancer as their first cancer during the period January 1993 to December 2015, and who were treated initially with surgery. Each woman entered the study three months after her date of diagnosis and remained in the study, contributing to the person-years, at risk until she died, emigrated, or reached either her 95th birthday or 31 December 2020, i.e. the end of follow-up. Annual breast cancer mortality rates (reported in percentages, i.e. 100 x number of deaths ÷ number of person-years at risk) were used to calculate cumulative five year percentage risks of death from breast cancer for women diagnosed during each of the calendar periods 1993-99, 2000-04, 2005-09, and 2010-15. Within each of these periods, the annual breast cancer mortality rate was highest during the five years after diagnosis and then declined. Mortality declined progressively across calendar periods. The overall cumulative five-year breast cancer mortality risk was 14.4% for women diagnosed during 1993-99 but was only 4.9% for women diagnosed during 2010-15. More detailed analyses provided separate estimates according to the woman’s age at diagnosis, whether the cancer was detected by screening, the number of involved lymph nodes, tumour size, grade, oestrogen receptor status and human epidermal growth factor receptor 2 status.
The aim of our study was to provide information to help clinicians estimate breast cancer mortality risk for each patient presenting in the clinic, taking into account the routinely-available patient and tumour characteristics, including whether or not the cancer had been detected via screening. Patients have identified that this information is needed in the clinic today to inform current and future patients about their prognosis (2). Our study was an observational cohort study and therefore cannot quantify the causal roles played by different factors in the observed decreases in the breast cancer mortality rate following a diagnosis of early breast cancer. Therefore, we did not assess the effects of treatment, which would need data from randomised trials. Neither did we quantify the effects of breast cancer screening on breast cancer mortality and some of the reasons for this have been summarised by Dr Robert (3).
Our study provides information for clinicians to use when estimating breast cancer mortality risks for patients who present in the clinic today, taking into account the characteristics of their tumour, including whether the cancer was screen detected. It provides no evidence either for or against breast screening.
References
1. Taylor C, McGale P, Probert J, et al. Breast cancer mortality in 500 000 women with early invasive breast cancer diagnosed in England, 1993-2015: population based observational cohort study. BMJ 2023; 381 e074684
2. MacKenzie M, Stobart H, Taylor C, Dodwell D. Risk of breast cancer death after a diagnosis of early invasive breast cancer. BMJ 2023; 381: p1355
3. Robert V. Rapid Response Re: Breast cancer mortality in 500 000 women with early invasive breast cancer in England, 1993-2015: population based observational cohort study https://www.bmj.com/content/381/bmj-2022-074684/rr-0
Competing interests: No competing interests
Dear Editor
Taylor and colleagues (1) published an interesting study describing breast cancer mortality between 1993-2015. The study confirmed a decrease in mortality over time. This study also compares mortality associated with screen detected cancers with mortality associated with not screen detected cancers.
Although this study does not reach any conclusions on the screening benefits, and despite explicit warnings in the publication, screening proponents utilize it to promote screening.
It is therefore important to clarify why comparing the mortality seen in women with screen detected cancers to those with not screened cancers does not allow for any conclusion on the value of screening.
1. Interval cancers.
The two groups of cancers diagnosed and not diagnosed by screening do not correspond to the split between screened and unscreened women. Cancers not detected by screening include interval cancers discovered between two screening rounds in women who participate in screening. These interval cancers are screening failures, and it makes no sense to evaluate screening performance by attributing these failures and resulting deaths to unscreened women. This is far from a minor issue: in England, interval cancers account for approximately one-third of all cancers in screened women (2). Furthermore, these interval cancers have a higher stage with unfavourable molecular features (3).
2. Overdiagnosis.
Overdiagnosis, corresponding to cancers found by screening but which would never have affected health if they had remained undiscovered, leads to an artificial increase of cancers without an increase in deaths.
Overdiagnosis, therefore, tends mathematically to lower the mortality rate and create the illusion of screening effectiveness, whereas, in reality, screening has not improved the prognosis of "real" cancers ("real" in the sense of cancers likely to impair health).
Again, this is not a minor problem since overdiagnosis may account for over 40% of cancers diagnosed by screening (4).
3. Non-comparability between groups.
Because the two groups corresponding to cancers diagnosed and not diagnosed by screening were not set up using randomization, the distribution of risk factors for death is probably not balanced between the groups.
In many cases, failure to respond to screening invitations reflects psycho-social problems or difficulties in accessing healthcare facilities, the consequences of which are not just limited to not accepting screening invitations but are also likely to affect cancer management and prognosis.
4. Lead time bias
Lead time bias should not play a significant role in late mortality. In contrast, it is likely to reduce early mortality (5-year mortality) of screen detected cancers.
5. Screening harms should also be considered
Assessing the benefits of screening is not enough.
Screening harms, such as stress due to false alarms, overdiagnosis with psychological and social repercussions, somatic consequences of unnecessary treatment caused by overdiagnosis, and radiation-induced cancers by repeated mammography, must all be considered and weighed up.
The study does not provide any information on these aspects (nor was this the aim of the study). However, only after considering all of these factors - the decrease in mortality and the cost of obtaining this decrease - can a judgment on the value of screening be made.
And it is up to each woman to decide for herself, without coercion or persuasion, whether or not she wants to be screened.
References
1. Taylor C, McGale P, Probert J, Broggio J, Charman J, et al. Breast cancer mortality in 500 000 women with early invasive breast cancer in England, 1993-2015: population based observational cohort study. BMJ 2023; 381 e074684
2. Bennet RL, Sellars SJ, Moss SM. Interval cancers in the NHS breast cancer screening programme in England, Wales, and Northern Ireland. Br J Cancer 2011;104(4):571-577
3. Ambinder EB, Lee E, Nguyen DL, Gong AJ, Haken OJ, Visvanathan K. Interval Breast Cancers Versus Screen Detected Breast Cancers: A Retrospective Cohort Study. Acad Radiol. 2023 Feb 3:S1076-6332(23)00020-X
4. Jørgensen KJ, Gøtzsche PC, Kalager M, Zahl PH. Breast Cancer Screening in Denmark: A Cohort Study of Tumor Size and Overdiagnosis. Ann Intern Med. 2017;166(5):313-323
Competing interests: No competing interests
Dear Editor
Taylor and colleagues [1] have published an interesting paper on the prognosis of early-stage breast cancer diagnosed in England in 1993-2015 and concluded that “The prognosis for women with early invasive breast cancer has improved substantially since the 1990s”. It is not obvious that the prognosis has improved substantially, because overdiagnosis bias (which is including both lead time bias and length time bias) has not been properly accounted for, I think.
The authors use the term mortality to describe survival after a breast cancer diagnosis. Usually, mortality is being used about number of deaths per 100,000 individuals at risk of death. I do not find any reference to exposure years of the population at risk in this paper, so it looks like the authors are studying mortality (or survival) after a diagnosis rather than mortality as normally defined in epidemiology. Survival after a diagnosis is a valid measure for comparing cancer therapies in randomized trials; however, changes in 5-year survival after a diagnosis over time bear little relationship to changes in cancer mortality per 100,000 at risk. Instead, changes in 5-year survival appears primarily related to changing patterns of diagnosis [2]. The changing in patterns of breast cancer diagnosis is termed overdiagnosis (defined as the detection of tumours that would never become clinical disease during the patients’ lifetime) [3].
The level of overdiagnosis when screening for breast cancer, was typically around 50 percent 20 years ago [3]. Overdiagnosis is not only restricted to screening per se but also related to the introduction of diagnostic methods with improved sensitivity, methods which are also used outside public screening programs. Mammography also detects many ductal carcinoma in-situ (DCIS), which many think is a premalignant lesion. If true, this should lead to less invasive breast cancer being detected, which has never been observed. The level of overdiagnosis is increasing over time due the introduction of new diagnostic methods with higher sensitivity [4] and so is the level of DCIS too. Invasive breast cancer and DCIS are positively correlated and not negatively correlated.
Suppose the level of overdiagnosis is 50 percent over a time period, there is no improvement in cancer treatment and the 5-year survival was 80 percent for early-stage breast cancer at the beginning of the period. Then the number of deaths among women diagnosed with breast cancer would decline from 20 per 100 breast cancer cases to 20 per 150 breast cancer cases – a reduction by 1/3 even when there is no improvement in breast cancer treatment.
References
1. Taylor C, McGale P, Probert J, Broggio J, Charman J, et al. Breast cancer mortality in 500 000 women with early invasive breast cancer in England, 1993-2015: population based observational cohort study. BMJ 2023; 381 e074684.
2. Welch HG, Schwartz LM, Woloshin S. Are increasing 5-year survival rates evidence of success against cancer? JAMA 2000; 283: 2975-8.
3. Zahl P-H, Strand BH, Mæhlen J. Breast cancer incidence in Norway and Sweden during introduction of nation-wide screening: prospective cohort study. BMJ 2004; 328: 921-4.
4. Bakker MF, de Lange SV, Pijnappel RM, Mann RM, Peeters PHM, et al. Supplemental MRI Screening for Women with Extremely Dense Breast Tissue. NEJM 2019; 381: 2091-102
Competing interests: No competing interests
Re: Breast cancer mortality in 500 000 women with early invasive breast cancer in England, 1993-2015: population based observational cohort study
Dear Editor
The aim of our recent study (1) is to provide information to help clinicians estimate the breast cancer mortality risk for individual patients presenting in the clinic today. Patients have identified that this information is needed to inform current and future patients about their prognosis (2). We derived these risk estimates by first calculating annual breast cancer mortality rates (reported in percentages, i.e. 100 x number of deaths ÷ number of person-years at risk) according to all the following factors simultaneously: time since diagnosis, age at diagnosis, whether the cancer was detected by screening, the number of involved lymph nodes, tumour size, grade, oestrogen receptor status and human epidermal growth factor receptor 2 status.
The case-fatality rate, which is the proportion of persons with a particular condition who die from that condition (3), can be a useful measure of disease severity in some circumstances. However, it would not have enabled us to achieve our study objectives, as it does not enable risks to be estimated according to time since diagnosis. This is particularly relevant for a disease where, despite treatment, patients continue to have an increased risk of death for several decades following diagnosis.
One of the most important findings in our study was that, for any given time since diagnosis, crude annual breast cancer mortality rates and risks decreased with increasing calendar period. It is likely that many factors played a role in the decrease. But the aim of our paper was not to quantify the contribution of individual factors. Other study designs, including studies of mortality in the general population, randomised trials of different treatments, and surveys of the extent to which effective treatments are being used would be more appropriate for achieving such aims (4).
References
[1] Taylor C, McGale P, Probert J, Broggio J, Charman J, Darby S C et al. Breast cancer mortality in 500 000 women with early invasive breast cancer in England, 1993-2015: population based observational cohort study BMJ 2023; 381 :e074684 doi:10.1136/bmj-2022-074684
[2] MacKenzie M, Stobart H, Taylor C, Dodwell D. Risk of breast cancer death after a diagnosis of early invasive breast cancer. BMJ 2023; 381: p1355
[3] Principles of Epidemiology | Lesson 3 - Section 3 (cdc.gov)
(https://www.cdc.gov/csels/dsepd/ss1978/lesson3/section3.html)
[4] Extramural Committee To Assess Measures of Progress Against Cancer, Measurement of Progress Against Cancer, JNCI: Journal of the National Cancer Institute, Volume 82, Issue 10, 16 May 1990, Pages 825–835, https://doi.org/10.1093/jnci/82.10.825
Competing interests: No competing interests