Breast cancer mortality in 500 000 women with early invasive breast cancer 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:e074684
Carolyn Taylor, professor of oncology and honorary clinical oncologist1 2,
Paul McGale, statistician1,
Jake Probert, statistician1,
John Broggio, cancer analytical lead3,
Jackie Charman, senior cancer analyst3,
Sarah C Darby, professor of medical statistics1,
Amanda J Kerr, systematic reviewer1,
Timothy Whelan, radiation oncologist4,
David J Cutter, senior clinical research fellow and clinical oncologist1 2,
Gurdeep Mannu, lecturer in general surgery1,
David Dodwell, senior clinical research fellow and clinical oncologist1 2
1Nuffield Department of Population Health, University of Oxford, Oxford, UK
2Oxford University Hospitals, Oxford, UK
3National Disease Registration Service (NDRS), NHS England, Birmingham, UK
4Department of Oncology, McMaster University and Juravinski Cancer Centre, Hamilton, ON Canada
June the 16th, 2023
Aim of the study
This is an observational cohort study (a group of subjects followed for the duration of the study) involving 512,447 women.
There are two objectives:
1°- Assessment of annual breast cancer mortality rates and cumulative risks by time since diagnosis for women diagnosed during each of the following calendar periods: 1993-99, 2000-04, 2005-09, and 2010-15.
2°- Examination of variations in these mortality rates according to several criteria: according to the calendar period of diagnosis, according to the time elapsed since diagnosis, according to whether or not the cancer was detected by screening, and according to the characteristics of the patients and the tumours they presented.
Overall, almost half the cancers in women in the age groups eligible for screening were detected by screening.
Main results:
Crude risks of breast cancer mortality decreased with increasing calendar period.
In other words, women in calendar periods closer to our contemporary period are more likely to survive long after a cancer diagnosis than women diagnosed in calendar periods further back in time, with a significant magnitude.
The cumulative five-year mortality risk from breast cancer was :
- 14.4% for women diagnosed between 1993 and 1999, and
- 4.9% for women diagnosed between 2010 and 2015.
These results correspond to the entire cohort of 512,447 women aged 18-89, including :
-women eligible for screening, with cancer detected as part of organized screening: 128,240 women (i.e., around a quarter of the cohort)
-women eligible for screening but not screened, with cancer detected outside screening: 133,427 women (i.e., around a quarter of the cohort)
-women not eligible for organized screening: 250,780 women (around half the cohort).
The composition of groups is shown in Table 1, extracted below:
Adjusted annual breast cancer mortality rates also decreased similarly with the advancing calendar period in almost all patient groups, by a factor of around three for estrogen receptor-positive cancers, which correspond to forms of cancer with a better prognosis, and by around two for estrogen receptor-negative cancers, which correspond to more pejorative forms of cancer. The mortality risk improves with the advancement of the calendar periods studied towards more recent years than earlier years.
The study's main aim was to use five-year breast cancer mortality risks for newly diagnosed patients. Indeed, say the authors, these mortality rates, which are now known, can be used to estimate breast cancer mortality risks for today's patients.
The study aims to inform patients and clinicians of the likely absolute mortality risks for patients treated for breast cancer today, considering, among other things, the characteristics of their tumor.
The study shows that, for women diagnosed with early breast cancer, the risk of dying within five years fell considerably between the 1990s and 2010-2015. For most newly diagnosed women, the risk of dying from breast cancer within five years was 3% or less. This is helpful information for women living today.
The authors conclude, "It should be noted, however, that the improvements in breast cancer mortality observed in women whose cancer was detected by screening were paralleled in women whose cancer was not detected by screening."
Detailed conclusions:
The prognosis for women with early invasive breast cancer has improved considerably since the 1990s. Most can expect to survive cancer in the long term, although the risk remains appreciable for a few.
Since the 1990s, the five-year cumulative risk of death from breast cancer has fallen from 14.4% to 4.9% overall, with reductions observed in almost all patient groups.
Indeed, the five-year cumulative mortality risk was 14.4% (95% confidence interval 14.2% to 14.6%) for women diagnosed between 1993 and 1999 and gradually decreased with increasing calendar period to 4.9% (from 4.8% to 5.0%) for women diagnosed later, between 2010 and 2015.
This shows that breast cancer mortality rates decreased with the calendar period of diagnosis over the study period.
But although decreases occurred in almost all patient groups, the magnitude of the mortality rate decreased, and the 5-year risk of cancer death varied considerably between women with different characteristics:
- the risk of mortality was less than 3% for 62.8% of women,
- but 20% or more for 4.6%, corresponding to particularly aggressive forms of cancer that are difficult to cure.
In our data," explain the authors, "the lack of mortality reduction in women aged 80 to 89 with estrogen receptor-negative breast cancer may be explained by the fact that these women generally do not receive systematic adjuvant treatment (treatment that complements the main treatment to prevent the risk of local recurrence or metastases, e.g., hormone therapy or immunotherapy), or radiotherapy, so any improvement in these treatments per se would not have affected mortality in this group of patients. ), or those who do not receive radiotherapy, so any improvement in these treatments per se would not affect mortality in this group of patients.
Patients under 40 years of age at diagnosis had a higher risk of breast cancer mortality than those diagnosed at 40, which is explained by the fact that breast cancers in younger women are inherently more aggressive than those in older women.
The authors found that breast cancer mortality always decreased as a function of the calendar period of diagnosis, irrespective of differences in tumour characteristics, and even the improvements in breast cancer mortality observed in women whose cancer was detected by screening were accompanied by improvements also in those whose cancer was not detected by screening.
This is summarized in the illustration below, which presents the results for all women (a quarter of whom are eligible and screened, a quarter are cancer cases in eligible but unscreened women, and half are women not eligible for screening):
It can be argued that screening has enabled the detection of smaller and smaller tumors over the years with significant technological improvements in mammography equipment, with tumors found of ever lower grades, but, say the authors, this decline in mortality cannot be attributed to changes in tumor size, number of positive nodes or tumor grade alone, as breast cancer mortality continued to decline according to the calendar period of diagnosis, even after adjusting for these factors.
Furthermore, screening and more sensitive breast imaging techniques are likely to have led only to earlier diagnosis and longer survival without altering the clinical course of the disease. Survival, it should be remembered, corresponds to the duration of life after cancer diagnosis and increases with improved treatment and overdiagnosis. The earlier in a person's life cancers are detected that were not destined to kill their host anyway, that are very low-grade and will remain so, the more survival data are artificially improved, without affecting life expectancy.
Relationship with screening
For patients diagnosed with screened or unscreened cancer, annual breast cancer mortality rates and cumulative breast cancer mortality risks showed similar downward trends to those for all women, depending on the calendar period of diagnosis.
The study shows that the improvements in breast cancer mortality observed in women whose cancer was detected by screening were also paralleled by improvements in those whose cancer was not detected by screening.
The increase in screening does not, therefore, explain the improvements in mortality.
The contribution of the study
Other studies have already shown the very marginal role of screening in the decline in breast cancer mortality since the 1990s.
We already know that the risk of breast cancer mortality following early invasive breast cancer diagnosis has decreased over the last few decades.
Bleyer and Miller's impact study concluded that the link between screening mammography and the degree of reduction in breast cancer mortality observed in recent years was increasingly controversial. Their comparison of eight countries in Europe and North America showed no correlation between the intensity of national screening and the timing or even the extent of reduction in breast cancer mortality.
The evidence from the three different approaches (temporal approach, magnitude approach, and comparative approach with other non-screened pathologies) and other additional observations did not support the hypothesis that mammography screening was the main reason for the reduction in breast cancer mortality observed in Europe and North America.
Similarly, P.Autier's study of the three pairs of countries compared suggested that screening had not played a direct role in reducing breast cancer mortality, given the contrast between the temporal differences in the implementation of mammographic screening and the similarity in mortality reductions between the pairs of countries.
In other words, countries that introduced screening earlier than other countries that introduced it later experienced a similar reduction in breast cancer mortality. In contrast, there should have been an amplifying phenomenon in mortality reduction due to the earlier introduction of campaigns. There is, therefore, no link between screening activity and reduced mortality.
And invasive metastatic cancer remains at the same level, as screening is unable to detect this aggressive form of cancer because of its intrinsic biologically aggressive characteristics and often because of its high velocity.
To conclude, we quote this study: Søren R Christiansen, Philippe Autier, Henrik Støvring, Change in effectiveness of mammography screening with decreasing breast cancer mortality: a population-based study.
Summarized here.
According to the authors, improvements in cancer therapies over the past 30 years have reduced mortality, which could erode the benefit-disadvantage balance of mammography screening.
Furthermore, future improvements in managing breast cancer patients will increasingly reduce the benefit-harm ratio of screening.
The benefit of mammography in terms of mortality reduction is diminishing, while the harms, such as overdiagnosis, are constant.
Screening leads to both over-diagnosis and over-treatment, at both human and economic costs,
What the study here provides is an estimate of the extent of the decline in breast cancer mortality rates observed since the 90s, and this is not linked to screening or any other factor related to the tumor or the woman carrying cancer since there is no difference in variations in mortality rates whatever these factors may be, whether the cancer is found by screening or not.
The reason for this is most likely to be found in the therapeutic improvements of recent decades.
Illustration: annual mortality rates and cumulative mortality risks
Cumulative risk is the sum of the various annual risks, present over 5 years. The mortality risk function describes the evolution as a function of time and cumulative factors of the instantaneous risk of death.
Summary by Cancer Rose
This is a descriptive epidemiological study. It aims to quantify the reduction in mortality observed since the 90s. This reduction is not a scoop, but it was interesting to quantify it globally and by sub-group.
It ranges from 14.4% to 4.9% at 5 years between the two periods examined, for all women, with a similar reduction depending on the group (screened or not).
Studies of the impact of screening (see our article) have already demonstrated this reduction in breast cancer mortality since the 90s. Still, the impact of screening is very marginal, or even non-existent, since this reduction is not synchronized with the introduction of screening campaigns.
In essence, the authors conclude that recent data show an improvement in breast cancer mortality risk compared with older data, which is confirmed by their results. They make several points: "... Therefore, increases in screening cannot solely explain the decreases in breast cancer mortality that we observed” and a little further on: " this observational study cannot determine the specific causes of these reductions in mortality.
And yet again, probably the most important: "... Notably, however, the improvements in breast cancer mortality seen in women with screen-detected cancers were paralleled by improvements in those whose cancers were not screen-detected.”
This study confirms (and, above all, quantifies) the downward trend in breast cancer mortality, but it does not conclude (nor does it enable us to conclude) the cause(s) of this decline.
What's important to understand is that in this study, we're not talking about the mortality rate but the cumulative RISK of mortality over 5, 10, or 15 years. These cumulative mortality risks depend on the time T0 chosen. Here, T0 is the date of cancer diagnosis. Therefore, the mortality risks presented in the study are influenced by the lead time (since T0 will be earlier for screened cancers than for unscreened cancers). They will give an apparent better success rate in the screened groups.
Lead-time bias is a well-known inherent bias in screening, giving the illusion of better cancer survival when we've just anticipated its 'date of onset'.
The prognosis for breast cancers is improving, but it is impossible to say how much of this is due to screening, therapeutic advances, and confounding factors such as early diagnosis bias, overdiagnosis in particular, and social and economic factors.
According to the studies already available (see article), the role of screening is probably marginal, and apparent success in screened groups is influenced by lead time.
Rapid responses in BMJ
Dr Vincent Robert is our statistician.
Per-Henrik Zahl is researcher on the Norwegian Institute of Public Health
https://www.bmj.com/content/381/bmj-2022-074684/rapid-responses
Opinion
Risk of breast cancer death after a diagnosis of early invasive breast cancer
BMJ 2023; 381 doi: https://doi.org/10.1136/bmj.p1355 (Published 13 June 2023)Cite this as: BMJ 2023;381:p1355
Mairead MacKenzie, patient advocate1,
Hilary Stobart, patient advocate1,
David Dodwell, senior clinical research fellow and clinical oncologist23,
Carolyn Taylor, professor of oncology and honorary clinical oncologist23
- 1Independent Cancer Patients’ Voice
- 2. 2Nuffield Department of Population Health, University of Oxford
- 3. 3Oxford University Hospitals, Oxford, UK
Two patient advocates reflect on how they helped to shape a research study into breast cancer
Mairead MacKenzie and Hilary Stobart were diagnosed with breast cancer some years ago. They’re just two of the half a million women who contributed their data to our study of women diagnosed with early breast cancer in England. As patient advocates, they also helped to shape the study.
Hilary and Mairead both feel that up-to-date information is needed on outcomes after a diagnosis of early breast cancer. They used their expertise as patients to highlight how data from women diagnosed with breast cancer in the past could help in the clinic today. Moreover, the study also gave them a chance to reflect on all that has changed since they were diagnosed with cancer.
“You don't have much grasp of having cancer until you've had it,” explains Hilary. “You suddenly join a club that you don't want to be part of, and you find you have an awful lot in common with the other people in the club. You have a different perspective on what's important.”
Our study was informed by that perspective.
The study provides risk estimates for individual patients. Both Hilary and Mairead stress that doctors need to help patients understand that breast cancer is “not all one thing.” Prognosis varies widely according to risk factors such as tumour size, lymph node involvement and whether the tumour was detected by screening.
“When I was diagnosed 20 years ago, I was not given a prognosis other than the fact that this is serious and we need to treat you quickly,” says Mairead. “But I think good, clear communication about prognosis can make a vast difference to a patient's quality of life, and how they can cope with things.”
“When people are diagnosed with breast cancer they may already know somebody who has died from breast cancer,” adds Hilary. “They might assume that their risk is the same, but many of them might only have less than 1% risk of dying from breast cancer at five years.”
“For the majority of women, the prognosis is good,” agrees Mairead. “This study backs that up and gives reassurance—because, initially, everybody thinks they're going to die.”
The study shows that, for women diagnosed with early breast cancer, the risk of dying from it within five years reduced substantially between the 1990s and 2010-15. For most women diagnosed recently their five year risk of breast cancer death was 3% or less.
Breast cancer patients have contributed to that improvement.
“I’ve yet to meet a cancer patient who isn’t happy for their data to be used for research,” says Mairead. “If there's a chance of doing something that might make it easier for those coming after, patients nearly always say yes.”
“And if people hadn't said yes, we wouldn't be where we are now, would we?” agrees Hilary “We know our treatment now is good because of all the work that was done earlier …the large numbers of trials and the thousands of women who were prepared to go into them.”
Our results are part of that legacy. They quantify decades of improvements and lay the foundation for more to come. Meanwhile, they can inform how doctors talk with patients about their prognosis today.
“It’s good news,” concludes Hilary. “It shows what we’ve done, and that we need to go on doing it. More studies like this one will be needed in the future. Breast cancer is still with us. There’s a lot more to do.”