The Evaluation of Breast Screening

A subgroup of the Policy Research Unit team at Queen Mary University of London (QMUL) require the baseline and screening exposure data of the breast screening cohort study (currently already held at QMUL) with follow-up data for cancer incidence and death (currently held to 2005, requesting to 31st October 2020). This is for the purpose of evaluating the breast cancer screening programme in terms of its effect on breast cancer mortality and in terms of harms of screening, notably overdiagnosis.

This work was instigated by the Institute for Cancer Research (ICR), Sutton, Surrey, to establish the effect of the NHS Breast Screening Programme on incidence of and mortality from breast cancer.

This work is research in the public interest because its purpose is to evaluate the breast cancer screening programme in terms of its effect on mortality and the potential harms due to identification of false positives. Notably, the number of people who would potentially benefit from this research is exceptionally high:

- Two million women undergo breast cancer screening every year

- One person is diagnosed with breast cancer in the UK every 10 minutes

- One in eight women will develop breast cancer in their lifetime

- Approximately 11,500 people – overwhelmingly women – die of breast cancer every year

Once received from NHS Digital, the data will only be used in identifiable form for the specific purpose of data linkage and will then be pseudonymised.

The cohort comprises approximately 2.7 million women born between 1923 and 1945 resident in the study area who were eligible for at least one invitation by the NHSBSP between January 1988 and December 1994. The study area covered 22 Local Health Boards (LHBs) in Wales and 137 PCTs in England, and the study therefore included approximately 38% of the population of England and Wales who were eligible for invitation to NHS breast screening between these dates.

Following the closure of the Cancer Screening Evaluation Unit at ICR and the establishment of the Policy Research Unit in Cancer Awareness, Screening and Early Diagnosis (PRU) by the Department of Health (now the Department of Health and Social Care), this work was moved to QMUL, the home institute of the PRU director. The purpose of the work remains the same, to evaluate the major benefits and harms of the NHS Breast Screening Programme. QMUL is the only organisation which will have access to record level data. The ICR is no longer involved in this project. Thus QMUL will be responsible for both data control and data processing.

The Department of Health and Social Care established the PRU, led by the applying organisation, QMUL, with a remit among other tasks to evaluate the NHS Cancer Screening Programmes, in terms of both positive and negative effects of the programmes.

The primary aims of this programme are to quantify in both relative and absolute terms:

(i) The reduction in breast cancer mortality associated with the policy of offering mammographic screening from breast cancer.

(ii) The corresponding reduction in breast cancer mortality associated with individual participation in the NHS Breast Screening Programme.

(iii) The risk of overdiagnosis associated with participation in the programme.

Secondary aims include the estimation of self-selection biases, including development of methods of estimation of such, to improve the accuracy of aims (i-iii) above. It is also planned to estimate benefits and harms within narrow age subgroups, and estimate effects of time since screen, to inform policy on the target population and on the frequency of screening.

The PRU shall also examine effects by area, to flag up areas where the screening has not worked as well, which will trigger investigation of the cause, and subsequent rectification, if necessary. This is a unique opportunity. Area-based process measures (screen detection rates, false positive rates etc) have long been available but there has been no data on the effect on clinical outcomes by area.

This study has been running for more than three decades, it is a stand-alone cohort study, the data has been supplied before, and the objectives remain the same (as noted above).

The data required for the aims listed above are

(1) Retention of the baseline demographic and screening data which are already held at QMUL;

(2) The death and cancer notification data up to 2005, which are also already held at QMUL; and

(3) further death and cancer notification data to 31st October 2020.

These data are requested as they will give the opportunity to estimate the effects of the NHS Breast Screening Programme on mortality from and incidence of breast cancer, with rigorous adjustment for potential biases. In particular, the further notification data to 2020 will give valuable information on the long-term effects. We had originally requested data to 2016, but the technical issues for this very large tracing/flagging job (2.7 million women) meant that by the time the process was in place, it was possible to have notifications to 2020. This was considered a worthwhile amendment as it would now give us rather longer follow-up than the 29 years published for the most mature trial of mammographic screening.

Further background to and details of objectives

The NHS Breast Screening Programme (NHSBSP) was introduced in England and Wales in 1988, following the publication of randomised controlled trials (RCTs) and population demonstration projects which reported that mammographic screening could reduce breast cancer mortality by 20-30%.

Mechanisms to monitor the NHSBSP in terms of quality assurance were built into the programme and using performance indicators the programme currently appears to be performing well compared with RCTs. However, monitoring the performance of the service does not directly address questions about its effect on breast cancer mortality and the question remains whether the reduced breast cancer mortality observed in RCTs can be achieved in a population programme, such as the NHSBSP.

In addition, mortality from breast cancer is falling in the United Kingdom and there is debate as to how much of the reduction is due to screening or improvements in treatment (or some other factor). The difficulty of producing quantitative estimates of the effect of a national screening programme on mortality is well recognised. Population level data have been used to estimate the effect of the NHSBSP on breast cancer mortality in England and Wales, but ecological studies suffer the disadvantages of using aggregated data and are complex. The impact of the NHSBSP on breast cancer mortality in the East Anglian region has been estimated using prognostic characteristics of breast cancers to predict breast cancer mortality. However, to assess the programme rigorously the PRU needs to use individual-level data and mortality from breast cancer as the outcome.

The NHSBSP now screens more than 1.3 million women per year and has an annual budget of approximately £52 million. The programme has a high public profile, and debate following recent scientific publications has again raised the issue of whether the Breast Screening Programme is an effective use of NHS resources, and underlined the need for proper evaluation of the programme in terms of its effect on breast cancer mortality.

The primary aim of this study is to evaluate the impact of the NHS Breast Screening Programme on mortality from breast cancer. The objective is to use individual-level data on a cohort of women to assess breast cancer mortality in relation to screening history; specifically to compare risks in those invited with a contemporaneous group of the same age who were not invited until a later date.

Also to assess:

a) mortality from breast cancer in relation to attendance for screening. Those who attend for screening do not represent the entire target population of the programme and bias is introduced by factors influencing the uptake of an invitation. However, mortality among women who attend for breast cancer screening is of considerable interest.

b) mortality from all causes, all cancers and vascular disease in relation to invitation to and attendance at screening. Previous literature has reported a relationship between breast cancer screening and non-breast cancer deaths, specifically from other neoplasms and from vascular disease.

Under v3.4 of the Data Sharing Agreement QMUL sought support to flag the remainder of the cohort who have not been flagged previously in order to obtain up-to-date cancer incidence and mortality follow-up to conduct analysis on a longer-term basis. The objectives of this follow-up phase are to further assess the effectiveness of the screening programme by comparing risk of death from breast cancer in women who were invited for screening with those who were not invited in the same period for follow-up over a 25-year period, and wider objectives related to the impact of the screening programme on diagnosis of breast cancer.

While it is understood that supply of updated notifications have now ceased, we are requesting retention of the data to complete the analyses for three years to September 2025. This is because the analyses are not routine and the dataset is a large and complex one. While we anticipate that the main analyses will be completed by December 2024, the further period will allow for clarifications and reanalyses prompted during and after the publication process.

The analyses will be carried out as the main tranche of a PhD project who is a substantive employee of Queen Mary University of London (QMUL).

This unique dataset will provide estimates of the major benefit (reduction in breast cancer mortality) and harm (overdiagnosis: identification of false positives) conferred by the NHS Breast Screening Programme. It includes subjects recruited at the initiation of the programme in the late 1980s. Despite substantial experimental and observational evidence there remains uncertainty over the magnitude of the benefit of the NHS breast screening programme in terms of breast cancer mortality and the magnitude of the harm in terms of overdiagnosis (identification of false positives). This extremely valuable data resource has the potential to provide answers to these questions with a precision unrivalled by any other study worldwide.

The post-2005 data is needed to answer two specific lines of research:

- First, there is a need to estimate the long-term benefit of early detection, in terms of breast cancer mortality. For many cancers, if recurrence does not occur within, say, five years, it will not occur at all. This is not the case for breast cancer, which can recur decades after the original diagnosis, and recurrence can prove fatal. It is therefore important to ascertain whether the benefit of early detection holds for later events as well as earlier.

- Second, there are conflicting results in terms of early detection, where some trials show the benefit being maintained up to thirty years and others (particularly in younger age groups) show it declining after ten years. Resolving this issue may have implications for other aspects of management than early detection- it may have implications for treatment of breast cancer occurring at earlier ages, for example.

The data controller requires baseline data as already present in the dataset held at Queen Mary University of London, plus notifications of breast cancers and all deaths within the cohort, including cause of death. The notifications of deaths and cancers at the moment only extend to 2005. Queen Mary University of London seek approval for reopening this and giving notifications to 31st October 2020 if available -, and request that this be part of the new data sharing agreement. In addition, Queen Mary University of London request that flagging be extended to the entire cohort, rather than only those recruited after 1991. This would give a longer-term and larger population incidence-based mortality effect for the screening programme than is available from any other dataset in the world. In addition, it will provide long-term data on incidence (indeed lifetime incidence in many cases), essential for estimation of the identification of false positives unbiased by lead time. Note that substantial additional funding is available for the further flagging and notifications.

This work was instigated more than 20 years ago by the Institute of Cancer Research and was transferred to Queen Mary University of London (QMUL) in 2015, following the award of the Department of Health Policy Research Unit in Cancer Awareness, Screening and Early Diagnosis to QMUL, and the closure of the Cancer Screening Evaluation Unit at ICR. The aims of the work remain the same, to quantify the benefits and harms of the NHS Breast Screening Programme.

The expected outputs pertain to methodology, knowledge creation in relation to the effect of breast screening, and potential healthcare policy changes. For this first, this dataset affords an opportunity to understand better and in a UK context the phenomenon of self-selection for screening, as it will contain women who have, contemporaneously, not yet been invited (e.g. someone recruited to the cohort in 1989 and first invited in 1994 will have spent 5 years not yet invited to screening) and women who have refused (i.e. turned down their invitation) screening, that is it can estimate self-selection unconfounded by time, and the potential changes in population behaviour and attitudes over time. In terms of knowledge, it will give estimates of the benefit of screening in terms of breast cancer deaths prevented and harms in terms of identification of false positives to an unprecedented level of precision. This will result in high-impact publications. Further, the size of the data resource will permit detailed analyses of subgroups and tactical issues, such as time between screens and age at screening with the potential to inform policy as to target population, interval between screens, etc.

In terms of the findings reaching decision/policy makers, researchers, clinicians, patients and the public, as with all PRU's research, the results are published in peer-reviewed medical journals.

The further follow-up and associated methodological development will enable estimation of the mortality reduction and the overdiagnosis (identification of false positives) associated with actually being screened rather than only the effect of invitation. It will also allow estimation of the long-term effects.

Again, results will be published in peer-reviewed medical journals. However, prior to publication, these will be shared with the relevant policy community. The PRU is represented on all the cancer screening advisory groups and on the National Screening Committee. These bodies will be apprised of the results in a timely manner, as will Cancer Research UK, Macmillan Cancer Care and Breast Cancer Now. In addition, the funding body (the Department of Health and Social Care, via NIHR) will be informed of the results prior to publication.

The results shall also be posted on the PRU website. PRU will take advice from the PRU patient and public advisory committee as to the most appropriate form of general public disseminations. In addition, the PRU’s scientific advisory committee will be consulted on options for further dissemination.

The study has already borne fruit in terms of publication of medium-term estimates of the reduction in mortality and the overdiagnosis - identification of false positives - associated with invitation to screening:

Johns LE, Coleman DA, Swerdlow AJ, Moss SM. Effect of population breast screening on breast cancer mortality up to 2005 in England and Wales: an individual-level cohort study. Br J Cancer 2017; 116: 246-52

Johns LE, Swerdlow AJ, Moss SM. Effect of population breast screening on breast cancer mortality to 2005 in England and Wales: A nested case-control study within a cohort of one million women. J Med Screen 2018; 25: 76-81

Results include a substantial reduction in medium term mortality from breast cancer, with little or no identification of false positives (Johns et al, Br J Cancer 2017; 116: 246-52). However, it would add considerable value to update these results to long-term effects and to extend the population flagged to the entire cohort.

All outputs will contain only data that is aggregated with small numbers suppressed in line with the HES Analysis Guide.

Note that a new Patient and Public Involvement Board has been set up to advise on:

- setting the research direction

- obtaining wider involvement from all the population this programme serves (ie, not just patients with lived experience of cancer)

- increasing public engagement

with the appointment of an academic PPI lead and a PPI co-investigator (Bishop Douglas Lewins, now retired and replaced by Ms Jennifer Bostock).

The PAB's first meeting was on 14th March 2019.

In the past, analysis of around 40% of the cohort with medium term follow-up has borne fruit in terms of estimates of the effect of the programme on incidence of and mortality from breast cancer (Johns et al, J Med Screen 2018; 25: 76-81; Johns et al, Br J Cancer 2017; 116: 246-52).

Analysis of the full cohort with long term follow-up will yield more precise estimates, enable detailed age-specific estimation, and give policy makers estimates of the long term effects of screening.

A subgroup of the Policy Research Unit team at Queen Mary University of London (QMUL) require the baseline and screening exposure data of the breast screening cohort study (currently already held at QMUL) with follow-up data for cancer incidence and death (currently held to 2005, requesting to 2016). This is for the purpose of evaluating the breast cancer screening programme in terms of its effect on breast cancer mortality and in terms of harms of screening, notably overdiagnosis (identification of false positives).

This work was instigated by the Institute for Cancer Research (ICR), Sutton, Surrey, to establish the effect of the NHS Breast Screening Programme on incidence of and mortality from breast cancer.

This work is research in the public interest because its purpose is to evaluate the breast cancer screening programme in terms of its effect on mortality and the potential harms due to identification of false positives. Notably, the number of people who would benefit from this research is exceptionally high:

- Two million women undergo breast cancer screening every year

- One person is diagnosed with breast cancer in the UK every 10 minutes

- One in eight women will develop breast cancer in their lifetime

- Approximately 11,500 people – overwhelmingly women – die of breast cancer every year

Once received from NHS Digital, the data will only be used in identifiable form for the specific purpose of data linkage and will then be pseudonymised.

The cohort comprises approximately 2.7 million women born between 1923 and 1945 resident in the study area who were eligible for at least one invitation by the NHSBSP between January 1988 and December 1994. The study area covered 22 Local Health Boards (LHBs) in Wales and 137 PCTs in England, and the study therefore included approximately 38% of the population of England and Wales who were eligible for invitation to NHS breast screening between these dates.

Following the closure of the Cancer Screening Evaluation Unit at ICR and the establishment of the Policy Research Unit in Cancer Awareness, Screening and Early Diagnosis (PRU) by the Department of Health (now the Department of Health and Social Care), this work was moved to QMUL, the home institute of the PRU director. The purpose of the work remains the same, to evaluate the major benefits and harms of the NHS Breast Screening Programme. QMUL is the only organisation which will have access to record level data. The ICR is no longer involved in this project. Thus QMUL will be responsible for both data control and data processing.

The Department of Health and Social Care established the PRU, led by the applying organisation, QMUL, with a remit among other tasks to evaluate the NHS Cancer Screening Programmes, in terms of both positive and negative effects of the programmes.

The primary aims of this programme are to quantify in both relative and absolute terms:

(i) The reduction in breast cancer mortality associated with the policy of offering mammographic screening from breast cancer.

(ii) The corresponding reduction in breast cancer mortality associated with individual participation in the NHS Breast Screening Programme.

(iii) The risk of overdiagnosis (identification of false positives) associated with participation in the programme.

Secondary aims include the estimation of self-selection biases, including development of methods of estimation of such, to improve the accuracy of aims (i-iii) above. It is also planned to estimate benefits and harms within narrow age subgroups, and estimate effects of time since screen, to inform policy on the target population and on the frequency of screening.

The PRU shall also examine effects by area, to flag up areas where the screening has not worked as well, which will trigger investigation of the cause, and subsequent rectification, if necessary. This is a unique opportunity. Area-based process measures (screen detection rates, false positive rates etc) have long been available but there has been no data on the effect on clinical outcomes by area.

This study has been running for more than three decades, it is a stand-alone cohort study, the data has been supplied before, and the objectives remain the same (as noted above).

The data required for the aims listed above are

(1) Retention of the baseline demographic and screening data which are already held at QMUL;

(2) The death and cancer notification data up to 2005, which are also already held at QMUL; and

(3) further death and cancer notification data to 2016.

These data are requested as they will give the opportunity to estimate the effects of the NHS Breast Screening Programme on mortality from and incidence of breast cancer, with rigorous adjustment for potential biases. In particular, the further notification data to 2016 will give valuable information on the long-term effects.

Further background to and details of objectives

The NHS Breast Screening Programme (NHSBSP) was introduced in England and Wales in 1988, following the publication of randomised controlled trials (RCTs) and population demonstration projects which reported that mammographic screening could reduce breast cancer mortality by 20-30%.

Mechanisms to monitor the NHSBSP in terms of quality assurance were built into the programme and using performance indicators the programme currently appears to be performing well compared with RCTs. However, monitoring the performance of the service does not directly address questions about its effect on breast cancer mortality and the question remains whether the reduced breast cancer mortality observed in RCTs can be achieved in a population programme, such as the NHSBSP.

In addition, mortality from breast cancer is falling in the United Kingdom and there is debate as to how much of the reduction is due to screening or improvements in treatment (or some other factor). The difficulty of producing quantitative estimates of the effect of a national screening programme on mortality is well recognised. Population level data have been used to estimate the effect of the NHSBSP on breast cancer mortality in England and Wales, but ecological studies suffer the disadvantages of using aggregated data and are complex. The impact of the NHSBSP on breast cancer mortality in the East Anglian region has been estimated using prognostic characteristics of breast cancers to predict breast cancer mortality. However, to assess the programme rigorously the PRU needs to use individual-level data and mortality from breast cancer as the outcome.

The NHSBSP now screens more than 1.3 million women per year and has an annual budget of approximately £52 million. The programme has a high public profile, and debate following recent scientific publications has again raised the issue of whether the Breast Screening Programme is an effective use of NHS resources, and underlined the need for proper evaluation of the programme in terms of its effect on breast cancer mortality.

The primary aim of this study is to evaluate the impact of the NHS Breast Screening Programme on mortality from breast cancer. The objective is to use individual-level data on a cohort of women to assess breast cancer mortality in relation to screening history; specifically to compare risks in those invited with a contemporaneous group of the same age who were not invited until a later date.

Also to assess:

a) mortality from breast cancer in relation to attendance for screening. Those who attend for screening do not represent the entire target population of the programme and bias is introduced by factors influencing the uptake of an invitation. However, mortality among women who attend for breast cancer screening is of considerable interest.

b) mortality from all causes, all cancers and vascular disease in relation to invitation to and attendance at screening. Previous literature has reported a relationship between breast cancer screening and non-breast cancer deaths, specifically from other neoplasms and from vascular disease.

The applicants are now seeking support to flag the remainder of the cohort who have not been flagged previously in order to obtain up-to-date cancer incidence and mortality follow-up to conduct analysis on a longer-term basis. The objectives of this follow-up phase are to further assess the effectiveness of the screening programme by comparing risk of death from breast cancer in women who were invited for screening with those who were not invited in the same period for follow-up over a 25-year period, and wider objectives related to the impact of the screening programme on diagnosis of breast cancer.

This unique dataset will provide estimates of the major benefit (reduction in breast cancer mortality) and harm (overdiagnosis: identification of false positives) conferred by the NHS Breast Screening Programme. It includes subjects recruited at the initiation of the programme in the late 1980s. Despite substantial experimental and observational evidence there remains uncertainty over the magnitude of the benefit of the NHS breast screening programme in terms of breast cancer mortality and the magnitude of the harm in terms of overdiagnosis (identification of false positives). This extremely valuable data resource has the potential to provide answers to these questions with a precision unrivalled by any other study worldwide.

The post-2005 data is needed to answer two specific lines of research:

- First, there is a need to estimate the long-term benefit of early detection, in terms of breast cancer mortality. For many cancers, if recurrence does not occur within, say, five years, it will not occur at all. This is not the case for breast cancer, which can recur decades after the original diagnosis, and recurrence can prove fatal. It is therefore important to ascertain whether the benefit of early detection holds for later events as well as earlier.

- Second, there are conflicting results in terms of early detection, where some trials show the benefit being maintained up to thirty years and others (particularly in younger age groups) show it declining after ten years. Resolving this issue may have implications for other aspects of management than early detection- it may have implications for treatment of breast cancer occurring at earlier ages, for example.

The data controller requires baseline data as already present in the dataset held at Queen Mary University of London, plus notifications of breast cancers and all deaths within the cohort, including cause of death. The notifications of deaths and cancers at the moment only extend to 2005. Queen Mary University of London seek approval for reopening this and giving notifications to 2016 - or 2017 if available -, and request that this be part of the new data sharing agreement. In addition, Queen Mary University of London request that flagging be extended to the entire cohort, rather than only those recruited after 1991. This would give a longer-term and larger population incidence-based mortality effect for the screening programme than is available from any other dataset in the world. In addition, it will provide long-term data on incidence (indeed lifetime incidence in many cases), essential for estimation of the identification of false positives unbiased by lead time. Note that substantial additional funding is available for the further flagging and notifications.

This work was instigated more than 20 years ago by the Institute of Cancer Research and was transferred to Queen Mary University of London (QMUL) in 2015, following the award of the Department of Health Policy Research Unit in Cancer Awareness, Screening and Early Diagnosis to QMUL, and the closure of the Cancer Screening Evaluation Unit at ICR. The aims of the work remain the same, to quantify the benefits and harms of the NHS Breast Screening Programme.

The expected outputs pertain to methodology, knowledge creation in relation to the effect of breast screening, and potential healthcare policy changes. For this first, this dataset affords an opportunity to understand better and in a UK context the phenomenon of self-selection for screening, as it will contain women who have, contemporaneously, not yet been invited (eg, someone recruited to the cohort in 1989 and first invited in 1994 will have spent 5 years not yet invited to screening) and women who have refused (ie, turned down their invitation) screening, that is it can estimate self-selection unconfounded by time, and the potential changes in population behaviour and attitudes over time. In terms of knowledge, it will give estimates of the benefit of screening in terms of breast cancer deaths prevented and harms in terms of identification of false positives to an unprecedented level of precision. This will result in high-impact publications. Further, the size of the data resource will permit detailed analyses of subgroups and tactical issues, such as time between screens and age at screening with the potential to inform policy as to target population, interval between screens, etc.

In terms of the findings reaching decision/policy makers, researchers, clinicians, patients and the public, as with all PRU's research, the results are published in peer-reviewed medical journals. The study has already borne fruit in terms of publication of medium-term estimates of the reduction in mortality and the identification of false positives associated with invitation to screening:

Johns LE, Coleman DA, Swerdlow AJ, Moss SM. Effect of population breast screening on breast cancer mortality up to 2005 in England and Wales: an individual-level cohort study. Br J Cancer 2017; 116: 246-52

Johns LE, Swerdlow AJ, Moss SM. Effect of population breast screening on breast cancer mortality to 2005 in England and Wales: A nested case-control study within a cohort of one million women. J Med Screen 2018; 25: 76-81

The further follow-up and associated methodological development will enable estimation of the mortality reduction and the overdiagnosis (false positives) associated with actually being screened rather than only the effect of invitation. It will also allow estimation of the long-term effects.

Again, results will be published in peer-reviewed medical journals. However, prior to publication, these will be shared with the relevant policy community. The PRU is represented on all the cancer screening advisory groups and on the National Screening Committee. These bodies will be apprised of the results in a timely manner, as will Cancer Research UK, Macmillan Cancer Care and Breast Cancer Now. In addition, the funding body (the Department of Health and Social Care, via NIHR) will be informed of the results prior to publication.

The results shall also be posted on the PRU website. PRU will take advice from the PRU patient and public advisory committee as to the most appropriate form of general public disseminations. In addition, the PRU’s scientific advisory committee will be consulted on options for further dissemination.

It is estimated that results will be available in 2020.

All outputs will contain only data that is aggregated with small numbers suppressed in line with the HES Analysis Guide.

Note that a new Patient and Public Involvement Board has been set up to advise on:

- setting the research direction

- obtaining wider involvement from all the population this programme serves (ie, not just patients with lived experience of cancer)

- increasing public engagement

with the appointment of an academic PPI lead and a PPI co-investigator (Bishop Douglas Lewins).

The PAB's first meeting was on 14th March 2019.

The study has already borne fruit in terms of publication of medium-term estimates of the reduction in mortality and the overdiagnosis - identification of false positives - associated with invitation to screening:

Johns LE, Coleman DA, Swerdlow AJ, Moss SM. Effect of population breast screening on breast cancer mortality up to 2005 in England and Wales: an individual-level cohort study. Br J Cancer 2017; 116: 246-52

Johns LE, Swerdlow AJ, Moss SM. Effect of population breast screening on breast cancer mortality to 2005 in England and Wales: A nested case-control study within a cohort of one million women. J Med Screen 2018; 25: 76-81

Results include a substantial reduction in medium term mortality from breast cancer, with little or no overdiagnosis (Johns et al, Br J Cancer 2017; 116: 246-52). However, it would add considerable value to update these results to long-term effects and to extend the population flagged to the entire cohort.