Study to investigate the accuracy with which breast cancer recurrence can be identified in women registered with invasive breast cancer using routinely collected data compared with recurrence information collected by the AZURE trial ( ODR1718_364 )

Breast cancer is the commonest cancer in the UK and around 55,000 women are diagnosed with the disease each year. Breast cancer incidence rates have been increasing since the mid-1970s and the disease now accounts for almost 1 in 3 of all newly diagnosed cancers in women. Although a large amount of population-based information is available on death and survival from breast cancer in the UK, much less information is available on breast cancer recurrence. Reliable information on breast cancer recurrence is crucial for decision making in several areas of breast cancer management, including:

i. comparing the outcomes of various treatments so that the best treatments can be identified for each woman diagnosed with breast cancer

ii. establishing the prognosis for individual breast cancer patients so that clinicians can advise patients appropriately.

iii. determining the burden of breast cancer so that appropriate policies for the management of the disease can be put in place.

The reason that there is a paucity of data on recurrence of breast cancer is that recurrence information has not traditionally been part of the standard NHS information flow. To rectify this, Public Health England (PHE) mandated that, from 1st January 2013, all women newly diagnosed with breast cancer after 1st January 2013 and who subsequently suffer a recurrence of their cancer should be ‘flagged’ in the new Cancer Outcomes and Services Dataset (COSD). Since then, there have been variables in COSD to indicate recurrence (2). However, they are not completed reliably. In addition, on 1st January 2013 there were around half a million women alive in the UK who had already been diagnosed with breast cancer. Little is known about how many of these women had a recurrence before 1st January 2013 and at present there is no mechanism in place to record any recurrences that they have developed since then, or may develop in the future.

To fill this information gap, researchers at the University of Oxford have been collaborating with National Disease Registration Staff to identify recurrences in women registered with breast cancer using a number of routinely collected data sources, such as Cancer Analysis System (CAS), Hospital Episode Statistics (HES) Cancer Waiting Times (CWT), Digital Imaging Dataset (DID), Radiotherapy Dataset (RTDS) and Systemic Anti-Cancer Therapy dataset (SACT). As part of this study, an algorithm has been developed to identify which women have had a recurrence of their breast cancer and, for those who did, the date of the recurrence.

The aim of this project is to undertake this external validation of the algorithm using data from the AZURE trial.

Data from the AZURE trial have been linked to routinely collected data sources within NHS England. This linkage was carried out using data already held by the Leeds Clinical Trials Unit and by NHS England. No study participants were contacted. The study comprises secondary use of information that had already been collected via trial follow-up and routine methods.

An initial validation of the algorithm demonstrated that it could successfully identify distant recurrences recorded in the AZURE trial. Further analyses are, however, needed to improve the ability of the algorithm to identify locoregional recurrence and contralateral breast cancer using routinely collected data and to confirm that the algorithm produces appropriate results when applied to all women registered with invasive breast cancer by the National Disease Registration Service.

NDRS Cancer Registrations

NDRS Cancer Pathway

NDRS Linked Cancer Waiting Times (CWT) Monitoring

NDRS Linked Diagnostic Imaging Dataset (DID)

NDRS Linked Hospital Episodes Statistics AE (HES AE)

NDRS Linked Hospital Episodes Statistics Admitted Patient Care (HES APC)

NDRS Linked Hospital Episodes Statistics Outpatients (HES OP)

NDRS Radiotherapy dataset (RTDS)

NDRS Systemic Anti- Cancer Therapy (SACT) dataset

The quantum of data requested is necessary to achieve the remaining objectives of this project as stated above. Data are required to investigate the accuracy with which the various types breast cancer recurrence can be identified in women registered with invasive breast cancer using routinely collected data compared with recurrence information collected by the AZURE trial.

The level of the data is pseudonymised and will be minimised as follows:

-Limited to the approximately 2200 women registered in the AZURE trial in England

-Limited to the approximately 2200 women mentioned above for whom Patient IDs (usually NHS numbers) have been previously submitted to the NDRS by the University of Leeds.

-Limited to women randomised to the AZURE trial between 2003 and 2006

-Limited to data between the date on which the woman was diagnosed with breast cancer and the date of her last follow-up in the AZURE trial. Follow-up can last for up to 10 years.

-Limited to variables in the dataset listed above that either do or may indicate that the woman has had a recurrence of her breast cancer.

The data will be processed by the University of Oxford in England.

University of Oxford relies on General Data Protection Regulation Article 6(1)(e) - the processing is necessary for you to perform a task in the public interest or for your official functions, and the task or function has a clear basis in law. The public interest in this circumstance is research to increase medical knowledge for the benefit of all and to improve public health.

University of Oxford relies on General Data Protection Regulation Article 9(2)(j) - processing is necessary for archiving purposes in public interest, scientific or historical research purposes. The legitimate need for processing special category data under 9(2)(j) is that it adheres to the UK Policy Framework for Health and Social Care Research and aims to produce generalisable and publicly available information to inform future decisions over patients’ treatments or care.

Before starting this project, a pilot project was carried out using just data from the West Midlands Cancer Registry. The results of this pilot were presented to several groups, including the Association of Breast Surgeons and the project was discussed with those present. The views of consumer representatives (Independent Cancer Patients Voice) were also sought in an advisory capacity in dedicated meetings.

Funding is provided by the University of Oxford and by Grants from Cancer Research UK to the University of Oxford. Cancer Research UK will have no ability to suppress or otherwise limit the publication of findings of this research. There are no other organisation(s) involved or accessing the NHS England data, including organisations acting in an advisory capacity or as part of an oversight or steering committee.

The AZURE trial was a randomised controlled trial conducted in women diagnosed with invasive breast cancer in which some women standard treatment only and some women received zoledronic acid in addition to standard treatment. The objective of the trial was to compare the rate at which recurrences occurred in the two groups. Women were recruited from 2003 to 2006. Following recruitment, the women were followed up by inviting them to attend the clinic at which they were treated at regular intervals for 10 years. The data were assembled and collated by the Clinical Trials Research Unit at the University of Leeds. For the present study, the trials office in Leeds assigned an anonymous Study ID to each woman in the AZURE trial and then submitted these Study IDs, accompanied by patient identifiers (usually NHS number) to Public Health England (PHE) using standard operating procedures.

PHE identified the women in the AZURE trial in the National Cancer Registration Database. They then, using the algorithm that had already been developed, they identified women likely to have had a recurrence of their breast cancer and the date and type of the recurrence to a sample of the women in the AZURE trial. PHE then notified the Chief Investigator at the University of Oxford of these events and dates, using the anonymous Study IDs as the only identifiers. This process was repeated several times until the algorithm appeared to be working as well as could be expected for the women in the sample. The final version of the algorithm was then applied to the remaining women in the AZURE trial to test its validity and the Chief Investigator at the University of Oxford was notified of these events and dates.

The data linkages between the AZURE trial and the Cancer Registry have already been carried out and completed. Follow-up is not being extended for the AZURE trial, but the studies analyses are not yet complete, so The University of Oxford wish to continue and complete them until 2029.

The Leeds Trials Office provided the data collected during the course of the AZURE trial to the Chief Investigator at the University of Oxford, with the anonymous Study IDs as the only identifiers. Thus, no identifiable information is accessible to anyone outside the Leeds Clinical Trials Unit and NDRS at Public Health England (now NHS England) (both of whom already have access to these data). Researchers at the University of Oxford received only de-identified data with anonymised IDs.

This study has been granted ethical approval (Reference 17/WM/0347) and a Clinical Data Disclosure Agreement between the University of Leeds and the University of Oxford is in place.

The data will remain on the servers at the University of Oxford at all times. The data will not be transferred to any other location. The University of Oxford provides IT support and all IT hosting services. Data will be stored within Nuffield Department of Population Health (NDPH) in Oxford University. Electronic data files are kept on password-protected network servers behind a local firewall. Servers are kept in a locked room with access restricted to IT staff. Backups are held on private internal servers spread over three locations within the Old Road Campus of the University of Oxford. Backed up data is encrypted both in transit and at rest. Access to data is restricted via user identification. No data are moved or copied from these servers.

Access to electronic data is restricted via user identification.The IT department have setup a special permissions compliant folder to receive and hold data securely. The PI of the study team controls who can access this folder and the list is reviewed every 6 months. Data are not copied but can be accessed by analysis programs. Receipt of data is recorded in an asset register within NDPH. When data are to be deleted a request is made to the NDPH IT department who provide a deletion certificate which the study team enter into the asset register (any backups resulting from the 28-day back-up cycle are also deleted).

The data will not leave England/Wales at any time.

The data will be accessed only by authorised personnel via remote access. Personnel are prohibited from downloading or copying data to local devices. The University of Oxford will ensure the correct supervisory and contractual requirements for personnel accessing the data are in place and that the University of Oxford’s organisational governance policies and controls are adhered to.

Whereby during the term of this agreement processing is required to be carried out by students affiliated with the University of Oxford (to potentially form part of studies ie: a DPhil degree) The University of Oxford, will only do so if beneficial to the research.

Whereby during the term of this agreement processing is required to be carried out by individuals with honorary contracts (to help introduce increased level of clinical expertise), The University of Oxford will only do so if beneficial to the research to help achieve the study objectives.

Data processing and access will be carried out only by personnel who have authorisation from the PI and who have been appropriately trained in data protection and confidentiality.

No other organisation is permitted to access the data, including the funder Cancer Research UK. The funder Cancer Research UK will not have influence on the outcomes nor suppress any of the findings of the research.

All data held by the University of Oxford are de-personalised and no linkages other than those described above will be carried out. Analysts and researchers from the Nuffield Department of Population Health at The University of Oxford will analyse the data for the purposes described above and there will be no requirement and no attempt to re-identify individuals when using the data in Oxford.

The research team have concluded the first stage of analysing these data.

The expected outputs of the data processing to date included:

• A peer-reviewed publication describing the process of deriving the algorithm, demonstrating its validity for identifying distant recurrences and describing its strengths and limitations.

The publication took place in 2025 and is available as follows:

Probert J, Dodwell D, Broggio J, Coleman R,Marshall H, Darby S, Mannu G.”Identification of recurrences in women diagnosed with early invasive breast cancer using routinely collected data in England. “ BJC Reports, 2025, https://doi.org/10.1038/s44276-025-00154-1

• The publication does not describe the technical details of the algorithm and accompanying computer code written by NDRS staff that implements it on data available within NCRAS.

The publication does not contain NHS England data, but rather contains only aggregated information with small numbers suppressed as appropriate in line with the relevant disclosure rules for the dataset(s) from which the information was derived.

In addition to the above, publication, presentations have been made at conferences relating to breast cancer and other meetings. Results have been presented to groups of clinicians and scientists whose work focuses on breast cancer, including the UK Breast Intergroup and the UK Breast Cancer Group.

The algorithm has been shown to work well in identifying distant recurrences among women included in the AZURE trial. However further work is required to demonstrate that it produces appropriate results when applied to all women registered with invasive breast cancer in England women and also to improve the ability of the algorithm to identify locoregional recurrence and contralateral breast cancer. Once this has been done it will be possible for the National Disease Registration Service to include recurrence as a linked variable in their datasets.

In addition, the algorithm will be available for use within NHS England allowing the National Disease Registration Service (NDRS) to make those curating the data for other cancers aware of the possibility of identifying recurrences in this way, thus leading to the development of algorithms to identify recurrence for other cancers.

The first requirement is to confirm that the algorithm functions satisfactorily on a nationwide basis. Once this has been done, then the outputs will be communicated to relevant recipients through the following dissemination channels:

• Journals

• Social media (e.g. press releases)

• Public reports (all journal publications are accessible by the public)

• Briefing documents provided to clinicians [details to be decided]

• Open source frameworks [it is planned that, when all are satisfied with its performance, the source code for identifying recurrences will be available within NCRAS for use by NCRAS staff and other researchers]

• Oral presentations and poster displays at conferences

• Patient Information leaflets [results will form the basis of revised patient information leaflets]

• Press/media engagement

Public promotion of the research [e.g. via press/media engagement]

In addition to the paper that has already been publishes, the research team expect further outputs to be available in during 2026-2029. Both Cancer Research UK and Independent Cancer Patients Voice are aware of this work. Other societies and charities will be informed in due course.

It is hoped the research will be able to provide a method of identifying when a patient who has been diagnosed with breast cancer has had a recurrence of that cancer. This would be a major improvement in NDRS capability and would allow them to provide better follow-up for randomised trials and more informative data for observational studies. This study will help provide reliable information on breast cancer recurrences for research to improve treatment and for decision making in different areas of breast cancer management in the future. It will also enable NDRS to comply with the stated National Cancer Plan of being able to define cases of recurrent breast cancer reliably in the near future.

The public benefits that are expected to be achieved as an outcome of this project include:

• Better determination of the burden of breast cancer so that appropriate policies for the management of the disease can be put in place.

• Catalyzation of policy improvement for the management of breast cancer patients

• Facilitation of scientific research into the field of outcomes after breast cancer

The specific benefits to patients are expected as an outcome of this project are that:

• It is hoped it will enable comparison of the outcomes of various treatments so that the best treatments can be identified for each woman diagnosed with breast cancer.

• It is hoped it will enable the prognosis for individual breast cancer patients to be established so that clinicians can advise patients appropriately.

It would allow follow-up of the women in randomised trials to be undertaken much more cheaply and easily in the future.

Positive feedback have yielded from recent presentations derived from development work so far.

July 2023: “Breast Cancer Recurrence Algorithm for Routinely Collected Data” Presentation of project by (University of Oxford), and the NDRS analysts (National Disease Registration Service) to the Health Quality Improvement Partnership (HQIP) commissioned National Audits of Primary Breast Cancer (NAoPri) and Metastatic Breast Cancer (NAoMe). These audits sit within the National Cancer Audit Collaborating Centre (NATCAN) established as a new national centre of excellence in October 2022 with the aim of strengthening NHS cancer services by looking at treatments and patient outcomes across the country. The members of these audits indicated how impressed they were with the project. They thought that the identification of recurrence is vital for both breast audits and that this work will inform parallel initiatives in other cancer sites.

August 2023: “Developing an Algorithm to Identify Breast Cancer Recurrences using Routinely Collected Data in England”. Oral presentation at the 44th Annual Conference of the International Society for Clinical Biostatistics: Joint Conference with the Italian Region of the International Biometric Society, Milan, Italy.

May 2025: The following peer-reviewed paper was published in the journal BJC Reports:

Probert J, Dodwell D, Broggio J, Coleman R,Marshall H, Darby S, Mannu G.”Identification of recurrences in women diagnosed with early invasive breast cancer using routinely collected data in England. “ BJC Reports, 2025, https://doi.org/10.1038/s44276-025-00154-1