Cardio-oncology: A high resolution national electronic health record investigation of the interplay between cancer and heart disease

VICORI (the Virtual Cardio-Oncology Research Institute) is a £1.4M 5-year programme funded by the British Heart Foundation and Cancer Research UK to undertake electronic health care record population research into the interplay between heart disease and cancer. VICORI's Programme Governance Board's terms of reference includes the role of an oversight committee and will:

• support the identification of current and future potential projects and partnerships to sustain developments

• ensure ongoing programme alignment with national policy and strategy

• complete work as individuals, or within sub groups, to assist with the delivery of the audit

• ensure the societies / organisations that are represented receive feedback on audit progress and outcomes

• ensure the programme maintains a patient focus.

The Programme Governance Board consists of representation from the three data controllers - Public Health England (PHE), Barts Health NHS Trust (the legal entity and data controller for the National Institute for Cardiovascular Outcomes Research/NICOR), and University of Leicester), patient representatives and academic partners (data processors for this agreement).

The work is delivered under the direction of the Programme Governance Board, specifically the Oversight Committee and is managed on a day-to-day basis by a programme manager (employed by the University of Leicester but holding honorary contracts with PHE and NICOR) who oversees and supports the delivery of the key research workpackages. The Office for Data Release (ODR) in PHE determine who can access the data (within PHE) and also with whom limited access pseudonymised data can be shared (subject to the limits of this agreement). This is supported by analytical experts and academic experts as required. This infrastructure provides close oversight of the programme including all aspects of data governance.

Funding is included for both the partner institutions (NICOR and Public Health England) as well as for the analytical team. The programme has been widely supported by professional organisations (including the British Cardiovascular Society, the British Cardiovascular Intervention Society, Macmillan Cancer Care, Cancer Patient’s Voice etc.). It has been developed in collaboration with the National Cancer Registration and Analysis Service (previously the NCIN) and the National Cardiovascular Intelligence Network in Public Health England. Permission is in place for linkage of NICOR data (approved by HQIP) and NCRAS data including cancer HES (which has been disseminated by NHS Digital) held under a data sharing agreement by PHE (approved by PHE-ODR) for VICORI.

VICORI will focus analyses on four key hypotheses of direct relevance to the outcomes of patients with Ischemic Heart Disease (IHD) and cancer and therefore of the highest potential for early impact. These hypotheses will provide the basis for focused analytical work-packages which will be progressed in parallel. Some of the questions which the research is hoping to answer include:

Workpackage 1: Hypothesis - Cardiac conditions are managed differently in cancer patients and cardiovascular outcomes are influenced by a prior cancer diagnosis

a) Are patients with a cancer diagnosis who subsequently develop acute myocardial infarction managed differently from patients without cancer and does this adversely affects their outcome?

• How does the procedural management of patients with cancer hospitalised with myocardial infarction (ST-Elevation Myocardial Infarction/STEMI, non-STEMI) differ from matched patients without a cancer diagnosis?

• How does the cardiovascular medical management of patients with cancer presenting with myocardial infarction differ from matched patients without a cancer diagnosis?

• How does cancer site, stage and grade (and stratified survival risk) affect this?

• How do any differences impact on clinical outcomes, such as death and major adverse cardiac and cerebrovascular events? (N.B. estimated Major Adverse Cardiovascular and Cerebrovascular Event (MACCE) rates will require reference to HES and mortality data to allow accurate time-to-event analysis)

• Is there regional variation in clinical practice?

• What is the opportunity to favourably modify practice and what is the potential impact of such a modification? (N.B. the study team recognise the analytical challenges of determining the clinical legitimacy of observed differences but have established experience of using national datasets for such analyses)

b) Are patients with cancer who develop advanced valvular and/or coronary artery disease managed differently from patients without cancer?

• What proportion of cancer patients with severe aortic stenosis are managed with surgery versus transcatheter aortic valve implantation versus palliative valvuloplasty?

• Likewise what are the rates of coronary artery bypass surgery (CABG), 3-vessel Percutaneous Coronary Intervention (PCI) or non-aortic valve surgery in cancer patients compared to matched non-cancer patients?

• How does cancer site, stage and grade affect this?

• How do these different management strategies impact on clinical outcomes (see note above)?

• Is there regional variation in practice?

• What is the opportunity to favourably modify practice and what is the potential impact of such a modification (see note above)?

Workpackage 2: Hypothesis - Cardiovascular treatments, interventions and surgery alter cancer risk or outcomes

• Does the X-ray dose administered to childhood congenital heart disease patients who undergo fluoroscopically guided cardiovascular interventions alter the subsequent cancer risk? (N.B. the congenital database dates back to 2000 but the study team estimate the follow-up period (up to 15 years) will be sufficient to demonstrate the expected effect. However the study legacy data resource will ultimately further enhance the power of these analyses)

• Does cumulative X-ray from cardiovascular fluoroscopic procedures in adult patients alter subsequent cancer risk? (N.B. this will intersect with existing work with NICOR partners on repeat procedures in coronary patients led by the Principal Investigator)

• Are specific cardiovascular procedures particularly high risk (e.g. Chronic Total Occlusion/CTO angioplasty)?

• Do other cardiovascular treatments (e.g. Angiotensin-Converting Enzyme/ACE inhibitors, beta blockers, statins) favourably or unfavourably alter cancer risk or cancer outcomes?

Workpackage 3: Null hypothesis - Cancer treatments cause long term adverse cardiovascular diseases states or cardiovascular events

• Do cancer treatments increase the risk of subsequent cardiovascular disease?

• If so, then which treatments are responsible for the increase and which types of cardiovascular disease are increased?

• How do any radiation-related risks vary with time since treatment? Do they only appear many decades later, or do they start within the first few years of treatment, as has recently been suggested?

• How do any chemotherapy (or other drug-related) increases vary with time since treatment? Are they limited to the first year or so after commencement of treatment, or do they cause long-lasting increases?

• How do any cancer-treatment related increases vary with the age and pre-existent cardiovascular risk factors of the patient when the treatment is administered?

• An increasing number of patients are receiving treatment with both radiation and systemic drugs. Does radiation modify the effect of chemotherapy and vice versa?

Workpackage 4: Null hypothesis - Pre-existent cardiac disease alter cancer management or outcomes

• Does previous coronary artery bypass surgery (CABG), Valve surgery, Percutaneous Coronary Intervention (PCI) or Acute Myocardial Infarction (AMI) reduce adoption of a curative (surgical, chemotherapy, radiotherapy or targeted therapy) strategy for and timeliness of cancer treatment?

• Are there regional differences in this interaction between prior cardiovascular disease and approaches to cancer treatment?

• How do these differences affect clinical outcomes?

• What is the opportunity to favourably modify practice and what is the potential impact of such a modification?

This request is to:

~ enable linkage and onward sharing of pseudonymised HES and mortality data linked to NICOR (HQIP) audit data (NIC-359940-W1R7B) to PHE. The NICOR audits consist of:

~ MINAP (Heart Attacks)

~ Percutaneous Coronary Intervention (Angioplasty)

~ Adult Cardiac Surgery

~ Cardiac Rhythm Management (Arrhythmia)

~ Congenital Heart Disease

~ Heart failure

~ TAVI (Transcatheter Aortic Valve Implantation)

~ Linked HES

~ Linked mortality data

~ enable linkage of pseudonymised PHE-ODR National Cancer Registration and Analysis Service (NCRAS) including data already linked to HES and mortality data according to the DSA between NHS Digital and PHE (NIC-343380-H5Q9K). NCRAS data consists of:

~ Site specific cancer

~ audits

~ Existing links:

~ Systemic Anti-Cancer Therapy Dataset

~ The National Radiotherapy Dataset

~ Linked HES

~ Linked mortality data

No further data will be disseminated from NHS Digital for this agreement.

The legal basis for processing falls under GDPR Article 6(1)( e) and GDPR Article 9(2)(i). Processing and dissemination is for research purposes, and also meets the public interest criteria as defined by the ICO for good decision-making by public bodies and securing the best use of public resources. The processing of data for this study is a task of public interest since the large patient numbers generated by national level data will enable outcomes on longer term incidence, outcomes and likely NHS impacts to be investigated.

A master-patient index will be generated containing the pseudo-IDs of patients with data present in NICOR and/or NCRAS together with a minimum set of data fields (which audits hold data on the patient and minimal demographic data (this will include non-identifiable data such as age, sex and ethnicity but will contain no identifiable data). Analysts will then use the master patient index to identify the patient cohorts required for a specific analysis. These will then be used to generate a bespoke pseudonymised dataset.

The dataset will then be used to understand the interplay between cancer and cardiovascular disease in a real world, using national population data and to quantify the potential opportunity to improve patient outcomes through a population-based linked multisource electronic health record national cohort study. Specific aims are:

• To investigate if cardiac conditions are managed differently in cancer patients and cardiovascular outcomes are influenced by a prior cancer diagnosis (workpackage 1)

• To investigate if cardiovascular treatments, interventions and surgery alter cancer risk or outcomes (workpackage 2)

• To investigate if cancer treatments cause long term adverse cardiovascular diseases states or cardiovascular events (workpackage 3)

• To investigate if pre-existent cardiac disease alters cancer management or outcomes (workpackage 4).

Specific bespoke datasets of pseudonymised VICORI data required for the above mentioned workpackages will then be shared with the listed data processors who will conduct the detailed analyses required. The following academic partners will receive data in this way, subject to approval by the Office for Data Release of Public Health England (the data controller), specifically; University of Leicester, University of Birmingham, University of Oxford, Royal Marsden Hospital, Royal Brompton and Harefield NHS Trust, University of Leeds, University College London and London School of Hygiene and Tropical Medicine.

Barts Health NHS Trust, Public Health England and University of Leicester are the data controllers. These three organisations have established a collaboration to jointly deliver a research database of patients with cancer and cardiovascular disease, they are determining the key research questions that will be addressed using the linked data and the purpose for which and the manner in which the data will be used. The key areas of research are:

Investigate the impact of a diagnosis of cancer on the management of ischaemic heart disease and acute coronary syndrome.

Study the impact of co-existent ischaemic heart disease and acute coronary syndrome on the management of cancer.

Report the longer-term cardiovascular sequelae of anti-cancer treatment in cancer survivors.

Undertake a health services research appraisal of provider based quality of care and clinical outcomes for patients with co-existence cancer and cardiovascular disease.

There are seven data processors who will work on one or more of the workpackages, under the direction of the data controllers. The range of research questions determined by the data controllers require a variety of analytical strategies and will include epidemiologists and statisticians with a considerable track record of the analysis of similar large datasets. The organisations providing the resource and expertise will be carrying out the processing on behalf of the data controllers, they have not determined the research questions or the purpose for which and the manner in which the data will be used, the data required to answer the research questions has already been predetermined by the data controllers.

The seven data processors are:

1) University of Oxford

2) University of Leeds

3) London School of Hygiene and Tropical Medicine

4) The Royal Marsden NHS Foundation Trust

5) Royal Brompton and Harefield NHS Trust

6) University of Birmingham

7) University College London

Imperial College London and South Tees Hospital are acting in an advisory capacity only.

Patient objections will be reapplied to the data to be processed under this agreement, as it is being used for a new purpose. This will take place prior to any other processing under this agreement. Once the objections are applied to the data, the HES IDs will be provided to the University of Leicester. HES IDs are not identifiable. No analysis or comparison may be undertaken between the two datasets.

Step 1: Conducted by NICOR staff: NICOR data already linked to HES and mortality data (NIC_359940-W1R7B) for the purposes of audit would first be modified to encrypt dates of death so that these cannot be used for the purposes of identification. Dates pertaining to each patient will either be used to generate derived survival intervals and expanded to within 1 week or, offset by having a number of days (0-7) added or subtracted such that no date can be identified but intervals between dates can measured. The number of days added or subtracted would be random between patients to prevent decryption.

Step 2: Conducted by NICOR staff: NICOR data already linked to HES and mortality (NIC-359940-W1R7B) would be encrypted using the common SALT key such that identifiable information (NHS number and date of birth) are used to generate a pseudo ID.

Step 3: Conducted by PHE staff: In parallel, NCRAS staff will encrypt NCRAS data already linked to HES and mortality (NIC-343380-H5Q9K) using the same encryption algorithm such that identifiable information (NHS number and date of birth) are used to generate a common pseudo ID.

Step 4: Co-ordinated by NICOR and PHE staff: Pseudonymised NICOR data linked to HES and mortality is securely transferred into PHE.

Step 5: Conducted by PHE staff: The pseudo IDs from NICOR and NCRAS datasets (both already linked to HES and civil registration data) are merged to generate a common master patient index consisting of the pseudo IDs, the registries in which those IDs appear (i.e. the NICOR audits as listed above or the NCRAS audits as listed above) and a minimal demographic dataset (this will not include identifiable data such as date of birth).

Step 6: VICORI analysts will use the master patient index to identify cohorts pertaining to specific research questions pertaining to the workpackages. No other study personnel will have access to the pseudonymised data or master patient index. VICORI analysts will use the data dictionaries from NICOR, NCRAS and linked HES to identify the data fields required for their analysis. VICORI analysts will then extract a bespoke dataset specific to each research questions for the purposes of analysis. This pseudonymised bespoke data will then be shared on a limited access basis with the lead investigators at the academic partners named as data processors, below (specifically: University of Leicester, University of Leeds, University of Oxford, University of Birmingham, University College London, Royal Marsden Hospital NHS Trust, London School of Hygiene and Tropical Medicine and the Royal Brompton and Harefield NHS Trust).

This bespoke dataset will be maintained only for as long as it is required for analytical purposes.

The linked data resource will therefore be generated and maintained by NCRAS within Public Health England under the terms of the Data Sharing Agreement. Analysis will be conducted either on-site at PHE or by secure remote access (using the same current standard procedures for off-site remote secure working by NCRAS analysts) by VICORI analysts with appropriate PHE honorary contracts and permissions.

Once generated the pseudonymised linked data would be stored within PHE. Access to the pseudonymised data and Master Patient Index would be through permission of the Office for Data Release (ODR) by PHE employees or persons with a suitably robust PHE honorary contract either on site or by a suitable secure and approved remote access link (VPN) as agreed with ODR. This will constitute the VICORI analytical team. Data processors will receive bespoke deidentified data appropriate to the specific research questions on a limited access basis and subject to VICORI oversight committee and ODR approval.

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

All organisations party to this agreement must comply with the Data Sharing Framework Contract requirements, including those regarding the use (and purposes of that use) by “Personnel” (as defined within the Data Sharing Framework Contract i.e: employees, agents and contractors of the Data Recipient who may have access to that data).

No attempt will be made to identify/re-identify anyone from the data once it has been pseudonymised.

The national linked datasets generated for this proposed programme will represent the largest cardio-oncology analysis to date. As such the potential for high impact outputs is considerable. The study team includes key opinion leaders from both cardiovascular and cancer epidemiology as well as the nascent cardio-oncology community and is therefore well placed to ensure study findings are given prominence at national and international meetings (including the British Cardiovascular Society) in relevant specialty areas. Findings will be published in journals with open access. The Lay study representatives will support the generation of ‘Plain English’ summaries of key outputs which will be posted on the study website and those of the partner institutions to ensure wider public understanding of the programme findings.

The following outputs will be produced:

~ Reports of interim results will be provided to the British Heart Foundation/Cancer research UK every 6-months.

~ The final report of results will be submitted to the British Heart Foundation/Cancer Research UK in 5-years upon programme completion. This will cover all findings of the study including: factors influencing planning and implementation. Throughout the programme study findings will be published in the open access, peer-reviewed journal(s) with a Lay summary of findings produced by our Lay Executive Members for publication on the study website.

~ For [each/a specific] paper published, a short presentation is developed to summarise the findings for a range of stakeholders, including clinicians, academics and patients.

~ Findings will be presented at appropriate international conferences such as the British Thoracic Oncology Group (BTOG) and the British Cardiovascular Society, national meetings and patient and public involvement (PPI) and patient meetings.

The primary focus of the work is to use national health records to advance a paradigm shift in our understanding of the interactions between heart disease and cancer. The secondary focus is to establish an accepted route for similar data linkage proposals so that the full potential of Big Data analysis in the UK can be achieved whilst maintaining the highest standards of data governance.

Although multi-morbid patients account for the largest proportion of NHS clinical activity, this population are under-represented in current clinical research. Such patients are, for example, usually excluded from randomised controlled trials. Population research with data linkage holds huge potential for better understanding of best practice in the real world of multi-morbid poly-pharmaceutical patients.

This research has considerable potential to lead directly to patient benefit both by enhancing current treatment approaches, developing new clinical trials (e.g. database follow-up RRCTs) and by enabling the early characterisation of any cardio-oncological interactions arising from the new treatment strategies. For example chemotherapy induced cardio-myopathy, especially following anthracyclines, is well established. However, the large patient numbers generated by national level data, will enable information on longer term incidence, outcomes and likely NHS impacts to be probed. This will enable refinements in cancer treatment to reduce adverse cardiovascular sequelae to be monitored in near real-time and allow the investigation of the potential of early protective strategies.

Likewise, coronary artery disease following thoracic radiotherapy is established. However as radiotherapy regimes are refined, rolling contemporary data is needed with the power to quantify the remaining effect sizes in this population. Equally, whilst it is known that X-ray doses increase cancer risk, this programme will enable the long term effects of fluoroscopy doses from cardiovascular procedures to be determined and interactions, for example with age and sex, to be studied. The risk of specific procedures (such as chronic occlusion angioplasty) can also be investigated. Hence as survival rates improve, these data will allow refinement of current treatment strategies based on a broader understanding of overall longer term morbidity and mortality as well as the development of new protective approaches to prevent adverse long term treatment effects. It is thus anticipated that this programme will generate research outputs with considerable potential to lead to changes in clinical practice.

Specific benefits for the workpackages include:

Workpackage 1 - benefit patients with cancer who develop cardiovascular conditions and clinical teams who care for those patients by providing data on whether current management is optimal and investigating outcomes in patients with cardiovascular diseases.

Workpackage 2 - benefit patients with cardiovascular disease who develop cancer and the clinical teams who care for those patients by providing data on how these conditions and their treatment alter future cancer risks and modify cancer outcomes.

Workpackage 3 - benefit patients requiring treatment for cancer and the clinical teams who care for those patients by providing data on the long term cardiovascular sequelae of these treatments and giving indications on how these might be minimised.

Workpackage 4 - benefit patients with both cardiovascular disease and cancer and the clinical teams who care for those patients by investigating how the co-existence of these conditions affects current clinical management and identifying how this might be improved.

Yielded Benefits is not a requirement for new applications.