Speedwell Study - Longitudinal Study of Ischaemic Heart Disease

Text removed is struck through; text added is underlined. Unchanged paragraphs are summarised rather than repeated.

Benefits reported

[3 paragraphs unchanged] Update 2025- Following a period of relative inactivity due to resource constraints, over the past 18 months considerable effort has been put into developing the policies and procedures to support the use of the Speedwell data by researchers. The study is one of several legacy cohort studies (i.e., studies with no ongoing participant involvement) maintained by the MRC Integrative Epidemiology Unit (MRC IEU) at the University of Bristol. A formal process is now in place for researchers to apply for access to data, including review of applications by a Data Access Committee. Oversight of the study is provided by a Steering Committee and a Public Advisory Group feeds into the maintenance, development and promotion of the study. Within the next REC review period (up to 2029) the Data controller hope to move the Speedwell data into the UK Longitudinal Linkage Collaboration (LLC), a trusted research environment (TRE), that will further enable Speedwell data to contribute to meta-analyses with other UK cohort studies. The controller has been approached by researchers who are interested in analysing Speedwell data when the Data controller can facilitate access to updated data either locally (for University of Bristol researchers) or via the TRE. Therefore, whilst the yielded benefits have been minimal in the period since the Data Controller's last application renewal, gaining approval to maintain current data is vital to secure this resource for the long term. Furthermore, with the future addition of new outcome data from NHS-England the Data Controller will have followed up almost the whole cohort to death which is very unusual in most studies. These data provide a window into the health of men from the earlier part of the twentieth century and will enable comparisons with more modern cohorts.

Unchanged: Objective for processing, Processing activities, Expected output, Expected measurable benefits.

Text removed is struck through; text added is underlined. Unchanged paragraphs are summarised rather than repeated.

Objective for processing

[8 paragraphs unchanged] The aim of collecting the data can only be achieved if it [23 words unchanged] specific deaths and cancer registrations) which can only be obtained from NHS Digital England Medical Research Information Service (MRIS) reports, as the study are no longer [12 words unchanged] the fact most will have died. The additional data provided by NHS Digital England is very valuable as it will increase the statistical power to look [20 words unchanged] continue processing identifiable data they already hold from 1982 to March 2016. [2 paragraphs unchanged] Current analyses being undertaken by the ERFC collaboration and how they use the data provided by NHS Digital England are described below: 1. Determining the reductions in life expectancy according to different ages at diagnosis of diabetes mellitus. This research will use NHS Digital England data on date and cause of death to calculate the negative impact on life expectancy for diabetics versus non-diabetics. 2. Development and validation of updated CVD risk prediction models for Europe [26 words unchanged] fatal and non-fatal CVD events and thus necessitates the use of NHS Digital England data. 3. Estimating the association between renal function and CVD incidence in the [11 words unchanged] to predict future fatal and non-fatal CVD events and thus requires NHS Digital England data. 4. Use of lifetime risk for cardiovascular disease risk assessment. Current risk [37 words unchanged] period. Data on vital status and age at death comes from NHS Digital. England. 5. Risk prediction for composite cardiovascular outcomes. This will expand the usual [11 words unchanged] events such as heart failure and pulmonary embolus and thus requires NHS Digital England data. 6. Sequential strategies for CVD screening, including health-economic evaluation. This work compares [14 words unchanged] to the prevention of future events (fatal or non-fatal) and requires NHS Digital England data. The University of Bristol has informed NHS Digital England that the data shared with the University of Cambridge was comprised of variables obtained from sources other than NHS Digital England and variables derived from NHS Digital England data. Until any further determination is made as to whether NHS Digital England are satisfied that the data shared is no longer considered NHS Digital England data, the University of Cambridge is listed in this Agreement as data [25 words unchanged] University of Cambridge would be removed from future versions of this Agreement. [1 paragraph unchanged] Researchers have access to individual level data from the baseline and subsequent assessments with outcome data from either participant self-report or hospital data and mortality from NHS Digital. England. NHS Digital England data covers all follow-up events to enhance statistical power and the geographical [147 words unchanged] best achieved by having cause of death data which maintains maximum flexibility. The data held and shared with researchers is the minimum required for [6 words unchanged] several additional steps are taken to mitigate the possibility of re-identification (NHS Digital England data dictionaries have been provided to enable researchers to assess for themselves any potential risk). [1 paragraph unchanged] The University of Bristol has shared data with a research group at [48 words unchanged] the University of Bristol for this purpose has been assessed by NHS Digital England as derived data. NHS Digital England has determined that data has been shared with the University of Oxford [37 words unchanged] is either aggregate (as in publication), anonymous, or pseudonymised containing no NHS Digital England data (such that additional data provided by the collaboration can be added [18 words unchanged] data and risk re-identification – this includes measures to suppress rare events). [7 paragraphs unchanged]

Processing activities

Identifying data was shared with the Office for National Statistics (ONS) to [49 words unchanged] ONS to the Health and Social Care Information Centre (HSCIC, now NHS Digital) England) in 2008, and this service was last supplied in March 2016. NHS Digital England data is currently held in a secure relational database where it has [131 words unchanged] not even seen as they are part of a much bigger average. [1 paragraph unchanged] No further linkages are undertaken to any other external or publicly available data. NHS Digital England data is linked to the patient data collected as part of the [67 words unchanged] signed consent forms which it needs to maintain to prove individual consent. [4 paragraphs unchanged]

Unchanged: Expected output, Expected measurable benefits, Benefits reported.

Text removed is struck through; text added is underlined. Unchanged paragraphs are summarised rather than repeated.

Objective for processing

[3 paragraphs unchanged] The processing meets the conditions of Schedule 1 Part 1 paragraph 4 of the Data Protection Act 2018 as the processing: (a)is necessary for archiving purposes, scientific or historical research purposes or statistical purposes, (b)is carried out in accordance with Article 89(1) of the GDPR (as supplemented by section 19), and (c)is in the public interest. [1 paragraph unchanged] The aim of collecting the data can only be achieved if it [91 words unchanged] and rarer outcomes. This is a request to permit The University of Bristo Bristol to continue processing identifiable data they already hold from 1982 to March 2016. [1 paragraph unchanged] Since 2002 The University of Bristol have shared a derived sub-set of data with The University of Cambridge, for the purpose of [98 words unchanged] currently working on 6 distinct research projects as part of their programme. [7 paragraphs unchanged] The University of Bristol has informed NHS Digital that the data shared [11 words unchanged] sources other than NHS Digital and variables derived from NHS Digital data. The University of Bristol does not consider Until any of further determination is made as to whether NHS Digital are satisfied that the data shared with the University of Cambridge to be data under this Agreement. is no longer considered NHS Digital requires additional information to enable a robust assessment to determine whether the data shared qualifies as ‘Derived Data’ and can be considered not to be data under this Agreement. Until such determination is made, data, the University of Cambridge is listed in this Agreement as data controller [24 words unchanged] University of Cambridge would be removed from future versions of this Agreement. The objectives of the ERFC are consistent with the aims and purpose [84 words unchanged] was withdrawn, thereby maintaining final control as to how the Speedwell data re are used. [11 paragraphs unchanged]

Processing activities

Identifying data was shared with ONS the Office for National Statistics (ONS) to carry out the linkage between the study data and civil registration data. Participants' records were ‘flagged’ with the Office for National Statistics (ONS). ONS. ONS notified the study team at University of Bristol of participants’ deaths [9 words unchanged] The ‘flagging for long-term follow up’ service transferred from ONS to the HSCIC Health and Social Care Information Centre (HSCIC, now NHS Digital) in 2008, and this service was last supplied in March 2016. Data NHS Digital data is currently held in a secure relational database where it has its [43 words unchanged] writing a script within a statistics package that can then derive or categorised categorise variables e.g., number of cigarettes smoked none, 1-14, 15-24, 25+. Once the [9 words unchanged] and regression models are run to quantify any associations as effect estimates (95% confidence intervals, p-values) in a variety of multivariable models. In the ERFC this is done [29 words unchanged] not even seen as they are part of a much bigger average. [1 paragraph unchanged] No further linkages are undertaken to any other external or publicly available data. Data NHS Digital data is linked to the patient data collected as part of the research [66 words unchanged] signed consent forms which it needs to maintain to prove individual consent. [1 paragraph unchanged] The data stored at the University of Bristol is encrypted on University [30 words unchanged] to have permissions for access. Permissions can only be obtained by the PI principal investigator formally submitting a request to the IT department for a University of [14 words unchanged] University of Bristol laptop that has been set up to create a VPN virtual private network (VPN) which would require authentication with a University username and password. [2 paragraphs unchanged]

Expected output

[2 paragraphs unchanged] Planned outputs for current analyses are as follows: The following are either on-going or completed pieces of work 1. Determining the reductions in life expectancy according to different ages at diagnosis of diabetes mellitus. The proposed journal This work is currently being written up for publication is circulation to the ERFC co-investigators for comments with the aim of submitting the paper to the Lancet. 2. Development and validation of updated CVD risk prediction models for Europe (in collaboration with European Society of Cardiology (ESC) 2021 guideline committee). The paper will be submitted to the European Heart Journal. 2. Development and validation of updated cardiovascular disease risk prediction models for Europe (in collaboration with European Society of Cardiology (ESC) 2021 guideline committee). This work is now completed and has led to a new SCORE2 risk prediction algorithm to estimate 10-year risk of cardiovascular disease in Europe, published in the European Heart Journal in 2021. This new simple algorithm is better at identifying at-risk individuals, has accounted for secular changes compared to the previous SCORE index, accounts for competing risks, and can be used for four distinct European regions defined by varying cardiovascular disease risk levels. This improved method should lead to more targeted and cost-effective interventions to reduce cardiovascular risk and improve public health. 3. Estimating the association between renal function and CVD cardiovascular disease incidence in the general population. The This paper will be submitted to has been accepted for publication in the Lancet. journal Circulation but is not yet in the public domain. 4. Use of lifetime risk for cardiovascular disease risk assessment. This paper will be submitted to the journal Circulation. assessment is currently being revised and is still work in progress. 5. Risk prediction for composite cardiovascular outcomes. The paper will be sent outcomes has not progressed due to the European Heart Journal. other competing demands. 6. Sequential strategies for CVD cardiovascular disease screening, including health-economic evaluation. The output will This work is being re-evaluated as to whether it should or should not be submitted to the British Medical Journal. taken forward. Data in such outputs is in aggregate form such as counts in [14 words unchanged] than 7 are either suppressed or aggregated up to a bigger category. Dissemination is usually through standard academic routes e.g. journals, conferences however depending on the newsworthiness the main findings may be disseminated through newspapers radio or television features, social media, websites and non-governmental organizations (e.g. British Heart Foundation) who may wish to highlight the work especially if they have funded this. This ensures that the public, medical professionals and other academics are aware of the work. Dissemination is usually through standard academic routes e.g. journals, conferences however depending on the newsworthiness the main findings may be disseminated through newspapers radio or television features, social media, websites and NGOs (e.g. British Heart Foundation) who may wish to highlight the work especially if they have funded this. This ensures that the public, medical professionals and other academics are aware of the work. For example, the recent work on cardiovascular risk charts was presented at the European Society of Cardiology (ESC) Congress 2019 - 31 August - 04 September 2019, Paris, France (Presenters: Lisa Pennells and Stephen Kaptoge,Methodology of the revised WHO CVD risk charts).

Expected measurable benefits

[1 paragraph unchanged] It is hard to quantify the benefits but as these are common diseases which are now transitioning into low middle-income countries the potential benefits are large. It is hoped that this sort of information will be used by government to guide expenditure and rational planning of services but there are many other fiscal factors that may impact on this. This makes it hard to give a specific time frame by which benefits can be expected as this is clearly outside the control of the researchers. The findings of this research study are expected to contribute to evidence-based decision-making for policy-makers, local decision-makers such as doctors, and patients to inform best practice to improve the care, treatment and experience of health care users relevant to the subject matter of the study. The use of the data could: • help the system to better understand the health and care needs of populations. • advance understanding of regional and national trends in health and social care needs. • advance understanding of the need for, or effectiveness of, preventative health and care measures for particular populations or conditions such as cardiovascular disease, age-related cognitive decline and macular degeneration. • support knowledge creation or exploratory research (and the innovations and developments that might result from that exploratory work). [1 paragraph unchanged] (a) The results suggest a novel aetiological mechanism. As the data are [26 words unchanged] develop a potential intervention or using existing therapeutics for a novel indication. For example, other work from the ERFC published in The Lancet Diabetes & Endocrinology in 2021 on vitamin D suggests a causal relationship between vitamin D concentrations and mortality for individuals with low vitamin D status. A benefit of vitamin D was only seen in participants with low levels. These findings have implications for the design of vitamin D supplementation trials, and potential disease prevention strategies. In this way it may translate into better health care, but this may take usually has a long time window (e.g. 10+ years). (b) Descriptive study looking at time trends. The data from Speedwell reflect the experience on men born in the 20-40s. 1920s to 1940s. As such this is important in relation to data from more recent [50 words unchanged] time and whether socioeconomic differences have widened, narrowed or stayed the same. (c) Diagnostic or prognostic information. Results from risk models can be used to derive algorithms to stratify individuals in low, medium, or high risk of future events or mortality. These are usually on well-established [21 words unchanged] doctors now use the QRISK algorithm routinely in counselling and managing patients. Data from the ERFC have contributed to the development of SCORE2 which should have clinical and public health benefit by better targeting individuals at future risk of cardiovascular disease in the next 10 years and thus maximising health care expenditure. Future policy may modify existing risk prediction scores such as QRISK in the light of newer research findings. There are several routes by which outputs from the study could lead to public benefits. (i) By enhancing understanding around disease aetiology, it is possible new preventative strategies will emerge. These may be pharmacological through drug development or non-pharmacological through the promotion of public health campaigns that alter an individual’s life style. The successful implementation of such strategies should result in future decline in mortality and morbidity (ii) By providing better prognostic algorithms such as SCORE2, patients will benefit as clinicians can better identify those patients at highest risk of future disease and manage these more aggressively. Similarly patients themselves will find this information of value and this may help incentivise them to adhere to life style changes. Whilst the original study was designed in the late 1970s when it was not conventional to incorporate patient and public involvement (PPI) and the current cohort members in the study have mostly now died, the intention is to to have PPI members on a future data access committee. They would give a lay perspective as to whether external requests for data access should or should not be accepted. Where research findings are published in e.g. the British Medical Journal or the Lancet, the communications department for these journals will press release lay summaries for wider coverage by the media where the findings are relevant to a wider audience. In addition, there is often interaction with leading charities e.g. British Heart Foundation and occasional public bodies such as the National Institute of Health and Care Excellence (NICE).

Benefits reported

The Speedwell study has contributed to several studies looking at different biochemical [10 words unchanged] and mortality. In particular, inflammatory blood markers and measures of blood stickiness. It has also published To date, the benefits of this research have been predominantly to encourage further research on cataract and macular degeneration, areas that are far less researched. It has been involved in large meta-analyses as part of the ERFC collaboration. For example, a recent paper from the ERFC group has updated the World Health Organization risk algorithm for the prevention of cardiovascular disease worldwide. This used data from the Speedwell as well as many other studies across both high income and low middle-income countries to develop and validate a how these biochemical measures can be incorporated into risk prediction model which has utility for all countries scores and has tremendous public developing therapeutics. It is too soon to have seen any direct impact on patient benefit and/or the health potential benefits. care system. It has also published on cataract and macular degeneration, areas that are far less researched. These and other data can be used to better plan the provision of eye care as it provides an evidence base for calculating regional variations in clinical need. It has been involved in large meta-analyses as part of the ERFC collaboration. For example, the new SCORE2 paper has improved risk prediction across 4 different regions of Europe with varying background risk of cardiovascular disease. It is too soon to see any uptake yet of this work because organizations such as the National Institute of Health and Care Excellence (NICE) will need to assess whether the new SCORE2 system has or does not have advantages over the existing QRISK2 prediction tool and whether it should therefore replace it.

Text removed is struck through; text added is underlined. Unchanged paragraphs are summarised rather than repeated.

Objective for processing

The following provides background information on the purpose of the original study and how the data are managed: The Speedwell study was set up by the University of Bristol as a prospective cohort study in the late 1970s to look at the determinants of cardiovascular disease, though over time other phenotypes of interest have been added. The study was set up to examine risk factors for cardiovascular disease such as elevated cholesterol and obesity amongst around 2500 middle aged men living in Bristol. It recruited men aged 45-59 years of age from 16 General Practitioners based at two health centres in Bristol. Men were invited to research clinics where they completed questionnaires, had clinical measures such as blood pressure and blood samples taken. Over time men were seen over an additional 4 clinics (known as phase 2, 3, 4, and 5) so that new data on non-fatal cardiovascular events could be collected as well as repeating the collection of data on risk factors such as blood pressure, body mass index etc. At phase 5, as the men were older, cognitive function and retinal photographs were added to look at age-related cognitive decline and eye diseases such as macular degeneration. The storage and processing of these data are in line with the GDPR principles for the following reasons. The primary aim of the data collection and follow-up of these participants is to understand the underlying risk factors behind a number of common chronic diseases and their potential consequences on health and mortality. The processing of these data are in the public interest (GDPR Article 6(1)(e)) as they contribute to a greater understanding of health related risks and how these may be tackled in terms of public health prevention and as such serve as public interest. Such research can identify new risk factors, be they environmental or genetic that may lead to new causes of disease with the potential for developing new interventions to prevent disease. Storage and archiving of these data is of value as external researchers may contact the principal investigator for the possibility of new analyses and this fulfills GDPR Article 9(2)(j) which states “is necessary for archiving purposes in the public interest, scientific or historical research purposes or statistical purposes…”. The researchers carrying out this study believe that the potential benefit of using these data far outweigh any potential harm. Steps are taken to reduce and mitigate any risk of individual identification and hence there is minimal risk of potential harm to the public by dissemination. Sadly most of the participants will have died by now as the study was started in 1979. The aim of collecting the data can only be achieved if it is actively used for medical research. With further follow-up and given the age of the cohort, there are a large number of further clinical events (cause specific deaths and cancer registrations) which can only be obtained from NHS Digital MRIS reports, as the study are no longer contacting the men in this cohort due to their age, frailty and the fact most will have died. The additional data provided by NHS Digital is very valuable as it will increase the statistical power to look at outcomes such as heart disease and rarer outcomes. At this stage the University of Bristol and the University of Cambridge are asking for permission to continue processing the data they already hold from 1982 to March 2016. The University of Cambridge co-ordinates the Emerging Risk Factors Collaboration (https://www.phpc.cam.ac.uk/ceu/erfc/) funded by the British Heart Foundation and Medical Research Council. This group host a consortium of >130 prospective studies from >30 countries that has collated and harmonised individual-participant data (IPD) from a total of ~2.5 million participants to study risk factors for cardiovascular diseases and cause-specific mortality in greater detail by IPD meta-analysis. They have previously published work in all the major medical journals and have contributed to new guidance with the World Health Organisation on how to operate the risk of diabetes mellitus. They are currently working on 6 distinct research projects as part of their programme. Current analyses being undertaken by the ERFC collaboration and how they use the data provided by NHS Digital are described below: 1. Determining the reductions in life expectancy according to different ages at diagnosis of diabetes mellitus. This research will use NHS Digital data on date and cause of death to calculate the negative impact on life expectancy for diabetics versus non-diabetics. 2. Development and validation of updated CVD risk prediction models for Europe (in collaboration with European Society of Cardiology (ESC) 2021 guideline committee). This looks at an existing risk prediction model and examines its ability to predict both fatal and non-fatal CVD events and thus necessitates the use of NHS Digital data.. 3. Estimating the association between renal function and CVD incidence in the general population. Baseline renal function from blood tests will be used to predict future fatal and non-fatal CVD events and thus requires NHS Digital data. 4. Use of lifetime risk for cardiovascular disease risk assessment. Current risk calculators such as QRISK look at 10 year CVD risk. This research will expand the time window to look at life time risk so each member of the cohort is followed up until death or end of the follow-up period. Data on vital status and age at death comes from NHS Digital. 5. Risk prediction for composite cardiovascular outcomes. This will expand the usual outcomes of heart attack and stroke to other fatal or non-fatal events such as heart failure and pulmonary embolus and thus requires NHS Digital data. 6. Sequential strategies for CVD screening, including health-economic evaluation. This work compares two different screening strategies, population-based versus high risk and compare the cost-effectiveness in relation to the prevention of future events (fatal or non-fatal) and requires NHS Digital data. The objectives of the ERFC are consistent with the aims and purpose behind the establishment and maintenance of the Speedwell study i.e. to look at risk factors for a wide range of chronic diseases especially cardiometabolic disease. ERFC works collaboratively with the Speedwell team who are involved in the drafting, interpretation and approval of any academic outputs through co-authorship of publications. If the Speedwell team have any concerns as to how the Speedwell data has been used they would request for this to be amended or in the worst case that the data form the cohort was withdrawn, thereby maintaining final control as to how the Speedwell data re used. Researchers have access to individual level data from the baseline and subsequent assessments with outcome data from either participant self-report or hospital data and mortality from NHS Digital. NHS Digital data covers all follow-up events to enhance statistical power and the geographical spread is limited to the cohort, hence the Bristol area (though some participants will have moved away over the years). Individual level data are required to enable sophisticated statistical models that can harmonise data coding and adjust for confounding factors through some sort of regression technique as appropriate. Cause of death data are essential as it is standard epidemiological practice to look at risk factors with specific causes that are thought to be caused by those risk factors. In some cases, researchers may wish to aggregate up specific causes e.g. all types of stroke but often need the more detailed sub-groups e.g. ischaemic versus haemorrhagic stroke as, for example high cholesterol level is associated with the former but not the latter. To maximise the scientific and public health value of the data, researchers need to have the flexibility to examine future research hypotheses and this can be best achieved by having cause of death data which maintains maximum flexibility. The data held and shared with researchers is the minimum required for the respective analysis and in addition, several additional steps are taken to mitigate the possibility of re-identification (NHS Digital data dictionaries have been provided to enable researchers to assess for themselves any potential risk). The University of Bristol and the University of Cambridge are joint data controllers and both organisations process the data for this study. No other organisations process the data for the purposes described. The University of Bristol has shared data with a research group at the University of Oxford which is examining how the body responds to allergens in terms of an immune response may contribute to the risk of cardiovascular disease. There are few cohort studies that have measures of immunoglobulins and long term follow-up such as the Speedwell study. Data shared by the University of Bristol for this purpose has been assessed by NHS Digital as derived data. NHS Digital has determined that data has been shared with the University of Oxford in a controlled way, bound by Data Sharing Agreements that confer the same level of adherence to protecting the rights of individuals involved in research and specifically covered by informed consent of subjects. All such data shared is either aggregate (as in publication), anonymous, or pseudonymised containing no NHS Digital data (such that additional data provided by the collaboration can be added to the sum of knowledge about subjects within the University of Bristol, but not allowing others to link data and risk re-identification – this includes measures to suppress rare events). Any data shared with the University of Oxford must be subject to the conditions that the University of Oxford: i. must not combine it with other datasets which could potentially increase the risk of reidentification for individuals in the dataset; ii. must not attempt to re-identify individuals in the dataset; iii. must not onwardly share the dataset; iv. must use the dataset for a defined purpose in support of the LLP’s aims defined within this Agreement, and v. must not publish the data. Under the terms of this Agreement, the University of Bristol and the University of Cambridge are responsible for ensuring compliance with the above conditions and for confirming destruction of the data by the University of Oxford once the data is no longer required for the purpose for which it was shared. The following provides background information on the purpose of the original study and how the data are managed: The Speedwell study was set up by the University of Bristol as a prospective cohort study in the late 1970s to look at the determinants of cardiovascular disease, though over time other phenotypes of interest have been added. The study was set up to examine risk factors for cardiovascular disease such as elevated cholesterol and obesity amongst around 2500 middle aged men living in Bristol. It recruited men aged 45-59 years of age from 16 General Practitioners based at two health centres in Bristol. Men were invited to research clinics where they completed questionnaires, had clinical measures such as blood pressure and blood samples taken. Over time men were seen over an additional 4 clinics (known as phase 2, 3, 4, and 5) so that new data on non-fatal cardiovascular events could be collected as well as repeating the collection of data on risk factors such as blood pressure, body mass index etc. At phase 5, as the men were older, cognitive function and retinal photographs were added to look at age-related cognitive decline and eye diseases such as macular degeneration. The storage and processing of these data are in line with the GDPR principles for the following reasons. The primary aim of the data collection and follow-up of these participants is to understand the underlying risk factors behind several common chronic diseases and their potential consequences on health and mortality. The processing of this data is in the public interest (GDPR Article 6(1)(e)) as they contribute to a greater understanding of health-related risks and how these may be tackled in terms of public health prevention and as such serve as public interest. Such research can identify new risk factors, be they environmental or genetic that may lead to new causes of disease with the potential for developing new interventions to prevent disease. Storage and archiving of these data are of value as external researchers may contact the principal investigator for the possibility of new analyses and this fulfils GDPR Article 9(2)(j) which states “is necessary for archiving purposes in the public interest, scientific or historical research purposes or statistical purposes…”. The researchers carrying out this study believe that the potential benefit of using these data far outweigh any potential harm. Steps are taken to reduce and mitigate any risk of individual identification and hence there is minimal risk of potential harm to the public by dissemination. Sadly, most of the participants will have died by now as the study was started in 1979. The aim of collecting the data can only be achieved if it is actively used for medical research. With further follow-up and given the age of the cohort, there are many further clinical events (cause specific deaths and cancer registrations) which can only be obtained from NHS Digital Medical Research Information Service (MRIS) reports, as the study are no longer contacting the men in this cohort due to their age, frailty and the fact most will have died. The additional data provided by NHS Digital is very valuable as it will increase the statistical power to look at outcomes such as heart disease and rarer outcomes. This is a request to permit The University of Bristo to continue processing identifiable data they already hold from 1982 to March 2016. DATA SHARING WITH THE UNIVERSITY OF CAMBRIDGE: Since 2002 The University of Bristol have shared a derived sub-set of data with The University of Cambridge, for the purpose of supporting the Emerging Risk Factors Collaboration (ERFC) (https://www.phpc.cam.ac.uk/ceu/erfc/), a programme of work funded by the British Heart Foundation and Medical Research Council. This group host a consortium of >130 prospective studies from >30 countries that has collated and harmonised individual-participant data (IPD) from a total of ~2.5 million participants to study risk factors for cardiovascular diseases and cause-specific mortality in greater detail by IPD meta-analysis. They have previously published work in all the major medical journals and have contributed to new guidance with the World Health Organisation on how to operate the risk of diabetes mellitus. They are currently working on 6 distinct research projects as part of their programme. Current analyses being undertaken by the ERFC collaboration and how they use the data provided by NHS Digital are described below: 1. Determining the reductions in life expectancy according to different ages at diagnosis of diabetes mellitus. This research will use NHS Digital data on date and cause of death to calculate the negative impact on life expectancy for diabetics versus non-diabetics. 2. Development and validation of updated CVD risk prediction models for Europe (in collaboration with European Society of Cardiology (ESC) 2021 guideline committee). This looks at an existing risk prediction model and examines its ability to predict both fatal and non-fatal CVD events and thus necessitates the use of NHS Digital data. 3. Estimating the association between renal function and CVD incidence in the general population. Baseline renal function from blood tests will be used to predict future fatal and non-fatal CVD events and thus requires NHS Digital data. 4. Use of lifetime risk for cardiovascular disease risk assessment. Current risk calculators such as QRISK look at 10-year CVD risk. This research will expand the time window to look at lifetime risk, so each member of the cohort is followed up until death or end of the follow-up period. Data on vital status and age at death comes from NHS Digital. 5. Risk prediction for composite cardiovascular outcomes. This will expand the usual outcomes of heart attack and stroke to other fatal or non-fatal events such as heart failure and pulmonary embolus and thus requires NHS Digital data. 6. Sequential strategies for CVD screening, including health-economic evaluation. This work compares two different screening strategies, population-based versus high risk and compare the cost-effectiveness in relation to the prevention of future events (fatal or non-fatal) and requires NHS Digital data. The University of Bristol has informed NHS Digital that the data shared with the University of Cambridge was comprised of variables obtained from sources other than NHS Digital and variables derived from NHS Digital data. The University of Bristol does not consider any of the data shared with the University of Cambridge to be data under this Agreement. NHS Digital requires additional information to enable a robust assessment to determine whether the data shared qualifies as ‘Derived Data’ and can be considered not to be data under this Agreement. Until such determination is made, the University of Cambridge is listed in this Agreement as data controller for the data it has received from the University of Bristol. Should the data be determined to be ‘Derived Data’ in the future, the University of Cambridge would be removed from future versions of this Agreement. The objectives of the ERFC are consistent with the aims and purpose behind the establishment and maintenance of the Speedwell study i.e., to look at risk factors for a wide range of chronic diseases especially cardiometabolic disease. ERFC works collaboratively with the Speedwell team who are involved in the drafting, interpretation and approval of any academic outputs through co-authorship of publications. If the Speedwell team have any concerns as to how the Speedwell data has been used, they would request for this to be amended or in the worst case that the data form the cohort was withdrawn, thereby maintaining final control as to how the Speedwell data re used. Researchers have access to individual level data from the baseline and subsequent assessments with outcome data from either participant self-report or hospital data and mortality from NHS Digital. NHS Digital data covers all follow-up events to enhance statistical power and the geographical spread is limited to the cohort, hence the Bristol area (though some participants will have moved away over the years). Individual level data are required to enable sophisticated statistical models that can harmonise data coding and adjust for confounding factors through some sort of regression technique as appropriate. Cause of death data are essential as it is standard epidemiological practice to look at risk factors with specific causes that are thought to be caused by those risk factors. In some cases, researchers may wish to aggregate up specific causes e.g., all types of stroke but often need the more detailed sub-groups e.g., ischaemic versus haemorrhagic stroke as, for example high cholesterol level is associated with the former but not the latter. To maximise the scientific and public health value of the data, researchers need to have the flexibility to examine future research hypotheses, and this can be best achieved by having cause of death data which maintains maximum flexibility. The data held and shared with researchers is the minimum required for the respective analysis and in addition, several additional steps are taken to mitigate the possibility of re-identification (NHS Digital data dictionaries have been provided to enable researchers to assess for themselves any potential risk). The University of Bristol and the University of Cambridge are the data controllers and both organisations process the data for this study. No other organisations process the data for the purposes described. The University of Bristol has shared data with a research group at the University of Oxford which is examining how the body responds to allergens in terms of an immune response may contribute to the risk of cardiovascular disease. There are few cohort studies that have measures of immunoglobulins and long-term follow-up such as the Speedwell study. Data shared by the University of Bristol for this purpose has been assessed by NHS Digital as derived data. NHS Digital has determined that data has been shared with the University of Oxford in a controlled way, bound by Data Sharing Agreements that confer the same level of adherence to protecting the rights of individuals involved in research and specifically covered by informed consent of subjects. All such data shared is either aggregate (as in publication), anonymous, or pseudonymised containing no NHS Digital data (such that additional data provided by the collaboration can be added to the sum of knowledge about subjects within the University of Bristol, but not allowing others to link data and risk re-identification – this includes measures to suppress rare events). Any data shared with the University of Oxford must be subject to the conditions that the University of Oxford: i. must not combine it with other datasets which could potentially increase the risk of reidentification for individuals in the dataset; ii. must not attempt to re-identify individuals in the dataset; iii. must not onwardly share the dataset; iv. must use the dataset for a defined purpose in support of the LLP’s aims defined within this Agreement, and v. must not publish the data. Under the terms of this Agreement, the University of Bristol is responsible for ensuring compliance with the above conditions and for confirming destruction of the data by the University of Oxford once the data is no longer required for the purpose for which it was shared.

Processing activities

Identifying data was shared with ONS to carry out the linkage between the study data and civil registration data. Participants' records were ‘flagged’ with the Office for National Statistics (ONS). ONS notified the study team at University of Bristol of participants’ deaths (date and cause) and cancer events when they occurred. The ‘flagging for long-term follow up’ service transferred from ONS to the HSCIC in 2008. Data was last supplied in March 2016. Data is currently held in a secure relational database where it has its own table. This can then be merged as required for specific data queries so it can be linked to the explanatory variables that have been collected as part of the research with the participants’ consent and knowledge. Further processing is usually done by writing a script within a statistics package that can then derive or categorised variables e.g. number of cigarettes smoked none, 1-14, 15-24, 25+. Once the data has been cleaned and derived the main tabulations and regression models are run to quantify any associations as effect estimates (95% confidence intervals, p-values) in a variety of multivariable models. In the ERFC this is done independently for every dataset they hold and the results are pooled (using meta-analysis) to get the most precise estimate so that individual results from a single study are often not even seen as they are part of a much bigger average. The University of Bristol shared data with the University of Cambridge in 2002 with updates in subsequent years. Data was originally transferred by secure email. Later extracts were encrypted and password protected and the data link was sent using secure data transfer software (FLUFF). The password was transmitted to the data manager by text to their mobile phone. These were pseudonymised individual level data (initially under the ONS accredited researcher scheme). On the advice of the Office for National Statistics, the date of any events had either (a) random noise added or (b) the age of the participant at the date of the event e.g. 76.4 years. Furthermore, any rare events which had fewer than 7 occurrences within the dataset were suppressed by aggregating up so the exact cause was not identifiable but a higher level category was still available. Any published outputs are also checked to suppress cell sizes less than 7. In most cases the results are shown as the average effect across many cohorts so data from the Speedwell study is not even identifiable. No further linkages are undertaken to any other external or publicly available data. Data is linked to the patient data collected as part of the research clinics and questionnaire with the event data. There have been no further data flows in either direction. The linkage is done through the study unique identifier so is pseudonymised. The University of Bristol currently hold within the database a link file which has identifiable data e.g. name of participant. This is not available to the University of Cambridge researchers. The University of Bristol also holds the signed consent forms which it needs to maintain to prove individual consent. All analyses are undertaken by research academics who are substantive employees based at the University of Bristol or the University of Cambridge. As such they are fully aware of the need to maintain confidentiality and not attempt to re-identify participants. It is normal practice for the Universities to ensure staff are fully aware of the GDPR principles as part of their mandatory training. The data stored at the University of Bristol is encrypted on University protected servers with appropriate access requirements (e.g. password protection) supported by their respective IT departments. The server has specific study folders that are in a managed group. This requires staff to have permissions for access. Permissions can only be obtained by the PI formally submitting a request to the IT department for a University of Bristol staff member to have access. Remote access is possible if it is a University of Bristol laptop that has been set up to create a VPN which would require authentication with a University username and password. At the University of Cambridge, the data is stored on University of Cambridge servers in a password protected, restricted access environment and is accessible by only the "Data Manager". When the "Data Manager" is requested to provide a dataset for analysis by a University of Cambridge researcher, the data manager makes that data available to the relevant researcher on a case by case basis. The subsequent analysis dataset is stored on the same University servers in a password protected, restricted access directory which is accessible by only the researcher and the "Data Manager". From this point on the data and performed analysis remain on the same University servers. No identifying data for study participants, such as name, address etc are held in either University of Oxford or University of Cambridge. Identifying data was shared with ONS to carry out the linkage between the study data and civil registration data. Participants' records were ‘flagged’ with the Office for National Statistics (ONS). ONS notified the study team at University of Bristol of participants’ deaths (date and cause) and cancer events when they occurred. The ‘flagging for long-term follow up’ service transferred from ONS to the HSCIC in 2008, and this service was last supplied in March 2016. Data is currently held in a secure relational database where it has its own table. This can then be merged as required for specific data queries so it can be linked to the explanatory variables that have been collected as part of the research with the participants’ consent and knowledge. Further processing is usually done by writing a script within a statistics package that can then derive or categorised variables e.g., number of cigarettes smoked none, 1-14, 15-24, 25+. Once the data has been cleaned and derived the main tabulations and regression models are run to quantify any associations as effect estimates (95% confidence intervals, p-values) in a variety of multivariable models. In the ERFC this is done independently for every dataset they hold, and the results are pooled (using meta-analysis) to get the most precise estimate so that individual results from a single study are often not even seen as they are part of a much bigger average. The University of Bristol shared data with the University of Cambridge in 2002 with updates in subsequent years. Data was originally transferred by secure email. Later extracts were encrypted, and password protected, and the data link was sent using secure data transfer software (FLUFF). The password was transmitted to the data manager by text to their mobile phone. These were pseudonymised individual level data, that the applicant deems to be derived (initially under the ONS accredited researcher scheme). On the advice of the Office for National Statistics, the date of any events had either (a) random noise added or (b) the age of the participant at the date of the event e.g. 76.4 years. Furthermore, any rare events which had fewer than 7 occurrences within the dataset were suppressed by aggregating up so the exact cause was not identifiable but a higher level category was still available. Any published outputs are also checked to suppress cell sizes less than 7. In most cases the results are shown as the average effect across many cohorts so data from the Speedwell study is not even identifiable. No further linkages are undertaken to any other external or publicly available data. Data is linked to the patient data collected as part of the research clinics and questionnaire with the event data. There have been no further data flows in either direction. The linkage is done through the study unique identifier so is pseudonymised. The University of Bristol currently hold within the database a link file which has identifiable data e.g., name of participant. This is not available to the University of Cambridge researchers. The University of Bristol also holds the signed consent forms which it needs to maintain to prove individual consent. All analyses are undertaken by research academics who are substantive employees based at the University of Bristol or the University of Cambridge. As such they are fully aware of the need to maintain confidentiality and not attempt to re-identify participants where pseudonymised data is being used. It is normal practice for the Universities to ensure staff are fully aware of the GDPR principles as part of their mandatory training. The data stored at the University of Bristol is encrypted on University protected servers with appropriate access requirements (e.g. password protection) supported by their respective IT departments. The server has specific study folders that are in a managed group. This requires staff to have permissions for access. Permissions can only be obtained by the PI formally submitting a request to the IT department for a University of Bristol staff member to have access. Remote access is possible if it is a University of Bristol laptop that has been set up to create a VPN which would require authentication with a University username and password. At the University of Cambridge, the data is stored on University of Cambridge servers in a password protected, restricted access environment and is accessible by only the "Data Manager". When the "Data Manager" is requested to provide a dataset for analysis by a University of Cambridge researcher, the data manager makes that data available to the relevant researcher on a case-by-case basis. The subsequent analysis dataset is stored on the same University servers in a password protected, restricted access directory which is accessible by only the researcher and the "Data Manager". From this point on the data and performed analysis remain on the same University servers. No identifying data for study participants, such as name, address etc are held in either University of Oxford or University of Cambridge.

Expected output

Data held by the University of Bristol and the University of Cambridge will result in academic publications that will add to scientific knowledge and understanding about the causes of diseases. The ERFC has a very well established track record of publication in high impact journals and collaborates with major organizations such as the World Health Organization or charities such as the British Heart Foundation. Results may also be presented at conferences or other meetings. Planned outputs for current analyses are as follows: 1. Determining the reductions in life expectancy according to different ages at diagnosis of diabetes mellitus. The proposed journal for publication is the Lancet. 2. Development and validation of updated CVD risk prediction models for Europe (in collaboration with European Society of Cardiology (ESC) 2021 guideline committee). The paper will be submitted to the European Heart Journal. 3. Estimating the association between renal function and CVD incidence in the general population. The paper will be submitted to the Lancet. 4. Use of lifetime risk for cardiovascular disease risk assessment. This paper will be submitted to the journal Circulation. 5. Risk prediction for composite cardiovascular outcomes. The paper will be sent to the European Heart Journal. 6. Sequential strategies for CVD screening, including health-economic evaluation. The output will be submitted to the British Medical Journal. Data in such outputs is in aggregate form such as counts in a table or measures of association e.g. correlation or regression coefficients. Cell counts less than 7 are either suppressed or aggregated up to a bigger category. Dissemination is usually through standard academic routes e.g. journals, conferences however depending on the newsworthiness the main findings may be disseminated through newspapers radio or television features, social media, websites and NGOs (e.g. British Heart Foundation) who may wish to highlight the work especially if they have funded this. This ensures that the general public, medical professionals and other academics are aware of the work. For example, the recent work on cardiovascular risk charts was presented at the European Society of Cardiology (ESC) Congress 2019 - 31 August - 04 September 2019, Paris, France (Presenters: Lisa Pennells and Stephen Kaptoge,Methodology of the revised WHO CVD risk charts). Data held by the University of Bristol and the University of Cambridge will result in academic publications that will add to scientific knowledge and understanding about the causes of diseases. The ERFC has a very well-established track record of publication in high impact journals and collaborates with major organizations such as the World Health Organization or charities such as the British Heart Foundation. Results may also be presented at conferences or other meetings. Planned outputs for current analyses are as follows: 1. Determining the reductions in life expectancy according to different ages at diagnosis of diabetes mellitus. The proposed journal for publication is the Lancet. 2. Development and validation of updated CVD risk prediction models for Europe (in collaboration with European Society of Cardiology (ESC) 2021 guideline committee). The paper will be submitted to the European Heart Journal. 3. Estimating the association between renal function and CVD incidence in the general population. The paper will be submitted to the Lancet. 4. Use of lifetime risk for cardiovascular disease risk assessment. This paper will be submitted to the journal Circulation. 5. Risk prediction for composite cardiovascular outcomes. The paper will be sent to the European Heart Journal. 6. Sequential strategies for CVD screening, including health-economic evaluation. The output will be submitted to the British Medical Journal. Data in such outputs is in aggregate form such as counts in a table or measures of association e.g., correlation or regression coefficients. Cell counts less than 7 are either suppressed or aggregated up to a bigger category. Dissemination is usually through standard academic routes e.g. journals, conferences however depending on the newsworthiness the main findings may be disseminated through newspapers radio or television features, social media, websites and NGOs (e.g. British Heart Foundation) who may wish to highlight the work especially if they have funded this. This ensures that the public, medical professionals and other academics are aware of the work. For example, the recent work on cardiovascular risk charts was presented at the European Society of Cardiology (ESC) Congress 2019 - 31 August - 04 September 2019, Paris, France (Presenters: Lisa Pennells and Stephen Kaptoge,Methodology of the revised WHO CVD risk charts).

Expected measurable benefits

The study was set up as a medical research project that may have benefits for clinical and public health around the determinants of common chronic diseases associated with ageing. The expected measurable benefits are variable depending on the nature of the actual research project and the results. It is hard to quantify the benefits but as these are common diseases which are now transitioning into low middle income countries the potential benefits are large. It is hoped that this sort of information will be used by government to guide expenditure and rational planning of services but there are many other fiscal factors that may impact on this. This makes it hard to give a specific time frame by which benefits can be expected as this is clearly outside the control of the researchers. The following describes different scenarios that could have benefits: (a) The results suggest a novel aetiological mechanism. As the data are observational they need to be reproducible and triangulated with other evidence e.g. animal studies. These sort of results may then lead onto further experimental work to develop a potential intervention or using existing therapeutics for a novel indication. In this way it may translate into better health care but this may take a long time (b) Descriptive study looking at time trends. The data from Speedwell reflect the experience on men born in the 20-40s. As such this is important in relation to data from more recent cohorts so, for example, researchers can synthesise different studies to track what is happening with obesity or hypertension as well as socioeconomic differences. These descriptive results are valuable to health care providers, commissioners and government to see whether the health of the UK population has or has not improved over time and whether socioeconomic differences have widened, narrowed or stayed the same. (c) Diagnostic or prognostic information. Results from risk models can be used to derive algorithms to stratify individuals in low, medium, high risk of future events or mortality. These are usually are on well-established risk factors. Such models are extremely valuable to help identify who may have the most benefit from an intervention. Primary care doctors now use the QRISK algorithm routinely in counselling and managing patients. The study was set up as a medical research project that may have benefits for clinical and public health around the determinants of common chronic diseases associated with ageing. The expected measurable benefits are variable depending on the nature of the actual research project and the results. It is hard to quantify the benefits but as these are common diseases which are now transitioning into low middle-income countries the potential benefits are large. It is hoped that this sort of information will be used by government to guide expenditure and rational planning of services but there are many other fiscal factors that may impact on this. This makes it hard to give a specific time frame by which benefits can be expected as this is clearly outside the control of the researchers. The following describes different scenarios that could have benefits: (a) The results suggest a novel aetiological mechanism. As the data are observational, they need to be reproducible and triangulated with other evidence e.g. animal studies. These sorts of results may then lead onto further experimental work to develop a potential intervention or using existing therapeutics for a novel indication. In this way it may translate into better health care, but this may take a long time (b) Descriptive study looking at time trends. The data from Speedwell reflect the experience on men born in the 20-40s. As such this is important in relation to data from more recent cohorts so, for example, researchers can synthesise different studies to track what is happening with obesity or hypertension as well as socioeconomic differences. These descriptive results are valuable to health care providers, commissioners and government to see whether the health of the UK population has or has not improved over time and whether socioeconomic differences have widened, narrowed or stayed the same. (c) Diagnostic or prognostic information. Results from risk models can be used to derive algorithms to stratify individuals in low, medium, high risk of future events or mortality. These are usually on well-established risk factors. Such models are extremely valuable to help identify who may have the most benefit from an intervention. Primary care doctors now use the QRISK algorithm routinely in counselling and managing patients.

Benefits reported

The Speedwell study has contributed to several studies looking at different biochemical [80 words unchanged] as well as many other studies across both high income and low middle income middle-income countries to develop and validate a risk prediction model which has utility for all countries and has tremendous public health potential benefits.