Using national electronic databases to validate cardiovascular outcomes in PATCH – a pilot study to assess the use of electronic databases for clinical trial follow up.

The Medical Research Council Trials Unit (MRC CTU) at University College London (UCL) requires pseudonymised Hospital Episode Statistics (HES) data for use in a sub-study of the PATCH (Prostate Adenocarcinoma: TransCutaneous Hormones, PR09, CRUK/06/001; ISRCTN70406718, ClinicalTrials.gov NCT00303784) trial. The sub-study is called “Using national electronic databases to validate cardiovascular outcomes in PATCH – a pilot study to assess the use of electronic databases for clinical trial follow up.”. UCL are the sole Data Controller who also process data for this project.

In a sub-study of the PATCH trial, MRC CTU propose to compare the coverage and accuracy of cardiovascular outcome reports derived from:

- electronically collected from the National Institute for Cardiovascular Outcomes Research (NICOR)

- electronically collected from HES datasets from NHS Digital.

- routine outcome ascertainment within the PATCH trial (manually collected data via the PATCH Case Report Forms (CRFs).

This sub-study will be for patients currently recruited into the PATCH trial (approx. 1600 patients). This will inform whether future event ascertainment within the trial could be supplemented with registry data (NICOR or HES and other nationally collected datasets) to improve data quality/completeness, or may even reduce or negate the need for trial follow-up (and therefore reduce research costs) because if the study can show that there is a good concordance between HES data and trial data then the team could in the future use registry data to supplement the trial data. This is the objective of this methodological study. Additionally the data source (i.e. data collection within PATCH trial CRFs vs routine data collection (HES/NICOR) may also affect the quality of data. This is perhaps more important, since trials are moving towards electronic data capture, but there may still be a benefit to supplementing it with registry data.

The justification of processing this data under the principles of GDPR, is Article 6(1)(e): ‘Processing is necessary for the performance of a task carried out in the public interest or in the exercise of official authority vested in the controller’ as this is a task within the public interest and as the research involves health data, Article 9(2)(j) is also applicable, as this details that processing is necessary for scientific and research purposes, subject to appropriate safeguards.

PATCH is an ongoing clinical trial assessing the safety and efficacy of replacing androgen suppression using Luteinising Hormone Releasing Hormone (LHRH) analogues with transdermal oestrogen patches, in men with locally advanced or metastatic prostate cancer. The study is funded by Cancer Research UK and currently recruiting at multiple centres across the UK. The cardiovascular safety of systemic oestrogen administration is one of the key toxicity concerns around its usage in PATCH. Although transdermal oestrogens bypass the liver first pass effect seen with oral administration and therefore should decrease cardiovascular risk, surveillance of these events within the study is an ongoing endeavour. This is challenging and time consuming, as it requires frequent exchanges of Serious Adverse Event (SAE) forms and documentation between centres and the MRC CTU. Furthermore the complexity of the event data is subtly different from the more usual cancer specific endpoints that form the main endpoints of the trial.

In terms of cardiovascular outcomes in PATCH, a previous 2013 report (Langley et al, Lancet Oncology 2013) reported data from the first 254 patients randomised (169 oestrogen patches and 85 LHRHa), using events reported via the investigators (CRFs). These were independently reviewed by cardiologists and classified as one of: heart failure, acute coronary syndrome, thromboembolic stroke, arterial thrombosis (other) or venous thromboembolism. Importantly, absolute numbers of events were small).

For large scale clinical trials, clinical outcome data, often collected over many years, is central to the satisfactory analysis of both primary and secondary endpoints but also subsequent hypotheses and interpretation. Conventional data collection via CRFs, whilst considered a gold standard, is time consuming, costly and, despite the efforts of NHS research departments throughout the UK, subject to patients moving away and being lost to follow up. In addition, clinical trials frequently collect information for a limited follow-up period, for example five years. There may well be events of interest which occur many years after the end of the trial which are currently not collected.

With the development of national clinical databases and audits, there is the potential for electronic data capture of specific outcomes via routinely collected information, improving the robustness and efficiency of trial follow up. However, information around the totality of event capture within the different available datasets and how this compares with conventionally collected trial data via CRFs is lacking. Initial comparisons of such databases show differences in event detection arising for a number of reasons including difficulties in disease coding (Herrett et al, BMJ 2013).

The cohort of patients will be those enrolled within the PATCH trial. This will be in line with the data availability from HES and NICOR. Only events occurring during time on study will be retrieved and compared. The PATCH trial database will be used as the template for this, and only events occurring post randomisation will be analysed. NICOR and HES event data will be cleaned in this way by the MRC CTU.

UCL are requesting that HES Admitted Patient Care and A&E datasets are linked to the cohort and the requested fields extracted. The HES data, including the associated unique study number, will be transferred in pseudonymised form to UCL.

UCL are providing a cohort. The patient identifiers being sent to NHS Digital are Study ID, NHS Number, Date Of Birth, Postcode and Gender (although all cohort are male). The identifying patient data (cohort) provided to NHS Digital to enable linking to HES data is held within a separate UCL data safe haven and will not be combined with the disseminated data.

UCL are not permitted to re-identify individuals under this agreement.

The subset of the outcomes collected in PATCH will also be collected by the National Institute for Cardiovascular Outcomes Research (NICOR), based at Barts Health NHS Trust which houses national databases on cardiac surgery (SCS), percutaneous revascularisation (BCIS), myocardial infarction (MINAP), heart failure, arrhythmia and congenital heart disease.

The research design and objectives are as follows:

Substudy Design:

1) request NICOR data on patients enrolled into PATCH study from MINAP and Heart failure audits (under patient consent model).

2) request HES Inpatient and A/E data (cardiovascular outcomes to include stroke and thromboembolic disease) on the same cohorts from NHS Digital.

3) Perform 3-way comparison of coverage and accuracy of outcomes as reported in NICOR, HES and manually collected data (via the PATCH CRFs).

4) The data may also be used for methodology research into the quality, accuracy and completeness of the datasets, including the reasons or causes of any discrepancies.

The aim of the study is to compare cardiovascular events between PATCH study data, NICOR and HES data and see if they are comparable.

NICOR data will be analysed within UCL data safe haven and has no identifiable data. NICOR will send to the MRC CTU all cardiovascular events that has occurred in the cohort of PATCH trial participants supplied to them. Each cardiovascular event has the participants study number attached to it. Although NICOR data could be re-identifiable, the team are not permitted to re-identify individuals in this study.

The NICOR and HES data will not be linked but using the study numbers of the PATCH trial participants researchers will compare cardiovascular events that occurred while on the trial to see if there are differences between the three different data sets.

Sponsorship for the PATCH trail is currently being transferred from Imperial College London to UCL. This sub-study will be solely managed, financed and run within UCL (MRC CTU).

The PATCH research team based in MRC CTU at UCL will provide to NHS Digital, NHS number, Postcode, gender (all males) and DOB for each individual to allow accurate and reliable identification of patients within the cohort. These will be associated with a unique study number. Patients have consented for data to be shared with researchers in an annoymised or linked annonymised form (this is where the 'Linked anonymised data' are anonymous to the people who receive and hold it (e.g. a research team) but contain information or codes that would allow the suppliers of the data to identify people from it). They have also consented that personal details can be used to obtain long term follow up information from national registries.

The pseudonymised HES datasets, along with the unique study number, will be transferred to the UCL PATCH team by NHS digital, this data will reside in the UCL data safe haven and will be identified by study number only – thus there will be no identifying personal data attached to a study number. Only defined members of the PATCH team and MRC CTU’s methodology team will have access to the data safe haven for analysis of the data. These will be substantive employees of UCL. All UCL substantive employees must have completed training in data protection and confidentiality and also have had appropriate training on the UCL data safe haven.

Outcomes of interest will be defined pragmatically, for example using both generic and more specific codes within the HES data hierarchy. The data retrieved will then be assessed on an individual event basis by a PATCH clinician to assess the concordance of clinical definitions between the data sources.

Analysis will be achieved within the UCL data safe haven using pseudonymised data. Analysts will not have access to the identifiable data which is held within a different data safe haven within the clinical trials unit.

The research will perform a three-way comparison using the PATCH data (collected via the PATCH CRFs) as the gold standard, as compared with NICOR and HES. Major cardiac outcomes should be common to all, though collected in 3 different ways. Cerebrovascular and thromboembolic event will be compared between PATCH and HES.

The NICOR and HES data will not be linked but using the study numbers of the PATCH trial participants researchers will compare cardiovascular events that occurred while on the trial to see if there are differences between the HES, NICOR and the manually collected data sets.

HES data will not be shared with NICOR or any other third party.

NICOR data will not be shared with NHS digital as per the data sharing agreement between NICOR and UCL.

Venn diagrams may be produced to assess concordance between databases (as per Herrit et al, BMJ 2013) and descriptive statistics used to detail the relative rates of event detection in the separate databases. The positive predictive value of each database in recording and classifying events subsequently collected via the CRF will be calculated. Additionally the reverse will be performed to assess the efficiency of CRFs in capturing events as compared with national electronic datasets.

UCL are not permitted to re-identify individuals under this agreement.

Outcomes of this sub-study will not be used as outcomes for the PATCH trial in the future.

NHS Digital reminds all organisations party to this agreement of the need to 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 ie: employees, agents and contractors of the Data Recipient who may have access to that data).

Data provided will be for all members of the cohort from 2010/11 onward. This is so the cohort of data is standardised for all participants to verify trial data and attempt to eliminate false positives and negatives from analysis.

The results of this study are expected be published in high-impact peer reviewed journals such as Lancet Oncology, or an appropriate methodological journal such as Trials.

The study will be published in peer reviewed journals to give the highest impact and broadest readership.

In addition, the results will be presented at scientific conferences related to cancer, cardiovascular, and trial methodology, and to key stakeholders. The conferences will be chosen depending on the key findings and also the target audience. This would be important to deliver the findings with the most impact possible. Possible conferences include International Clinical Trials Methodology Conference, European of cancer registries conference and National institute of cancer research conference. This would allow for coverage of key stakeholders in the setting of registry data, clinical trial design and clinicians involved in cancer research.

The results will also be discussed at meetings within the clinical trials community in the use of electronic health records to allow for further development of this resource in trials.

Through these mechanisms the results will be widely disseminated around the clinical trial community, including clinical trial centres who run the majority of clinical trials in the UK. The results of this study will establish the usability of HES data to identify important adverse events such as heart-attacks, by comparing HES data with a robust, gold-standard event data capture conducted as part of standard protocol in the PATCH trial. This will inform on the general quality of HES data for this purpose, and more specifically to inform whether this method of data collection is suitable in the longer-term for the PATCH trial and for other clinical trials who may wish to use this data.

Publications will be published based on UCL open access policy. This would include publication in open access journals and access to the results on the MRC clinical trials website which is freely open to the public.

Outputs will be anonymised to the level required by the ISB anonymisation standard and will contain aggregate data only.

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

A challenge for several clinical trials is the long-term follow up of patients over many years, required to determine the risk/benefit of an intervention in the longer-term. This is important to understand:

- whether early effects are retained in the longer-term.

- to measure any delayed effects (benefits and harms) from interventions that cannot be captured in the short term.

With traditional methods of follow-up, there is generally:

- attrition of participants over time which creates missing data

- a high cost associated with ongoing trial visits and data capture.

This loss-to-follow-up can also result in bias, for example:

- participants from lower socioeconomic classes may be more likely to withdraw from trials

- those with disease progression might be more or less likely to attend follow up (depending on the setting).

A potentially more cost-effective way of continuing follow-up in the long term is to use routine electronic health records. These records can contain demographic, clinical, and patient-centered data.

The use of routine electronic health records within multicentre trials has the potential to:

- reduce the burden of additional trial-specific clinic visits and assessments for both participants and clinical research teams.

- significantly reducing the costs of central trial coordination and data collection.

To see if this is a viable option for the future there needs to be a body of work to provide evidence of the suitability of this data. This study will add to this body of evidence to allow for more effective clinical trials in the future which in turn will benefit patients who participate in trials and also who benefit from the trial result.

The beneficiaries of this research could be:

1. Patients - as this could allow for trials to be streamlined so they would not need such intense follow up. This would mean less visits to hospital which can often be time consuming and costly.

2. Clinicians and researchers involved in clinical trials in NHS hospitals. The decrease of work load of providing evidence of serious adverse events which could allow for more time with patients.

3. Clinical trial staff work load would be decreased with lack of paper work needed to process serious adverse events forms. This would decrease cost and improve efficiency of large clinical trials.

4. This could allow for innovative designs of clinical trials that could be developed in the future in partnership with national registries.

Yielded Benefits is not a requirement for new applications.