Assessing the utility of healthcare systems data for trials: data utility comparisons in the STAMPEDE trial (DUCkS)(previously: ODR1718_094)

The Data will be used for the purpose of a research study: "Assessing the utility of healthcare systems data for trials: data utility comparisons in the STAMPEDE* trial (DUCkS)"

*STAMPEDE (Systemic Therapy in Advancing or Metastatic Prostate Cancer: Evaluation of Drug Efficacy) is a clinical study. STAMPEDE aims to identify new treatments for prostate cancer.

Data for this study was previously obtained when the data were controlled and managed by Public Health England (PHE). PHE facilitated data release via its Office of Data Release service (ODR). ODR was responsible for providing a common governance framework for responding to requests to access PHE data for secondary purposes, including service improvement, surveillance and ethically approved research. All requests to access data were reviewed by the ODR and were subject to strict confidentiality provisions. The responsibility for the management of the National Disease Registration Service of which the National Cancer Registration and Analysis Service is a part, transferred from PHE to NHS England on 1st October 2021. The STAMPEDE study previously accessed data via Public Health England under the reference: ODR1718_094.

The Medical Research Council Clinical Trials Unit at University College London (MRC CTU at UCL) is the Data Controller, who also processes data for this study. Herein forward the Data Controller shall be referred to as University College London or UCL. UCL is the trial’s Sponsor and delegates sponsorship responsibilities to MRC CTU at UCL.

The STAMPEDE study has received funding from Cancer Research UK and the DUCkS study has received funding from Health Data Research (HDR) UK. HDR UK and Cancer Research UK do not determine the purpose(s) the data will be processed for nor undertake any role which defines them as a Data Controllers. HDR UK and Cancer Research UK will also have no access to NHS England data other than in the form of aggregated and suppressed tabulations (as per the HES analysis guide).

In order for trials to use healthcare system data, UCL are requesting record level NHS England data in order to assess the utility of the data by comparison with trial collected data. The study team intend to do this for data from approximately 10,500 participants from within the STAMPEDE trial.

STUDY AIMS

University College London aim to assess the concordance agreement between traditional trial-specific data collection and healthcare systems data (“routinely-collected healthcare data”) in approximately 10,500 STAMPEDE participants. The analyses will involve assessment of five objectives:

1. Assessment of survival,

2. Chemotherapy treatments,

3. Radiotherapy treatment,

4. Second-line treatment,

5. Toxicities.

UCL are therefore requesting NHS England record level health data linked to specific participants of the STAMPEDE trial. The study team will curate the data to be analysed against the five objectives above, working through each in turn.

NEW DATA REQUESTED

UCL are requesting access to the following NHS Hospital Episode Statistics (HES) data sets:

- HES Outpatients (OP)

- HES Admitted Patient Care (APC)

- HES Accident and Emergency (A&E)

- Emergency Care Dataset (ECDS)

The HES data provided from NHS England provides additional fields to those previously received from PHE therefore this will enrich the data previously held. To date, approximately 10,500 people have joined STAMPEDE within England and Wales, and recruitment continues (though the study Arms A, H-L are due to close their recruitment around September-October 2022) but this agreement is for a one-off extraction of record-level data, and the study does not anticipate returning for health data on those recruited after the cohort has been provided to NHS England under this agreement.

UCL are also requesting further access to the following National Cancer Registration and Analysis Service (NCRAS) National Disease Registration Service (NDRS) datasets (formerly available via Public Health England (PHE):

- NDRS Radiotherapy Data Set (RTDS)

- NDRS Systemic Anti-Cancer Therapy (SACT)

- NDRS Cancer Registry

These data sets, holding information on patients treated with radiotherapy or chemotherapy, allow the audit build of the full picture of the treatment provided to cancer patients, in-depth analysis of specific regimens and changes to prescribed treatments. It allows the exploration of whether the radiotherapy and chemotherapy data items, collected by the Audits from hospitals, are appropriate and necessary. Should particular data be available in existing, national data sets, these data items could be removed from the data collection, to ease the burden on data providers (hospital staff). RTDS data is linked to Cancer Registry data using the ‘Patient ID’ pseudonym; it is not possible to request RTDS data in isolation without this linkage, therefore NDRS Cancer Registry has also been requested.

In order to establish a full picture of the data, the study team are requesting HES and NDRS data from 2005/06 to 2021/22 (where datasets will allow) for approximately 10,500 participants from the STAMPEDE trial.

The study team currently hold data which was collected via PHE (under ODR1718_094), for the period 2005/06 to 2017/18 for a historic (previously submitted) cohort. This includes:

- NDRS RTDS

- NDRS SACT

- NDRS Linked HES A&E Data

- NDRS Linked HES Inpatient Data

- NDRS Linked HES Outpatient Data

PATIENT AND PUBLIC INVOLVEMENT AND ENGAGEMENT (PPIE)

PPIE was sought to discuss the use of routine data of trial participants recruited before 2013, where it was unknown if they also consented to data linkage (optional participation). Two focus groups of twelve and seven people met in July and September 2021 respectively. They were recruited by Prostate Cancer UK and Cancer Research UK. All who attended have been affected by cancer in some capacity, either a patient or a family member.

Overall, the groups expressed agreement that it was acceptable to access routine data of trial participants without consent provided there was transparency, i.e. information stating that this was being done in line with the reasons given on the trial website. As only a few trial participants (up to 12) did not agree to linkage of their routine data before 2013 and it was not known why, the groups felt it was acceptable to use the routine data as most of the trial cohort would have agreed, and it was not practical to go back to them to ask for retrospective consent. Some felt involvement in the clinical trial should automatically allow for the use of routine data, so participants would have to opt-out or withdraw from the trial if they didn’t agree. Consequently, the privacy notice was revised on the trial website to clearly explain the use of routine data without explicit consent to linkage.

Members discussed how the patient information sheet and consent form should clearly state that routine data will be accessed and that participation in the trial would permit access to the records held by NHS England and other data providers. Both documents have provided this information about data linkage since 2013 when all patient-facing documentation was updated.

The Data Controller will process the data under GDPR 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 a higher education establishment, the University conduct research to improve health care and service and the linkage requested is necessary for the performance of a task carried out in the public interest; i.e. improving the treatment offered to men with prostate cancer.

Additionally, under GDPR Article 9(2)(j) processing of Special Category Personal Data is necessary for archiving for research purposes. Data minimisation process is being followed and only data that is required specifically for the purposes of this study has been requested, to protect the rights of the data subjects.

OVERVIEW

Project Aims:

1. To identify better ways of obtaining clinical trial outcome data to (A) Repeat reported STAMPEDE analyses, to validate the PhD project's algorithm capabilities and the trial results; (B) Perform new, secondary analyses not possible with conventionally-collected trial data, including but not limited to, rates of neutropenic sepsis when chemotherapy is given at different times in their disease or cardiac events or hormone therapy.

2. To develop a clinically useable tool (e.g. algorithm), for use in a clinical trial, to accurately identify disease-driven events and trial outcomes.

3. To help reduce the burden of collecting clinical trial data from traditional patient and clinician contact. By using data that has already been accurately collected by the NHS, it may be possible to improve timeliness, reduce costs and save resources that can be used elsewhere.

Objectives:

1. Develop, enhance, and validate a methodology to calculate trial outcomes from EHRs. Can the new tool accurately detect disease-related events and trial outcome events, and therefore successfully identify treatment effect differences? The tool must:

a. Either be able to use HES data alone or use non-HES-based data in addition

b. Be clinically useable for planned applications

2. Identify if the routine data can be used directly for clinical trial follow-up through comparison assessments of specific clinical events and treatments recorded in routine data and trial data (survival, chemotherapy, radiotherapy, second-line treatment, toxicities including safety events).

a. If yes, then utilise this data for routine trial follow-up.

COMMERCIAL BENEFIT AND TRANSPARENCY

For transparency it is noted here that the following pharmaceutical companies provided discounted or free drugs used in the STAMPEDE Trial and funding of Educational grants in return for accessing the final primary outcomes report and record-level trial-relate data (NOT ODR/NHS England data), but had no influence over how the trial was conducted nor how outcomes were reported.

> Sanofi Aventis, Novartis, Pfizer, Janssen, Astellas and CLOVIS.

These pharmaceutical companies have no involvement with the DUcKS analysis in this agreement.

The data flow will be as follows:

1. University College London will send two cohort files (of approx. 10,500 patient records in total) each with the identifiers listed below to NHS England securely via Secure Electronic Transfer (SEFT). One cohort file will contain individuals processed under Section 251 and one cohort file will contain individuals processed under consent.

2. The NHS England data production team will link patient identifiers to the NHS England datasets (HES OP, HES APC, HES A&E, ECDS, NCRAS Cancer Registrations, SACT and NCRAS RTDS)

3. NHS England will apply National Data Opt-Outs to the cohort of participants received under s251 (approx. 1,600).

4. The NHS England production team will then remove identifiable patient information from the linked data.

5. The NHS England production team to securely send the de-identified linked data, containing study ID to the data recipient at the University College London via SEFT.

To facilitate the linkage of the STAMPEDE cohorts to HES and NCRAS data, the study team at University College London will securely transfer the following identifiers to the NHS England data production team:

FOR CONSENTED COHORT

• Study ID (STAMPEDE trial participant identifier)

• NHS Number

• First Name

• Last Name

• Postcode

FOR SECTION 251 COHORT

• Study ID (STAMPEDE trial participant identifier)

• NHS Number

• Date of Birth

• Postcode

There is no linkage field for gender as the entire trial cohort is male.

These pseudonymised datasets with the Study ID will be sent to the MRC CTU at UCL using the specified transfer method above. The data will reside in UCL’s Data Safe Haven and will be identified by Study ID only, thus there will be no identifying personal data attached to a study number. Only defined members of the DUCkS study team and MRC CTU’s methodology team will have access to Data Safe Haven for data analysis - all are either substantive employees of UCL or Chief Investigator, Comparison Chief Investigators or clinical delegates involved in the blinded review of clinical information for this project only who are under a contract with UCL. All UCL substantive employees have completed training in data protection and confidentiality, and users of Data Safe Haven receive appropriate training before being granted access.

MRC CTU at UCL is not permitted to re-identify individuals under this agreement.

DATA OPT OUT

>National Data Opt-Out

NHS England record level data sought under s251 under this agreement is subject to National Data opt-out. If an individual has evoked their right to opt-out from the use of their data for research or planning purposes (the National Data Opt-Out) their data will not be released under this Agreement. This will not apply for participants in the cohort who have provided informed consent.

> NDRS specific Data Opt-Out

NCRAS data is subject to both the NDRS Opt-Out as well as National Data Opt-Out. If an individual has indicated that they wish to be excluded from the national cancer registry, their data will be permanently removed from all NCRAS datasets within 20 days of receipt. This means that for this agreement, for both the cohort under s251 and the consented cohort, no NCRAS data will be released for any person who has registered an NDRS specific opt out which has been completed.

DATA STORAGE AND ANALYSIS

The data will be held on UCL’s Data Safe Haven using UCL approved computers. The Data Safe Haven is UCL’s technical solution for transferring and storing research information that is highly confidential. It meets the requirements of the NHS England DSP Toolkit and ISO 27001 Information Security standard. Access is controlled by the Information Asset Owner, and all UCL staff complete training in confidentiality and data protection, which is renewed annually.

Statistical data analysis will be carried out via UCL owned devices connected to the secure UCL network remotely, using an appropriate statistical package. To remotely access the server requires a secure 2-factor authenticator (VPN) and users are then able to securely access the secure server on the University’s IT framework. All data analysis will be conducted within the confines of the University’s secure Data Safe Haven, and will not be downloaded to remote devices for storage or processing. The export of record-level data is not permitted from the Data Safe Haven under access restrictions.

HES and ECDS DISCLOSURE CONTROL / SMALL NUMBER SUPPRESSION

In order to protect patient confidentiality, when presenting results calculated from HES record level data, outputs will contain only aggregate level data with small numbers suppressed in line with HES Analysis Guide. When publishing HES data, data processors must make sure that:

• National-level figures only may be presented unrounded, without small number suppression

• cell values from 1 to 7 (inclusive) are suppressed at a sub-national level to prevent possible identification of individuals from small counts within the table.

• Zeros (0) do not need to be suppressed.

• All other counts will be rounded to the nearest 5.

Data will not be made available to any third parties other than those specified except in the form of aggregated outputs with small numbers suppressed in line with the HES Analysis Guide.

The aim was to complete the analyses by April 2023 with results being written up (containing aggregate data with small numbers supressed) for publication and dissemination and papers ready for publication around Autumn 2024.

The MRC CTU at UCL plan to communicate the methods used and the results of these data comparison studies via:

- Presentations at major international and national scientific conferences

- Publications in high-impact peer-reviewed journals

- Training workshops disseminating study results to trialists, funders and other key stakeholders.

- Results may also be made available via the HDR Innovation Gateway to enable researchers to decide data relevance for their studies.

The study team also aim to communicate the main STAMPEDE trial results via a booklet to trial participants and to share findings via the trial's website.

The study team plan to write a paper of the methodology used and a paper of the results. The aim was to disseminate the results via workshop around March 2023.

Benefits type: Knowledge about use of healthcare systems datasets for clinical trials.

The DUCkS trial is in place to assess the utility of healthcare systems data for trials.

The results of the DUCkS project aims to support trialists, funders and NHS England in the better use of HES, ECDS, and NCRAS data in clinical trials. Dissemination of the results from the DUCkS study aims to enable the trials community to understand if centrally-collated national datasets can replace trial-specific data collection of important outcomes such as chemotherapy and radiotherapy treatments and the occurrence of serious safety events (for example cardiovascular events and toxicities)

Should the DUCkS study prove that data collection through national datasets is effective, this has the potential to improve efficiency of trials, saving time and money. Although the exact savings for the NHS cannot be predicted, it is hoped this will be achieved through research nurses/practitioner/doctors spending less time on data collection for clinical trials. Consequently, trials could potentially be completed more quickly with less missing data and fewer participants lost to follow-up. This supports innovation and faster research and development of better treatments.

It is hoped that the project could form a basis for HDR UK research activities in the "Transforming Data for Trials" programme over the 5 years from 2023. One of the key aims is to transform the utility of healthcare systems data for trials.

The DUCkS project hopes to pave the way for others to conduct data comparison studies which could determine the utility of other datasets to support clinical trials. Additional studies should provide the trials community with further evidence of which healthcare datasets can be used for specific trial outcomes. This will allow more trials to be designed using these data as the main resource, minimising time and effort by healthcare workers to collect this information. Instead, they would have more time to provide care to patients.

Cancer progression and recurrence is not routinely recorded within patient health records, and so it is difficult for clinical trialists to accurately estimate time to cancer progression. Time to cancer progression is often used as a primary trial outcome to determine if a trial intervention is effective or not. The development of an algorithm that uses routinely collected health data to better estimate fact and time of cancer progression should allow trialists to test if a cancer treatment is effective. The data previously obtained from PHE (obtained under ref: ODR1718_094) has enabled MRC CTU at UCL to develop and test an algorithm to estimate cancer recurrence. The results have been written up in a PhD thesis and are also expected to be published as a peer-reviewed research article in 2023.

Additional yielded benefits (for 2025 DSA renewal):

The following three studies have yielded benefit: 1) direct patient benefit by improving clinical practice in the management of prostate cancer, and 2) indirectly through using health data to provide timely and comprehensive data for the primary outcome, overall survival.

The STAMPEDE trial team demonstrated that HSD can help identify survival outcomes, for example, by using Civil Registrations of Death (CRD) data. Specifically, in the STAMPEDE Metformin Comparison, overall survival data from CRD allowed the trial to report its primary outcome sooner than would have been possible through traditional data derived from case report forms (CRFs) (https://doi.org/10.1016/j.annonc.2024.08.2313). This highlights the ability of HSD to improve trial efficiency and disseminate vital cancer research rapidly.

In the STAMPEDE Radiotherapy comparison study, investigators used HES to demonstrate that prostate radiotherapy reduced the long-term risk of upper urinary tract obstruction in trial participants (https://doi.org/10.1016/j.annonc.2023.09.2732). Consequently, patients with prostate cancer may now receive radiotherapy for this indication.

Another example was a sub-study within the STAMPEDE Zoledronic Acid Comparisons. This study used HES data to show that patients with metastatic disease who received Zoledronic acid had a reduced risk of long-term fractures (https://doi.org/10.1097/JU.0000000000003226.13). As a result, many patients with prostate cancer are now offered bone protection treatment with Zoledronic acid to prevent fractures, illustrating the ability to utilise HSD in trials to directly optimise the care of future patients with prostate cancer.