Why linking data effectively is crucial for analysing clinical trial outcomes – NHS England Digital
The SDE Network supports researchers by tackling key challenges, such as public confidence and trust in researchers using NHS data, and difficulties accessing data across a fragmented system with over 7000 entry points.
The 10 regionally based and 1 England-wide SDE in the SDE Network provide secure data and research analysis platforms (see Research Powered by Data). The SDE gives approved researchers with approved projects secure access to NHS data. To ensure patient safety and confidentiality, this data is:
- de-identified (pseudonymised), so that any identifying (or personal) patient information such as names, addresses or NHS numbers are replaced with artificial, or ‘pseudo’, information
- minimised, so that only relevant patient information is provided for research projects
The data in SDEs remains securely under NHS oversight, and researchers can only take away their research findings, not the actual data itself. This approach ensures sensitive health data is accessed safely and securely, and it is being echoed across the UK’s nations and internationally.
My colleagues and I have just published a new study — The Need for Fully-Effective Federated Analytics of Data Sources for Clinical Trials — in the International Journal of Population Data Science (IJPDS). We examined what could happen if information about clinical trial participants is held across geographically separate environments and what if those environments did not allow those data to be brought together for analysis. We compared 2 approaches. First, we used a 1-stage analysis, which assumes that data from all participants could be analysed together within, or as if they all were in, a single environment. This is just like the approach typically used in clinical trials. Second, we used a 2-stage approach, which assumes individual participants could not be brought together and, instead, the data is analysed separately within each environment and then their separate sub-results combined; this approach is effectively a ‘meta-analysis’.
Our study found that using a 2-stage approach could introduce limitations. For analyses with quantitative measurements, like haemoglobin or prostate specific antigen (PSA), the results were reassuringly similar to those in a 1-stage approach. However, for analyses that examine how long it takes for participants to experience a specific outcome, such as recovery, relapse, or hospitalisation (known as ‘time-to-event data’), the result was more uncertain and less precise. Only being able to work with separated data also makes it tough to investigate whether treatments work differently for particular groups of patients (‘sub-group analyses’).
These findings are especially timely in light of the changes to make ‘data access’ (researcher comes to data) the default route to NHS data for research rather than the more familiar ‘data sharing’ (data goes to researcher) approach. Recognising this important challenge, the SDE Network has been working to ensure its technical and information governance processes can support a single stage approach.
