Sharing clinical trial data

New research published today in Trials details the clinical trial data sharing approach of the Clinical Trials Unit at University College London. In a time when how best to share clinical trial data is hotly debated, the authors of this article discuss its importance and why their process works.

Randomized controlled trials are the best way of testing whether new health interventions work, but they are often expensive and time-consuming.

Clinical trials collect huge amounts of high-quality data, which enables us to answer our primary question. But this data also has the potential to answer many other questions. It makes sense to make the most of the data we collect in trials, whether that’s by carrying out additional analyses ourselves or sharing it with others.

Data from trials can be combined with data from other trials in meta-analyses, linked with other data sources (such as tissue samples or death registers), used to answer methodological questions, or to help other researchers design new studies. This all provides extra value, a principle which we strongly support and funding bodies like.

Often, this means sharing trial data with other groups, but we need to think carefully about how we do this.

Some people have suggested that all data from a trial should be made available to anyone, the so-called “open access” model, either on request or via an open repository. The European Medicines Agency came very close to this approach in working up their guidelines (but they later parked this with updated guidance). We think there are some dangers in a completely open approach.

Clinical trials collect huge amounts of high-quality data, which enables us to answer our primary question. But this data also has the potential to answer many other questions.

Dangers in a completely open approach

Some problems are about whether people consented to have their trial data accessed in this way. It might seem to be obvious that people would be happy for their data to be shared for a good scientific aim, even if they hadn’t agreed to this when joining the trial. But we cannot assume this.

Sure, most of us will have littered the internet with personal data at some time, but usually this has been our own choice. There are healthcare matters we might never wish to tell people about. And children join trials before they are old enough to give informed consent. What happens when they grow up?

In the future, we can ask people before they join a trial. But we cannot realistically go back to re-consent people in older trials.

Some people will say that you can anonymize data then make it open. But it is possible to link datasets and identify people, or for people to identify themselves (or people they know). Methods for unlinking data are becoming more sophisticated, but so are the methods for connecting data.

Privacy aside, the bar is set high when deciding who can formally analyze clinical trials data. With more eyes on the data there are always risks of wrong analyses, false positives, and over-interpretation.

Why our approach works

Here at the Medical Research Council Clinical Trials Unit at University College London, and published today in Trials, we have developed controlled access approach to sharing our clinical trial data. Researchers apply for access to data for projects and these are considered by independent reviewers. Formal, legal data sharing agreements can mitigate against most privacy concerns.

This method is not without challenges. Most notably, these are time and cost. Funders do like data sharing, but data sharing requests are usually late in the life of a trial or after it has ended. Are funders interested in supporting data sharing activities by extending grants or providing core resources? Going forward, they will have to be.

Processing data sharing requests, taking them through reviews, preparing shareable datasets, and supporting people who have received dataset is a surprisingly large task. Unfunded at the moment, these activities compete for time with our current trials and priority research questions.

Through our controlled access approach, we have shared data from our trials more than 100 times in the 3 years from 2012 to 2014, which is nearly once a week. There is obviously a clear demand. Funders will need to address robust financial investment.

This guest blog is by Matt Sydes, Tony Johnson, Sarah Meredith, Mary Rauchenberger, Annabelle South and Mahesh Parmar of Medical Research Council Clinical Trials Unit at University College London.

View the latest posts on the On Medicine homepage