NEJM: Bringing Together Ideas for Sharing Clinical Trial Data
Interview with Jeffrey Drazen, M.D., Editor-in-Chief, The New England Journal of Medicine
July 11, 2017
Can data from clinical trials be shared responsibly, fairly, and effectively? That was the question addressed by clinical trialists, data analysts, and others at the recent NEJM Data Summit — a candid, two-day event that initiated dialogue among those with differing points of view on this hot topic.
Key constituents listened and learned from each other, as attendees at the New England Journal of Medicine gathering focused on the promise and pitfalls of sharing data from clinical trials. NEJM and Editor-in-Chief Jeffrey Drazen, M.D., have been at the forefront of public discussion on the topic.
To explore the potential of data sharing leading up to the April summit, NEJM hosted the SPRINT Data Analysis Challenge. Individuals and groups were invited to analyze the dataset underlying the SPRINT article published in NEJM, A Randomized Trial of Intensive versus Standard Blood-Pressure Control. Data analysts were challenged to identify a novel scientific or clinical finding that advances medical science. Of 200 teams or individuals from around the world who passed the initial qualifying round, 143 submitted full entries. An expert panel combined with online crowd voting chose three winners: first place from Tel Aviv’s Clalit Research Institute, second place from Boston University, and third place from Stanford University.
The two-day summit in April, Aligning Incentives for Sharing Clinical Trial Data, brought SPRINT Challenge winners together with trialists, patients, data scientists, and representatives of data repositories and funding agencies. With 140 participants in the room and nearly 5,900 people registered to view the live webcast, speakers addressed the question, “How can we create an effective, sustainable data research eco-system?” Amidst the many views, one point of agreement emerged: small experiments such as the SPRINT Challenge can help generate experience and confidence for moving forward.
Dr. Drazen reflects here on what was learned from the SPRINT Challenge and the Data Sharing Summit, and what to expect next for data sharing.
What are the biggest takeaways from the Data Sharing Summit?
This is a controversial area in transition. You have three major constituencies: Patients who put themselves at risk by participating in trials; trialists who design the trials and gather the data; and data analysts who look for signals in data that someone else has collected. Bringing these three groups together in one place helped us to define key issues in unique and illuminating ways.
We made real progress. We learned that patients want data to be used widely and respectfully. Trialists heard others’ views on why they shouldn’t be allowed to hold onto a dataset solely for their own use for an extended period of time. Data analysts heard that while some datasets can be shared easily and quickly, there may be good reasons for other data not to be shared or used right away.
But the key thing was having patients there. Patients who had been in clinical trials were truly surprised at the degree of conflict between trialists and analysts. Trialists and analysts need to remember that disease is the real enemy. One of the patients who had been in the SPRINT trial essentially said, “You are the experts, you understand the issues, you know your fields. Work together — figure this out.” This was a reminder to everyone that the real goal is to improve human health.
Participation in the SPRINT Data Analysis Challenge was impressive – 143 completed entries. Did anything surprise you about the response or about the entries themselves?
On the downside, I lost a bet with [Summit Co-chair] Isaac Kohane. I predicted the SPRINT Challenge would draw fewer than 30 entries; he predicted more than 100.
The response was heartwarming. There was a big range in participants. Not all were data analysts. We had ordinary citizens, pharma, and medical students. Entries came from around the world. Some teams had representation from both the clinical and the data camps. Good surprises came in the sophistication of some of the analyses and seeing the novel techniques used to advance questions.
I did hope for more entries that combined multiple datasets, in creative ways. Maybe the timeframe hindered that. I also was surprised that no one tried to predict the final rate of events from the SPRINT Trial’s close-out data. The trial was stopped early and published in November 2015. But the trial’s closeout visits were not until July 2016. It would have been fun to predict what the closeout numbers actually were compared to the data available in the fall of 2015.
Interestingly, a number of entries were decision-support tools. The tool from the Clalit Research Institute was very easy for physicians and patients to use and understand as they decide whether to aim for intensive treatment of hypertension. This may reflect a key message: It’s really about caring for the patient.
At the Data Sharing Summit, did the level of disagreement surprise you?
No, we expected the conflicting views. And we encouraged people to express their real opinions. Progress is made when we listen respectfully to the views of others. We wanted people with varied viewpoints to speak with each other and with the wider world. One speaker said afterward that this was the beginning of a dialogue — people talking with each other instead of at each other.
Bernard Lo summarized the feeling in the room. He said that each group — trialists, data analysts, repositories — may need to give up something for the common good. There also was realization that opposing sides may need to be incentivized if we are to get together and find a way to make data sharing work. For example, there was a strong call for academic and promotion committees to recognize data analysis, and not only published research, as a measurable professional achievement.
What comes next in discussion of Data Sharing?
At the Summit, there was a call for small, short-term experiments, to “get granular” as one speaker said, and to create some evidence base as to what happens when data is shared. Some of these are already underway, with some data sets being widely used. The NIAID TrialShare, with clear guidelines, has been up for more than two years. The effect has been only positive. I think we’ll see more experimentation.
Our SPRINT Challenge was one such experiment, showing how contests can make people aware of the possibilities and confident about moving forward.
This month, the International Committee of Medical Journal Editors announced that as of July 1, 2018, manuscripts will be considered for publication only if they include a data sharing statement. And, trials that begin enrolling patients on or after January 1, 2019, must include a data sharing plan. These steps will move data sharing forward, which is explained more in this interview [starting at 6:40].
You’ve said unequivocally that NEJM supports data sharing. What future do you see?
We have a generation of young people who are used to sharing tools and information. When I pose a problem to a class of graduate students, they automatically break into groups to solve it. There is a desire to take, use, and share information. It’s the openness of the next generation that I think will make a difference.
We have a new science of bioinformatics. How do we use it to maximum benefit? Instead of trying to figure out who is right and who is wrong on the ground, trialists and data scientists need to work together and ask, How do we get true information that makes people’s lives better? What is the best way to help people through the best science?
Putting data out there — available and open — will increase public trust. Doing so is more believable than if you say, “Trust me, I’m a doctor or I’m a trialist.”
Sharing data has the potential to teach us things we did not know, to maximize the value of clinical trials, and to improve health outcomes. When I am at the bedside, I want to tell patients not what I think, but what I know.
NEJM LibraryHub would like to hear from you: What role do you expect to play in data sharing? Are you currently storing or managing clinical trial data? How might medical librarians prepare now for any changes ahead?
Email your thoughts to firstname.lastname@example.org to inform the next LibraryHub update on data sharing.