Inadequacy of remote desktop interface for independent reanalysis of data from drug trialsBMJ 2014; 349 doi: https://doi.org/10.1136/bmj.g4353 (Published 09 July 2014) Cite this as: BMJ 2014;349:g4353
- Jon N Jureidini, child psychiatrist1,
- John M Nardo, clinical assistant professor of psychiatry, Emory University, retired2
- 1Paediatric Mental Health Training Unit, University of Adelaide, North Adelaide, SA 5006, Australia
- 2Atlanta, GA 30322, USA
GSK granted our RIAT team access to the data from the Paxil Study 329 to do an independent reanalysis, and we’ve now had months of experience working with their “remote desktop” interface as a portal to their information. It is a single windowed multiple document interface, totally self contained, which cannot be accessed by software other than that provided inside the window. Multiple passwords are needed multiple times for access, and some days we are frequently thrown from the system without warning, necessitating repeated logins. Another internal window from SAS contains the data and allows us to run SAS programs to analyse the data. The data are also provided as text based CSV files that can be moved outside the internal SAS window and displayed in spreadsheets (Open Office is provided). The open source statistical program “R” is also provided. Because none of us is fluent in SAS procedures and programming, we are using R, which has the necessary statistical functions. Getting the data into the format required by R means using multiple spreadsheets, which is difficult in the cramped space provided—it can take days and involve many “startovers.” Once the data are analysed, only non-data containing files with results can be exported, and that’s only by application for approval. It is difficult to use the primitive graphing functions of Open Office or R in the remote desktop, and the graphs cannot be exported. We therefore copy the information for the graphs by hand to get it into the computer proper and create our images. The process of using this interface for analysis is unnecessarily maddening and a severe obstruction to the task.
The original hand written case report forms are provided as PDFs. In our case, there are 275 participants, many with several PDFs, each containing 200 plus pages. In the interface provided, we can see only one of the nearly 60 000 pages at a time. Tallying the adverse events from the PDFs is hard enough, but doing it without being able to make printed copies and referring back and forth is a double nightmare. It results in days of wasted time and delegation is impossible. Working inside this little window brings home how important it is to have a workspace that allows simultaneous display of lots of information. One of our team has lost one assistant already over this interface.
The European Medicines Agency’s decision to consider using this kind of interface is, at best, ill advised, and a negation of the initial promise of data transparency. If we didn’t have a retired member and some members with assistants, we couldn’t do it. It is impossible to imagine that anyone contemplating this decision who knows the ins and outs of data analysis has attempted it in this kind of environment. We would liken it to going to sea to see the world in a submarine, looking through a periscope.
Cite this as: BMJ 2014;349:g4353
Competing interests: None declared.