At Cytel, we are very often asked to get involved in DMCs ( Data Monitoring Committees) in a variety of capacities. Our statistical experts are recognized for their work in many of the areas related to DMCs including-group sequential and adaptive designs, multiplicity, missing data, and decision sciences. One key operational management issue when producing analysis for DMCs is that rules must be defined so that trial unblinding is not compromised and bias remains controlled. In this blog, we will explore an innovative approach developed by members of our statistical programming team to improve the efficiency of this process, while maintaining the highest levels of data security.
One traditional way of handling DMCs is to have two separate teams ( (often in two different locations) - blinded versus un-blinded Statistical Study Team, responsible for the DMC (blinded) and Final Analysis (un-blinded) respectively. From an operational efficiency point of view, this solution can require a heavy use of resources since an entirely separate Statistical Programming Team in addition to the blinded one, needs to be assigned to the project as per the figure to the right.
Last year at PhUSE Angelo Tinazzi and Dean Schults of Cytel presented an alternative solution which they have developed in order to improve efficiency in this process. In this approach, we would still have two separate teams, sitting in different locations, however the Blinded Statistical Team is responsible for both DMC and Final Analysis activities. All outputs are programmed and quality controlled by the Blinded Statistical Team using blinded data and dummy treatment schedules as applicable.
In addition to the use of dummy treatment codes a “data masking” process is applied to the real data. Different masking criteria can be agreed between the blinded and unblinded Statistical Team, approved by the Sponsor, and applied by the unblinded Statistical Team prior to the transfer of data to the blinded Statistical Team. For example shuffling patient information, blinding medical coding information through random assignment of occurring medical terms, etc. With such an approach the blinded statistical team will still be working with 'real data' changed in a way that retains scientific meaning.
Consistency checks are done to ensure the correct treatment assignments are being used and outputs transferred to the DMC members via a secure file transfer system.
The Trial Blinded Statistical Team will then transfer the final validated programs to the un-blinded Statistical Team. The un-blinded Statistical Team will the rerun the programs using blinded data on the unblinded study folder, and conduct a comparison between the provided blinded outputs and the blinded outputs run on the unblinded platform, in order to confirm that the transferred programs produce the same blinded results. After this consistency check the unblinded Statistical Team will update flags to use actual randomization assignments, and create the unblinded outputs.
This process operates as per the figure below.
Cytel’s proprietary software ACES also helps to facilitate our DMC work providing a secure, centralized repository for storing and accessing data, information, and documents related to a clinical trial and associated interim analyses. ACES allows both internal trial sponsors and external parties (e.g., DMC, ISC, and IWRS) to access information in a controlled and consistent way using a role-based security model defined at the trial level.
While this solution requires careful planning and selection of the appropriate variables to transfer and/or blind, the approach has been able to deliver efficiencies where it has been used. In addition it streamlines the re-use of DMC Analysis Programs for Final Analysis since the same programming team is used.
For further details of this approach and to download the associated poster presentation click below.