INTERIM ANALYSIS AND DMC
Blinded outputs were programmed internally by a sponsor and we had an independent statistician who attended the meetings and gave statistical advice. The sponsor created blinded outputs outside of their standard reporting environment so that we could easily recreate outputs using a standalone SAS environment. We were supplied the DMC charter for review prior to finalization.
A limited number of statisticians and programmers were involved and we held an internal kick-off meeting at the start of the project, discussing, amongst other things, the issues around ensuring we retained the blind for the rest of the study team.
The analysis included a Cox Proportional Hazards model; the assumptions of this analysis were checked at every analysis run. In practice, we did need to customize a few lines of our AUTOEXEC.SAS to match the sponsor’s SAS environment. After the first DMC dataset transfer, we had a number of questions and clarifications on the outputs, which were resolved by discussion with the sponsor. We created the minutes from the DMC meetings and shared these with the DMC team. We also scheduled subsequent meetings using web-based scheduling tools. The DMC sent us this note after we delivered minutes and scheduled one of the meetings: “Many thanks and also thanks for the organization of the teleconference and the minutes. You are very competent.”.
We continued in this manner and had six DMC meetings for the project. We did have to discuss some issues during the conduct of the project with the sponsor, e.g.:
- We received randomization codes for patients who were not in the randomized population.
- There were occasional mismatches between the production and QC code supplied by the sponsor. These were always discussed with the sponsor in a way that would not unblind anyone from the sponsor.
At the third DMC meeting, the DMC requested additional outputs. It was decided that if we requested these to be produced by the sponsor, it may indicate the area of some of the DMC concerns. We requested the raw data from the sponsor and programmed an additional set of five outputs for the subsequent DMC meeting. In preparation for the 4th DMC meeting, when we received DMC programming from the sponsor, we had errors running the graphs. Our solution was to import into our system the actual SAS device driver that the sponsor was using: this solved the problem.
Outputs were put on our secure Windows server which can be accessed remotely over the internet with a login ID and password specific to each individual DMC committee member.