Randomisation works in blocks e.g. blocks of 4 or 6 etc. Lets say in this case randomisation was in blocks of 6, where 3 were randomly given placebo and 3 iPPS. The ratio is 1:1, so yes we should expect 15 were given placebo and 15 iPPS. In terms of pull outs, these numbers will have to be worked in to the sensitivity analyses, although I think in this case they may just be excluded.
It is intriguing that only 30 were enrolled before 008 was halted. This was despite people still trying to register for the trial. Was the trial halted for poor recruitment, and if so, why was it halted at exactly 30 participants? To me, this appears to have been deliberate. At the time 008 was halted, PAR could've had access to results from around half of the participants. As of yesterday, PAR could've had access to day 56 results from all 30 participants.
A valid critique is, if PAR are aware of such results then why haven't they disclosed them to the market? A possible answer is, because analysis of the results are incomplete and the trial is ongoing. We know that PAR will be receiving information about these trials, what we don't know is what type of information.
Looking at the statistical analysis plan for 008, PAR does not pay a statistical penalty for looking at the results and making changes to the observation period, since the trial is exploratory in nature and the changes only apply to the secondary endpoints. These changes are essentially "a free shot on goal".
Also, if we look at the primary endpoint we can see that PAR only needs to observe a significant change in ONE biomarker for the primary endpoint to be achieved, even though ten biomarkers are specified. No type I error rate has been specified in the SAP and a restricted maximum likelihood (REML)-based mixed model for repeated measures (MMRM) approach will be used to assess the change from baseline in biomarkers in synovial fluid at Day 56.
This approach allows each biomarker to be assessed separately for the untreated/treated groups without applying a penalty for the overall number of biomarkers being assessed. In other words, it is similar to placing your chips on red at the roulette table with 10 spins in hand and you only need to land on red once to win. And if we achieve the primary endpoint then all of the secondary endpoints that reach significance are also valid.
Expand
(20min delay)