Golf,

Agreed.  See my speculations below.  It is important to remember that topline was not just about the %RRR that was achieved, but when it was achieved.

Iconoclast,

I am interested in how you modeled time-to-event in your analysis.  We know that RVX entered the trial believing that apabetalone worked better with Crestor than with Lipitor and that they expected an overall RRR of about 30%.  This could be achieved if they expected a MACE reduction of 50% in the Crestor group and of 10% in the Lipitor group [i.e. for the treatment group (0.5 x 0.5) + (0.5 x 0.1) =0.3.] They may or may not have achieved this.  And, of course, it could also have been achieved if the MACE reduction in both groups was equal at 30% and with a myriad of other combinations (see Golf's post for example).

However, the primary endpoint was actually to achieve a significant increase in time-to-first-MACE.  This involved not only the percentage of apabetalone-treated patients that avoided a first MACE event, but the temporal distribution of those that did have a MACE in both the control and treatment groups. For a significantly increased time-to-event, it is important for apabetalone to have prevented MACE events relatively early in the trial, giving treated patients who avoided a MACE (for a while or completely) lots of extra time before the end of trial.  RVX must have modeled a distribution of MACE events in order to estimate the power they had for detecting increased time-to-MACE, but have not revealed their expected distributions of MACE events in the control and apabetalone groups.  Have you tried this?  And have you used one of the free Kaplan-Meier programs to take a look at the modeled data?  This would be very informative!

We know that it took longer than expected to meet the 250 MACE target and can now reasonably conclude that MACE events took longer than expected to occur in both the treated and control patients.  We may have missed topline because the extra time that treated patients who avoided a MACE got before the end of trial was just too short (and if the trial had gone for another 6 months or so, giving them more MACE-free time, we would have hit topline as well as the overall 30% RRR). 

I would be interesting to see if your models have addressed the time parameter, since it may be the factor that was not predicted well in the pre-trial modeling.

Jupe