Hi Davisite ….. I think this issue is much more complicated than...

Hi Davisite …..

I think this issue is much more complicated than to simply characterise things along this primary / secondary outcome type of debate.

If we forget about primary / secondary outcomes for a second.

Ask what accounts for the difference between the negative results from the recurrent model – where the treatment was associated with an increased hazard (HR=1.2, p=0.406) and the time to first event (TTFE) models which are showing positive results – although MSB (in typical fashion) don’t report the HRs.

Now this is not a simple question to answer. Some reading about how to interpret inconsistent results between recurrent event models and TTFE models is here:

https://link.springer.com/article/10.1007/s00392-018-1205-7

https://www.ema.europa.eu/en/docume...nterpretable-treatment-effect-measures_en.pdf

I think SI makes a fair point when he says the first goal of any treatment is to keep people alive. The secondary aim is to keep people out of hospital / and reduce disease burden (which is what the recurrent event model picks up on).

The difficulty he has is that in 2016 following extensive negotiations with the FDA the primary outcome was switched from the TTFE approach to the recurrent event approach. This enabled the sample size to be halved from 1165 to 600.

A later further change was to target Class 3 patients (more severe) in lieu of the Class 2 patients (less severe).

MSB also claimed that they would conduct a sister parallel trial with the new trial configuration based on the recurrent event model with a sample of 600.

Now of-course it turns out that it seems it is the Class 2 patients who enjoy better outcomes (the opposite to what MSB found in their post-hoc analysis of the P2 trial) and that it is the time to first event approach (or so it seems) that is more sensitive to treatment effect than the recurrent event model.

But net effect of the switch in models is that the TTFE analyses cannot be considered to be robust because they are based on a much smaller sample size than was originally considered necessary. Which unfortunately is compounded by the patient enrichment strategy to Class 3 patients where the treatment benefit is less.

So imo the TTFE results are methodologically weak which ever you look at it; even before you go to primary / endary outcome arguments.

I suspect this wont be difficult for the FDA. The answer is well why don’t you go away and do the sister trial like you said you were going to and then we will have a look at the results from both trials.

Or alternatively do a P3 confirmatory study based on a time to first event approach like you originally had. But the sample size will be much larger than 1200 because you are now targeting Class 2 patients.