SRPT - Another cut at the p value. Earlier I calculated the p value based upon loss of ambulation data (not a good p value), but with more time I tried to disentangle the individual patient data and run a p value with the entire set of patients.
However, before getting to that, some reminders:
a) small single arm trials with a small number of centers almost always look a lot better than historical. Often clinically meaningfully different - an illusion presumably caused by better SOC, better patient selection (and these were healthier patients than typical for their 6 MWD - e.g. see FEV/FVC data), ... . This is a bias that no statistics can remove.
b) Past behavior from Sarepta is that they appear to have used 'Difference of Means' to calculate p values - and it should be expected that that calculation would be very misleading for data like that in DMD. (Note that Sarepta is hardly alone in their apparent touting of 'Dif of mean' statistics when they know perfectly well the FDA will calculate p using a very different method).
Given #b and my p value calculation from the non-ambulatory patients I expected to find a p value not as good as they claimed. However their p value looks valid (at least as far as it is possible to tell given the difficulties of reading their individual patient data).
Why is the p value so much better than the non-amb p value and previous p values?
1) It is better than the non-amb p value because even comparing the still-amb patients the Etep patients didn't degrade anywhere near as much.
2) It is better than my past calcs because, of course, previously I was using Mercuri (the dataset that Sarepta themselves had previously chosen to compare to) as the comparator and now their data/claim is that P51 patients do *much* worse than overall Mercuri. Credible? Depends upon how well those patients match the Etep patients (baseline, SOC, ... this is a component of the post hoc issue mentioned previously in these threads)
With regards to approval... previously I would have said little chance. But if the P51 comparison set withstands scrutiny by the FDA (e.g. does it match Biomarin's natural history data and do the historical patients match other Etep baseline parameters), there are no SAE surprises and the FDA chooses to look the other way on problem #a above then it may be approvable.
That said, I am willing to double down on my earlier 'bet' at 1:1 odds that they are not stat sig on the ITT primary endpoint of any of their on-going phase 3s and further that one or more trends the wrong way. I.e. they cheaped out on their trials and the lessons learned that were evident today have not made it into their trials.
I am also willing to bet that in a head to head trial with Dris, Dris would have meaningfully more efficacy.
AI
