Anavex Life Sciences Corp (AVXL)

[Joseph_K]

There are ways to control for multiplicity, but I don't think the 0.025s are one of the ways. It helps the trial results overall, but the multiplicity problem resides within the multiple alternatives specifically for ADL, where Anavex almost certainly relied on a not-first-listed statistical approach.*

Pre-specifying does not eliminate multiplicity. I read about controlling for multiplicity, and it's over my head at this point (ways you calculate some variable alpha) but it's clear Anavex would have to account for the multiplicity with an approach that takes into account each and all of the alternative statistical tests it pre-specified for ADL, and I don't think it would concern itself in this instance with the other endpoints.

Besides, I think this quote, from the message you were replying to, contradicts the idea that problems of multiplicity and reliance on post-hoc analysis don't arise where there is pre-specification:

Therefore, this approach [the use of multiple, alternative statistical models] to pre-specification is ineffective, as it still allows investigators to analyse the data in many different ways before deciding on a final approach.

If Anavex listed 10 alternative statistical tests for ADL in the SAP, and then chose one of them (other than the first-listed) to rely on, that choice was the result of a post-hoc analysis.


* See my reply to Hoskuld that I posted about half an hour ago.