The sample size was based around the idea that you would need a...

The sample size was based around the idea that you would need a very large positive effect size to justify the invasive treatment.

But this has an unintended consequence – it restricts analysis to less powerful methods – a one way ANOVA.

So I can see the two pronged strategy being proposed here.

The clinical experts need to take another look at just where they have set the MCID in terms of the risk / benefit equation. Perhaps it was too conservative – the surgery might now been seen as less risky than anticipated justifying a smaller benefit.

Meanwhile the statistician has to be talked into applying post-hoc a more powerful test such as a ANCOVA which will take into account all data time points. Even though there will good statistical reasons for not doing this. But at the end of the day this trial is really not proving anything in a confirmatory sense – it is just guiding what to do next.

Its hardly pretty … and with the benefit of hindsight a “weaker” than hoped for result should have been planned for and a series of more exploratory analyses pre-specified.

Particularly when the 26 week result didn’t pan out and the company was hoping for a “sleeper” effect. This was always a long shot … and Plan B should have been in place.

Anyway that’s my 2c worth from a quick flick through.