I think if you're going to amend/update/change your SAP from the one that you started out with in your protocol, then it has to be done before you have access to unblinded data.
You might wish to change an endpoint. Or change the weighting of alpha spend between two or more endpoints. Or you might wish to change the statistical method or methods that you will employ to analyse whether that endpoint has been 'met', in terms of statistical significance.
Any such amendments should be made before seeing unblinded data.
I doubt at this stage that they will fully change an endpoint.
They would probably needed to have PR'd it, if that was their intention.
They might have, or intend to, change alpha spend arrangements.
They may have already done this perhaps.
They probably will change the statistical method for analysing the data from the one that they originally intended to use.
That's what probably gives this task a 'ginormous' adjective.
This would be, because they have recognized that non-proportional hazards have played a big part in this trial.
And because they need to employ a recognized stat method to calculate and allow for any perceived crossover confoundment of the OS endpoint.
They don't want to use standard Cox regression analysis, with a single derived HR for OS, because that is only appropriate for PH, and does nothing to elucidate long tail survival (which they need to demonstrate is well above anything that TTF can muster..)
And they need to employ a recognized model to further calculate estimated long-term survival beyond the actual study period.
And on PFS, they will surely be looking for other methodology to accurately assess the point of real disease progression, by using other clinical parameters in addition to radiographic images, such as overall patient clinical presentation, and possibly even using a biomarker demonstrating an effective immune response, so to further distinguish between disease progression and treatment response.
So I think they must state in the final SAP what alternative analyses they will employ. And for OS, I guess they will be looking at RMST and multi-point milestone survival analysis.
So all of that before unblinding.
Just my opinion.
