Biotech Values

[iwfal]

Why couldn’t (shouldn’t) a futility test be based on a linear combination of: i) the a posteriori power of the trial based on the actual data to date and the original powering assumption for the data yet to be reported; and ii) the a posteriori power of the trial based on the actual data to date and the assumption that the data yet to be reported are consistent with the data to date?

Sure - some Baysean weighting of apriori and experienced would probably be optimal in accuracy. But note that I would be very surprised if, in aggregate, the accuracy of the experienced efficacy weren't 2x the accuracy of the apriori estimate.