Given that nobody I know of has references to actuall data from the P1/2s (just summary), we have no clue on what this is the earlier trials.
I guess if clark want'd to waste some time he could he could make a much better comment on this.
Some random comments:
a) My guess is that although the ph ii was probably not using a bar to calculate p value (they were probably, as you suggest, calculating p value using std dev), the ph iii probably is using a bar (my guess is 1.01g/DL of hemoglobin improvement). The reason is that the FDA dislikes p values using standard deviations because it assumes a curve shape.
b) I could back into standard deviations, then assume a curve shape and then calculate powering. But too time consuming and a bunch of assumptions (I don't like assumptions much more than the FDA). Better to just ask the company (which is more transparent than 99% of companies) what they saw in the phase iib using the same endpoint - e.g. 10% of placebo patients saw 1.01 improvement and 80% of treated... .
If someone does ask that question they should also ask about enrollment criteria differences between the ph iii and the ph iib.
Clark
Market Data 
Markets 
AI
