If BAVI works, using the average would be very high. The bad data had 1mg patients living nearly twice as long, os combining them and reworking the KM graph, I wouldn't expect much of a change from the original control group.
One question I would have to FTM... if there are 80 patients in the combined new "conrol" group, and the MOS is only slightly better than the original control, and the 3mg MOS at worst stays the same, could the 3mg ever become statistically significant? Basically, does the control group patien number have any affect on statistical signifcants?
JBAIN