From todays AFR, page-11

joe the Primary Outcome of SARAH is very clear cut. Which to my mind is a good thing. Let's look at the trial design:

"The trial have 80% power to detect a clinically meaningful increase in median survival time of 4 months between sorafenib (expected median survival time 10.7 months) and Y90 RADIOEMBOLIZATION (expected median survival time 15 months) with a two-tailed type I error risk of 5%"

As I read this, I interpret the design to be that the trial will reach its Primary Outcome if, with 80% certainty, it can be shown that SIR-Spheres has a improvement in OS of 4 months over sorafenib.

So, as I see it, an improvement in OS of 3 months would be a failure.

Not (and lets be very specific about this) a failure of SIR-Spheres, but a failure of the particular test that SIR-Spheres was set against sorafenib. Further sub-analysis may show that SIR-Spheres is still superior ( in some instances etc) to sorafenib, or indeed that a combination of the two is better still.

Given Gilman's track record, I expect him to report that the trial "failed its primary objective". Various secondary objectives may have been achieved.