As others have mentioned, it's important to note that this is not a trial of ECMO vs no ECMO, but of immediate ECMO vs usual care including ECMO if needed, and that 39% of the control arm wound up receiving ECMO, which could have the effect of reducing apparent treatment benefit. Also, the 28% reduction in the primary endpoint could well reach statistical significance if the trial numbers were larger. Finally, this was for patients with cardiogenic shock, most likely predominantly heart attack patients. They likely represent a smallish portion of the patients who might be considered for ECMO, so one can't generalize to all scenarios.