Placebo response here as improvement in clinical response was 46%. 66 patients in high dose means 24 did well (36.4%). If you take out the 9 deaths your way, you get 24/57 or a 42% response, not quite as good as placebo. Give Paion the benefit of the doubt and say all 9 deaths would have been responders as they each got hit by busses on the way home from a successful rehab, you get a 50% response (33/66) vs 46% for placebo. Assume a generous 7/9 were responders and you get 46% vs 46%.