> EMIS – Can you explain what a "non-standard binary-responder metric" is?<
They called patients with an A1c reduction >=1.1% responders and called everyone else non-responders. Then they tallied the % of responders in each trial arm, which gave a p-value of 0.037 for the highest-dose insulin arm vs the control arm.
However, the above is a post hoc metric designed to allow EMIS to say that something had a p-value <0.05. By sheer chance, almost any trial of this size this will have some metric in some subgroup that looks good.
Market Data 
Markets 
AI
