My favourite part: "The upper end of the confidence interval for the difference between Afrezza and Novolog was 0.36%, which didn't exceed the 0.40% difference that would have caused the study to fail. Therefore, MannKind can claim Afrezza Dreamboat is statistically non-inferior to Novolog -- hence a "positive" study -- but just barely".

Conviently side stepping the basic law of statistics that probability of one extreme of CI is the same as the other extreme.