Statistical analyses Hierarchical testing procedure Hierarchical testing was used to control the type I error rate (false positive results) for the confirmatory secondary endpoints. This scheme included a fixed ordering of the confirmatory secondary endpoints (Number of confirmed hypoglycemic episodes; Change in FPG after 52 weeks of treatment; Withinsubject variability in pre-breakfast PG; and Responders for A1C <7% at end of trial without confirmed hypoglycemia who had been exposed to trial drug for at least 12 treatment weeks) to be tested for the superiority of insulin degludec to insulin glargine, if insulin degludec was shown to be non-inferior to insulin glargine in changing A1C from baseline to a non-inferiority limit of 0.4%. The non-inferiority criteria was determined as per recommendation in the 2008 FDA guidance (22) which states, “Typically, we accept a noninferiority margin of 0.3 or 0.4 HbA1c percentage units provided this is no greater than a suitably conservative estimate of the magnitude of the treatment effect of the active control in previous placebo-controlled trials.” Superiority could only be confirmed for endpoints when all previous hypotheses were confirmed.