Figure 2.

Decision Tree for Statistical Analyses. To control for type-I effect, assessment of each outcome used a hierarchical gatekeeping strategy as illustrated. First, the treatment-by-visit interaction term was tested; if significant, between-group differences at each time point were tested at α=0·05. If not significant, the treatment main effect was tested next; if significant, between-group differences at each time point were tested at α=0·05. If neither of the treatment-by-visit interaction or the treatment main effect were significant, Bonferroni correction was applied at each time point, with p values adjusted by multiplying the nominal p value by the number of tests (truncated at 1·0).