The graphical testing system used to preserve the overall study-wise type 1 error is depicted in Fig. S2. First, the dual primary endpoints are tested, with the alpha split in 1% for weight change and 4% for change in KCCQ-CSS. The tests follow the weighted Holm-Bonferroni procedure (with weight one), such that if one of the two endpoints is superior then the full alpha can be recycled for the other endpoint and hence the remaining primary endpoint will be tested at the 5% significance level (two-sided).