Abstract
Fu (Citation2018) developed a parametric consonant closed testing procedure for correlated endpoints in a group-sequential design. In their approach, correlations between endpoints are exploited leading to a proportional up-scaling of local significance levels from Holm’s weighting scheme. We challenge the claim that the proposed calculation of the up-scaling factor leads to a test that protects the type I error in the group-sequential setting and outline two alternative approaches which do lead to exact error control.