A modified Williams-type approach is proposed, which controls the family-wise error rate globally over both correlated endpoints and over dose comparisons. This order-restricted procedure is formulated for ratio-to-control comparisons, where the comparison of difference-to-control is included directly. Endpoint-specific variances are allowed, and even a modification for dose-group-specific variance heterogeneity is proposed. Therefore, this approach can be used simply for evaluation of real data by means of the R package SimComp. Both multiplicity-adjusted p-values and simultaneous confidence intervals are provided, whereas confidence intervals are clearly recommended to allow the interpretation for clinical relevance.