We consider general adaptive group sequential designs for trials comparing three independent arms, say test group T, reference group R, and placebo group P. The goal of the trial is to establish noninferiority of T compared to R while T is also superior to P. Thus, we have a hierarchical testing procedure of a priori ordered hypotheses, that is, first we should establish T > P, and then T > R - Delta, where Delta is a given nonnegative noninferiority margin. We derive nested confidence intervals on the trial parameters for normally distributed response variables with unknown variances in each interim analysis keeping the predefined confidence coefficient. In each interim analysis, the sample size for the next part of the trial can be determined in a completely adaptive way based on all the unblinded data collected so far. Exact formulas for sample size updating are provided. Moreover, in each interim analysis, it is possible to switch, in the sample size planning, from showing noninferiority of T compared to R to showing superiority of T, that is, we set Delta = 0 in the second testing problem.