Fig 4. Illustration of a frequentist multi-arm multi-stage design with 3 treatment arms and 5 planned (interim) analyses.

Standardized test statistic Z(t) as a function of time t for 3 different treatment arms together with upper (efficacy) and lower (futility) stopping boundaries. In this example the trial would stop at the fourth interim analysis with one treatment crossing the upper boundary (square). One treatment (triangle) would be dropped at the second interim analysis, because it crosses the lower boundary. The third treatment (circle) would be declared ineffective at the time the trial stops.