Fig. 3
From: When to keep it simple – adaptive designs are not always useful
Properties of the TAILoR trial assuming different delay lengths in the endpoint. The actual endpoint was assessed 24 weeks after randomisation. Plotted trial properties were simulated using 10,000 simulation replicates for each potential endpoint delay length between 1 and 48 weeks. a Expected sample size averaged over 10,000 simulation replicates for each delay length. Blue dashed line represents the properties of the trial under the null scenario, when all experimental doses have the same efficacy as control; red solid line represents the properties under the alternative scenario, when one dose has a standardised effect of 0.545 and the others have the same efficacy as control. b Proportion of patients who were allocated to the effective dose in the alternative scenario (as in a, one dose had a standardised effect of 0.545 and the others 0)