Figure 1: Drop-the-loser design (with intervening response-adaptive randomization)