Flexible trial design in practice – dropping and adding arms in STAMPEDE: a multi-arm multi-stage randomised controlled trial

Methods The trial recruits men with locally advanced or metastatic prostate cancer starting standard long-term hormone therapy. There are 5 research arms and 1 control arm. The trial has a pilot stage assessing safety and feasibility, 3 intermediate “activity” stages (I-III) where the outcome measure is failure-free survival (FFS) and one final “efficacy” stage (IV) with overall survival as primary outcome measure. At the end of each stage, each research arm is formally compared pairwise to the control arm. Accrual of further patients is discontinued early for any research arm either not showing sufficient evidence of activity or with adverse safety considerations; accrual continues to arms showing activity with acceptable safety. The stopping guideline compares the treatment effect against a pre-defined cut-off value using the hazard ratio when the hazards are proportional and restricted-mean survival time otherwise. This interim hurdle becomes increasingly stringent stage-by-stage. The addition of new research arm(s) can be actively considered when sufficiently interesting agents emerge. New research arms are compared only to contemporaneously-recruited control arm patients using the same intermediate guidelines in a time-delayed manner. The addition of new research arms is independent of any of the original research arms stopping accrual early subject to adequate recruitment to support the overall trial aims.