Fig 1.
Schematic of a two-stage adaptive trial design with delayed response using the independent increments assumption.
Fig 2.
Schematic of a two-stage adaptive trial design with delayed response using patient-wise separation.
Table 1.
Worst case type I error for various choices of weights and information fractions.
Nominal level α = 0.025 one-sided.
Fig 3.
Expected total number of events as a function of time based on exponential survival with hazard rates λC = 0.05 and λE = 0.035.
Slow recruitment: 8 patients per month for a maximum of 60 months. Fast recruitment: 50 patients per month for a maximum of 18 months. Vertical lines are at Tint, Tend and Tmax.
Fig 4.
Difference between the noncentrality parameters of the adaptive test statistics Z(T*) and Z(Tend) as a function of the time extension T* − Tend ∈ [0, Tmax − Tend].
Horizontal lines are drawn at k* − Φ−1(0.975), where k* denotes the cut-off value of the full-data guaranteed level-α test, and Φ denotes the standard normal distribution function.
Table 2.
Trade-off involved in choosing between methods when extending the follow-up period of a survival trial.
Methods considered: (A), data is combined assuming independent stage-wise increments; (B), patient-wise separation with pre-fixed end of first-stage follow-up; (C), naive patient-wise separation without pre-fixed end of first-stage follow-up; and (D), patient-wise separation using the full-data guaranteed level-α test.