Skip to main content

Table 2 Interim stopping probability of superiority or futility with long term survival and delayed clinical effect

From: Statistical issues and challenges in immuno-oncology

Interim sample size 520 540 480 500
Number of events 256 256 256 256
P E T a (superiority) 0.25 0.25 0.06 0.06
PET a (futility) 0.01 0.01 0.08 0.08
  1. Under the same design assumption, we incorporated an interim analysis at the information fraction of 50%. Based on the same accrual rate of 20 patients per month, the total number of randomized patients for the four models ranged from 480 to 540 when 256 events were reached. When a superiority interim analysis using O’Brien-Fleming boundaries was built into the study design, the probability of early termination when the agent was active (PET a ), i.e., true positive rate, at the interim analysis under exponential decay was 0.25. The PET a reduced to 0.06 when the delayed clinical effect of 3 months was present. If a futility interim analysis was incorporated, the PET a (false negative rate) increased from 0.01 to 0.08.