|
Abstract:
|
In many early-phase clinical trials, rather than characterizing patient outcome as a binary variable, it may be more natural to construct decision rules based on time-to-event variables and event rates. This approach often makes better use of available information and avoids logistical problems and loss of information due to replacing an event time by an indicator of whether the event occurred within a fixed observation period. Monitoring event times may involve various complications, however. We describe Bayesian methods to deal with certain complications that may arise when monitoring event times, including interval censoring and effects of intermediate events on survival time. We present several designs, each based on the times to a severe adverse event, disease progression, and death. Simulation results for each design are presented.
|
