TTEDesigner User’s Manual John D. Cook Department of Biostatistics, Box 447 The University of Texas, M. D. Anderson Cancer Center 1515 Holcombe Blvd., Houston, Texas 77030, USA cook@mdanderson.org September 3, 2007 1 Introduction The natural end point of clinical trials is often the time to an event. The event under consideration may be undesirable, such as disease progression or death, in which case the purpose of a trial would be to maximize the time to this event. Alternatively, the event being studied may be desirable, such as time to transplant engraftment, in which case the trial would seek to minimize the time to the event. Statisticians often dichotomize time-to-event outcomes by picking some time window and monitoring only whether the event occurred during the window. For example, rather than monitoring the safety of a trial based on survival time, one might consider only whether patients survived six months. Clearly there are several drawbacks to this approach. First, the choice of observation window is arbitrary. Why pick six months rather than, say, eight months? Second, information is thrown away. The dichotomized outcome does not distinguish a patient who died immediate after treatment from a patient who survived five months. Finally, partial information is not used. In a dichotomized model, if we know a patient has survived five months, we have no useable information. With a time-to-event model, every day the patient survives provides a little more information. This software will guide the user in designing a safety monitoring rule for a single-arm trial with a time-to-event endpoint using a Bayesian statistical model. Until this software was written, such designs required trial and error sim- ulation to achieve the operating characteristics desired by researchers. This software inverts the design process by asking the user for desired operating characteristics and searching for the parameters that best satisfy these desires. 1