Abstract:
|
Treatment-emergent adverse events (TEAE) are usually reported by incidence and rarely by duration; however, the knowledge of the duration is crucial sometimes. When a large percentage of event-stopping date are not available given the length of observation, the non-parametric approach by Kaplan-Meier method will not be able to provide the estimated median duration of the event. In this talk, a parametric approach is demonstrated to overcome the problem, which assumes an underlying distribution and estimates the parameters via a MLE method, using all the data including the censored. This approach was successfully used in characterizing the feature of hot flashes, a TEAE observed in the clinical trials for raloxifene, an approved agent for the treatment and prevention of postmenopausal osteoporosis. In the analysis, an exponential distribution was assumed, and the assumption was checked by testing the constant hazard function against the alternative hypotheses that specify monotone hazard functions. The results eased the concern for the slightly higher occurrence in raloxifene-treated patients: The median durations were very similar between raloxifene and placebo treatment groups.
|