|
Abstract:
|
Late onset toxicities are a serious concern in phase I trials in oncology. Since most toxicities occur soon after the start of therapy, conventional dose-finding methods rely on a binary indicator of toxicity occurring within a relatively short initial time period. If an agent causes late onset toxicities, such methods may allow an undesirably large number of patients to be treated at toxic doses before any toxicities are observed. We use a Bayesian dose-finding method based on toxicity time data. We use a set of rules, based on predictive probabilities, which may suspend accrual if the risk of toxicity is too high. If additional follow up data later reduce the predicted risk of toxicity to an acceptable level, accrual is re-started, and the process is repeated several times. An extensive simulation study shows that our proposed method provides a measure of safety over other methods.
|
