Analysis of Time-to-Event Data with Categories of Response

• Activity Number: 515
• Type: Contributed
• Date/Time: Wednesday, August 1, 2012 : 10:30 AM to 12:20 PM
• Sponsor: Biopharmaceutical Section
• Author(s): Beiying Ding*+ and Chunlei Ke
• Companies: Amgen, Inc. and Amgen, Inc.
• Keywords: time-to-event ; proportional odds ; severity

Abstract

Time-to-event endpoints are frequently encountered in clinical trial data. When event times are observed, additional information concerning the type or severity of event is often recorded. For example, in patients with non-metastatic cancer, the first bone metastasis at time of detection could be asymptomatic (i.e. bone lesion only present in bone scan) or also accompanied by symptoms such as bone pain. When evaluating different treatment options, this level of information may prove to be valuable especially if treatment is more effective in preventing event of certain level(s) of severity that is clinically more meaningful. Different analysis strategies are explored in this context including time-varying proportional odds model, continuation ratio model and competing risk models with ordinal response categories. Pros and cons of different approaches under various assumptions of the treatment effect pattern related to the severity of the event as well as potential generalizations will also be discussed.