Some Design Considerations for Time-to-Event Trials in the Context of Comparative Efficacy Needs for Payer Evidence Groups

• Activity Number: 218
• Type: Topic Contributed
• Date/Time: Monday, August 4, 2014 : 2:00 PM to 3:50 PM
• Sponsor: Social Statistics Section
• Author(s): Stephen Lane*+
• Companies: GlaxoSmithKline

Abstract

Sample sizes for Phase III studies with time-to-event endpoints as primary are derived using a targeted hazard ratio along with the desired error probability constraints, with trial reporting then driven by attainment of a specific number of events. While the hazard ratio is the appropriate measure of overall treatment effect, estimates of median times are equally important for conveying benefit to the majority of physicians and payer evidence groups. Reporting trial results based only on the required number of events may result in median estimates which are highly affected by the amount and timing of censoring, especially in areas where new agents are providing dramatic benefit over previous standards of care. We will present both a real-life example from a recent Phase III oncology trial and results of simulations to illustrate the impact of enrollment patterns and analysis timing on median estimation, with the intent to illustrate the need for good statistical counseling when designing trials in the face of rapidly changing treatment landscapes where the facilitation of comparative efficacy among all possible treatments is of increasing importance.