Can Larger Sample Size Provide More Power?

• Activity Number: 338
• Type: Contributed
• Date/Time: Wednesday, August 11, 2004 : 10:30 AM to 12:20 PM
• Sponsor: Biopharmaceutical Section
• Author(s): Mingxiu Hu*+ and Gordon Lan and Rebecca J. Benner
• Companies: Pfizer Global Research & Development and Aventis Pharmaceuticals and Pfizer Inc.
• Keywords: clinical trial design ; amount of information ; sample size ; power ; logrank statistics ; treatment effect

Abstract

In clinical trials, we recruit more patients to achieve higher power under the assumption that all recruited patients have similar treatment effects. However, in practice, we may experience deteriorating treatment effects with the patient population, i.e., patients recruited earlier (probably with severer conditions) may have more improvement than patients recruited later (probably with milder conditions). In this case, recruiting more patients will not necessarily lead to more power. This presentation examines how power changes with sample size under some deteriorating patterns and shows when the power reaches its maximum. We will extend this concept to survival analysis to investigate how logrank statistics changes over time assuming the new therapy truly delay the occurrence of death. Under a survival setting, we follow patients longer to obtain more events to increase power. However, the logrank statistics may not increase in a monotone fashion as events accumulate, which often leads to misinterpretation of long-term treatment effect. Examples will be provided to demonstrate and explain this phenomenon.