Practical Study Design Methods for Interval-Censored Failure Time Data

• Activity Number: 74
• Type: Contributed
• Date/Time: Sunday, July 29, 2012 : 4:00 PM to 5:50 PM
• Sponsor: Biometrics Section
• Author(s): Daniel Tancredi*+ and Danielle J Harvey and Laurel Beckett
• Companies: University of California at Davis School of Medicine and University of California at Davis and University of California at Davis
• Keywords: interval censoring ; panel data ; accelerated failure time ; Fisher information ; power and sample size calculations ; missing information principle

Abstract

In many longitudinal studies with multiple inspection times the timing of a failure event of interest is impossible or impractical to observe exactly but one can observe the inspection times between which the event occurred. Examples of interval censored failure times include the onset of Alzheimer's disease in multiwave panel studies, the failure time of a hidden component in an industrial reliability study and the timing of HIV transmission in clinical trials. Key study design decisions include the number and timing of inspections and the sample size required to achieve suitably precise estimates of location parameters (e.g. means and regression coefficients) for failure time models. We propose practical methods for making these decisions, including graphical methods for understanding and minimizing information loss due to interval censoring and power and sample size calculation methods for interval-censored accelerated failure time regression models.