Analysis of Clinical Data with Varying Time Intervals

• Activity Number: 339
• Type: Contributed
• Date/Time: Wednesday, August 11, 2004 : 10:30 AM to 12:20 PM
• Sponsor: Biopharmaceutical Section
• Author(s): Cong Chen*+
• Companies: Merck & Co., Inc.
• Keywords: GEE ; longitudinal data ; mixed effect

Abstract

In clinical studies, repeated measures of a response variable collected at pre-specified clinical visits are routinely analyzed as visit-driven longitudinal data. However, such an analytic strategy always has a difficulty in dealing with data taken at a rescheduled or unscheduled visit. In practice, based on a time-window for each visit, these data are often either excluded from the analysis or carried (forward or backward) to a pre-specified visit in the same neighborhood. Examples in textbooks on longitudinal analysis are often based on such "cleaned" data. Both approaches have problems. The exclusion of otherwise admissible data leads to the loss of power for the detection of treatment effect; valuable time information on data measurements is lost when data is carried to a different time point. More sophisticated approach may consider it a complicated missing data problem, and deal with it accordingly. We instead propose a simple analytic strategy that is supported by standard procedures in SAS, and consistent with common practice of clinical data reporting. It is easily acceptable to the medical community, and readily accessible to statistical practioners.