Analytic Alternatives in Assessing Sustained Clinical Response Based on Repeated Continuous Outcomes: A Clinical Trial Example

• Activity Number: 74
• Type: Contributed
• Date/Time: Sunday, August 7, 2005 : 8:00 PM to 9:50 PM
• Sponsor: Biopharmaceutical Section
• Author(s): Ilya Lipkovich*+ and Shuyi Shen and Craig Mallinckrodt
• Companies: Eli Lilly and Company and Eli Lilly and Company and Eli Lilly and Company
• Keywords: repeated measures ; multiple imputation ; binary outcome ; sensitivity analysis

Abstract

In clinical trials, superiority (noninferiority) of a new treatment over an old treatment or placebo control often is defined in terms of treatment difference in clinical response (e.g., response = 20% improvement in an outcome score from a pretreatment condition) achieved at a predefined time point (t) and sustained until the end of the study. Evaluating treatment differences in sustained response may be a challenging task, especially in the presence of dropouts. While clinicians would be inclined to dichotomize data first and compare treatments at time t (or study endpoint) using a simple univariate test unadjusted for dropouts, many other analytic approaches are available to a clinical statistician. Such alternatives include methods for repeated categorical measures; multiple imputation of underlying, repeated continuous outcome followed by a complete data analysis of categorical outcome at time t; and probability of sustained clinical response at any time point evaluated from a linear mixed model fitted to the underlying continuous outcome. Data from a recent, large multicenter depression trial are used to illustrate the pros and cons of these approaches.