Rank-Based Adjustment of Baseline Imbalance in Assessing BOTOX® Dose-Dependent Response of Reducing Urinary Urge Incontinence in a Phase II Trial

• Activity Number: 233
• Type: Contributed
• Date/Time: Monday, August 3, 2009 : 2:00 PM to 3:50 PM
• Sponsor: Biopharmaceutical Section
• Author(s): Jihao Zhou*+ and Thomas Lin and Corneliar Haag-Molkenteller and Catherine Thompson and Grace Daniell and Anand Patel and Brenda Jenkins
• Companies: Allergan Pharmaceuticals, Inc. and Allergan Pharmaceuticals, Inc. and Allergan Pharmaceuticals, Inc. and Allergan Pharmaceuticals, Inc. and Allergan Pharmaceuticals, Inc. and Allergan Pharmaceuticals, Inc. and Allergan Pharmaceuticals, Inc.
• Keywords: analysis of covariance ; baseline adjustment ; rank analysis of covariance ; semi-parametric regression ; nonparametric regression ; clinical trial

Abstract

ANCOVA is commonly used for baseline adjustment. It may improve precision and to mitigate bias of treatment effect estimation conditionally on the observed baseline difference. We analyzed the urinary incontinence data in a double-blind, 6-dose group trial of BOTOX in idiopathic overactive bladder patients. A simple analysis of change from baseline and the ANCOVA initially didn't show a conventional dose response. The data showed baseline imbalance and non-normality. Therefore we adopted a rank-based semiparametric approach. Mean rank regression residuals revealed a response trend, confirmed by the Jonckheere-Terspstra test. The Mantel-Haeszel score statistic was used for testing between-dose difference. Comparisons with GEE-based Poisson/negative binomial regressions and generalized Wilcoxon test were made. Only the rank-based approach fully detects the delicate ordered dose response.