Finding unrecognized information in overactive bladder clinical trial data: A new approach to understanding placebo and treatment effects§

Authors


  • Disclosure information: Norman Zinner has received research grants from Astellas Pharma, is on the speaker's bureau and has consulted for Astellas Pharma, Ferring, and Watson, and has participated in trials for Astellas Pharma, Ferring, Watson, Warner Chilcott, and Nymox. Larry Ammann does not have any disclosure to declare. Stan Bukofzer and Gabriel Haas are employees of Astellas Scientific and Medical Affairs, Weizhong He is an employee of Astellas Pharma Global Development, Inc., and Stephen Janning is an employee of GlaxoSmithKline.

  • Author roles: Larry Ammann: trial design, organization, statistical analysis, and execution of the study, drafting of publication, and final approval of publication. Norman Zinner: concept, trial design, analysis and interpretation of data, drafting and revision of publication, and final approval of publication, Stan Bukofzer, Gabriel Haas, Weizhong He, and Stephen Janning: trial design, analysis and interpretation of data, drafting and revision of publication, and final approval of publication.

  • §

    Karl-Erik Andersson led the peer-review process as the Associate Editor responsible for the paper.

Abstract

Aims

To identify combinations of variables among overactive bladder (OAB) clinical trial subjects that allow prediction of those who are more—or less—likely to respond strongly to placebo, or to medication.

Methods

Data from two Phase IIIb clinical trials of solifenacin in OAB were combined. Predictive models for placebo and treatment responses were constructed using baseline variables including individual items from the OAB questionnaire. These models were reduced to an essential subset of predictor variables. Two outcome measures are reported: urgency and incontinence.

Results

In placebo subjects, 14 selected variables permitted distinction between those who responded with significant reductions in urgency and those who did not. A subset of nine variables in treated subjects permitted distinction between those more—or less—likely to respond to medication. Data for urgency were combined from both placebo and actively treated subjects to identify those who had one of the previously identified clusters of variables. It was possible to predict, among all subjects, who would be likely to experience a strong placebo or active treatment response and who would not. This process was also applied to incontinence data.

Conclusions

We have developed a new process to help understand placebo and treatment responses and their relationships to baseline conditions. The effectiveness of these methods was indicated using data from two solifenacin clinical trials and would benefit from future validation using other data sets. Methods used here are suitable for predicting the placebo effect in other clinical trials. Neurourol. Urodynam. 32: 308–313, 2013. © 2012 Wiley Periodicals, Inc.

Ancillary