Using Correlation Patterns of Study Findings to Assess Data Quality in Clinical Trials

• Activity Number: 85
• Type: Contributed
• Date/Time: Sunday, August 9, 2015 : 4:00 PM to 5:50 PM
• Sponsor: Biopharmaceutical Section
• Author(s): Richard Zink*
• Companies: SAS Institute
• Keywords: Data Quality ; Good Clinical Practice ; Graphical Approaches ; Misconduct ; Risk-Based Monitoring ; Visualization

Abstract

Recent U.S. and E.U. guidance on risk-based approaches in clinical trial oversight has generated renewed interest in statistical and graphical approaches to identify anomalous data. In a classic example, Bailey (1991) describes how unexpected or implausible relationships among several variables helped to identify an instance of misconduct in a multicenter animal study. In order to identify quality issues or potential misconduct, I summarize statistical and graphical approaches to screen the study database for unusual correlation patterns between the study procedures performed at each clinical site. Methods to account for multiplicity and the impact of data standards will be addressed. An analysis of a multicenter clinical trial is presented to illustrate methods and available software.