Dealing with Missing Data in Clinical Trials: A CSP Perspective

• Activity Number: 605
• Type: Topic Contributed
• Date/Time: Thursday, August 2, 2012 : 8:30 AM to 10:20 AM
• Sponsor: Section on Government Statistics
• Author(s): Kousick Biswas*+
• Companies: CSPCC
• Keywords: Missing Data

Abstract

Missing values, a very common occurrence in clinical trials, have been an analysts' nightmare. In modern clinical trials, it is a common practice to deal with missing values after its occurrence and it includes quantification, proper characterization, identification of patterns, selection of proper analytic methods, and finally, demonstration of the impact the missing values have on the primary findings. Depending on the type, handling of the missing values statistically can be very challenging and the sophisticated analytic methods that are designed to deal with non-ignorable missing data are not only difficult to implement but also difficult to interpret the results. The other problem with missing values is the degree of missingness and particularly when the amount of missing outcome values is very high (50% plus). In these situations, the reliability of trial findings is in question irrespective of the sophistication level of missing data handling procedures. This presentation is geared towards introducing discussion points about potential strategies to deal with missing data in clinical trials conducted by Cooperative Studies Program (CSP).