|
Activity Number:
|
294
|
|
Type:
|
Invited
|
|
Date/Time:
|
Wednesday, August 14, 2002 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
Biopharmaceutical Section*
|
| Abstract - #300093 |
|
Title:
|
Investigation of Drop Out Rates in Clinical Trials via Data Mining
|
|
Author(s):
|
Richard De Veaux*+ and Robert Small and Rafe Donahue
|
|
Affiliation(s):
|
Williams College and GlaxoSmithKline, Inc. and GlaxoSmithKline, Inc.
|
|
Address:
|
Bronfman Science Center, Williamstown, Massachusetts, 01267, USA
|
|
Keywords:
|
Missing Data ; Imputation ; Empirical Models ; Informative censoring ; Intent to Treat
|
|
Abstract:
|
Drop outs offer special challenges for the correct analysis and interpretation of clinical trials. Although the reasons for these dropouts are usually unknown, their occurrence is unlikely to be random. If the dropouts can be linked to other data collected on the patients, such as baseline or demographic data, the result may be a more appropriate analysis of the data. The resulting model may also be used to mitigate the problem with dropouts in the design of subsequent trials involving similar patients and conditions. Even when the data are missing at random, imputation techniques often require models relating missing values with existing data. Using data mining techniques we investigate the pattern of dropouts in a number of clinical trials with substantial missing data and attempt to form an empirical model for various aspects of the missing data.
|
