|
Activity Number:
|
246
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Tuesday, August 8, 2006 : 8:30 AM to 10:20 AM
|
|
Sponsor:
|
Biopharmaceutical Section
|
| Abstract - #307022 |
|
Title:
|
Imputing Nonignorable Missing Data on Clinical Laboratory Assessments
|
|
Author(s):
|
Kapildeb Sen*+ and Chen-Sheng Lin and Kannan Natarajan and Jun Xing
|
|
Companies:
|
Bristol-Myers Squibb Company and Novartis Pharmaceuticals Corporation and Bristol-Myers Squibb Company
|
|
Address:
|
311 Pennington-Rocky Hill Road, Pennington, NJ, 08543,
|
|
Keywords:
|
non-ignorable missingness ; longitudinal analysis ; laboratory assessments ; single imputation ; multiple imputation
|
|
Abstract:
|
A common problem in clinical trials is the missing data on the direct measures of clinical laboratory assessments (e.g. Glomerular Filtration Rate, low-density lipoprotein cholesterol) at various time points due to improvement or decline in the system function parameter itself that is being measured. In such cases, an approximate calculation of this measure can be made indirectly. We will investigate an imputation method that models the missing data mechanism based on information from the indirect measurements. Using simulations, this method will be compared to single imputation methods such as linear regression and hot-deck imputation and to multiple imputation methods. These methods will be evaluated with respect to the accuracy and precision of the trend estimates in a longitudinal analysis model.
|
