|
Activity Number:
|
343
|
|
Type:
|
Topic Contributed
|
|
Date/Time:
|
Tuesday, August 2, 2016 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
Biopharmaceutical Section
|
|
Abstract #320297
|
View Presentation
|
|
Title:
|
Novel Missing Data Imputation Methods
|
|
Author(s):
|
Peter Mesenbrink*
|
|
Companies:
|
Novartis Pharma
|
|
Keywords:
|
Non-inferiority ;
Missing at random ;
Multiple impuation ;
competing risks
|
|
Abstract:
|
It has been documented in Fleming (2008) and by others that there are many issues that impact the design and analysis of non-inferiority trials. Missing data in the experimental treatment or active control can potentially confound interpretation of results and sometimes bring into question scientific validity of the between-treatment comparisons. Often in immunology, Phase III trials are conducted as non-inferiority trials against standard-of-care treatment. Many of these clinical trials involve a composite binary endpoint where it is possible that the missing data patterns may be different for the components that make up the overall composite endpoint (e.g. missing at random (MAR) may be a reasonable assumption for some endpoints and for other components not-missing at random (NMAR)may need to be assumed. Methods of handling missing data in composite endpoints will be examined for testing the primary hypotheses as well as appropriate sensitivity analyses that should be performed to evaluate the robustness of assumptions made. Traditional methods such as LOCF and non-responder imputation (NRI) will be compared to multiple imputation methods and other likelihood-based approaches.
|
Authors who are presenting talks have a * after their name.
Back to the full JSM 2016 program
|
