|
Abstract:
|
Missing data are unavoidable in oncology clinical trial especially in the biomarker analysis and if ignored, missing data may undermine the validity of trial results which has often been overlooked in the medical literature. At missing data imputation, it is necessary to characterize individualize treatment decisions based on patient characteristics, accurate estimation of the interaction between biomarkers and treatment. When auxiliary information including patient characteristics and prognostic variables are available, multiple imputation can be used . In this paper, extensive simulations are conducted to characterize the performance of complete-case analysis and the multiple imputation under the MAR and MNAR assumptions in the context of oncology clinical trials to explore the impact of patient characteristics and prognostic variables. The simulation suggests the importance of incorporating auxiliary information and selecting the right imputation methods when one estimates the treatment and biomarker interaction effects, which will be further illustrative with an example from an Oncology Phase III Study.
|
