Abstract:
|
To meet rigorous regulatory requirements, design and analysis of pivotal clinical trials, including missing data handling, must be pre-specified, and adhere to effectiveness and Intent-To-Treat (ITT) principles. Understanding the influence of study characteristics, such as the extent and nature of missing data, is critical to ensure statistical rigor and trial success. The impact of very high (>50%), differential, informative missing data on accuracy, precision, type I error and power of the primary endpoint analysis was examined in various fixed and adaptive design and analysis scenarios, using single or repeated measure models, as well as imputation or likelihood-based missing data handling. Accuracy, precision, type I error and power varied considerably with design, analysis and missing data handling choices, as well as with effect size, variance, sample size, and the proportion of differential and informative missing data, as illustrated by design simulation and data analysis examples. Simulation studies generated useful insights into these associations and revealed the optimal design and analysis methods for each specific situation.
|