This talk considers the use of multiple imputation in a noninferiority clinical trial using the method of fully conditional specification (FCS). Two major considerations are inclusion of covariates in the models for imputation and the imputation size. We study the effect of the imputation size in a simulation study, which considers the following experimental factors: sample size, proportion of missingness, and number/type of covariates for imputation. For the study design, we assume two equal-sized arms, an effect size of zero, and a standard deviation of 2.5. ANCOVA is used as the substantive model, and the FCS method of regression with predictive mean matching for continuous covariates is adopted in the imputation model. Assessment over the experimental parameters includes computational burden as measured by time for imputation runs and a variety of statistical operating characteristics. such as bias and variance of the estimate, stability with respect to random seed, and power to reject the inferiority null hypothesis (based on the lower confidence bound) for a variety of noninferiority margins. Finally, we present a motivating example with an optimal imputation size of 50.