Direct likelihood analysis versus simple forms of imputation for missing data in randomized clinical trials | Synapse