RL27 Missing Data Handling
*Yongman Kim, U.S. Food and Drug Administration 


An important issue in both regulatory agency and sponsor of new drug development is how to handle ‘missing data’ due to dropouts in clinical trials submitted for NDA. In a typical clinical trial assessing outcomes repeatedly over time, dropouts before trial endpoint are inevitable mainly due to intolerable adverse events or treatment failure and when the dropout rate is high, different methods of imputing missing data may yield conflicting results leading to a difficulty in interpretation of trial data and a difficulty in regulatory decision making. Even though many statistical methods were developed in the past, it is believed that there is no definitely best method. Each method is often based on unverifiable assumptions upon which it is shown unbiased and efficient. Current regulatory position in especially chronic pain trials is that sponsor should not use imputation method assigning ‘good’ scores for ‘missing’ data due to dropouts in the primary analysis whatever the dropout reason is. We may share our experience in dealing with the issue and discuss whether there exist methods that satisfy both regulatory agency and sponsor of drug development.