|
Abstract:
|
In clinical studies, data may be missing for a number of reasons, such as after a patient discontinues treatment. In studies of glucose-lowering agents for the treatment of diabetes, missing data are also introduced by the use of rescue therapy when a patient's glycemic parameters become too elevated. Often, data gathered on a patient after initiation of rescue therapy are censored and are consequently missing, to avoid the confounding effect of the rescue medication on the efficacy analyses. In this presentation, missing data mechanisms are summarized for diabetes studies. Different statistical approaches to handling missing data, including the last observation carried forward (LOCF) and repeated measures ANCOVA, are compared with the constrained longitudinal data analysis (cLDA) approach. The bias and precision of the different techniques will be compared.
|
