Abstract:
|
In randomized clinical trials patients may switch to a rescue medication or to another treatment than the initial randomized treatment due to ethical reasons if pre-defined criteria for the switch are fulfilled. As a consequence patients are withdrawn at the timepoint of their switch from the study, i.e. before the planned assessment of the primary endpoint. Information after this timepoint is not available in most cases. A variety of imputation approaches is possible, but different approaches may lead to discrepant results and thus to ambiguous decisions. Measurement of treatment effect is impacted by the number and timing of such early withdrawals. Appropriate estimands and imputation approaches (e.g. multiple imputation) will be defined and compared for the example in which the patients themselves decide how much rescue medication they need to control the symptoms of their disease.
|