In Oncology trials with survival endpoints large imbalances in the dropout rates can lead to a biased interpretation of study results. Such dropouts happen when patients withdraw from the study, often due to underlying side effects of the drug or desire to switch over other cancer therapies. In such cases an often used sensitivity analysis is to consider such dropouts as events to assess their impact on the study results. However such imputations may be too extreme by assuming events would occur at the time of censoring, thus fail to do a realistic impact assessment.
In this talk we present a statistical framework to assess the robustness of the time to event endpoints in oncology clinical trials based on sensitivity analyses under different scenarios of the dropout mechanism as an alternative to the extreme case scenario. A tipping point framework is also proposed to explore the extreme boundary conditions of the dropout mechanism that could lead to the study results to loose statistical significance. We will demonstrate the method based on a case study from an Oncology trial, assessing the inpact of such dropouts on the the robustness of study results .
|