Overall survival (OS) is the gold standard endpoint in oncology clinical trials. Missing data due to early dropout is an important issue. In the spirit of minimizing missing data, public records are often consulted for patient survival status, but usually death events are available. We will demonstrate the challenges of incorporating the public record data in the analysis. While an open-label trial is not uncommon, there are early dropouts mainly in the control arm due to randomization disappointment. We will discuss sensitivity analysis methods for randomization disappointment, particularly issues with conservative or worst cases analysis. We will also present a multiple imputation based method to deal with missing data for OS.