A summary measure that depends on the censoring
distribution is the biggest problem
•In
a survival study, we typically have a different censoring
distribution at successive analyses
•Hence,
different summary measures are being tested at different
analyses