Evaluating and adjusting for premature censoring of progression-free survival

The intent-to-treat principle, grouping subjects as they were randomized and following all subjects to the endpoint or the end of study, allows valid statistical comparisons. Progression-free survival (PFS) has been used as a decision-making endpoint in oncology. It can be difficult to have a meaningful intent-to-treat analysis of PFS as some studies have extensive loss to follow-up for PFS. In the analysis, subjects lost to follow-up for PFS have their PFS times censored, with the censoring treated as noninformative. We use remaining overall survival to investigate whether premature censoring for PFS is informative and the potential bias in treating such censoring as noninformative.