Practical guidance for collecting reproducible time-to-event endpoints in retrospective observational oncology studies

Time-to-event endpoints such as progression-free survival, overall survival, and recurrence-free survival are among the most commonly reported outcomes in cancer research, yet the practical decisions required to collect and structure these data appropriately in retrospective observational studies remain poorly described in the literature. While substantial regulatory and methodological guidance exists for prospective clinical trials, investigators working with data collated retrospectively must navigate consequential decisions about index date selection, event definition, censoring, and follow-up time without equivalent support. Decisions at the data extraction stage can bias results and undermine reproducibility regardless of the rigor of the statistical analysis that follows. In this paper, we provide a practical, decision-focused guide to collating and updating time-to-event endpoints in retrospective oncology studies. Specific topics include how to define the index date, event date, and follow-up end date for different endpoints; what counts as an event and how competing events should be handled; how to manage missing and interval-censored data; how to update outcomes over time; and how to identify potential data quality issues through exploratory analysis and sensitivity testing. We illustrate key concepts using a publicly available lung cancer dataset and a worked example that follows endpoint creation from source records to analysis-ready variables, and we provide a practical checklist that researchers can apply directly to their own work. Our goal is to help epidemiologists, clinicians, and statisticians build time-to-event analyses on a foundation of rigorously collected data, resulting in more reliable and reproducible inference.