BACKGROUND: Non-interventional studies (NIS) using prospective RWD collection have gained prominence as a primary data source, aiming to overcome limitations of secondary data sources. Patients are typically enrolled based on cancer diagnosis, start a treatment (Tx), etc. and by signing informed consent forms (ICF) in routine clinical practice. The sequence of eligibility for data collection and the requirement for patients to be alive at the time of ICF signature may introduce bias in overall survival (OS) analysis.

OBJECTIVES: The study aims to quantify potential bias introduced by the time gap from eligibility to ICF signature in different scenarios and propose a robust analysis to address this bias.

METHODS: Two situations were considered for oncologic patient data collection: (a) ICF signed then start Tx, (b) start Tx then ICF signed. Two scenarios were simulated: (1) ICF signature independent of Tx outcome, (2) ICF signature associated with Tx outcome. Median OS (mOS) was estimated using conventional Kaplan-Meier methods (KM) and multi-state modeling (MSM) in each scenario. The sensitivity analysis was conducted on group (a) patients.

RESULTS: For scenarios (1) and (2), data for 1000 patients were simulated, with median follow-up times of 41.6 and 35.2 months from Tx, and 25.5% and 33.3% of patients died during follow-up. The median time from Tx to ICF signature in group (b) patients was 11.2 and 17.6 months, respectively. In scenario (1), both KM and MSM methods showed non-significant difference of mOS with true mOS in main and sensitivity analysis, while in scenario (2), KM significantly underestimated the mOS in sensitivity analysis, analyzing only patients who started Tx after the ICF.

CONCLUSIONS: Using all patient data and applying MSM to estimate the OS is superior to the current K-M approach in prospective RWD analysis, particularly if signing the ICF is associated with the Tx outcome.