OBJECTIVES: Budget impact analysis (BIA) guidelines do not report specific methods for the inclusion of time-to-event (TTE) outcomes and mean estimates are commonly used. Our objective was to compare methods in the context of BIA oncology modelling, and to identify when it is appropriate to use simple or more complex TTE incorporation methods.

METHODS: An incident cohort BIA was developed over 5 years, including pre- and post-progression treatment acquisition costs, disease management costs, and end-of-life costs. Three TTE methods were compared, whereby pre- and post-progression times were incorporated based on crude mean estimates (overall mean method; OMM), mean estimates adjusted on the year of entry and time horizon (incremental mean method; IMM), and means based on the outcomes calculated from a three-state partitioned survival model (yearly health state occupancy method; HSOM). Median pre-progression survival (PFS) and overall survival (OS) times were extrapolated using exponential distributions to derive mean estimates. Mto assess the yearly and cumulative difference of each cost type between the three TTE methods.

RESULTS: The OMM was found to overestimate cumulative first line acquisition costs by more than 5% compared to the HSOM when the median PFS was at least 6 months, and by up to +17% when PFS was . Post-progression acquisition costs were also overestimated compared to HSOM, from +9% to +32% across various PFS and OS times. Reducing the time horizon to 3 years produced similar results. Although 5-year cumulative costs obtained with the IMM were equivalent to the HSOM, yearly estimates were found to differ by more than 40% compared to the HSOM.

CONCLUSIONS: While the OMM and IMM produced similar results to HSOM for very short PFS and OS values, BIAs should rely on HSOM, when possible, to produce reliable cumulative and yearly cost estimates.