Presentation (CTI)

OBJECTIVES: Survival extrapolation and partitioned survival model (PSM) based cost-effectiveness analysis are usually performed in separate software, requiring users to export extrapolated curves and re-import them elsewhere. This fragmentation introduces transcription error and limits reproducibility. We aimed to develop and demonstrate a module that extends extRpolateS, a cloud-based R Shiny survival-extrapolation platform, into a PSM-based economic evaluation within a single reactive pipeline.
METHODS: The module consumes the progression-free survival (PFS) and overall survival (OS) extrapolations produced upstream, without an export step, and derives health-state occupancy using the partitioned survival framework, consistent with NICE DSU Technical Support Documents 14, 19, and 21. Four analytic components structure the workflow. Curves and Health States links Kaplan-Meier, parametric, spline, cure, or piecewise extrapolations for PFS and OS across treatment and comparator arms under independent or dependent structures. Cost-Effectiveness Analysis combines drug, administration, monitoring, adverse-event, and health-state costs with utilities, discounting, half-cycle correction, and treatment discontinuation. Deterministic Sensitivity Analysis varies parameters over user-defined ranges, and Probabilistic Sensitivity Analysis samples from Gamma, Beta, Lognormal, and Normal distributions.
RESULTS: The module reproduces the full PSM cost-effectiveness workflow without requiring the analyst to write code. It generates incremental cost-effectiveness ratios, net monetary benefit, cost-effectiveness planes, and cumulative cost and outcome plots. Deterministic analysis adds tornado and one-way plots, and probabilistic analysis adds cost-effectiveness plane scatter plots and cost-effectiveness acceptability curves. The economic model reads directly from the upstream extrapolations, so a change to any survival specification propagates automatically to every downstream estimate.
CONCLUSIONS: The PSM module brings survival modeling and economic evaluation into a single application, which consolidates a workflow that conventionally spans multiple tools and reduces opportunities for error. The reactive pipeline supports reproducible, transparent cost-effectiveness analysis and lowers the technical barrier for intermediate R users. Planned extensions include state-transition models and value-of-information analyses.