4D-03 - extRpolateS: A Shiny-Based Interactive Platform for Time-to-Event Data Modeling in Health Technology Assessments

Background: Clinical drug trials often have limited follow-up durations, resulting in incomplete time-to-event data. While these studies are primarily designed for regulatory approval, accurate long-term estimates (e.g., survival extrapolations) are crucial for health technology assessments (HTAs). Reliable extrapolation is essential, yet model selection significantly impacts projected treatment benefits. Although several tools exist, many lack advanced modeling capabilities and require substantial technical expertise. A user-friendly, interactive platform can streamline time-to-event data modeling for HTAs.

Objectives: To develop an intuitive, Shiny-based application (extRpolateS) that facilitates time-to-event analyses required for HTAs.

Methods: extRpolateS is a cloud-based or desktop platform developed in R Shiny, leveraging the survival and flexsurv packages for time-to-event modeling, and plotly/ggplot2 for interactive visualizations. The modular interface includes an input module that accepts three types of data: (1) a Kaplan–Meier figure digitized in-app, (2) pre-digitized Kaplan–Meier data—both converted into pseudo-individual patient data using the Guyot algorithm—or (3) empirical trial data. Users can store uploaded data in an integrated database for future analyses. The output module provides descriptive analytics (e.g., Kaplan–Meier curves, proportional hazard checks) and supports a range of modeling approaches, including general parametric (e.g., Weibull), spline-based (normal, hazard, or odds scale), mixture and non-mixture cure, relative survival and piecewise models. All modeling follows the guidelines outlined in the National Institute for Health and Care Excellence (NICE) Technical Support Documents 14 and 21.

Results: Pilot testing with researchers and pharmaceutical company representatives demonstrated that extRpolateS effectively generated time-to-event analyses suitable for integration into health economic models. Preliminary feedback emphasized the platform’s ease of use, particularly its ability to perform statistical analyses accessible to non-experts. Users recognized its potential to streamline survival extrapolation but expressed interest in expanding its capabilities to incorporate health economic models for early economic evaluation.

Conclusions: extRpolateS is a comprehensive solution for time-to-event analysis, accessible to both researchers and non-experts. It generates robust outputs for health economic models while ensuring ease of use. With its advanced modeling capabilities, extRpolateS enhances decision-making in HTAs. Future enhancements, including the integration of health economic modeling, could further expand its impact.