Survival extrapolation

This page is under construction. We are developing materials to support the use of expert elicitation for long-term survival outcomes, motivated by applications in health technology assessment. A Technical Support Document (TSD) is in preparation for the NICE Decision Support Unit. Some draft materials are available below; these will be updated following the recommendations in the published version of the TSD. Please check this page in March 2025 for updates.

Software

The SHELF R package now includes an app and supporting functions for survival extrapolation. Please use the development version on GitHub if you want to try them. 

# Install devtools first
install.packages("devtools")

# Windows users may need to install RTools from
# https://cran.r-project.org/bin/windows/Rtools/rtools44/rtools.html

# Then
devtools::install_github("OakleyJ/SHELF")

# To run the main app

SHELF::elicitSurvivalExtrapolation()

The app SHELF::elicitSurvivalExtrapolation() will be hosted online in due course, though we expect most users to run it offline on their own R installation, as the app requires data to be uploaded.

Draft SHELF templates

Please refer to the main SHELF download for the full set of templates and advice, including powerpoint slides to walk the experts through the process of making probability judgements. A full documentation of an elicitation session would use the SHELF 1 and SHELF 3 Survival Extrapolation templates.

• SHELF 3 template with notes

• Blank SHELF 3 template

Draft training slides

• Powerpoint slides with presenter notes

Draft training exercise

• Training exercise (pdf)

The above pdf is a handout, intended for use a training exercise in surival extrapolation. It is based on a study reported in Koshiaris, C., Aveyard, P., Oke, J. et al. Smoking cessation and survival in lung, upper aero-digestive tract and bladder cancer: cohort study. Br J Cancer 117, 1224–1232 (2017). https://doi.org/10.1038/bjc.2017.179.

Synthetic data

• Synthetic_data_smokers.csv

This .csv file contains synthetic data, based on Koshairis et al. (2017) cited above above. There is no real patient data in this file. Synthetic data has been generated by digitising the survival curve in Figure 2a in this paper, using the WebPlotDigitizer tool (A. Rohatgi, WebPlotDigitizer. [Online]. Available: https://automeris.io).

The data can be used with the R shiny app SHELF::elicitSurvivalExtrapolation(), and are also the basis for the training exercise.