Survival extrapolation incorporating general-population mortality using excess-hazard and cure models (Sweeting et al. tutorial)
This method helps predict long-term survival in health economic models by combining patient data with general population mortality rates. It uses 'excess hazard' models to separate disease-specific risks from background risks, and can include a 'cure' component where some patients may no longer face elevated disease risk. This reduces uncertainty when projecting survival over long periods, especially in diseases like cancer.
At a glance
Use when
Projecting long-term survival in cost-effectiveness models, especially in oncology or chronic diseases where cure or diminishing hazard over time is plausible; when standard parametric models show high discordance
Avoid when
In diseases with strong interaction between disease risk and background mortality; when population mortality data are unavailable or not representative; when no plausible biological basis for cure exists
Inputs
Individual patient survival data, general population mortality rates (e.g., lifetables), disease-specific covariates, and optional cure fraction assumption
Outputs
Long-term survival extrapolations, restricted mean survival time (RMST), cure fraction estimates, goodness-of-fit statistics (e.g., AIC), and comparative survival curves
How it works
The method applies excess hazard (EH) models, which decompose all-cause hazard into population background mortality and disease-specific excess hazard. EH models are fitted with and without a cure parameter, allowing for the possibility that a proportion of patients may be cured. Parametric models are compared using Akaike information criterion (AIC), restricted mean survival time (RMST), and long-term survival extrapolations up to 30 years. The approach leverages general population lifetables to anchor extrapolations and reduce model uncertainty. Sensitivity to lifetable misspecification is also evaluated.
- HTA domains
- Clinical Effectiveness, Costs & Economic Evaluation
- Assumptions
- Disease-specific hazard is independent of background mortality; population lifetables are representative of the study cohort; cure, if modeled, follows a mixture or relative survival framework; proportional excess hazards may be assumed depending on model
- Strengths
- Reduces extrapolation uncertainty by anchoring to population mortality; improves plausibility of long-term survival projections; particularly useful in cancer settings where cure is possible; robust to moderate lifetable misspecification
- Limitations
- Requires reliable population mortality data; assumes independence between disease and background mortality; cure models may be overly simplistic if the biological cure assumption does not hold; limited external validation in diverse disease areas
- Also known as
- Excess Hazard Model with Cure, EH Cure Model, General Population Mortality-Adjusted Survival Extrapolation
Questions this answers
- › How can long-term survival be reliably extrapolated in cost-effectiveness analyses?
- › How can general population mortality data improve survival predictions?
- › When is it appropriate to assume a 'cure' fraction in survival modeling?
- › How do excess hazard models reduce uncertainty in survival extrapolation?
- › How sensitive are excess hazard models to errors in population mortality data?
- › What are the differences in long-term survival projections between standard and excess hazard models?
References & sources
Similar by meaning
Beta record. Generated from the primary source via AI extraction and independent audit, pending final human review.