Survival Extrapolation Incorporating General Population Mortality Using Excess Hazard and Cure Models: A Tutorial.

Area of Science:

• Health economic modeling
• Biostatistics
• Survival analysis

Background:

• Parametric survival models in cost-effectiveness analyses can yield discordant extrapolations and decision uncertainty.
• Excess hazard (EH) methods offer a potential solution by incorporating general population mortality data to reduce model uncertainty.
• This review focuses on practical considerations for using EH methods in long-term survival estimation.

Approach:

• A case study involving 686 patients with low and high-grade breast cancer was analyzed.
• Seven standard parametric survival models were compared with EH models, both with and without a cure parameter.
• Survival extrapolations, restricted mean survival time (RMST), and cure fraction estimates were compared, alongside sensitivity analyses for lifetable misspecification.

Key Points:

• Standard models showed extensive variability in 30-year RMST (7.5 to 14.3 years).
• EH cure methods significantly reduced model uncertainty compared to standard models and EH models without cure.
• Long-term treatment effects approached the null across most models at varying rates, with minimal impact of lifetable misspecification on RMST differences.

Conclusions:

• EH methods are valuable for survival extrapolation, especially in cancer where excess hazards may decrease over time.
• EH cure models are beneficial when cure is plausible, leading to reduced extrapolation variability.
• EH methods demonstrate robustness to lifetable misspecification.