NICE DSU Technical Support Document 14
This document provides guidance on how to perform survival analysis for economic evaluations in health technology assessments, especially when using data from clinical trials. It emphasizes the importance of carefully choosing and justifying statistical models used to predict long-term survival beyond the observed data, to avoid biased or inconsistent results.
At a glance
Use when
Conducting economic evaluations requiring extrapolation of survival data from clinical trials, particularly in submissions to HTA bodies like NICE
Avoid when
Survival data is short-term and no extrapolation is needed; non-parametric methods are sufficient for the evaluation context; real-world evidence is available to directly inform long-term survival
Inputs
Patient-level survival data from clinical trials, candidate parametric models (e.g., Weibull, exponential, log-normal), statistical fit metrics
Outputs
Justified choice of survival model, extrapolated survival curves with plausible long-term trends, transparent documentation of model selection process
How it works
The guideline reviews survival analysis practices in 45 NICE appraisals in oncology and identifies common shortcomings, such as insufficient model comparison and lack of justification for extrapolation choices. It proposes a structured process for model selection, including assessing the plausibility of extrapolated survival curves and comparing multiple candidate models using statistical and clinical criteria to improve robustness in economic evaluations.
- HTA domains
- Clinical Effectiveness, Costs & Economic Evaluation, Safety
- Categories
- Cost-effectiveness ModellingData Exploration & VisualizationEvidence SynthesisReporting Standards & Best Practice
- Assumptions
- Long-term survival trends can be reasonably extrapolated from trial data using parametric models; clinical knowledge can inform model plausibility; multiple model comparison reduces bias
- Strengths
- Provides a systematic, transparent framework for model selection; reduces risk of bias and inconsistency in HTA submissions; grounded in real-world review of past HTAs
- Limitations
- Focuses on commonly used methods in NICE submissions, not all possible survival analysis techniques; limited discussion of uncertainty quantification in extrapolations; primarily applicable to oncology and similar chronic conditions
- Also known as
- TSD 14, NICE DSU TSD 14
Questions this answers
- › How should survival models be selected for economic evaluations in HTAs?
- › What are common flaws in survival extrapolation in HTA submissions?
- › How can extrapolated survival curves be assessed for plausibility?
- › Why is model choice important in cost-effectiveness analyses?
- › What steps should analysts take to justify survival model selection?
- › How can inconsistency in survival analysis across HTAs be reduced?
References & sources
Similar by meaning
- Survival extrapolation validation-based case study (Bullement et al.)
- Structured expert elicitation for long-term survival outcomes (NICE DSU TSD 26)
- Survival extrapolation incorporating general-population mortality using excess-hazard and cure models (Sweeting et al. tutorial)
- Partitioned Survival Analysis (NICE DSU TSD 19)
- survextrap
Beta record. Generated from the primary source via AI extraction and independent audit, pending final human review.