Concept Architecture
Concept
Theoretically, the Generalized Gamma Model is a fully parametric survival model that assumes survival times follow a generalised gamma distribution. It extends conventional parametric survival models by allowing highly flexible hazard functions capable of representing increasing, decreasing, constant and non-monotonic risks over time. In health economics, the generalised gamma model is widely used for survival analysis and long-term extrapolation when standard models such as the Weibull or Gompertz distributions do not adequately describe observed survival data.
Mathematically, the generalised gamma model specifies that the logarithm of survival time follows a generalised gamma distribution characterised by location, scale and shape parameters. These parameters determine the probability density, survival and hazard functions and are estimated using maximum likelihood estimation. The model includes several commonly used survival models as special cases, allowing formal comparison through likelihood-based methods.
In practice, the generalised gamma model is fitted to individual patient survival data using specialised statistical software. Candidate survival models are compared using statistical goodness-of-fit measures, visual assessment and clinical plausibility before selecting the most appropriate model for extrapolation. The fitted model is then used to estimate long-term survival, life expectancy, quality-adjusted life-years and healthcare costs within health economic evaluations.
Purpose
Used to model complex survival data, accommodate diverse hazard function shapes and generate long-term survival projections for health technology assessment and cost-effectiveness analysis.
Mathematical Formulae
Primary Formula
ln(T) ~ GG(?, �, Q)
where:
- T = survival time
- ? = location parameter
- � = scale parameter
- Q = shape parameter
Supporting Formulae
Survival function:
S(t) = 1 ? F(t)
Hazard function:
h(t) = f(t) / S(t)
Maximum likelihood estimation:
?? = arg max L(?)
Related Mathematical Methods
- Maximum likelihood estimation
- Parametric survival modelling
- Generalized gamma distribution
- Generalized F model
- Weibull model
- Log-normal model
- Log-logistic model
- Akaike Information Criterion (AIC)
- Bayesian Information Criterion (BIC)
Example
An oncology trial with five years of follow-up exhibits a changing hazard that cannot be adequately represented by a Weibull model. A generalised gamma model is fitted to the patient-level survival data and provides the best fit according to AIC while producing clinically plausible lifetime survival projections for estimating quality-adjusted life-years and incremental cost-effectiveness ratios.
Excel Implementation
| Function | Example Formula | Health Economics Application |
|---|---|---|
| LN | =LN(A2) | Calculate log survival times for exploratory survival modelling. |
| EXP | =EXP(B2) | Transform predicted log survival times to the original time scale. |
| LINEST | =LINEST(Y_range,X_range,TRUE,TRUE) | Perform exploratory regression prior to formal model estimation. |
| SUMPRODUCT | =SUMPRODUCT(CoefficientRange,CovariateRange) | Calculate the linear predictor for the survival model. |
VBA (Optional)
Automate comparison of alternative parametric survival models and generate long-term survival projections using the fitted generalised gamma model.
Sources
- Stacy EW. A Generalization of the Gamma Distribution.
- Cox C, Chu H, Schneider MF, Mu�oz A. Parametric survival analysis and taxonomy of hazard functions for the generalized gamma distribution.
- Lawless JF. Statistical Models and Methods for Lifetime Data.
- Briggs A, Claxton K, Sculpher M. Decision Modelling for Health Economic Evaluation.
- NICE. Health Technology Evaluation Manual.
Related Concepts (2)
Library
Tools & Resources
1
survHE — Survival Analysis for Health Economic Evaluation (R package) — Gianluca Baio, R package ed., 2023 (CRAN)
An R package for fitting and comparing parametric survival models for health economic evaluation, including Bayesian estimation, and for extrapolating time-to-event data to inform cost-effectiveness models.
Software (R package)View source →
Frequently Asked Questions (6)
What is the generalised gamma model?
A survival model based on the generalised gamma distribution, nesting simpler distributions, such as the Weibull and log-normal, within one framework.
Source: Stacy 1962
Why might a flexible model like the generalised gamma still be rejected for extrapolation?
Although the generalised gamma fits observed data flexibly, a model that hugs the trial data closely does not necessarily extrapolate well, and its several parameters can be poorly determined when follow-up is short, producing wild long-term projections. Analysts therefore judge it by fit and, just as importantly, by whether its extrapolated tail is clinically plausible, sometimes preferring a simpler distribution whose projection is more stable. Flexibility within the data does not guarantee sensible behaviour beyond it. Latimer (2013) cautions on this point.
Source: Latimer 2013
Why use the generalised gamma model?
The generalised gamma model is used because its flexibility lets it fit varied hazard shapes, and its nesting of common distributions allows the data to indicate which simpler form is appropriate, rather than assuming one in advance. This is valuable in survival extrapolation, where the choice of distribution strongly affects long-term estimates and no single simple form may be clearly best. By fitting the general model and examining which nested distribution its parameters approach, analysts can select a suitable survival model in a principled way.
Source: Latimer 2013
How does the generalised gamma model help select a distribution?
The generalised gamma model helps select a distribution because its parameters, when estimated, indicate which simpler nested distribution the data favour: particular parameter values correspond to the Weibull, gamma, exponential, or log-normal, so the fitted generalised gamma can point to the most appropriate simpler form. Comparing the general model with its special cases, using fit statistics, guides the choice. This makes the generalised gamma model a useful starting point for identifying a suitable survival distribution rather than assuming one from the outset.
Source: Stacy 1962
What hazard patterns does the generalised gamma model allow?
The generalised gamma model allows a wide range of hazard patterns, including monotonically increasing or decreasing hazards and non-monotonic hazards that rise then fall or fall then rise, owing to its three parameters. This lets it fit complex survival data that simpler models cannot, and it is particularly useful where the hazard shape is uncertain. The flexibility to capture turning hazards distinguishes the generalised gamma model from the exponential, Weibull, or gamma, whose hazards are monotonic.
Source: Collett 2015
What are the limitations of the generalised gamma model?
The generalised gamma model requires estimating three parameters, so it needs sufficient data and can produce unstable estimates or fail to converge with limited data, and its flexibility risks overfitting. Its complexity makes it harder to fit and interpret than simpler distributions, and its extrapolation, like that of any parametric model, depends on the assumed form and is uncertain. These limitations mean it is used where data support its estimation, frequently to identify an appropriate simpler nested distribution for the final survival model.
Source: Latimer 2013
Trust Record
Verified by Dr Darrin Baines
British health economist
Professional identity: darrinbaines.org
Verification date: 20 Oct 2025
Content version: 1.0.0
Canonical Identity
- Term code
- HE-EM-SM-030
Stable URI · Machine-readable · Resolvable · CC BY 4.0