F. J. Rubio,H. Putter,A. Belot
Abstract:Unobserved individual heterogeneity is a common challenge in population cancer survival studies. This heterogeneity is usually associated with the combination of model misspecification and the failure to record truly relevant variables. We investigate the effects of unobserved individual heterogeneity in the context of excess hazard models, one of the main tools in cancer epidemiology. We propose an individual excess hazard frailty model to account for individual heterogeneity. This represents an extension of frailty modelling to the relative survival framework. In order to facilitate the inference on the parameters of the proposed model, we select frailty distributions which produce closed-form expressions of the marginal hazard and survival functions. The resulting model allows for an intuitive interpretation, in which the frailties induce a selection of the healthier individuals among survivors. We model the excess hazard using a flexible parametric model with a general hazard structure which facilitates the inclusion of time-dependent effects. We illustrate the performance of the proposed methodology through a simulation study. We present a real-data example using data from lung cancer patients diagnosed in England, and discuss the impact of not accounting for unobserved heterogeneity on the estimation of net survival. The methodology is implemented in the R package IFNS.
What problem does this paper attempt to address?
### Problems the paper attempts to solve
This paper aims to solve the problem of Unobserved Individual Heterogeneity (UIH) in cancer epidemiology research. Specifically, the paper focuses on how to deal with unobserved individual heterogeneity caused by unrecorded important covariates or model misspecification in Excess Hazard Models. These problems may lead to parameter estimation bias and affect the estimation of Net Survival.
### Background and motivation
1. **Unobserved individual heterogeneity**
- Unobserved individual heterogeneity means that important covariates are not recorded or are unavailable in the sample, which may lead to model misspecification.
- Ignoring the impact of UIH may lead to bias in parameter estimation and affect the interpretation of epidemiological indicators (such as risk functions and incidence rates).
2. **Excess risk model**
- The excess risk model is a major tool in cancer epidemiology, used to separate cancer - related risks from risks due to other causes.
- The model usually assumes that the individual risk function can be decomposed into the risk due to other causes and the excess risk related to cancer:
\[
h(t; x)=h_P(\text{age}+t; \text{year}+t, z)+h_E(t; x)
\]
where \(h_P(\text{age}+t; \text{year}+t, z)\) is the expected population death risk based on the life table, and \(h_E(t; x)\) is the excess risk related to cancer.
3. **Limitations of existing methods**
- Existing methods have deficiencies in dealing with UIH. For example, the related frailty model proposed by Zahl [42] encounters inference problems when jointly modeling two competing risks.
- Other studies have also attempted to use frailty models to deal with UIH, but these methods usually assume that the frailty term has the same impact on different risks, which may not be applicable to real - data.
### Solutions
1. **Individual frailty excess hazard model**
- The paper proposes an Individual Frailty Excess Hazard Model, which deals with unobserved individual heterogeneity by introducing random effects (frailty terms).
- The model assumes that the excess risk is affected by the frailty term:
\[
\tilde{h}(t|\lambda; x)=h_P(\text{age}+t; \text{year}+t, z)+\lambda h_E(t; x)
\]
where \(\lambda\sim G\) is a random variable with a mean of 1, representing the frailty distribution.
2. **Properties and inference of the model**
- The paper derives closed - form expressions for the marginal risk and survival functions and discusses the interpretation of these expressions.
- By choosing an appropriate frailty distribution (such as the gamma distribution), the model can provide an intuitive interpretation, that is, the frailty term induces the selection of healthy individuals at the marginal level.
3. **Flexible parametric model**
- The paper uses a flexible parametric model to model the excess risk, allowing the incorporation of time - dependent effects and covariates acting on the risk scale.
- The model structure is rich, including the proportional hazards model, the accelerated hazards model, and the accelerated failure time model as special cases.
4. **Simulation studies and practical applications**
- The paper evaluates the performance of the model in different scenarios through simulation studies, especially when there are unobserved covariates.
- The practical application part uses data of lung cancer patients in England to show the performance of the model in real - data.
### Conclusion
The individual frailty excess hazard model proposed in the paper can effectively deal with unobserved individual heterogeneity and improve the estimation accuracy of net survival rate. The model shows good performance in both theoretical and practical applications, providing strong support for cancer epidemiology research.