Abstract:We propose a novel frailty model with change points applying random effects to a Cox proportional hazard model to adjust the heterogeneity between clusters. Because the frailty model includes random effects, the parameters are estimated using the expectation-maximization (EM) algorithm. Additionally, our model needs to estimate change points; we thus propose a new algorithm extending the conventional estimation algorithm to the frailty model with change points to solve the problem. We show a practical example to demonstrate how to estimate the change point and random effect. Our proposed model can be easily analyzed using the existing R package. We conducted simulation studies with three scenarios to confirm the performance of our proposed model. We re-analyzed data of two clinical trials to show the difference in analysis results with and without random effect. In conclusion, we confirmed that the frailty model with change points has a higher accuracy than the model without the random effect. Our proposed model is useful when heterogeneity needs to be taken into account. Additionally, the absence of heterogeneity did not affect the estimation of the regression coefficient parameters.
What problem does this paper attempt to address?
The main problem that this paper attempts to solve is how to handle heterogeneous (heterogeneity) data more accurately in survival analysis. Specifically, the author proposes a new frailty model, which combines change points and random effects to adjust for the heterogeneity between different clusters. This method is particularly suitable for clinical trial data of immune checkpoint inhibitors (such as nivolumab), where the proportional hazards assumption may not hold before and after certain time points.
### Main Research Questions
1. **Handling Heterogeneous Data**: The traditional Cox proportional hazards model assumes that all individuals have the same risk characteristics, but heterogeneity often exists between different groups in actual data. The author adjusts this heterogeneity by introducing random effects.
2. **Detecting Change Points**: In some clinical trials, the treatment effect may change at a specific time point, causing the survival curve to show different trends before and after this point. The author proposes a frailty model that includes change points in order to estimate this change point more accurately.
3. **Improving Model Accuracy**: By introducing random effects and change points, the author hopes to improve the model's fit and prediction accuracy for survival data.
### Method Overview
- **Model Definition**: The author defines a frailty model that includes change points, where the coefficient parameters between each change point can be different. The form of the model is as follows:
\[
\lambda(t_{mi}; x_{mi}, \beta_k, v_{km}, \tau_k) = \lambda_0(t_{mi}) v_{km} \exp(\beta_k^T x_{mi}) I(\tau_{k - 1} < t_{mi} \leq \tau_k)
\]
where \(\lambda_0(t)\) is the non - parametric baseline hazard function, \(v_{km}\) is the random effect of the \(m\) - th cluster in the interval \((\tau_{k - 1}, \tau_k]\), \(\beta_k\) is the regression coefficient vector in this interval, and \(\tau_k\) is the unknown change point.
- **Parameter Estimation**: Since the model includes random effects, the expectation - maximization (EM) algorithm is used for parameter estimation. The E - step calculates the expected value of the random effects, and the M - step maximizes the conditional likelihood function to estimate the parameters.
- **Simulation Study**: The author conducted simulation studies in three scenarios to verify the performance of the model in different situations. The results show that the model considering random effects has higher accuracy and lower mean squared error (MSE) in most cases.
- **Practical Application**: The author re - analyzed the data of two clinical trials, involving primary biliary cirrhosis (PBC) and malignant glioma (MG) respectively. The results show that the model considering random effects can estimate change points and regression coefficients more accurately.
### Conclusion
The frailty model with change points proposed by the author has higher accuracy in handling heterogeneous data. Even in the absence of significant heterogeneity, this model does not affect the estimation of regression coefficient parameters. Therefore, this model has important application value in survival analysis where heterogeneity needs to be considered.