Related Experiment Videos
Survivorship analysis when cure is a possibility: a Monte Carlo study
Statistics in Medicine
|April 1, 1984
Summary
This study introduces a new statistical model for clinical trials, improving survival analysis by accounting for a proportion of patients potentially cured by cancer treatments. This method enhances the accuracy of estimating treatment effectiveness.
Area of Science:
- Biostatistics
- Clinical Trial Design
- Survival Analysis
Background:
- Parametric survival analyses often assume a constant hazard function over time.
- Empirical data may show survival curves leveling off, suggesting limitations of constant hazard models.
- Existing models like Gompertz or Weibull distributions account for decreasing hazards, but may not capture cure fractions.
Purpose of the Study:
- To develop and evaluate a statistical model that incorporates a cure fraction in parametric survival analyses.
- To estimate the proportion of patients with a zero hazard (cured) and the proportion with a constant high risk.
- To assess the performance of the likelihood ratio test for detecting a cure fraction.
Main Methods:
- Maximum likelihood estimation (MLE) for the cure proportion parameter (pi).
- Likelihood ratio test (LRT) to statistically compare models with and without a cure fraction.
- Monte Carlo simulations to evaluate the power of the LRT and the accuracy of MLE under various scenarios.
- Illustration of the proposed methods using empirical clinical trial data.
Main Results:
- The proposed mixture model provides a more biologically plausible representation when cure is suspected.
- Maximum likelihood estimation effectively estimates the cure proportion (pi) and associated risk parameters.
- The likelihood ratio test demonstrates power in detecting the presence of a cure fraction, particularly with larger sample sizes and higher cure rates.
- Simulations confirm the robustness and performance of the estimation and testing procedures.
Conclusions:
- The mixture cure model offers a valuable extension to standard survival analysis in clinical trials, especially for diseases with potential cures.
- Accurate estimation of the cure proportion (pi) is crucial for correctly interpreting treatment effects and patient outcomes.
- The likelihood ratio test provides a statistically sound method for identifying trials where a cure fraction model is more appropriate than traditional models.