Mixture Cure Models in Oncology: A Tutorial and Practical Guidance.

Area of Science:

• Biostatistics
• Cancer Research
• Health Economics

Background:

• Novel cancer therapies can lead to survival patterns where a fraction of patients achieve statistical cure.
• Traditional survival analysis may not accurately capture long-term survival in such cases, impacting health economic evaluations.
• Mixture cure models address this by estimating a 'cure fraction' within the patient population.

Purpose of the Study:

• To provide a practical introduction and implementation guide for mixture cure models.
• To demonstrate two approaches: an 'uninformed' model estimating cure fraction from data and an 'informed' model using external data.
• To facilitate the accurate estimation of long-term survival for health economic analyses.

Main Methods:

• Development and implementation of two mixture cure models in R.
• Utilizing maximum likelihood estimation for fitting the models.
• Step-by-step guidance on data handling, model fitting, and results interpretation.
• Providing freely available R code on GitHub for reproducibility.

Main Results:

• Mixture cure models provide different long-term survival estimates compared to standard methods, especially when a cure fraction exists.
• The 'uninformed' approach estimates the cure fraction from trial data.
• The 'informed' approach incorporates external data for the cure fraction, offering flexibility.
• The models are presented as straightforward to use and interpret.

Conclusions:

• Mixture cure models enhance the accuracy of long-term survival estimates for cancer therapies associated with statistical cure.
• These models are crucial for reliable health economic evaluations, particularly with emerging cancer treatments.
• The tutorial provides accessible tools and guidance for implementing and interpreting mixture cure models in R.