Cure Models: What is Meant by a Survival 'Plateau', and Do Experts Agree on What Constitutes One?

Area of Science:

• Biostatistics
• Survival Analysis
• Mathematical Modeling

Background:

• Cure models are increasingly used for survival data with plausible long-term survival.
• A plateau in Kaplan-Meier curves is a criterion for fitting cure models, but lacks a clear mathematical definition.
• Visual inspection of survival curves is subjective, leading to ambiguity.

Purpose of the Study:

• To investigate and clarify the mathematical definition of a plateau in survival probability.
• To explore the features of data required for mixture cure models.
• To assess the suitability of survival curves for cure modeling.

Main Methods:

• Expert ranking of survival curves from oncology trials for suitability in mixture cure modeling.
• Mathematical exploration of data features for positive cure fraction estimation using exponential and Weibull mixture cure models.
• Case study application of the developed methods.

Main Results:

• Weak correlations were found between expert rankings of survival curves.
• Mixture cure models require specific data features, like decreasing event rates, but are highly model-dependent.
• Statistical issues were highlighted in the case study.

Conclusions:

• A precise mathematical definition for a plateau in Kaplan-Meier curves is elusive.
• External evidence and subject matter expertise are essential for validating cure plausibility.
• Interpretation of cure fraction estimates requires careful consideration of model dependency.