Imputation in cure models

Background

Mixture cure models are advanced statistical tools designed to analyze survival data when a fraction of the population achieves cure and will never experience the event of interest. These models are particularly relevant in cancer research, where distinguishing between curative and life-prolonging effects of treatments is essential. By modeling cure probability (incidence) and survival of uncured patients (latency) separately, mixture cure models provide insights into the factors influencing both outcomes. However, when key covariates are missing, traditional approaches, such as complete-case analysis, may yield biased results or reduce statistical power. To address this, multiple imputation methods can integrate missing data more effectively, ensuring robust and unbiased estimates.

Aims

This project aims to extend the methodology of mixture cure models by incorporating a multiple imputation framework. The proposed approach differentiates between covariates affecting cure probability and those influencing survival, accommodating scenarios with missing covariates. Simulation studies and an application to osteosarcoma data from the BO06 clinical trial will illustrate its benefits and practical implications.

Relevance for cancer research

The developed methodology allows researchers to handle incomplete data in survival analyses more effectively, leading to more accurate estimates of cure rates and treatment effects. Its application to the BO06 trial highlights its utility in assessing the curative impact of dose-intensive chemotherapy for osteosarcoma. Beyond this case, the method is adaptable to other cancers and clinical scenarios where missing data and cure fractions are relevant, improving personalized treatment planning and prognosis evaluation.

Team

• Prof.dr.  M. Alfò, Department of Statistical Sciences Sapienza University of Rome (Italy)

• M. Cipriani, PhD candidate at Department of Statistical Sciences Sapienza University of Rome

• Prof.dr. M. Fiocco, Mathematical Institute Leiden University, Department of Biomedical Data Sciences Leiden University Medical Center & Princess Máxima Center for Pediatric Oncology 

• Dr. E. Musta, Korteweg-de Vries Institute for Mathematics University of Amsterdam