Unveiling the Multifaceted Dynamics of Breast Cancer: A Copula Regression Approach to Modeling and Predicting...

Area of Science:

• Biostatistics
• Oncology
• Data Science

Background:

• Breast cancer is a leading cause of death, necessitating advanced analytical methods for understanding clinical variable interdependencies.
• Traditional multivariate analyses struggle with the complex, nonlinear, and asymmetric dependencies common in clinical data.
• Copula models offer enhanced flexibility for modeling mixed-type outcomes and their intricate relationships.

Purpose of the Study:

• To apply flexible copula regression models for analyzing interdependencies among clinical variables in breast cancer.
• To jointly model overall survival and age at diagnosis using copula-based approaches.
• To compare different copula families and assess their fit against independent marginal models.

Main Methods:

• Exploration of multiple copula families to jointly model binary overall survival and continuous age at diagnosis.
• Utilized the METABRIC dataset for analysis.
• Employed goodness-of-fit metrics and Probability Integral Transform (PIT) diagnostics for model selection and validation.

Main Results:

• The Gumbel copula demonstrated superior performance in capturing upper tail dependence, linking younger age and improved survival.
• Copula models significantly improved model fit compared to an independent margins baseline (likelihood ratio test, p < 0.0001).
• PIT diagnostics confirmed the adequacy of the marginal models used within the copula framework.

Conclusions:

• Copula regression models provide a more nuanced understanding of breast cancer progression by effectively modeling complex dependencies.
• These models enhance risk assessment accuracy and support data-driven decision-making in oncology.
• Integration of copula models is recommended for clinical research to improve patient outcome analysis.