PERSONALIZED RISK PREDICTION FOR CANCER SURVIVORS: A GENERALIZED BAYESIAN SEMI-PARAMETRIC MODEL OF RECURRENT EVENTS...

Area of Science:

• Oncology
• Biostatistics
• Epidemiology

Background:

• Improved cancer survival leads to an increased incidence of multiple primary cancers.
• The characteristics of the first primary cancer significantly influence the risk of developing subsequent cancers.
• There is a need for robust, individualized risk assessment models for cancer survivors to inform healthcare policy and clinical decisions.

Purpose of the Study:

• To develop and validate a Bayesian semi-parametric framework for characterizing the risk of developing second primary cancers in cancer survivors.
• To provide covariate-adjusted, age-to-onset penetrance curves for various cancer types.
• To support personalized health management strategies for cancer survivors.

Main Methods:

• A Bayesian semi-parametric framework using independent non-homogeneous Poisson processes for competing cancer types.
• Adjustment for covariates including the type and age at diagnosis of the first primary cancer.
• Application to a historical cohort with a high prevalence of multiple primary tumors and diverse cancer types.
• Validation using Receiver Operating Characteristic (ROC) curves and calculation of Area Under the Curve (AUC) values.

Main Results:

• Derived age-to-onset penetrance curves for cancer survivors, including specific estimates for second primary lung cancer.
• Validated the model's predictive performance for second primary lung cancer (AUC=0.89), sarcoma (AUC=0.91), breast cancer (AUC=0.76), and all other cancers combined (AUC=0.68).
• Demonstrated the framework's ability to provide quantitative, covariate-adjusted risk assessments.

Conclusions:

• The proposed Bayesian framework offers a robust method for quantitative risk assessment in cancer survivors.
• The model's predictive accuracy supports its potential utility in guiding personalized cancer screening and management decisions.
• This approach advances personalized health management for the growing population of cancer survivors.