Likelihood adaptively incorporated external aggregate information with uncertainty for survival data

Area of Science:

• Biostatistics
• Cancer Epidemiology
• Oncology

Background:

• Population-based cancer registries offer valuable aggregate survival data but often lack detailed biomarker information.
• Primary cohort studies may have limited statistical power, particularly for rare cancer subtypes.
• Integrating registry data with primary cohorts can improve treatment impact evaluation and survival outcome prediction.

Purpose of the Study:

• To develop a statistical approach for integrating primary cohort data with external aggregate survival data from cancer registries.
• To address the challenge of modest sample sizes in registries for rare cancers and account for aggregate data variability.
• To enhance the evaluation of treatment effects and survival predictions by leveraging complementary data sources.

Main Methods:

• Proposed an externally informed likelihood approach to link primary cohort data with aggregate registry data.
• Accounted for the variability inherent in aggregate survival statistics from external sources.
• Established asymptotic properties of the proposed estimators and evaluated performance through simulation studies.

Main Results:

• The developed method successfully integrates primary cohort data with aggregate survival data.
• Demonstrated the utility of the approach in a real-world application using inflammatory breast cancer (IBC) data.
• Enabled the appraisal of tri-modality treatment effects on survival across diverse IBC tumor subtypes.

Conclusions:

• The externally informed likelihood approach provides a robust framework for combining disparate data sources in cancer research.
• This method enhances statistical power and improves the accuracy of survival outcome predictions, particularly for rare cancers.
• Facilitates a more comprehensive understanding of treatment efficacy across different tumor subtypes by leveraging population-level data.