Comparison and multi-model inference of excess risks models for radiation-related solid cancer

Area of Science:

• Radiation epidemiology
• Risk assessment
• Biostatistics

Background:

• Current radiation risk assessments often rely on single models, neglecting model uncertainty.
• This can lead to potentially biased estimations of health risks.

Purpose of the Study:

• To analyze model uncertainty in radiation risk assessment for all solid cancer incidence.
• To apply model averaging techniques to recent data from Japanese atomic bomb survivors.

Main Methods:

• Utilized data from the Life Span Study of atomic bomb survivors.
• Included seven plausible risk models for all solid cancer incidence.
• Employed a model averaging procedure and Monte Carlo simulations for uncertainty estimation.
• Calculated model-averaged Excess Relative Risks (ERR) and Excess Absolute Risks (EAR).

Main Results:

• Identified three models that most strongly influenced weighted risks based on baseline and information criteria.
• When fitting models with a common baseline, one model consistently dominated.
• Quantified uncertainty in excess risk estimations, including parameter correlations.

Conclusions:

• Model averaging provides a more robust approach to radiation risk assessment than single-model reliance.
• It is recommended to incorporate model uncertainty into future radiation risk analyses.
• The findings contribute to a better understanding of cancer risks from ionizing radiation exposure.