Limitations associated with analysis of cytogenetic data for biological dosimetry.

The scientific literature concerning cytogenetic biodosimetry has been reviewed to identify the range of scenarios of radiation exposure where biodosimetry has been carried out. Limitations in the existing standardized statistical methodology have been identified and categorized, and the reasons for these limitations have been explored. Statistical problems generally occur due to either low numbers of aberrations leading to large uncertainties or deviations in aberration-per-cell distributions leading to over- or under-dispersion with respect to the Poisson model. A number of difficulties also stem from limitations of the classical statistical methodology, which requires that chromosome aberration yields be considered as something "fixed" and thus provides a deterministic estimate of radiation dose and associated confidence limits (because an assignment of a probability to an event is based solely on the observed frequency of occurrence of the event). Therefore, it is suggested that solutions to the listed problems should be based in the Bayesian framework. This will allow the investigator to take a probabilistic approach to analysis of cytogenetic data, which can be considered highly appropriate for biological dose estimation.