Objective Bayesian Analysis of Micronucleus Biodosimetry Data for Radiation Protection and Low-Dose Medical Applica-Tions of Ionizing Radiation

Jeffreys‑rule Bayesian inference for the Poisson mean was retrospectively applied to cytokinesis‑block micronucleus (CBMN) data from four studies conducted in a single laboratory (2001–2026), covering ionising‑radiation exposures in human lymphocytes and an ¹³¹I biodosimetry study (29 experimental groups). The CBMN assay is widely used for cytogenetic biodosimetry and for evaluating radioprotective agents, but classical analysis of net micronucleus (MN) yield becomes unreliable when effective radioprotectors reduce MN frequencies towards background, leading to near‑zero or negative net counts with poorly characterised uncertainty. In our reanalysis, the Bayesian correction systematically modified magnitude-of-protection and dose-reduction estimates in low-count, high‑protection conditions in comparison with the classical approach. Using Bayesian net counts to recalculate standard radioprotective metrics revealed that, while the overall ranking of radioprotective potency was preserved, Bayesian magnitudes of protection and dose reduction factors were consistently lower than their classical counterparts for the most effective radioprotectors administered before irradiation, providing more conservative efficacy estimates with quantified uncertainty. In the ¹³¹I biodosimetry cohort and in patients undergoing diagnostic nuclear medicine procedures, post‑exposure MN yields remained within the spontaneous background variability of the assay, and the Bayesian framework is fully compatible with the practical absence of detectable chromosomal damage at the doses used in routine clinical practice. Taken together, these results demonstrate that Jeffreys‑rule Bayesian inference provides a statistically coherent and practically feasible framework for CBMN analysis in the low‑count regime, stabilising net MN estimates, avoiding artefactual negative values and supplying confidence intervals for radioprotective efficacy metrics. As a concrete methodological recommendation, Bayesian analysis should be adopted as the default approach whenever post‑exposure MN counts are within approximately a factor of two of background, with classical methods remaining adequate at higher count levels.