Illustrative Bayesian Reanalysis of Published Micronucleus Frequencies: A Retrospective Methodological Comparison of Classical and Jeffreys-Rule Estimates in Selected Radioprotection Groups

Le résumé fourni par la source

Published cytokinesis-block micronucleus (CBMN) frequencies, normalised to 500 binucleated cells (MN/500 BC), were used to illustrate how the choice of statistical parametrisation affects background-subtracted radioprotection metrics. An illustrative Jeffreys-rule transformation was applied to published normalised frequencies from eight representative experimental groups, drawn from two previously published human-lymphocyte CBMN datasets generated by a single laboratory, within the broader retrospective context of four source studies (2000–2026): an irradiated control, four pre-irradiation radioprotective treatments, one post-irradiation treatment, one radiosensitiser, and an iodine-131 biodosimetry cohort. It was not used as a statistically valid Bayesian inference based on the original experimental counts. Because the original integer micronucleus counts and the exact numbers of binucleated cells scored per condition could not be recovered, the analysis is presented as a descriptive methodological comparison. On the normalised scale, the transformation increased the net MN value when observed frequencies were close to background, with a maximum correction of 2.38 MN/500 BC, and approached the analytically expected asymptotic offset of 0.5 MN/500 BC under the illustrative working parametrisation, as the observed values substantially exceeded the background. Consequently, within the selected groups analysed, the apparent magnitude of protection calculated from the transformed values was lower than the classical estimates for the selected pre-irradiation treatments, showing the largest classical MP values—for example, 79.5% versus 93.8% for rosmarinic acid. The ordering of the selected radioprotective-treatment groups by MP was unchanged. These results illustrate the sensitivity of background-subtracted MN metrics to statistical parametrisation under low-signal conditions. They must not be interpreted as valid experimental uncertainty estimates, as dose estimates, or as clinical decision criteria.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.