A study of threshold-like behaviour in low-dose radiation risk using improved statistical evaluation methods and...

A large number of epidemiological cohort studies have evaluated radiation risks at low doses; however, the applicability of the linear no-threshold model to the low-dose range (LDR) remains an unresolved issue. This study argues that conventional statistical methods used in cohort analyses may not be suitable for assessing radiation effects in the LDR. Major limitations include the use of broad attained-age stratification, the dominance of high-dose observations in single-model regression, and inconsistent treatment of strongly correlated confounders such as smoking. To address these limitations, we propose improved statistical procedures optimised for the LDR and apply them to the publicly available dataset of the Radiation Effects Research Foundation's life span study (LSS-2009). The proposed method derives baseline estimates exclusively from non-exposed groups, excludes individuals of unknown smoking status, and varies the upper boundary of the LDR while maintaining fixed baseline coefficients. Under these analytical conditions, the excess relative risk coefficient appears to change sign within a certain dose range, which may suggest the possibility of threshold-like behaviour. Even within the constraints of limited data, the present analysis indicates that adopting alternative analytical perspectives may allow such patterns to emerge. These findings suggest the importance of making additional datasets available and encouraging broader examination by the research community, so that conclusions can be refined through open and collective scientific discussion.