If You Torture Your Data Long Enough, It Will Confess to Anything: On the Epidemiological Basis of the LNT Model.

This note deals with epidemiological data interpretation supporting the linear no-threshold model, as opposed to emerging evidence of adaptive response and hormesis from molecular biology in vitro and animal models. Particularly, the US-Japan Radiation Effects Research Foundation's lifespan study of atomic bomb survivors is scrutinized. We stress the years-long lag of the data processing after data gathering and evolving statistical models and methodologies across publications. The necessity of cautious interpretation of radiation epidemiology results is emphasized.