Uncertainty, low-dose extrapolation and the threshold hypothesis.

Risk-based radiation protection policy is influenced by estimated risk and by the uncertainty of that estimate. Thus, if the upper limit, at (say) 95% probability, of risk associated with a given radiation dose is at an 'acceptable' level, it is unlikely (or not credible) that the true level of risk associated with the dose is at an unacceptable level. Central estimates presented alone, in the absence of probability limits, lack this safety factor. Estimating cancer risks from low doses of ionising radiation involves extrapolation of risk estimates based on high-dose data to the much lower dose levels that characterize the vast majority of exposures of regulatory concern. Proof of a universal low-dose threshold, below which there is no radiation-related risk, would revolutionise radiation protection. Available data fail to provide such proof and, in fact, leave considerable room for the possibility that DNA damage from a single photon can contribute to the carcinogenic process. Allowing for the possibility of a threshold would, however, remove very little of the regulatory burden associated with the so-called linear, no-threshold hypothesis, unless that possibility were a virtual certainty.