The risk to health from exposure to low levels of ionising radiation

Article Outline

• Copyright

The shape of the dose–response relationship describing the excess risk of stochastic health effects (cancer and hereditary anomalies) following low levels of exposure to ionising radiation has been the subject of heated debate. The standard approach for the purposes of radiological protection is that the radiation-induced risk is directly proportional to the dose received [the linear, non-threshold (LNT) model], but some have argued that this approach underestimates the actual risk (i.e. the relationship is properly described by a supralinear curve), or that, in reality, there is a threshold dose below which either no effect, or even a beneficial (hormetic) effect, exists. Certain groups hold strong and entrenched views on this issue, and are vociferous in their criticism of the LNT model. This dispute between the ‘radiological protection establishment’ and its critics tends to leave those without particular expertise in the subject, including policy makers, bemused and perplexed, and it is difficult to avoid the thought that obfuscation might be an objective of some of the more campaigning of the dissenting groups. The present report of an ICRP Task Group is a timely review of the available evidence on the carcinogenic effect of low-level exposure to low linear energy transfer radiation, and collates and examines the findings from a range of relevant scientific studies.

Of course, the ideal solution to the problem of the nature of the dose–response relationship at low doses would be to derive the curve from fundamental biological principles, and basic radiobiological mechanisms do provide a rationale for the LNT model: at low doses and (for sparsely ionising radiations) low dose rates, the pertinent damage to DNA is caused (either directly or through free radical production) by independent particle tracks, so that the probability of non-lethal cellular modification is directly proportional to the number of tracks traversing cell nuclei (i.e. the dose). At higher doses and dose rates, the likelihood of track interactions increases to produce an upward turn in the dose response (although this does not occur for densely ionising radiations, a single track of which generates sufficient damage to DNA by itself). However, this simple and reassuring radiobiological picture is challenged by novel mechanisms: the bystander effect and genomic instability imply that damage occurs in cells that have not directly experienced a particle traversal, and the adaptive response suggests that cellular defence processes may modify the effects of protracted, relative to acute, irradiation. Just how these mechanisms, which undoubtedly exist under particular experimental conditions, might affect the risk of radiation-induced cancer and hereditary disease in humans is, of course, the primary question, but it is not a question that may be answered with conviction on present radiobiological evidence. Hence, there is a need to revert to epidemiological studies, with all their complications, in an attempt to derive an appropriate dose–response relationship; epidemiological data will incorporate all the relevant radiobiological mechanisms that have led to the specific health outcomes under study.

Unfortunately, epidemiological studies bring their own interpretational problems. Epidemiology is principally an observational (i.e. non-experimental) science that is based upon data generated by the uncontrolled conditions of everyday life, since randomised controlled trials are unacceptable for the study of (actual or potential) hazardous exposures. Further, the excess risk predicted by the LNT model to be produced by low doses of radiation is small. Consequently, any signal of an effect of low-level irradiation will be easily hidden by the background noise of statistical and systematic deviations from expectation, and epidemiological data for low doses will inevitably be consistent with a number of curves describing possible dose–response relationships. All is not completely lost, however, since the broad range of epidemiological evidence may be capable of constraining the dose–response relationship to lie within an envelope of curves. Ultimately, scientific judgement is also required in deriving the most plausible dose–response relationship. For example, it is inevitable that at some dose, the overall risk of a certain health effect will be compatible (at some conventional level of statistical significance) with the absence of a radiation-induced excess risk. What is to be made of this? Can we reasonably conclude that no excess risk exists below this dose? My view coincides with that of the late Sir Richard Doll, who dryly observed in 1997 in an opening conference address that he believed that ‘a linear dose–response relationship will not suddenly dive to zero immediately below the lowest level at which a statistically significant excess is observed’.

There is epidemiological evidence, mainly from studies of those medically exposed to x rays for diagnostic purposes, that the risk of cancer is raised following the receipt of doses of around 10 mGy, and that this increase is broadly consistent with the predictions of the LNT model. This evidence points away from a threshold dose, in particular because a cancer induced by a dose as low as ∼10 mGy of x rays is very likely to have been caused by the passage of a single electron through a cell nucleus. Further, if the risk from low-level exposure has been seriously underestimated by the LNT model, this should be apparent from the overall results of low-dose studies that are presently available; however, no such consistent pattern emerges. Of course, the evidence allows room for manoeuvre away from the LNT model at low doses, although only to an extent, and one might expect that different types of cancer have somewhat different dose–response curves; leukaemia is an obvious example. Nonetheless, the parsimonious choice of relationship for low-level exposures on the basis of the current evidence covering the generality of cancer induction, and one that has the decided advantage of practicality, is an excess risk that is directly proportional to the dose; the LNT model.

The evidence reviewed in the present report – the sophisticated treatment of uncertainties is especially impressive – and the inferences drawn from it should be paid serious attention by those arguing against the LNT model. Clearly, the future accumulation of additional information is highly likely to lead to further debate, but this must be evidence based rather than mired in dogma. One can only hope that this report will help to provide a firm foundation from which constructive discussion can progress.