Since its discovery in 1789 by Martin Klaproth, uranium has been the subject of numerous toxicological and health impact studies. The first of these was performed in 1828 by Gmelin [1], and among the most recent was a mortality and cancer epidemiological study by McGeoghegan and Binks of nearly 20000 individuals employed at the Springfields uranium processing facility, which appeared in the June 2000 issue of this journal [2], along with a companion study by the same authors of more than 12000 uranium enrichment workers at Capenhurst, which appeared in the December 2000 issue [3]. In between, there have been numerous studies of uranium toxicity and health hazards, the most extensive of which were carried out during the Manhattan District programme during World War II, and which have been well documented in the epic historical compilation by Stannard [4]. These and subsequent studies indicate that for low enrichments of uranium, chemical toxicity, approximately that of lead, predominates, and is the overriding consideration. By comparison, the radiological hazards are trivial, and radiation protection standards for natural and depleted uranium intake are based on potential chemotoxic effects on the kidney, inferred from animal studies. It is significant to note that there has never been a study, epidemiological or otherwise, that has demonstrated deleterious kidney effects in humans, although a recent Canadian study showed biomarkers, but no functional diminution, in older members of a population whose water had elevated concentrations of natural uranium [5]. And, as an aside, it is of at least passing interest to note that in the early 1900s, uranium was used therapeutically as a treatment for diabetes mellitus; there appear to have been no reports in the literature of kidney damage from this therapy.

So why, more than 200 years after its discovery, are we still studying the potential health risks of exposure to uranium? Clearly the answer is that we still have a great deal more to learn in this regard. And, the revelation that depleted uranium (DU) was used in munitions in the Gulf War a decade ago, and more recently in Kosovo, has led to renewed and intensified interest in the potential health effects of DU, and whether DU exposure is the cause of the so-called Gulf War Syndrome, as was discussed in the previous issue of this journal by Wakeford [6]. Reduced to its scientific basics, the problem devolves to a rather simple two-part question: first, whether these claims of illness are in fact valid, and, if so, whether it is indeed uranium that is the cause. Clearly in the vast majority of the claims, there is a confirmed medical diagnosis of illness which would seem to resolve the first part of our two-pronged question. The answer to the second part, however, is not so easily gained, and has implications for radiation protection practice generally as well as the more specific situation outlined above.

Thus the mortality and cancer morbidity studies of workers at Springfields and Capenhurst reported by McGeoghegan and Binks [2, 3], which spanned a half century of experience of two specific well-identified cohorts, represent a potentially important contribution to understanding of the potential long-term bioeffects of low-level uranium exposure in humans based on actual human exposure. As such these studies, which hold the promise of additional refinements based on organ doses, are a natural extension and enhancement of the numerous classical studies of uranium toxicity that have been carried out over the years. Particularly relevant to the inquiries into the hazards of depleted uranium is the earlier of the two referenced studies [2]. That the results are generally consistent with those of numerous other radioepidemiological studies that have been carried out over the years would seem to afford it a measure of credibility. However, doubtless this study, like every other radioepidemiological study, is imperfect and thus can be criticised for various shortcomings. Equally doubtless is that this study, regardless of its merit and its passage through the scientific peer-review process, will be criticised on methodological or other technical grounds in an attempt to discredit it or to reinterpret the data through additional and different, sometimes arcane and unconventional, forms of statistical analysis so as to support different conclusions, perhaps more in keeping with the bias or other expectations of the secondary analyser. (Parenthetically I might add that despite the absence of hard data to support this conclusion, there seems to be an increasing trend among scientists generally, and some radioepidemiologists specifically, that if the first statistical test does not support the conclusion that is wanted or expected, perform another statistical test, and then another until you get the desired result. And, should the desired result still evade you, consult another statistician. And, if even that fails, then obviously the data are flawed and should be discarded! Clearly this is not the case with the referenced papers by McGeoghegan and Binks, who are to be commended for their clear and concise evaluations and inclusion of tabulations of their raw data.)

A better and more constructive approach, at least initially, would be to first examine these studies and the results and conclusions drawn by the authors in a holistic fashion, employing the tried and tested old-fashioned Razor of Occam. In so doing, the reader should bear in mind some caveats. Epidemiological investigation is subject to numerous confounders, and thus has very real and often unknown limitations in discerning low-level effects in a relatively small population group. Ferreting out possible small effects in one group as compared with another is an exercise that is fraught with potential for error, and there is often a tendency to give too much credibility to a single unreplicated finding, particularly by those in the media and other non-scientists who may have a particular axe to grind. And, the reader should also bear in mind that statistical significance (or near statistical significance, a term well used in the studies) is an arbitrary construct, which needs to be interpreted with some caution. Had the authors, for example, chosen the 90 per cent confidence level, or the 99 per cent confidence level, instead of the more or less commonly accepted 95 per cent, they unquestionably would have come to additional and more definitive conclusions one way or another, depending on which value they chose. But, even so, the basic conclusion derived from the Occam's Razor analysis will not change.

The principal observation that jumps out at the reader using the holistic Occam's Razor approach is that there is clearly no wholesale or large-scale excess mortality or cancer morbidity in either the study cohort of radiation workers employed at Springfields or that at Capenhurst, as compared with their non-radiation worker peers and the general population of England and Wales. Indeed, the healthy worker effect was observed, with radiation workers faring better than the non-radiation worker cohorts with respect to cancer. The effects of uranium exposure in the study groups are, therefore, likely to be small if they are in fact present at all. Thus, by logical extension, it would seem that there is no imminent danger of an epidemic of uranium-induced suffering and death from the use of DU-containing munitions, and that the current protection standards for workers and the general public are at least reasonably adequate. To be sure, there are differences in mortality and cancer morbidity among the cohorts and subcohorts studied by McGeoghegan and Binks, but some of these observed differences may in fact be attributable to unidentifiable Type I statistical errors. It is also likely that many or even most of the observed differences are in fact real, and that some actual differences may have been missed (Type II statistical errors), but whether these are due to uranium exposure is another matter. Clearly the apparent excess morbidity and mortality of Hodgkin's disease and morbidity of non-Hodgkin's lymphoma in the Springfields radiation workers bears additional study and scrutiny, and this is the intent of the authors. So too does the apparent slight excess relative risk of renal tumours among the Springfields radiation workers, and bladder tumours among the Capenhurst cohort, perhaps especially in light of the knowledge that uranium is known to be renotoxic, at least in animals, and, by extrapolation, believed to be so in humans. The authors of the study have been appropriately cautious in the wording of their conclusions, and the organ-specific dose trend analysis they propose to carry out will more likely than not shed additional light and lead to more definitive conclusions with regard to the potential health impacts of low-level, low-enrichment uranium exposure.

Returning to the allegations of significant health effects from the DU residuum of the Gulf War and Kosovo actions, it strikes this observer as odd that there seemingly have been no systematic long-term epidemiological follow-up studies of the potentially or allegedly affected populations, directed towards determining if in fact DU exposure is linked to the alleged excess of illness and discomfiture in these populations. But that is another story, and perhaps a suitable topic for another editorial. Certainly the two articles in recent issues of the journal, along with the follow-on work proposed by its authors, will serve as a welcome addition to our understanding of potential bioeffects from low-level exposure to uranium, and in addition should provide a rational scientifically-based yardstick by which to measure the likelihood that DU is the cause of the so-called Gulf War Syndrome (politicians and journalists take note!), as well as a measure of solace to those who worry that their illnesses are in fact the result of low-level exposure to DU.