Table 1. Increasing specific activity (MBq/kg) of DU due to ingrowth of daughters.

Over centuries, the specific activity of U-234 should be the same as that of the parent U-238, and thus the environmental concentrations of these isotopes is generally the same if the source is natural. The specific total activity is thus about 37MBq/Kg. It should be pointed out that DU material recently found in battlefields in Europe contains small quantities of isotopes of Plutonium, Neptunium and other fission products: thus the source of this DU is refinement of nuclear reactor waste. However, the quantities are very small and are not considered by the authorities to be of serious radiological significance.

            Owing to the high density of Uranium, (19g.cm-3 metal and 10.96 g.cm-3 for the dioxide) and the fact that the metal is pyrophoric (burns in air) the substance is used in the manufacture of armour piercing shells, missile nose cones and penetrators and certain ballast materials in some aircraft (e.g. helicopter rotors, commercial aircraft counterweights). A single Abrams 120mm tank shell contains about 3kg of DU (111MBq of radioactivity) and there is 275g in a 30mm GAU3A A-10 Thunderbolt Gatling Gun round.

The military penetrators explode on impact with hard targets with about 80% conversion to micron diameter Uranium Oxide particles of a ‘ceramic’ nature. These particles are highly mobile and extremely long lived in the environment, owing to the very high degree of insolubility of Uranium Oxides UO2 and U3O8. They can be inhaled and the sub-micron diameter particles are translocated from the lung to the lymphatic system, building up in the tracheobronchial lymph nodes and potentially able to circulate everywhere in the body. Alpha and beta disintegrations from these particles cause very high and repetitive doses to cells local to the range of the disintegration i.e. about 30microns for the alpha and 450 microns for the beta tracks.

3. Errors in the ICRP low level radiation risk model

The model used by the risk agencies and the military to predict the health consequences of such exposure is that of the ICRP and is based on the cancer yield of the Hiroshima bomb. The group of survivors of this single large acute irradiation exposure (in which many people were killed) have been collected into the Life Span Study or LSS and their cancer rates have been compared with controls from the same town who were shielded or outside the town at the time of the bomb. The cancer yield in this LSS cohort has been used as a basis for predicting cancer risk and other health detriments for all types of radiation exposure. It has been assumed that the relationship between dose and cancer yield is linear and so low levels of exposure have been assumed to carry no significant risk on this basis.

            This approach has been criticized extensively, and considerable evidence has become available in the last twenty years to suggest that the increases in cancer and leukemia near nuuclear sites are examples of a failure of the model to adequately address risk from internal radiation. However these arguments have always been countered by the risk agencies on the basis that other possible causes for the observed phenomena exist. However, very recently, two unequivocal pieces of evidence have defined errors of between 100 and 2000-fold in the ICRP risk models as applied to internal radiation risk. This evidence has forced the UK government to set up a new committee to examine the situation and assess the failures of the ICRP risk model applied to internal radiation exposure [CERRIE, 2001].

 In addition, the European Parliament has called for a similar process to be undertaken by the European Commission [EU,2001], and the recent WHO conference on Chernobyl in Kiev in 2001 came to a similar conclusion [WHO 2001].

3.1 The Chernobyl Infants

Following the Chernobyl accident in 1986, in five different countries, the cohort of children who were exposed in their mother’s  womb to radioisotopes from the releases suffered an excess risk of developing leukemia in their first year of life. This ‘infant leukemia’ cohort effect was first reported in Scotland [Gibson et al, 1988], and then in Greece [Petridou et al, 1996], in the United States [Mangano, 1997] and in Germany [Michaelis, et al. 1997]. We first reported increases in childhood leukemia in Wales and Scotland following the Chernobyl accident in 1996 [ Bramhall, 1996] but more recently examined the specific infant leukemia cohort in Wales and Scotland [Busby and Scott Cato 2000].

Unlike the earlier researchers, who merely showed the existence of a significant rise in infant leukemia, we decided to examine the relationship between the observed numbers of cases and those predicted by the present radiation risk model. This was an invaluable opportunity since the specificity of the cohort enabled us to argue that the effect could only be a consequence of the exposure to the Chernobyl fallout. There could be no alternative explanation, like the ‘population mixing hypothesis’ advanced to explain away the Sellafield childhood leukemia cluster. However implausible such theories may be, they have acquired popularity, and their proponents status, as a consequence of their utility to the nuclear lobby. However, population mixing may not occur at Sellafield but it cannot occur in the womb.

 Because the National Radiological Protection Board had measured and assessed the doses to the populations of Wales and Scotland and because they themselves had also published risk factors for radiogenic leukemia based on ICRP models it was a simple matter to compare their predictions with the observations and test the contemporary risk model. The method simply assumed that infants born in the periods 1980-85 and 1990-92 were unexposed, and defined the Poisson expectation of numbers of infant leukemia cases in the children who were in utero over the 18 month period following the Chernobyl fallout. This 18 month period was chosen because it was shown that the in utero dose was due to radioactive isotopes which were ingested or inhaled by the mothers and that whole-body monitoring had shown that this material remained in the bodies of the mothers until Spring 1987 because silage cut in the Summer of 1986 had been stored and fed to the cattle in the following winter.  The result was startling. First, there was a statistically significant 3.8-fold excess of infant leukemia in the combined Wales and Scotland cohort (p = 0.0002). Second,  the leukemia yield in the exposed ‘in utero’ cohort was about 100 times the yield predicted by the model. Table 2 compares the effect in the three main studies.  In passing it should be noted that this number, 100, is very close to the error required to explain the Sellafield childhood leukemia cluster.

            It should be noted that the possibility of the effect being due to chance may be obtained by multiplying the p-values for the null hypothesis that the effect was due to chance in each of the separate countries and studies to give an overall p-value less than 0.0000000001. Thus it was not a chance occurrence: it was a consequence of the exposure to low-level radiation from Chernobyl.

And since the World Health Organization has given approximate exposure levels in Greece, Germany and the United States, it was also possible to examine the leukemia yield in the infant ‘exposed cohort’ reported by the several other studies and establish a dose response relationship. This is shown in Fig 1. It is a curious shape and goes up, down and up again, and this shape should be noted. I will return to it below.

3.2 Minisatellite DNA in Chernobyl children

Since the discovery of the DNA minisatellite characterisation method, ‘DNA testing’ it has been increasingly applied to those who were exposed to the fallout from the Chernobyl accident. In a series of papers, Dubrova et al. showed an association between exposure of children in Belarus [Dubrova et al.1997] found a doubling in the mutation rate in children from the high exposure territories of Belarus compared with controls from low exposure territories. This discovery was astonishing to those who adhered to the ICRP risk model for genetic mutation since this was based on the belief that the Hiroshima exposures, which were hundreds of times higher than the average dose in Belarus, had produced no genetic effect on any offspring of those exposed. A doubling of the mutation rate thus pointed to an error of some 1 x 105 . Others pointed out that even if the minisatellite DNA was mutated, this was not an effect which had any significance since there were no phenotypical changes associated with minisatellite DNA. Shortly after this, it was reported that barn swallows which migrated to Belarus had similar changes in their minisatellite DNA and these were assocated with plumage pattern alterations which destroyed their camouflage and thus might be harmful. [Ellegren et al. 1997]

The question of proper controls and the reality of the effect was answered very recently in an elegant study by Weinberg et al.[2001]. They examined minisatellite DNA changes at various loci in the offspring of the Chernobyl ‘liquidators’ who were born after the accident and compared their DNA to their siblings born before the accident. Results showed that there was a significant difference of up to seven-fold. The dose response relationship appeared to be biphasic. Based on the natural mutation rate in the minisatellite DNA, the finding showed an error in the ICRP risk factor for mutation of between and 700-2000 fold. This series of studies thus demonstrates finally and unequivocally that the ICRP risk model for internal exposure is wildly inaccurate.

I must ask how it is that some fifty years after the atom bomb, and following a huge amount of research into the subject, we can have discovered such a huge error in the science of radiation risk. To understand the answer, we must look at the scientific method a little more closely.

4. Radiation Risk and Scientific Method

The classical exposition of the scientific, or inductive method (originally due to William of Occam) is what is now called Mill’s Canons, the two most important of which are:

·    The Canon of Agreement which states that whatever there is in common between the antecedent conditions of a phenomenon can be supposed to be the cause, or related to the cause, of the phenomenon.

·    The Canon of Difference which states that the differences in the conditions under which an effect occurs and those under which it does not must be the cause or related to the cause of that effect.

In addition, the method relies upon the Principle of Accumulation which states that scientific knowledge grows additively by the discovery of independent laws, and the Principle of Instance Confirmation, that the degree of belief in the truth of a law is proportional to the number of favourable instances of the law.

Finally to the methods of inductive reasoning we should add considerations of plausibility of mechanism.

These are the basic methods of science [Mill, 1879; Harre, 1985; Papineau, 1996]

            Let us first define our question. It is this. What are the health consequences of exposure to novel internal radioisotopes at whole organ dose levels below 2mSv?  Because we are looking at battlefield DU, we should add that in this case, although the element is ‘natural’, the exposure is novel, and due to internal sub-micron Uranium Oxide particles embedded in tissue.

Although risks from exposure to high levels of ionizing radiation are generally accepted, since they are fairly immediate and graphic, the situation with regard to low-level exposure is curious. There are now two mutually exclusive models describing the health consequences of exposure to low-level radiation. There is a nuclear establishment one, which is that which is presently used to set legislation on exposures and argue that DU is safe, and a radical one, which is espoused by the anti-nuclear movement and its associated scientists. I show these two models schematically in Fig 2.

                        The two models arise from two different scientific methods. The conventional model is a physics-based one because it was developed by physicists prior to the discovery of DNA. Like all such models it is mathematical, reductionist and simplistic, but because of this is of great descriptive utility. Its quantities, dose,  are average energy per unit volume or dE/dV and in its application, the volumes used are greater than 1kg. Thus it would not distinguish between the average energy transferred to a person warming themselves in front of a fire and a person eating a red hot coal. In its application to the problem at hand, the internal, low-level, isotopic or particulate exposure, it has been used entirely deductively. The basis of this application is that the cancer and leukemia yield has been determined following the external acute high-dose irradiation by gamma rays of a large number of Japanese inhabitants of the town of Hiroshima. Following this, arguments based on averaging have been used (quite spuriously) to maintain that there is a simple linear relationship (in the low-dose region) between dose and cancer yield. This Linear No Threshold (LNT) assumption enables easy calculations to be made of the cancer yield of any given external irradiation.

By comparison, the radical model shown in Fig.2 arises from an inductive process. There have been many observations of anomalously high levels of cancer and leukemia in populations living near nuclear sites, especially those where the measurements show that there is contamination from man-made radioisotopes, e.g. reprocessing plants. In addition, populations who have been exposed to man-made radioisotopes from global weapons tests, downwinders living near nuclear weapon test sites and those exposed to these materials because of accidents (like the Chernobyl infant leukemia cohort) or because of work in the nuclear industry or military. A review of these findings is available [Busby, 1995] and a more recent literature review of studies showing these effects if published by the Low Level Radiation Campaign [LLRC, 2000]. In addition, the radical model is based on biological considerations and considers each type of exposure according to its cellular radiation track structure in space and in time. It is not, therefore, possible to employ this model to predict risks from ‘radiation dose’ to ‘populations’ but only from microscopically described doses from specific isotopes or particles whose decay fractionations are considered to interact with cells which themselves respond biologically to the insults and may be in various stages of their biological development.  The dose-response relationship following from this kind of analysis might be expected to be quite complex.

These models are mutually exclusive: which one is correct? What considerations can we use to choose?

The answer is that the  conventional LNT model must be rejected because it is not scientific. Its conclusions are based on deductive reasoning. It falsely uses data from one set of conditions, high-level, acute, external exposure to model low-level, chronic, internal exposure. It is scientifically bankrupt, and were it not for political considerations, would have been rejected long ago. On the other hand, it should be clear that the radical model conforms to all the requirements of the scientific method listed above. Man-made radioisotopes, often in the form of ‘hot particles’ are common contaminants to the areas near nuclear sites where there are cancer and leukemia clusters, and to the downwinders, and to the fallout-exposed populations. This satisfies the Canon of Agreement. The contingency analysis tables with control populations for such studies show that the Canon of Difference is also satisfied: people living in more remote regions than the downwinders show lower levels of illness. We must by now also have some faith in a Principle of Instance Confirmation, since so many studies have shown that increases in cancer and leukemia follow these exposure regimes at low dose. Indeed, the Gulf War Syndrome, might be considered as such an instance confirmation. We are left only with ‘Plausibility of Mechanism’, which will be addressed briefly below.

5. Mechanistic Considerations

 Averaging Dose

I want to look more closely at the averaging model and its predictions at low dose.  It is essentially what used to be called a colligative model: the dE/dM formulation of dose requires that energy transfered from absorption of the consequence of a radioactive disintegration is averaged over the target site, usually the whole body or organ. Whatever lip service is made to considerations of what is now called ‘microdosimetry’, close examination of calculations done to establish risk near nuclear sites shows this to be the case. The documents NRPB R-276, Risk of Leukemia and other Cancers in Seascale from All Sources of Radiation published in 1995 is a good example. In this document, doses to the lymphatic system were calculated by modelling it as ‘liver, lung, kidney, spleen, pancreas, uterus and intestines’. A physiologist would not recognise this list as the ‘lymphatic system’, so why was it used? The answer is that breathing introduces the particles of plutonium that exist in the air near Sellafield into the lungs of the children who live there. From the lungs, these particles are scavenged to the two small tracheobronchial lymph nodes which have a combined mass of perhaps one gram. If NRPB had divided dE by 1 gram, the resultant dose to this part of the lymphatic system would have been extremely high. Given that this organ has been identified as a source of lymphoma and leukemia in animals, this sounds very like the cause of the Sellafield leukemia cluster. But dilution of the plutonium decay energies into the whole mass of guts used for dM reduces the ‘dose’ to an acceptable small level. This process, incidentally, is very relevant to the DU exposures.

            Figure 3 shows a phantom used by ICRP to calculate doses from external radiation fields. This is the model that is presently used to calculate internal doses. Of course, in the low dose region, cells are either hit or not hit, so the cell dose is very different from the tissue dose. Nevertheless, the model is valid as a means of establishing a quantity, ‘dose’ which can be correlated with some health consequence like cancer, so long as each cell in the body, or target region, has an equivalent probability of being hit (or more properly intercepted by a track). Dudley Goodhead has written of the low-dose region [Goodhead, 1988]:

Most situations of practical interest are characterised by cells receiving occasional single tracks well separated in time from any other tracks which may impinge on the same cell. From Natural Background, there is, on average, about one track per year through each cell nucleus. Therefore it is highly unlikely that there will be multiple tracks in short times (< 1 day) over which repair of radiation induced damage within cells is usually observed to take place.

 It is these (essentially external irradiation) considerations that enable the model to assume the linear dose response relationship that is the basis for radiation risk. But there are two situations of practical interest that Goodhead’s arguments do not address. The first is that a cell’s response to radiation damage is not constant over its lifespan: cells are very sensitive to radiation when they are in their repair and replication cycle. The second is that for internal radionuclide decays, either from sequential emitters or from ‘hot particles’ the microscopic local radiation flux, or energy density, may be very high, even though the average dose may be low. For internal exposure, these are common situations. Here the concept of ‘dose’ no longer applies and the conventional model breaks down. I will address these in turn.

Cellular responses to radiation: the Burlakova dose response

It has been known from almost the beginning of the radiation age that rapidly replicating cells are more sensitive to radiation damage [Bergonie and Tribondeau, 1906].  Indeed, this is the basis of radiotherapy for cancer where it is the rapidly proliferating cancer cells that are preferentially destroyed. Most cells in a living organism are in a non-replication mode, sometimes labelled G0. These cells are contributing to the organism as part of the normal living process and do not need to replicate unless there is some signal requiring this, perhaps because of tissue growth, damage or senescence. Throughout the growth and lifespan of individual organisms, there is a constant need for cellular replication, and therefore there are always some small proportion of cells which will be replicating: the magnitude will naturally  depend upon the type of cell.  When cells receive the signal to move out of stasis or G0, they undertake a fixed sequence of DNA repair and replication, labelled G0-G1-S-G2-M, with various identifiable check points through the sequence which ends in replication M or Mitosis. The period of the repair replication sequence is about 10 to 15 hours and the sensitivity of replicating cells to damage including fixed mutation is extremely high at some points during this sequence. This has been known for some time: Fig 4 shows the results of early experiments on Chinese hamster cells indicating up to 600-fold variation in the cell radiation sensitivity over the whole cycle. [ Morton and Sinclair, 1966] If we display this response variation on a scale that shows the normal cell lifespan in the organism, rather than just over the cell cycle in vitro, the window of opportunity for cell mutation at high sensitivity becomes apparent Fig 5.

            So the picture of isotropic dose to equivalent cells, the ‘bag of water’ phantom model outlined by Goodhead has to be reviewed. Perhaps 1 percent of these cells are actively dividing and are in repair replication sequences that we will assume, for argument, are 600 times more sensitive to being ‘hit’ by a track. What would we expect the dose-response to look like? Well as the dose was increased from zero, the sensitive cells would begin to be damaged and a proportion of these hits would results in fixing a mutation and increasing the possibility of cancer. As the dose increased further, eventually this rise in response would peak as these sensitive cells were killed. The mutation yield would then begin to fall. However, at some point, the insensitive G0 cells would begin to be damaged and the whole process would begin again, with a rise in cancer. Ultimately there would be a second fall, but this level of exposure would probably result in the death of the organism (although such considerations have been used to explain an observed fall-off in effect from alpha emitters at high dose). So the dose response would look like that in Fig 6.  This type of response was shown to occur in several experiments by Burlakova, although she gave a different explanation for it, involving a combination of increasing damage and induced repair curves.

She showed that such an effect can be seen by plotting the results of a large number of separate radiation and leukemia studies, a graph reproduced in Fig 7.

            The results of animal studies on beagle dogs and mice also show these biphasic effects in the low-dose region [Busby, 1995] . Note that this type of curve is seen in the Chernobyl infant studies collected together in Fig 1.   

The Second Event Theory

 There is large variation in sensitivity over the cell lifespan. Although naturally dividing cells may accidentally receive a ‘hit’, this process can be modelled by averaging over large masses of tissue, even if the dose response curve is not linear, as thought. However, unplanned cell division, preceded by DNA repair can be forced by a sub-lethal damaging radiation track: this is one of the signals which push the cell out of G0 into the repair replication sequence. It follows that two hits, separated by about eight hours, can generate a high sensitivity cell and then hit this same cell a second time in its sensitive phase. This idea, the ‘Second Event Theory’ is described and supporting evidence advanced in Busby 1995 and its mathematical description has been approached slightly differently in Busby 2000. It has been the subject of some dispute by NRPB (Cox and Edwards, 2000, Busby, 2000a) 

Very recently, developments in micro techniques have enabled some new evidence that supports the two hit idea to emerge. Miller et al., [1999] in a consideration of Radon exposure risks, have been able to show that the measured oncogenicity from exactly one alpha particle hit per cell is significantly lower than for a Poisson distributed mean of one alpha particle hit per cell. The authors argue that this implies that cells traversed by two alpha particles or more contribute most of the risk of mutation, i.e. single hits are not the cause of cancer.

There are two types of internal exposure for which there would be expected to be an enhancement of risk from this Second Event source. The first, due to sequentially decaying radioisotopes like Strontium-90 has been discussed in Busby 1995, Cox and Edwards, 2000 and Busby, 2000.  Following an initial decay from an Sr-90 atom bound to a chromosome, the second decay from the daughter, Yttrium-90, whose half-life is 64hrs can hit the same cell in the induced replication sequence with a probability that is simple to calculate. The same dose from external radiation has a vanishingly small chance of effecting the same process. The second type of  Second Event exposure, referred to in Busby 2000a, is from micron or sub-micron sized ‘hot particles’. If lodged in tissue, these will decay again and again increasing the probability of multiple hits to the same cell inside the 10 hour repair replication period. It is this process that is relevant to the Depleted Uranium problem.

Second Events from DU particles.

 The US Defence Department commissioned research into the levels of Uranium Oxide particulates produced by the impact of Abrams M1A1 Tank ammunition at the Nevada test site in 1986 [USBRL 1986]. The impact on armour of Depleted uranium penetrators results in about 80% conversion to Uranium Oxides UO­2 and U­­3O8 in the form of ceramic particles of diameters in the micron region. These aerosol particles are very mobile and can clearly be inhaled. In this regard the hazard is of a similar nature  to that from the Plutonium oxide particles resuspended from Sellafield discharges to the Irish Sea which were considered as a possible cause of the Sellafield leukemia cluster by COMARE and NRPB and referred to earlier where it was recorded that the ICRP66 models used to estimate doses did so by diluting the particles energy into large masses of tissue.

            For particles below 1 micron diameter, self absorption of the alpha particle decays may be considered second order and the dose to tissue in the range of these alpha decays calculated.  Table 3 shows the calculated doses in spheres of tissue within the 30micron range of the alpha decays. Also tabulated is the number of hits per day to this sphere of tissue. The table shows that for particles as small as 0.2 microns diameter, average annual alpha dose to the (lymphatic) tissue surrounding the particles is about the same as the total annual average background dose of 2mSv. For larger particles the dose rapidly increases. Between 0.5 and 5 microns, Second Event processes are stochastically likely. This is shown by Fig 8 where the number of hits per day is plotted against the particle diameter.

            These ‘hot particle’ processes have been known about for a long time: Fig 9 shows a radiographic photomicrograph of a plutonium oxide ‘hot particle’ in lung tissue. Overlapping tracks can be seen.

Energy density and risk

The consequence of aggregating decays into a small sphere around a ‘hot particle’ is, of course, that the number of different cells capable of being hit elsewhere is necessarily reduced: we have converted a number of tracks well separated to the same number of tracks close together. If all tracks carry the same risk of mutation in cells in the track, i.e. all hits are equivalent, then there should be no hazard enhancement. The hazard enhancement proposed arises not from some ‘hot coal’ type of energy concentration process but from the fact that cells may be triggered into a sensitive repair replication sequence which carries a very high sensitivity weighting. It may, of course be true that there would be other reasons why concentrated irradiation of a small cluster of cells could produce unstable cell replication or cell communication fields such as those recently proposed by Sonnenschein and Sato [1999] and this itself may lead to a tumour promotion advantage but this is another matter.

Beta emissions from DU

Before collecting together these considerations there is one further matter which may have been overlooked in the case of DU.  It was pointed out that Uranium-238 is an alpha emitter but depleted Uranium is also a beta emitter: indeed in the solid form the two beta-emitting daughter isotopes, Thorium-234 (beta; 0.26MeV, 24 days) and Protoactinium-234 (beta 0.23MeV, 6,75 hrs) are in equilibrium with the parent after 20 weeks (Table 1). These beta emissions are the main radiological hazard in handling the bulk material. In Iraq, I recently measured 24,000 counts per second at the surface of a stray A-10 30mm penetrator which was just lying on the ground. This represented a dose of about 1mSv/hour to the hands of anyone holding the penetrator. However, most of the beta (and alpha) decays were absorbed inside the bulk material, and only surface disintegrations were emerging to be absorbed in the scintillation counter head.

The equilibrium beta activity of DU is about 37MBq/kg. But most of this energy is absorbed in the bulk material: oxidation of the material on impact to produce some 1014 1 micron diameter Uranium Oxide spheres per kilogram would enable all of the decay energy to be potentially available for human exposure. The enhancement of efficiency in release of beta radiation is thus greater than 1000-fold.

Environmental Mobility of the DU particles

In order to be define the population at risk, it is necessary to know the fate of the Uranium particles subsequent to impact. At the Nevada test site, the atmospheric concentration at 100m from impact exceeded the UK NRPB Generalized Derived Limit for Uranium in Air by a factor of about 5 [Busby 1999]. Dietz  has reviewed data which establishes that DU particles are able to travel at least 100km from their impact source [Dietz, 1997]. I recently made measurements of alpha radiation levels in Iraq in three areas, the southern battleground near tanks destroyed by DU fire, the same area remote from the tanks, the town of Al Basrah and the city of Baghdad. Results showed that the alpha activity in the battleground area was more than five times higher than in Basrah and ten times higher that in Baghdad. In addition, and remarkably, levels on the surface of the ground near the damaged tanks did not generally  show high  levels of alpha or beta signal from Uranium and its daughters except in the case of one tank where a yellow contaminant, probably UO3, showed high levels of beta activity. In addition, the insides of tank turrets which had radioactive holes in them from A10 hits, did not show high levels of beta or alpha activity. The generally higher alpha levels in the whole area, coupled with these observations suggest that the Uranium particles has been efficiently dispersed by some mechanism. I believe that this mechanism is the repulsion of charged particles by themselves and by the earths permanent electric field of 150V/m. I have argued elsewhere that this effect operates in the Kennet Valley near the Atomic Weapons plant at Aldermaston and results in the preferential concentration of charged radioative particles near electrostatic discontinuities between strata with different conductivity [Busby, 1997] . A similar effect near high voltage power lines was recently found by Henshaw et al. [1999].

Conclusions on Mechanism   

Thus we can conclude that the external bag-of-water model is not an accurate representation of the kind of processes that occur at the cellular level and that the physics-based descriptions do not apply to internal irradiation. The Uranium Oxide particles are capable of travelling very large distances [ Deitz, 1997].  They may then be inhaled and will become trapped in the lymphatic sytem where they may be transported to any part of the body. Here they may cause sequential moderate dose irradiation of local tissue volumes where the risk of mutation is far higher than is suggested

The enhancement of mutation efficiency that follows from exposure to inhaled Uranium oxide hot particles is capable of explaining the ‘anomalous responses to low dose exposure’ found near Sellafield and other nuclear sites and also ‘Gulf War syndrome’ etc. We are not, however, reduced to looking only at the Gulf War Syndrome and the Iraqi children for supporting evidence though I shall return to these later. There are other indicators, and our springboard for these is the 1983 observation of a childhood leukaemia cluster at Sellafield. In the last four years Green Audit been funded by the government of the Republic of Ireland  to study cancer incidence close to the Irish sea. The study has used both Wales Cancer Registry and Irish Cancer Registry data to examine and explain variations in cancer risk with distance from the sea. The results of this work will be published elsewhere but since they cast considerable light on the DU problem, some of the findings will be briefly reviewed here.

6. Sea coast cancer risks and resuspended hot particles.

In three separate investigations between 1997 and 2000, Green Audit discovered profound and statistically significant evidence of excess risk of cancer incidence and mortality in coastal  populations in Wales, Ireland and Somerset. The excess risk has been found for most of the cancer types and sites and in the following data:

·    Incidence data for small areas in Wales from Wales Cancer Registry from 1974-89

·    Incidence data for small areas of Ireland from the Irish National Cancer Registry for 1994-1996.

·    Mortality data for census wards in  Somerset from the Office for National Statistics for 1995-1998

In each area the trend with distance from the sea shows a sharp rise in the group of people living within 800m of the sea coast. It is driven by proximity to areas of intertidal sediment known to be contaminated with radioisotopes from Sellafield discharges. In the case of the Somerset study, which was investigated as a hypothesis test the drying, offshore, mud bank, known as the Steart Flats, was contaminated by historic releases from the adjacent Nuclear Power site at Hinkley Point.

As one example of the type of result found, the all malignancy relative risk for populations of all ages in Wales from 1974-89 is shown plotted against distance from the Irish Sea in Fig 10. Note the sharp rise in risk near the coast.  Sufficient evidence has now accumulated from these studies to support the hypothesis that this cancer risk is a consequence of an exposure route involving inhalation of resuspended radioisotopes, particularly Plutonium Oxide particles. The trend in concentration of Plutonium with distance from the sea in Cumbria has been established and is shown in Fig 11. The radioactivity is brought inland by seaspray scavenging mechanisms which are quite  well understood: indeed, the ocean is the source of about 30% of all PM10 particles in the UK.   It is therefore not surprising that NRPB workers found Plutonium in the tracheobronchial lymph nodes of autopsy specimens from all over the UK in proportion to their distance from the west coast, particularly Cumbria [Popplewell, 1986]. Table 4 shows some results of these studies, which were, incidentally, omitted from the considerations in the COMARE report on the Sellafield leukemia cluster. Nor is it surprising that Plutonium is found in children’s teeth in the UK at levels which reflect a similar trend with distance from the Irish Sea [Priest et al, 1996]

7. Recent evidence on DU exposure risks and response by UK government

DU, leukemia, cancer and birth defects  in Iraq

There have been reports from within Iraq of serious health problems emerging after the Gulf War. These problems are apparent in the soldiers, in civilian adults living in the south near the war zone and also in children. They take the form of a range of conditions similar to those categorised as ‘Gulf War Syndrome’ in the US and UK veterans and also in large and significant increases in cancer and leukemia in adults and children and also birth defects including novel types of birth defect. I visited the country in september 2000 with Al-Jazeera TVand toured the hospitals in Baghdad and Basrah, speaking to senior doctors and health service researchers. Cancer registry data reflect the increases in cancer and show that the main increases are also in the parts of the country, south and north of Baghdad, where DU ammunition was mainly used. Significant pieces of evidence are the first, the geographical pattern of cancer and second  the cohort effect in childhood leukemia which shows the main excess in the cohort aged 5-9 in 1998. This is an unusual finding for childhood leukemia which normally peaks in the 0-4 age group and indicates that it was the war birth cohort that showed the greatest leukemia effect. The geographical pattern of cancer also broadly correlates with the measurements I made of alpha activity in air in the country, which again reflects the distribution of DU based on the areas where the material was mainly used.

DU in Kosovo

 No cancer data is available in Kosovo owing to the large changes which have been taking place there after the war. I was able to visit western Kosovo in January 2001 with Nippon TV and we used UN maps supplied by the Italian Army to locate areas where DU had been used. Using a survey scintillation counter I found areas where high beta counts indicated the presence of significant amounts of DU and took samples for analysis by alpha and gamma spectroscopy and also thermal ionisation mass spectrometry. Two main conclusions could be drawn from the results, which are shown in Table 5.

            First, some 18 months after its use, significant quantities of DU either were resuspended in or remained suspended in the atmosphere to be precipitated with snow and to pool under the snow when it melted. The ratio of daughter isotopes to parent U-238 was remarkable. Instead of there being a 1:1:1 equilibrium ratio, the activity of U-238 in the sample was much smaller than the activity of the daughter isotopes. Since Uranium is largely insoluble (or would not have been there if it were soluble) this result shows that the Uranium particles had become resuspended between the time the snow melted and the time I measured the activity (about 2 weeks)

UNEP report on Kosovo

Following concerns about the possible health effects of radioactive contamination from Depleted Uranium weapons used by NATO in the actions in Kosovo in 1999,  a number of scientists and experts were assembled under the auspices of the United Nations Environment Programme to visit Kosovo  between 5-19th November 2000 to investigate levels of contamination and report on possible health hazards. Details of the expedition and its protocols and findings are to be found in the report [UNEP, 2001]. I have analysed their findings in a presentation to the European Parliament in Strasbourg in 2001 [Busby 2001, www.llrc.org]  but will  briefly outline UNEPs findings and their conclusions.

UNEP made three main claims relating to their findings.

a.   There was no widespread dispersion of DU in areas of Kosovo where the shells were fired. DU measurements showed only local contamination, i.e. there was no evidence of DU further than 10-50 metres from a direct hit site.

b.   There was no contamination of water sources.

c.   There was no health hazard to humans anywhere with the possible exception of some slight danger from handling shell fragments for a long period.