|
Chapter Seven Health Survey of Former British Atomic Tests Participants: The Australian Commonwealth government initiated this investigation at the request of Australian nuclear veterans’ organisations. They requested the study because veterans for a long time had believed that their complaints of radiation inflicted injuries and disease in themselves and their offspring were not being given sufficient consideration by Federal Government authorities. However veterans have claimed that the findings of the investigation are flawed and as will be seen, these claims are supported by arguments given in this chapter. It was decided that the in order that the reader be given an opportunity to decide the issue, the report should be presented in its entirety and separately, rather than be incorporated into another chapter, with arguments of invalidity presented in the conclusion at the chapter’s end. It is of course, the reader’s prerogative to decide whether claims of flawed results are reasonable however, an affirmative answer will beg the question of why any responsible government would allow a report containing invalid conclusions be published on so important and controversial an issue as nuclear veterans’ health.
In November 1983 two reports, (Donovan et al 1983) contained in the same booklet were released, one, “ Survey Of Health Of Former Atomic Test Personnel” which was based on findings of a survey of 2,536 partially or fully completed questionnaires returned by atomic test veterans, and the second, “Causes Of Death Of Former Atomic Test Personnel” used death certificates of 1560 deceased veterans. The aim of the reports which were compiled by Dr J W Donovan of the Commonwealth Department of Health, was to “to identify any associations between atomic test involvement and subsequent illness” in the case of the former and “identify any associations between participation in the test programs and diseases causing death” with the latter. Because of the difference in purposes and the methods used – modelling techniques for the first report and modelling and control group comparison for the second, the reports will be discussed separately. An estimated total of 15,363 Australians were involved in the atomic tests. |
|
Survey of Health of Former Atomic Test Personnel The aims of the study were (Donovan et al 1983, p. 3): (a) To identify any associations between atomic test involvement and subsequent illnesses among Australian personnel, and in particular to compare the pattern of known radiation-related illnesses among the different groups involved; (b) To compare the illness experience of test personnel with that of the Australian population at large, to the extent that this is possible; (c) To identify test personnel seeking information and counselling; (d) To report in a form suitable for tabling in Parliament by mid to late 1983.
Because of the “the healthy soldier effect” (Rabbitt-Roff 1997, p. 15) difficulties were encountered in finding a control group suitable for the study, making (b) unattainable. This effect is a phenomenon stemming from the bias of choosing particularly fit and healthy candidates for arduous duty and/or military service, which contaminates results, the argument being that very healthy people are more able to shrug off threatening diseases than one of average or even poorer than average health. It followed then, that any control group chosen needed to consist of members of the same above average health and fitness over the tests period, for the comparison to be valid. Consequently, the researchers resorted to an analytical survey using linear-logistic modelling techniques to make comparisons among sub-groups of veterans in order to discover differences from the expected norm, and find explanations as to why they occurred. Factors that may influence the occurrence of particular diseases, for example age or whether there was a previous history of diabetes in the case of cataracts, were taken into consideration by the model. As previously stated, the data for the survey were obtained from 2440 fully or partially completed questionnaires of confirmed medical conditions. When there was no established proof of a claimed disease ie in cases of self-diagnosis, professional confirmation where possible was sought, otherwise the claim was excluded.
Prior assumptions by the researchers was that: (a) Sub-groups with a history of radiation exposure have no greater prevalence to any of the following diseases than a sub-group without such a history of radiation exposure, (Donovan et al 1983, p. 5) where the diseases are: All malignant diseases Lung cancer Leukemia and allied disorders Thyroid cancer Malignant melanoma Other skin cancers Cataracts Infertility where no cause is documented in wife
(b) It is assumed that concern over health does not vary with radiation exposure (Donovan et al 1983, p. 3).
A questionnaire was mailed to each member of both the main and self-identified groups resulting in 2536 acceptable survey candidates who had partially or wholly completed the questionnaire form. Table 7.1 shows the response categories of the total 8255 recipients. (Donovan et al 1983, p. 15). |
|
Table 7.1 Response Status to Questionnaire
Response Main Group (%) Self-identified (%) Total Returned to sender 467 6 21 9 488 Completed 2440 30 96 41 2536 Not involved 3835 48 35 15 3870 Blank 46 1 1 0 47 Dead 109 1 12 5 121 Refusal 19 0 2 0 21 No part in tests 49 1 2 1 51 Duplicate 14 0 28 12 42 No response 1039 13 14 17 1079
Total 8018 100 237 100 8255
Where “not involved” are those who claimed not to have been irradiated as opposed to the non-attendance of those who had “no part in tests”. Several in the category “no response” had died before being able to respond.
The Survey Participants Australian Government records available on 15 October 1982 showed that 15,364 Australian personnel had participated in some way in the atomic test program. From these 8018 survey candidates, the “main group” originally identified by the researchers from official records eg RAAF service records, RAN lists and medical records, was formed from those still living for whom addresses could be found. To this figure was added an additional 1079 people the “self-identified” group, for whom no records existed but who had identified themselves with the tests through some form of contact with the government authorities for example, an inquiry concerning compensation. But because of their self-selection, researchers feared that a disproportionate number of the second group could expected to be ill, and because their inclusion in the main group would have biased the results, they were treated separately (Donovan et al 1983, p. 9).
Ways in which respondents believed they might have been exposed to radiation Three questions among others asked of veterans relating to their duties before and during the tests were, the particular test series in which the respondent participated (question 2), the particular task(s) carried out by the respondent eg decontamination of personnel, (question 3) and the ways in which respondents believed they may have been exposed to radiation eg witnessed explosions (question 5). But for the purpose of this dissertation only the results of question 5 (Donovan et al 1983, p. 22) are regarded as relevant and these are shown in table 7.2.
Table 7.2
WAYS IN WHICH RESPONDENTS BELIEVE THEY MIGHT HAVE BEEN EXPOSED TO RADIATION Nature of activity Main Group (%) Self-identified (%)
Worked in blast area 540 (22) 28 (29) Worked near blast area 801 (33) 34 (35) Handled radio-active materials 691 (28) 27 (28) Flew through radio-active clouds 139 ( 6) 2 ( 2) Witnessed explosions 899 (36) 52 (54) Visited blast areas 1115 (46) 42 (44) Fallout from tests 482 (20) 26 (27) Other 438 (18) 25 (26)
Note that many of the respondents worked in more than one of the stated activities, and that this table (7.2) results from the responses of both the 2440 and 96 personnel of the main and self-identified groups.
The authors remark that there may have been some confusion regarding “blast areas” and perhaps they should have been referring to “radioactive areas” instead, since “radioactive areas particularly upwind of explosions, are very much smaller than blast areas”, “smaller” also meaning lower radiation levels. Important also perhaps is that the radiation level is higher downwind. Noted also is that 5% of the main group and 3% of the self-identified group held positions that may have required them to encounter radiation from time to time (Donovan et al 1983, p. 22).
The Search for other Variables Influencing Disease Prevalence In order to ensure that causes other than radiation exposure were considered by the study, the respondents were polled as to their life histories including employment history prior to and after the tests, and these were used to generate pre-study customised conditions tailored to the respondents histories. Following are the diseases and their non-radiation variables considered significant (Donovan et al 1983, p. 30).
Each respondent was asked: 1. Whether each had lived in the tropics 2. How much time each was required to work outside in five gradings ranging from always outside to always inside. 3. How much of holidays had been spent outdoors in five gradings as for 2. 4. Whether work had required exposure to radiation.
Malignant Melanoma The researchers found that the logarithm of age (as distinct from the age itself) and working outdoors was a significant factor as a cause of melanoma since according to the data all of those with malignant melanoma had spent a grade 3 (in a grading system of 1-5) or more working outdoors. They also concluded that “living in the tropics, whether holidays were taken indoors or outdoors and other work with radiation were not significantly associated with the prevalence of malignant melanoma” (Donovan et al 1983, pp. 30-31). Skin Cancer other than Malignant Melanoma The results for this section reflected those for malignant melanoma. Age factors and holidays spent outdoors were found to contribute significantly to the prevalence of the disease, whereas living in the tropics was by itself not a significant factor. The question as to whether working time was spent mostly indoors or outdoors and whether work was associated with radiation were not associated with the prevalence of skin cancer (Donovan et al 1983, p. 31).
All Other Cancers The logarithm of the respondent’s age and his smoking history were significant, eg whether he “never smoked”, “used to smoke” or “smokes now” and whether he uses cigars, a pipe or cigarettes. In the case of cigarettes, the analysis showed that compared to present smokers, former smokers have a higher prevalence to cancers, whereas pipe and cigar smoking and other work with radiation was of no significance.
Cataracts Considered were the logarithm of respondent’s age, any history of diabetes and other work history with radiation. The conclusions reached were that age factors and a history of diabetes were significant but radiation work other than at the tests was not.
Infertility For couples where the wife was shown to be fertile the variables considered were age factors and whether or not the respondent had worked in a radiation environment apart from the tests. It was concluded that neither was a precondition factor for the study (Donovan et al 1983, p. 33).
The Survey results Analysis of some 130 tests for associations between radiation indicators and the prevalence of various diseases and 156 tests for disease interactions required by the study, is prone to error and an estimated 5% error or 14 associations or interactions are to be expected. Since the model detected only 23 associations, the authors admit less than complete confidence in the results (Donovan et al 1983, p. 41). The authors pointed to several difficulties of a statistical nature which could be expected to be encountered in the study, including associations of undefined levels (ie where particular grading is not specified) and where data cross-classification causes normally ignored deviancies to become statistically significant. Using linear-logistic modelling techniques, the findings by the study of the previously discussed diseases are:
Malignant Melanoma Age factors considered and in a classification according to tasks performed as part of the test program, it was found that those employed on the construction of support facilities had an incidence of malignant melanoma that was below the incidence level predicted by the model. However the authors concede that “apparent protective effect” afforded by the constructions “was barely statistically significant and may perhaps be due to chance alone”. No melanoma sufferers were identified with the clean-up of radioactive areas, which the authors suggest could also be due to chance (Donovan et al 1983, pp. viii, 41). Personnel participating in the task of decontamination operations, including of people, aircraft, vehicles and equipment, were found to have 2.6 times the model’s predicted incidence of malignant melanoma but since only nine workers were involved the authors, because of so low a sample, expressed a lack of confidence in this result (Donovan et al 1983, pp. viii, 41-42).
Other Skin cancers
All Other Cancers Because respondents with particular forms of cancer were too few for their cancer to be considered as a separate entity, all those with cancer were lumped together under this one heading. The model found for this combined category of disease, that an association existed with the task of operating support services, producing a lower than expected prevalence for any type of cancer. However the authors cannot explain why this particular task type should be so protective (Donovan et al 1983, pp. viii, 45-46).
Cataracts
Infertility
Nuclear Veterans’ Criticisms of the Report
In the Sydney Daily Telegraph of March 22, 1984 an article “Army men up front in A-bomb explosion” quotes the Labor Defence Minister Mr Gordon Scholes who described how in the 1950s, twenty-three Australian servicemen were sited 1.6 kilometres downwind from an atomic explosion giving rise to the claim that they were exposed to radiation for experimental purposes. Similarly, an article from the Adelaide Advertiser of March 12, 1984 cites a then recently declassified British document “Memo ‘shows soldiers used in A-Bomb Tests’” that describes how British soldiers were deliberately exposed at the Australian and Pacific atomic tests in 1956 in order to “discover the detailed effects of various types of explosion on equipment, stores and men with and without various types of protection”. The clear implication of these articles is that the existence of deliberate, repetitious exposure would surely lead to a higher incidence of radiation induced illnesses than was found by this report.
Causes of Death of Former Atomic Test Personnel (a) To identify any associations between atomic testing with causes of death among Australian personnel at the atomic tests, in particular to compare the pattern of known radiation-related illnesses among the work units involved. (b) To compare the causes of death of test personnel with those of the Australian population as a whole.
In order to have a valid control population for the study; each former veteran was age matched on the basis of his relevant characteristics, eg his cause of death, with two members of the non-nuclear veteran population. A further need was that the two control associates’ histories were taken from the same registry, with their times of death corresponding as close as is possible to the time of death of the nuclear veteran. A secondary aspect of the study was where the cause of death of former test personnel was compared proportionally with the incidence in the Australian general male population.
Thus in simplistic terms we have first a comparison of veterans’ deaths with non-veteran associates that have similar disease histories and secondly, a general comparison with the male population at large for the same age-adjusted cause of death. Since the previous postal-survey study was restricted to addressing radiation induced diseases of the living, and sought no information on disease conditions that preceded a rapid health decline and subsequent death, an additional intended role of this study was to complement the previous one by rectifying that limitation.
Specific Hypotheses (a) all malignant diseases (b) lung cancer (c) leukemia and allied disorders (d) malignant melanoma (e) other skin cancer (f) thyroid cancer (g) hypertension (h) diabetes (i) cirrhosis of the liver (j) Chronic obstructive airways disease
Hypertension and diabetes were included as checks on the analysis because of their high prevalence in the community and independence from radiation as a cause. Cirrhosis of the liver and chronic obstruction of airway disease were included because of the purported high incidence of alcoholism and smoking amongst service personnel.
Table 7.3 Probability (tenths) Number of Registrations 1 20 2 13 3 10 4 11 5 22 6 10 7 11 8 24 9 37 10 1,402 Total 1,560
Demographic characteristics of the Test Participants
Deaths Available for the Study
Table 7.4 Years of Death of Test Participants Years Deaths % 1953 to 1957 37 2 1958 to 1962 130 8 1963 to 1967 190 12 1968 to 1972 313 20 1973 to 1977 396 25 1978 to 1982 494 32
Total 1,560 100
Table 7.5 Ages at Death of Test Participants Age Group Deaths % 45 to 54 449 29 55 to 64 471 30 65 to 74 271 17 75 to 84 71 5 85 and over 7 0
Table 7.6 Employment of Diseased Test Participants Employer Number % Royal Australian Air Force 313 20 Australian regular Army 138 9 Other Commonwealth Government 544 35 South Australian Government 20 1 Kwinana Constructions 345 22 Other Specified 70 4 Not Stated 47 3 Total 1560 100
The Prevalence of influencing variables of Each Disease
Table 7.7 Death Crude Adj Participants Controls Risk Risk All malignant diseases 355 659 1.10 1.10 Lung cancer 120 194 1.26 1.26 Leukemia and allied disorders 28 74 0.75 * Malignant melanoma 10 24 0.83 0.83 Other skin cancer 2 8 0.50 * Thyroid cancer 1 1 2.00 * Hypertension 11 29 0.76 * Diabetes 3 13 0.46 * Diseases caused by alcohol 54 60 1.83 1.83 Chronic obstructive airways disease 37 87 0.85 *
Importantly however, the report noted that “malignant diseases are not commonly mentioned on death certificates except as an underlying cause of death” meaning that a dying cancer sufferer who because of his lowered immunity contracts pneumonia, would have pneumonia stated as the cause of death! This will be discussed later under the reliability of death certificates as a means of determining the cause of death for research purposes.
Table 7.7 shows the numbers of deaths of test participants and controls from each of the diseases considered and the corresponding risk ratios, the crude risk and the adjusted risk. The crude risk ratio has no adjustment for age at death and year of death whereas the adjusted risk ratio was where applicable produced by the model with compensation for ages and years of death. A value greater than 1 for either of these ratios indicates a higher than normal risk. Adjusted risk ratio values replaced by an asterisk in the table either needed no adjustment from the crude risk ratio, or there was insufficient data available for adjustment. Note that in Table 7.7, there is double the number of controls as participants. As can be seen from the table, a slightly higher proportion of test participants suffered from malignant diseases, with a higher adjusted risk ratio of 1.1 than the control group. Lung cancer was also more prevalent with an adjusted risk ratio of 1.26.
|
|
Conclusion As stated previously Veterans were disappointed with the results of the two surveys as they had hoped that they would confirm their claims and of equal importance, their credibility. However were the study results biased as the veterans claim?
In the first survey, that of past test participants, the researchers resorted to computer modelling techniques as they had decided that no suitable control group was available. However since no test results or explanation of the methods used (for comparison see Roach and Vallis’s model in previous chapter) were supplied in order for the reader to determine and confirm the validity of the model, the results given in this report must be considered to lack credibility. In fact the report’s authors also express a lack of complete confidence with their results (Donovan et al 1983, p.41) albeit because of large inherent errors. Further, the first study modelling was based on the data supplied in 2440 completed or partially completed questionnaires out of the original 15,364 Australian service personnel test participants, where the nuclear veterans see the 2440 (15%) as being too small to be significant. Important though was that 91 self-identified veterans – those more likely to be suffering the effects of radiation as a reason for coming forward, were deleted from the test sample. Finally, it can be argued that those who died from radiation induced diseases should also have been included in the first study since they were originally part of the test sample.
The second study, that of former participants now dead, was based on the use of a control group to match the history of each veteran test subject, where each former veteran was age-matched on the basis of his relevant characteristics with two members of the non-nuclear veteran population. Log-linear modelling was again used as in the first study. The second study produced higher than normal (ie >1) probabilities (ie crude death rate – age not considered) in all malignant diseases (1.1), lung cancer (1.26), thyroid cancer (2.0) and alcohol induced disease (1.83). The remainder had lower than normal probabilities – eg malignant melanoma (0.83) and hypertension (0.76). However, the results can be construed as invalid because the cause of death in test subjects was determined by death certificate, a method that has already been established as unreliable (Roff December 1997, p.11). This is because for example, a patient dying of cancer may actually die of cancer-induced pneumonia, which is recorded on the certificate as the cause of death. Understandably, in so sensitive a study it is important to know that cancer was the predominating reason for death. More specifically stated, two researchers, Ms Sue Rabbit-Roff (December 1997) of Dundee University and Dr Carol McDonald (McDonald 1997) a New Zealander, are in accord regarding the use of death certificates to verify the cause of death in nuclear veterans. Roff sees studies based on death certificate data as “underestimating the specific causes of mortality” (Roff December 1997, p. 11). She cites a study (Ron et al 1994, pp. 485-486) based on 5000 necropsies (autopsies) which examined death certificate accuracy in twelve categories of disease where the actual cause was later determined by autopsy. The results showed the overall agreement between certificate and autopsy to be only 52.5% where almost 25% of cancers diagnosed at autopsy were missed on death certificates (Roff December 1997, p. 12). Only for neoplasms (obvious abnormal tissue growth) and external causes of death was there above 70% correlation between certificate and necropsy. The study also showed that death certificate error tended to increase with increasing age and hospital remoteness. |
| 11 |