Did Chemical Exposures of Servicemen at Porton Down Result in Subsequent Effects on their Health?

The 2005 Porton Down Veterans Support Group

 Case Control Study.

 First Report.

Chris Busby

 Saoirse Morgan

Occasional Paper 2006/2

Green Audit Aberystwyth

Summary

The Porton Down Veterans Support Group mailed 505 questionnaires to members in order to investigate any possible connection between exposures at Porton Down and subsequent ill health. Members were asked to obtain controls of the same age. There were 261 questionnaires which were returned by veterans together with 123 control questionnaires. This enabled us to conduct a case-control analysis of health in the veterans compared with the controls. For some conditions like cancer, birth outcomes or health in the children we were able to compare the rates in the veterans with national data. The mean age of the 261 veterans who responded was 69.0 (SD = 7.1) and of the 123 controls was 67.6 (SD = 6.1). Most of the experiments were conducted between 1950 and 1960 although some had been carried out as recently as 1989.  Cases reported being exposed to Mustard Gas, Sarin and other nerve gases, CS gases and riot agents, to mind altering substances like LSD and to a range of other toxic substances and unknown compounds.  Respondents recorded the effects on them at the time and in the period after. The experiments and the immediate effects were bizarre, frightening and often clearly involved major clinical symptoms. Some volunteers were blinded, others woke up in hospital, some took considerable time to recover.

            We examined cancer in the cases relative to the controls and national data. Levels of cancer in the controls were lower than expected on the basis of national rates. For all malignancy combined, using cancer diagnosed in the period 2001 to 2005  Standardised Incidence Ratio (SIR) based on England and Wales 2002 data was 1.02 in the veterans, equal to the expected level. Significantly high levels of prostate cancer were reported by the veterans (15 cases, SIR 2.6, CI 1.57<RR<4.26; p = 0.0001) and also malignant melanoma (3 cases, SIR = 8.0, CI 1.93<RR<26.8; p = 0.008). The melanoma excess is interesting as it is a biologically plausible cancer following skin exposure to a carcinogen like mustard gas. Comparison of cases with controls for current treatment for on-going conditions showed a significant excess of heart disorders (OR 2.37; CI 1.17-1.89, p= 0.01), arrhythmias (OR 6.7; p = 0.04), respiratory disorders (OR 10.3, 1.44-207, p = 0.01) and diabetes (OR = 4.7 (1.01-7.2; p = 0.03). Cases showed significantly higher levels of many other conditions and illnesses than controls but for some of these e.g. strokes, comparisons with the Welsh Health Survey showed that the levels were not significantly different from those that might be expected in the age group.

We examined miscarriages, infertility and child health in the 549 children born to the veterans and the 272 children born to the controls. There were significantly more miscarriages (OR = 2.23, 1.07<OR<4.75, p = 0.02) in the veterans and significantly more infertility (OR infinity based on 28 vs. 0). In addition, the children were significantly more ill with 4.39 times the reported health problems (OR 4.35; 2.08<OR<9.43, p = 0.00001). There were twice the number of child deaths (OR = 1.86 on 38 deaths), mental health problems (11 vs.0), epilepsy (3 vs. 0) and brain damage (5 vs. 0). There was an astonishingly high rate of child cancer reported in the offspring of the veterans including two child leukemias. This represented and SIR of 153 for leukaemia 0-14 (based on 2 cases, p = 0.00000) and 80 for all cancers 0-14 (based on 4 cases; p = 0.00000). Amazingly, there was one child leukaemia reported in the controls, the only child cancer in the controls. Overall, the infertility, the miscarriages and the health of the children of the veterans and their cancer rate, is consistent with genetic damage in the parents.

The data supplied on smoking and alcohol was unable to explain any of the observed effects. Veterans were not significantly more likely to have smoked or drunk alcohol than controls and the effects on cancer were second order to the status of the cases. Questions about emotional problems and physical illness or pain and its effect on social behaviour showed the veterans to be significantly more limited than controls. Logistical regression analysis showed such an effect at the p = 0.0001 level for classification of status as a veteran or a control on the basis of effects of both pain and health on social behaviour.

Examination of a range of 27 symptoms and conditions (e.g. loss of memory, chest pain, forgetfulness, sleeping difficulties, etc) all showed significantly higher response rates in the veterans than controls with OR ranging from 2.5 to 6.5.  

Finally the question of whether these effects found in the veterans and their children could be a consequence of selection bias was approached on the basis of analysing the range of diseases manifested in the veterans compared with the range in the controls and in average men of the same age. It is suggested that the peculiar range of illnesses and conditions exhibited by the veterans imply that at least some of their illnesses are the consequence of exposures at Porton. It is argued that if the increases in ill health shown in the veterans were due entirely to self selection then the range and spectrum of conditions would be the same as in the controls, merely the magnitude would be different.

It is therefore concluded that this preliminary analysis shows that the Porton Down Veterans have suffered some serious long term health problems as a result of their exposures. The exact extent to which selection bias has affected the results is difficult to asses at this stage. Further statistical work on the data might illuminate the situation.

Introduction

The study is an attempt to examine the question of the long term effects of exposures to various chemical agents used in experiments carried out on army volunteers at the Porton Down establishment in Wiltshire. These experiments on ‘human volunteers’ occurred in the UK mostly between 1950 and 1965 although some experiments were carried out before and after this period and some as recently as the 1990s. The experiments involved exposures to lethal chemical warfare agents including nerve gases and mustard gas and also behaviour altering agents like LSD and crowd control agents, tear gases and to antidotes to these substances. The chemicals were delivered by inhalation, by droplet, by skin contact with impregnated cloths or impregnated clothing. Results were often alarming, resulting in hospitalisation and on at least one occasion in death. Those who were experimented upon in the period of National Service before 1960 have complained in this survey that they were not told the nature of the substances they were being exposed to, and in many cases were told that they would be engaged in research into the common cold.

The majority of the living survivors of these experiments are currently aged between 65 and 75. Many are suffering various illnesses and are subject to various conditions which they believe may have been the result of their exposures. This is biologically plausible since some of the agents that were used at Porton Down are known to have serious long term effects. Mustard Gas, for example, is a potent DNA alkylating agent, mutagen and carcinogen; nerve gases may have long term effects on brain function. In order to examine the hypothesis that the exposures may have had any long term effects, the Porton Down Veterans Support Group discussed with Green Audit the possibility of conducting a questionnaire study of their members. It was clear that such a questionnaire had to be short enough for the veterans to be likely to fill it in, contain sufficient questions to bracket the types of health conditions and illnesses that were likely consequences of any chemical exposure, and also be sufficiently open ended enough to pick up any conditions that might define a consequence of exposure that was uncommon in the normal population. In addition, the questionnaire allowed respondents to write about their experiences at the time and shortly after the exposures: it was thought that such a record might be valuable for historians of the events.

In order to define the group that was being examined, and in the absence or normative values in the national population for the age group involved, it was necessary to obtain a control group with which to compare the cases. This posed something of a problem given the resources available, yet without a control group, many of the questions being answered would be difficulty to interpret. It was decided to ask each of the cases to obtain a control. In order to avoid the argument that the cases would be choosing particularly healthy controls the cases were asked to choose controls of the same age on the basis of a list of choices that was pre-defined i.e. brother-in-law, nearest neighbour of same age, friend and so forth. Whilst it is conceded that this is not ideal, we believe that this inexpensive approach will give a group of men of the same age that may be used to compare the cases with on some level and that a certain amount of analysis can be employed to examine the general health of this group with that of the exposed cases and obtain meaningful results. For certain illnesses and conditions, e.g. cancer, the cases can, of course, be compared with national data. There remains one major problem with a study of an organisation that has formed in response to concerns about exposure and health, and that is the selection bias that this involves. Clearly it may be argued that such a group may already be selected for ill health, as those veterans of exposures at Porton Down who feel themselves to be healthy may be less likely to join such an association. This is a fair argument. In addition, there is also scope for elective bias in the way in which the survey was conducted. Thus of the 500 questionnaires which sent out only 250 were returned, and it may be argued that those who returned these were from the members who were more ill and that therefore had some need to explain their illness. However, there may be way in which the elective bias in the sample can be analysed. For example, if there is elective bias, it is difficult to see how it could select for certain specific illnesses and conditions. It might be that the veterans represent a more sick group than the controls, for all the reasons given, elective bias, choice of controls etc, but if the spectrum of rates of various illnesses in the controls and the veterans were significantly different, then this would suggest that there is a real effect, a real difference between the controls and the study group, that is, the veterans who were exposed. So these are the problems faced by anyone studying the health of the veterans, and must inform caveats regarding any conclusions that are drawn.

2. The survey

The questionnaire (Appendix B) was posted to 505 members of the Porton Down Veterans Support Group in September 2005 and responses returned over the next two months. The questionnaire was accompanied by a letter (Appendix A) which explained its purpose and also explained how each veteran was to try to obtain two controls from among their friends and relations. These controls of roughly the same age and sex were to be picked in the following order:

1. Brother in law or friend of sister
2. Friend who lives second closest
3. Friend who lives closest
4. Colleague or ex-colleague from work
5.  Friend who lives third closest

This process was intended to ensure that veterans were less likely to choose controls who were well, though of course, we have to take on trust the choices of those who returned the questionnaire, and indeed, many veterans were unable to find any controls. In the event, some 123 controls were obtained, less than the number needed to ensure complete balance for a case control study of 261 cases. Nevertheless the number of controls was sufficiently high to ensure that some useful conclusions could be drawn about the health of the exposed group, as the results will show.

The questionnaire itself together with the relevant extract of the accompanying letter is attached in Appendix C. It can be seen that in order to make the questionnaire reasonably short, yet obtain the maximum amount of useful information, some questions are open ended and allow the respondent to explain the existence and nature of any health condition that they feel is relevant.

3. Results

3.1    Age groups and period of testing.

Returned questionnaires were coded into a database. The numbers and ages of those responding are shown in Table 1 and the distribution of the ages of the cases are plotted in Fig1 and Controls in Fig 2.  The groups were fairly similar in distribution of ages with the cases having mean ages slightly higher than the controls. There were 9 cases under the age of 50 years with the youngest being 43. There were only two controls under the age of 50 years; both were aged 49.

Table 1 Ages of those filling out questionnaire

Fig 1 Age distribution of Cases. Mean age 69.08; Standard Deviation 7.1

Fig 2 Age distribution of controls; mean age  = 67.65, Standard Deviation = 6.1

Fig 3 Distribution of years that cases were experimented on.  Mean = 1956

Cases were experimented upon between 1939 and 1989 but the main period of the experiments was the post war period when there was national service, between 1950 and 1960. This is clear from the histogram of the results given in Fig 3.

3.2 Exposure substances reported (Question 12)

Many of the responses contained claims that the respondents had been told that the research was associated with a cure for the common cold. The types of exposure reported were divided by us into three categories: Mustard Gas, Sarin and Other. Not all the respondents remember what the exposures were, but Table 2 shows the number of cases by category of exposure as reported. Table 3 lists the other exposures reported.

Table 2. Numbers of volunteers reporting exposures by category.

Table 3 Some of the substances in the ‘Other Exposures’ category reported in responses.

3.3 What sort of tests were reported (Question 13)

This open ended question was generally answered but the answers have not been yet coded into the database as there are many different answers. Tests involved exposure to gases, to droplets on the skin, to substances saturated in cloths and clothing. Men were exposed naked and photographed, were exposed in full protective suits, had to wear underpants saturated with unknown liquids which burned their skin, the procedures were often bizarre and frightening.

3.4  Clinical and physical responses to the tests at the time (Question 14 and 15)

These responses have been entered into the database and are listed in the Appendix B Table B1. They show a wide range of clearly major and frightening clinical effects caused by the exposures immediately after the tests. Some volunteers were temporarily blinded. Some woke up in hospital after passing out. We know that at least one volunteer died. In addition there were significant effects in the first few months after the tests and these are listed in Table B2.

3.5  Cancer

Cancer reported by the cases and the controls as having been diagnosed in the last five  year (2001-2005) is listed in Table 4 below and in Table 5 is listed all the numbers of individuals reporting cancers diagnosed. For Table 5, if there are two cancers reported in the same individual both cancers are included. The raw data is reproduced in Appendix B Tables B3 and B4. In Table 6 are given the Relative Risks for the main cancers reported in the last 5 year period 2001-5. These are calculated as Standardised Incidence Ratios based on England and Wales National incidence rates for 2002 (Office for National Statistics Series MB 1 No 32).

Table 4 Cancers reported in the cases and controls in 5 years between 2001 and 2005; where the year of diagnosis was not give the possible total is bracketed.

Table 5. All cancers reported in the cases and controls by site irrespective of when diagnosed. If an individual reports more than one cancer, the cancer is included.

NMSC = Non melanoma skin cancer.

Table 6 Relative Risks 2001-2005. Standardised Incidence Ratios are calculated on England and Wales rates for 2002 and represent the risk of cancer relative to the national average after allowing for age. Thus for melanoma, there are eight times the number in the cases that would be expected on the basis of the national rates in 2002 after allowing for age. Significant cancer incidence sites are given a *.

There is clearly a statistically significant 2.6-fold excess of prostate cancer in the veterans, both relative to the controls and to the national population. There is also a significant eight-fold excess of malignant melanoma reported together with a small, non-significant excess of colorectal cancer. Looking at the overall reported cancer, it seems that the cancers reported are from regions of the body that are near the lower abdomen. Thus we see 6 bladder cancers, 4 colorectal cancers, 2 testicular cancers and 18 prostate cancers. Prostate, lung and colorectal cancer are the most common cancer in men of this age group. We see here a preponderance of prostate cancer with no reports of lung cancer in either the cases of the controls. This may be because of the fatality of lung cancer, but it still seems unusual and suggests that the prostate cancers are significant in the group. The melanoma incidence rate of 8-times the national average is unusually high, but this is based on three cases and it is suggested that a larger study concentrate on this, especially since many of the exposures were to carcinogenic substances applied to the skin.

3.6  Current treatment for any health conditions Question 20-21

One open ended question asked if respondents were currently being treated for any chronic or on-going health condition by a doctor. Responses were reduced to the database and are listed in Table B5. In order to compare the prevalence in the cases and controls we gathered the conditions into groups which were roughly based first on the standard classifications of causes of death and second on recognised illnesses. Comparisons of the cases and controls was then carried out and the results given in Table 7. In Table 8 are listed conditions reported by the cases that are not shown in the controls. From these results it seems that the cases more generally sick than the controls and are being treated for various condition that separate them from the controls. These conditions are listed in Table 9.

Table 7 Comparison of cases and controls for current chronic and on-going health conditions which are manifested in both groups. Conditions where the excess reported prevalence is statistically significant are shown * in the Odds Ratio column.

Where there are values less than 5, 2-tailed Fisher exact is used. All others are Mantel- Haenszel, apart from * Yates corrected

Table 8 Conditions which were not reported in the controls

Table 9 The main chronic and on-going conditions being treated by a doctor which are significantly more reported by the cases than the controls. We have standardised for age and compared these with results given in the Welsh Health Survey 1998 for the same ages for some of the conditions which are comparable.

3.7  Admission to hospital; Questions 22 and 23.

There were 61 cases reporting admission to hospital in the last 5 years as opposed to 31 controls. This was roughly the same percentage.

3.8  Pregnancy and children. Questions 24-30

Questions 24 to 30 were about pregnancy outcomes, fertility and the health of the children. Results are given in Table 10. Veterans had roughly the same proportional number of children as controls. They reported more miscarriages than controls and reported that their children suffered from more than 4 times more health problems. One case reported four miscarriages and the one child that lived had learning difficulties and autism. The miscarriage reported rate difference was statistically significant.  Veterans also reported problems with being able to have children. 28 reported such a problem compared with none of the controls. Cases also reported twice as many child deaths in relation to the numbers and also increased rates for various childhood conditions and illnesses. There were two cases of childhood leukaemia reported in the cases and one in the controls which is a high rate for both groups, representing in less than 1000 children an excess risk of more than 100-fold (see below). In summary, there does seem to be a significant difference between cases and controls in the area of pregnancy and child health. This should be further investigated. One possible approach will be logistical regression methodology to examine relationships between the response parameters, including the period of the tests, the health outcomes in parent and child and  the substances used etc.

Cancer and leukemia in the children

It was of interest to examine the possibility that genetic damage in the parents may have resulted in effects in the children, specifically cancer. There was an astonishingly high rate of leukaemia in both the cases and the controls, based on three children. Child leukaemia is a very rare disease. The 1972 rate for all leukemias in the 0-14 age group was 2.4 per 100,000 yet in the 549 children of the veterans there were two cases and one in the 272 children of the controls. There were two other cancer cases reported in the veteran’s children with none in the controls. The Standardised Incidence Ratios for these are given in Table 11. For the leukemias the risk in the veterans children is 153 times the national rate, for cancer in the veteran’s children it is it is 80 times the expected rate based on 1972 national rates. These cancers and leukemias in the veteran’s children should not be dismissed as chance because it is biologically plausible since the parents were exposed to a known mutagen. The one case of leukaemia in the control children may be a change occurrence.

Table 10 Questions 24-30; Fertility, pregnancy and children’s health

*based on numbers of fathers

Table 11. Cancer and leukaemia in 549 veteran’s and 272 control’s children aged 0-14 at diagnosis (calculations based on England and Wales rates 0-14 for 1972).

3.9 Smoking and drinking; Questions 31-33

Smoking

The rate for reported currently smoking in the 261 cases was 0.04. For those who developed cancer it was twice this at 0.07; 4 of the 55 cases with cancer were current smokers.  The level of current smoking in the controls was higher at 0.08; none of the individuals with cancer among the controls was a current smoker.

On the basis of ‘ever smoked’ the rate among all 261 cases was 0.43 (114/261) compared with 0.36 (45/123) in the controls. For those who reported cancer it was 0.5 in the cases (28/55) and 0.3 (3/9) in the controls. Roughly half of the cases and one third of the individuals reporting cancer had smoked at one time in their life.

Alcohol

Using the mean number of total number of units of alcohol per week reported and averaged per individual, the levels in the controls were slightly higher at 9.4 compared with 8.0 in the cases. In general there was no evidence that smoking and drinking could account for the differences observed between the cases and the controls.

3.10 Questions 34-39 Physical health, emotional problems pain effects on social activity.

 Fig 4 and Fig 5 below shows the distribution of responses to the question No 34, which asks on a scale of  four points 0,1,2,3 from ‘not at all’ to ‘extremely’ the extent to which physical, health and emotional problems interfered with social activity.

The responses to Question 35, which asks about pain and how much it interferes with work or leisure  (on a scale of three points) are shown in Figs 6 and Fig 7. It is clear that there are significant difference between the veterans and the controls (logistical regression on <STATUS> gave p<0.0001 for covariate <HEALTHSOCIAL> and P = 0.007 for <PAINSOCIAL>).

Fig 4 Responses to question No 34 about effect of ill health upon social activities (see text). ‘0’ represents the answer ‘not at all’ (see Appendix C for the full question).Status 1= veterans, 2= controls.

Fig 5 Responses to Question 35 about effect of bodily pain on social activity. ‘0’ = not at all (see Appendix C for the full question).Status 1= veterans, 2= controls.

3.121 Question 40: conditions and symptoms.

Question 40 (see Appendix C) displays tick boxes for 27 common conditions and symptoms. Results of case-control comparisons are shown in Table 12.

Table 12 Case control comparisons of symptoms and conditions reported by veterans and controls in Question 40.

Where there are values less than 5, Yates corrected is used, all others are Mantel-Haenszel

3.11 Question 41. Open ended comment

These comments have been reduced into Appendix B, Table B7. There were also several letters which were sent with the completed questionnaires and these have been reduced into the table where possible or useful.

4. Discussion.

The veterans complain that they were experimented upon and exposed to a wide range of dangerous substances in bizarre, frightening and embarrassing episodes which made them extremely sick at the time and also in the period after the tests. Many say that they were told that the procedures were presented to them as ‘common cold’ research. We received letters which complained bitterly about what was carried out. Many of the substances that were used are known to be extremely dangerous and to have long term fatal effects including cancers and heritable genetic damage. They feel betrayed, and it is hard to avoid concluding that they were indeed betrayed and lied to. 

First, there is no question that on the basis of the responses, the veterans are significantly more ill than the controls. In addition, their children are also more ill. This latter finding is interesting since it seems unlikely from the responses that that the elective bias would operate though the children. On the other hand, if the cases were already more ill than the average population, it might be likely that their children would also be more ill. It is, however quite extraordinary that there is an 80-fold excess of cancer in the children of veterans (150-fold excess of child leukemia). Although this is based on only 4 cases we have to recall that the Sellafield leukemia cluster which resulted in a Public Enquiry was also based on four children and a relative risk of 10-fold. What both the veterans and the Sellafield cases have in common is exposure of the fathers to a carcinogen. In the case of the Porton fathers this is mustard gas; possibly other substances. In Sellafield it is radioactive isotopes. The veterans also appear to be significantly associated with more miscarriages than the controls, with considerably greater infertility and with a statistically significant (p < 0.0001) four fold excess of child health problems including brain damage, epilepsy, skin problems, mental health, spina bifida, learning difficulties and twice as many child deaths.

The veterans show a statistically significant 2.7 fold excess of prostate cancer relative to the national average, and also a statistically significant 8-fold excess of melanoma.  In addition we see in the overall cancer reports that the cancer sites that were common in the veterans were all in the area of the lower abdomen. The patterns of melanoma sites on the body and the diseases connection with ultraviolet light suggests that it is caused by genetic damage to skin cells in the deep epidermis. Thus the skin exposure to mustard gas could be seen as a plausible biological cause.

A whole spectrum of illnesses and conditions cluster in the area of brain stem or lower brain functional disturbance. Such a spectrum of illnesses are plausible biologically as related to nerve gas or nerve gas antidote exposure outcomes.

The essential problem we are faced with is deciding whether the significantly greater ill health in this group of veterans is an effect of their exposures or if it is an artifact due to their selection into the Porton Down Veterans Association, and their choice to respond to the questionnaire itself. This is not an epistemologically insoluble problem for two reasons. First, we can examine the biological plausibility of the spectrum of illnesses found. Second we can examine the spectrum of illnesses and conditions in the cases and the controls and ask if the veteran group is just generally more ill than the controls, or if there are specific illnesses in the cases which are not there is the controls and which might be plausibly connected with the exposures.

Let us, for the sake of argument, take the view that the veterans joined the Association and/or filled out the questionnaire because they are a more ill section of the community than the average and are looking to blame someone. Then we should expect that they would have all the usual illnesses and conditions of men of their own age group i.e. the controls, only they have them to a far greater extent, or that at very least, the illnesses would not cluster. So we can test this statistically to see if it is so. If on the other hand, the veterans show a cluster of illnesses or conditions that are similar, and which separate them from the controls not only in magnitude but also in spectrum, then we might feel that this is evidence that their exposures were causally linked. A spectrum of illnesses and condition in the veterans and the controls is shown in Fig 6. Note that for reasons of space, not all the bars in the spectrum are labeled.

There is another way we can use the data to approach this problem. We can take some time to examine relationships between the various factors, the types of exposure, the ages of the veterans, the types of disease and condition and so forth. The procedure involves logistical regression Factor Analysis and Principal Component Analysis. This initial study does not attempt to conduct these analyses which are costly in time and effort.

5. Conclusions

We can draw certain tentative conclusions about the health consequences of the Porton Down experiments from the initial analysis of the responses. First, there appears to be evidence of genetic damage in the veterans. This is shown by the cancer level and spectrum, and also by the infertility rates and the high rates of illness and cancer in the children. There is a statistically significantly high rate of prostate cancer and also a high rate of melanoma skin cancer. The latter is a biologically plausible response to exposure by skin absorption of a carcinogenic substance. We know from the responses that the veterans had been exposed to mustard gas and we know that this is a carcinogenic substance. The spectrum of cancers is also unusual relative to national rates and also to the controls, although there were too few controls to draw any firm conclusions.

Additional evidence for genetic damage comes from the high rates of miscarriage, high rates of infertility and high levels of congenital and other illnesses in the 549 offspring of the veterans. The level of childhood cancer in these children was astonishingly high, more than 50 to 150 times the expected level. Taken together these findings argue strongly for the existence of a genetic damage effect in the Porton Down group.

Second, there is no doubt, from the responses that the Porton veterans group members are significantly more ill in almost every way (except asthma) than the controls. Their quality of life is far worse: they complain of a wide range of mental and emotional problems and also of pain. It might be easy to dismiss these findings as due to selection bias effects, but we know that these individuals were exposed to dangerous mind-state altering substances like LSD and also to nerve agents and their antidotes, both of which have serious and permanent effects upon the brainstem and nervous system. These brainstem effects of nerve agents and other organophosphate agents used by farmers have been shown to produce all the long terms effects shown by the Porton veterans. Studies reported in the literature using focused Magnetic Resonance Imaging with P32 probes have confirmed significant brainstem damage in individuals exposed to nerve agents in the first Gulf War.

            It is difficult to quantify the degree of ill health which can be ascribed to selection bias in the initial study. Further analysis of the results of the questionnaire would perhaps enable us to examine this issue more fully. However we can say that the evidence examined so far shows that the Porton Down veterans who responded exhibited a range and spectrum of serious effects which may be plausibly ascribed to their exposures.

Finally, we have not attempted to present or quantify the sheer level of fury, pain and betrayal which shines out from most of the completed questionnaires and the letters and photographs which were sent with them. We feel privileged to have been allowed work on this issue and feel with every questionnaire we read and reduce to the database the experiences of the individual, or sometimes, when the individual has died, the pain of the wife or friend who has written on his behalf. We found it hard to believe that anyone in a civilized country could take young men of seventeen and experiment upon them, stripping them naked and exposing them to terrifyingly dangerous substances, to lie to them about the exposures and their effects and to delude them into thinking that they were helping with research into the common cold. We try to see this dispassionately, as scientists, as epidemiologists, weighing the evidence. But behind this, and following our experience of having read these 261 questionnaires, we have built up a level of sadness and amazement which is hard to describe. The experiments were criminal, and our results suggest that they had serious long term effects on those who were treated as the laboratory animals.

            This study was funded in part by the Porton Down Veterans Support Group but we would like to thank the Joseph Rowntree Charitable Trust for core funding without which this study would not have been possible. It may be of interest that the Medical Research Council has funded a study of the Porton Down Veterans for in excess of £300,000. We await the results of this MRC funded study with interest.