1. Ann. Occup. Hyg., pp. 1–10
Ó The Author 2007. Published by Oxford University Press
on behalf of the British Occupational Hygiene Society
doi:10.1093/annhyg/mem017
Reconstructing Exposures from the UK Chemical
Warfare Agent Human Research Programme
T. J. KEEGAN1*, M. J. NIEUWENHUIJSEN2, T. FLETCHER3, C. BROOKS1,
P. DOYLE3,4, N. E. S. MACONOCHIE3,4, L. M. CARPENTER1 and
K. M. VENABLES1
1
Department of Public Health, University of Oxford, Oxford OX7 3LF, UK; 2Department of
Epidemiology and Public Health, Imperial College, London SW7 2AZ, UK and CREAL, Barcelona
08003, Spain; 3Department of Public Health and Policy, London School of Hygiene and Tropical
Medicine, London WC1E 7HT, UK; 4Department of Epidemiology and Population Health, London
School of Hygiene and Tropical Medicine, London WCIE 7HT, UK
Received 8 January 2007; in final form 23 February 2007
The UK government has carried out a research programme studying military capability under
conditions of chemical warfare at a facility at Porton Down, Wiltshire, since World War I. In 2001
the Ministry of Defence commissioned a cohort study to investigate the long-term health effects
on military veterans of their participation in this programme. We assessed the availability and
quality of exposure assessment data held in the archive at Porton Down for the purpose of this
study. This involved looking in detail at exposure data in a sample of 150 veterans and undertak-
ing a general review of all available records held in the archive. These sources suggested that the
Porton Down records were largely complete and included sufficient identifying information for
linkage with service personnel data and with national mortality and cancer registration records.
Servicemen usually had multiple tests so data were most readily available in a test-wise format,
allowing subsequent aggregation of tests by individual. The name of the chemical used in each test
could be determined for most tests and most of the named chemicals could be categorized into
major groups for epidemiological analyses. For the major groups (vesicants and nerve agents),
quantitative data were available on exposure and on acute toxicity. Standardization will be re-
quired of the several different units which were used. Based on this study, exposure assessment
for the cohort study of Porton Down veterans will involve abstraction of the name of the chemical
used in each test, with quantitative data on exposure and acute toxicity for vesicants and nerve
agents. Our results here show that experimental records at Porton Down offer a unique and valu-
able resource for reconstructing the chemical exposures used in this research programme. The
resulting cohort study has the potential to provide information which will assist in understanding
the long-term health impact of chemical warfare agent exposure on these veterans.
Keywords: chemical warfare agents; exposure; mustard gas; sarin
INTRODUCTION during combat in WWI (Case and Lea, 1955; Beebe,
1960; Norman, 1975) and also among workers en-
Chemical warfare agents, first used in World War I
gaged in sulphur mustard gas manufacture (Wada
(WWI), are designed to have lethal, incapacitating or
et al., 1968; Easton et al., 1988) but not in people
unpleasant physical effects in humans and much is
exposed to sulphur mustard in human experiments
known about their acute toxicity (Evison et al., 2002).
Less is known about their long-term health effects. (Bullman and Kang, 2000). In addition, reports of
The small number of epidemiological studies with clinical follow-up in people exposed to sulphur mus-
long-term follow-up have shown some evidence for tard during the Iran–Iraq war have been published
an increased risk of respiratory cancers in the US (Balali-Mood et al., 2005). While persisting symp-
and UK servicemen exposed to sulphur mustard toms have been reported in civilians exposed to sarin
in the Tokyo terrorist incident (Nishiwaki et al., 2001;
*Author to whom correspondence should be addressed. Miyaki et al., 2005), no excess of cause-specific
E-mail: thomas.keegan@dphpc.ox.ac.uk mortality was seen in them or in people exposed to
1 of 10

2. 2 of 10 T. J. Keegan et al.
sarin in human experiments (National Research tail to establish what information could be used in an
Council, 1982; Page, 2003). Detailed exposure infor- exposure assessment for the cohort study. The ques-
mation is typically lacking after the use of chemical tions we asked were the following: are records of
warfare agents in warfare or by terrorists and is varia- the human trials available over the whole period
bly available for occupational studies, meaning that 1939–1989? Are the records complete? Can individual
only limited exposure–response analyses can be car- servicemen be identified so that linkage is feasible with
ried out. However, with human experiments, consid- their military personnel records and, ultimately, na-
erable exposure information may be available. Very tional death and cancer incidence records? Did service-
few developed countries have had research pro- men have one exposure at Porton Down or several; if
grammes involving collection of detailed exposure several, to one chemical or several? Is the name of the
information as part of human experiments (Pechura chemical clearly documented so that servicemen with
and Rall, 1993; Evans, 2000). similar exposures can be amalgamated into exposure
A British governmental research facility has been groups for analysis? Is the date of exposure recorded?
located at Porton Down since WWI. Along with lab- Is there information on the intensity of exposure and
oratory-based research and animal toxicology, the fa- on its duration, for use in quantitative exposure–re-
cility runs a programme, mainly involving members sponse analyses? Is there information on acute toxicity
of the armed services, of human research relating to (e.g. skin reddening or change in blood cholinesterase
military capability under conditions of chemical war- level) which can be used in dose–response analyses?
fare (Carter, 2000; Evans, 2000; Ministry of Defence, Is there information on other important factors which
2006). It is not known how many servicemen were may be potential exposure or effect modifiers, such as
recruited to this programme since WWI, but in the route of exposure or wearing of protective clothing?
50-year period following the outbreak of WWII
(1939–1989) up to 20 000 individuals are thought
to have taken part (Ministry of Defence, 2006). The METHODS
programme is known as the ‘human volunteer pro-
gramme’ or the ‘service volunteer programme’ and Availability of records
groups of servicemen (and some servicewomen), A collection of archival material dating back to
known as ‘observers’, were recruited for visits that, 1920, mainly books of attendance records and experi-
in general, were expected to last up to 2 weeks. In mental record books, was assembled at the Defence
practice, however, many veterans stayed for less than Science and Technology Laboratory at Porton Down
a week and very few veterans stayed for more than in the 1970s, both as a historical collection and as an aid
two (Bramwell, 1959; Ministry of Defence, 2006). in responding to queries from veterans who had been
Some people made repeat visits and may have been participants in experiments. In July 1999, Wiltshire
involved in many tests during each visit. Some tests Police initiated enquiries into allegations of illegal hu-
involved exposure to chemical warfare agents or riot man experiments at Porton Down (these enquiries
control agents; others involved trials of antidotes or were not subsequently taken forward to prosecution).
of protective equipment without chemical exposure. These enquiries were known as ‘Operation Antler’
A US Institute of Medicine committee reviewed and covered the period 1939–1989 (Wiltshire Police,
the comparable US human experiments involving 2004). The archive was catalogued in 2000 by the
mustard gas and lewisite (Pechura and Rall, 1993) police. The same archive was made available to us
and commented that ‘. . . morbidity and mortality when this epidemiological research started in 2002.
studies should be accomplished . . . comparing . . . The title of each item in the archive was noted, with the
cohorts to appropriate control groups in order to date of each first and last entry. Books containing lists of
resolve some of the remaining questions about the names and dates of attendance but no, or limited, exper-
health risks associated with exposure to these imental information were categorized as ‘administrative’
agents’. Two cohorts have been followed in the US while those containing principally experimental infor-
(National Research Council, 1982; Bullman and mation were categorized as ‘experimental’. All archival
Kang, 2000; Page, 2003). In the UK, an important material that contained information on tests between
step has been taken with the start of a cohort study 1939 and 1989 was included as a possible data source
which includes 20 000 veterans who attended Porton in this pilot phase. The archival material was briefly ex-
Down between 1939 and 1989 together with 20 000 amined and key personnel were interviewed to obtain
military controls, followed up for cause-specific a general view of the completeness of the archive and
mortality and cancer registration. of exposures during the human volunteer programme.
The study reported here was a precursor to the co-
hort study and, in it, we undertook a general review Comparison of subject numbers with
of the archive of historical documents made available contemporaneous annual reports
to us at Porton Down and studied the records of A database of participants in the human volunteer
150 randomly sampled individual servicemen in de- programme from 1939 to 89 compiled earlier by the

3. Exposures from the UK chemical warfare programme 3 of 10
Ministry of Defence from these records was made only the occurrence of the test was documented
available to us. We also searched the Porton Down because of the availability of exposure information
library for annual reports relating to the human from other sources.
volunteer programme for the same period (Porton Each chemical was classified into an extended cat-
Down, 1963–1988) and consulted two Porton Notes egorization derived from that in the NATO Handbook
and one Porton Medical Committee minute about on the Medical Aspects of NBC Defensive Operations
the history of the human volunteer programme (US Department of the Army, 1996) with assistance
(Bramwell, 1959; Bradshaw, 1965; Kemp, 1974). from Chemfinder (Chemfinder.com, 2004) and the
Bradshaw’s (1965) document, ‘The service volunteer Merck Index (O’Neill, 2001). This resulted in 16 cat-
observer scheme’, was an addendum to Bramwell’s egories which were: vesicants, nerve agents, riot con-
(1959) document, ‘History of the service volunteer trol agents, incapacitants, lung-damaging agents, blood
observer scheme at CDEE’, and these two Porton agents, choking agents, smokes, fuels and incendiary
Notes were treated as one. We compared the number devices, herbicides, irritants, treatments, rubber mix
of people in each calendar year from these four sour- tests, mustard sensitivity tests, other and unknown.
ces. We noted the years in which the Ministry of Dermal patch tests of different rubber compounds
Defence database recorded !10 persons fewer than were carried out in great numbers over the study pe-
the smallest number noted in the contemporaneous riod, we understand to assist in the specifications for
annual and other reports relating to that year. protective equipment. Mustard sensitivity tests were
routinely administered to exclude those with a known
Sample selection procedures sensitivity to mustard from being re-exposed. Chem-
A stratified random sample of 150 military service icals whose purpose was known but which did not
numbers was chosen from the database compiled by fit the categories listed above were categorized as
the Ministry of Defence. The database was stratified ‘other’ (e.g. dye, antiperspirant). Chemicals whose
(in order to obtain 30 veterans per decade) according purpose was unclear were categorized as ‘unknown’
to the decade of the first visit to Porton Down, (e.g. those listed only by a code number).
1939–1949, 1950–1959, 1960–1969, 1970–1979 or
1980–1989. This sample of 150 service personnel en-
sured that adequate exposure information was pro- RESULTS
vided for tests carried out in the later decades, when
fewer veterans attended Porton Down. It was known General review of the archive
that most veterans attended Porton Down in the early For the period 1939–1989 the Porton Down histor-
part of the study period. The military service numbers ical archive provided to us comprised 111 record
of 30 people were randomly selected from each books, categorized as administrative (n 5 14) and
decade. The selection by means of service numbers experimental (n 5 97) (Table 1). There were no gaps
necessarily excluded any civilians, although small between the finish date of one administrative book
numbers of Porton Down civilian staff were known and the start date of another, suggesting that this se-
to have participated in the programme. The adminis- ries of administrative books for this period was com-
trative and experimental books were then searched plete. It is known that experimental books covering
for all the information relating to each person’s first part of 1963 and 1964 are missing (Ministry of
visit to Porton Down; any second or subsequent visits Defence, 2006). It was clear that supplementary
were not included. material (in the form of Porton Notes and other
documents) may be available to assist indirectly by
Extraction of the exposure data providing background information for the exposure
Each test during the first visit was reviewed. Tests assessment, but it was unclear in this study if any
were categorized as chemical or non-chemical. further direct information on the exposures of indi-
Where there was no exposure to chemicals (such as viduals was available.
for personality tests, human performance tests and Most administrative books were handwritten and
tests of new equipment under varying conditions), all contained partial or complete information on
no data regarding the test were abstracted. some or all of the following: surname, forename, mil-
For chemical tests, the name of the chemical was itary service number, military rank, date of birth or
noted and chemicals with one or more synonyms age, military unit, and first and last dates of atten-
were recoded to the most common name. The title dance. Some also recorded a number assigned to
of the test, use of protective equipment, use of decon- each volunteer on arrival, a laboratory number for
taminants, prophylactic agents or antidotes, the level blood tests, height and weight, and records of pay
and duration of exposure, and any quantitative acute relating to the visit.
toxicity data were recorded. The route of exposure The experimental books were handwritten and
was abstracted later. For the standardized mustard contained extensive information whose content, style
sensitivity tests and rubber patch tests (see below), and quality varied widely between books, within

4. 4 of 10 T. J. Keegan et al.
Table 1. Books held in the Porton Down archive which related to the study period (1939–1989), by book category and
date range of content
Book type Category First date Last date n
Summary book Administrative 01/08/1938 01/12/1997 10
Alphabetical list Administrative 01/12/1945 01/01/1983 4
Mustard tests Experimental 01/06/1931 01/05/1978 7
Miscellaneous experimental records Experimental 01/01/1934 01/05/1988 8
Ward report Experimental 01/05/1938 01/01/1971 7
Chamber experiments Experimental 01/10/1938 01/10/1954 10
Day book Experimental 01/12/1938 01/05/1956 14
Physiological tests Experimental 01/01/1952 01/12/1957 1
Observers’ workbook Experimental 01/01/1956 01/01/1972 7
Psychological tests Experimental 01/06/1960 01/10/1974 2
Notebook Experimental 01/10/1966 01/11/1983 6
Official diary Experimental 01/01/1972 01/01/1988 16
Rubber mix tests Experimental 01/01/1974 01/07/1984 1
Human volunteer studies Experimental 01/02/1979 30/01/1992 18
Total 01/06/1931 30/01/1992 111
books, and over time (Fig. 1). Many results involving 1987, the Porton Notes the period 1946–1965 and
dermal exposures to vesicants were recorded in the the Medical Committee minute the period 1965–
form of pictograms (Fig. 1a), some results were 1987 (Bramwell, 1959; Bradshaw, 1965; Kemp,
recorded in table form (Fig. 1b), while others were 1974). Where data from two or more sources were
in a diary format (Fig. 1c). Records made in the available, all points are plotted. The Ministry of De-
1940s or 1950s were, in general, more legible than fence database shows that, over this period, the great-
those made in later decades. Experiments in the est numbers were recorded in WWII, with a second
books were usually, though not always, recorded peak of visits in the 1950s. Broadly, the numbers
chronologically. Individuals were identified in exper- from annual reports and other documents agree well
imental books by surname only and other identifying with the numbers in the Ministry of Defence data-
information, such as forename or service number, base, although in the years 1946, 1953, 1955, 1956,
was rarely documented. However, it was possible to 1958, 1960, 1961, 1962, 1963, 1968, 1969, 1973,
link individual entries with those in the administra- 1974, 1975 and 1981 the number from the database
tive books using the date of the test and the personal is !10 persons less than the number from other sour-
identifiers available. ces, possibly raising questions about the complete-
ness of the historical archive.
Types of test
The types of test were varied. Most took place in Experimental data for sample of subjects
a laboratory or environmental chamber on the Porton It proved very time consuming to abstract data re-
Down site. Others took place on the Porton Down lating to 150 specific individuals because the data for
ranges, an adjacent large area of chalk grassland. any one could be recorded in several different books.
The range trials were intended to simulate conditions The 150 sampled persons underwent 1075 tests and
likely to be encountered on active service. Many tests Table 2 shows the types of test carried out. Of the
were designed to test the protection afforded by, or 150 persons 69 were involved in more than one type
decontamination of, normal uniform or protective of test. A total of 120 persons were involved in tests
equipment. Other tests were of the efficacy of de- with chemicals and 72 with more than one chemical
contamination procedures, prophylactic agents and type. The median number of days on which tests
antidotes. were carried out on an individual was 2 (range 1–13).
In this sample, more tests took place on Mondays
Comparison with annual reports than on any other day (Monday, n 5 423; Tuesday,
The annual number of persons recorded in the n 5 176; Wednesday, n 5 197; Thursday, n 5 167;
Ministry of Defence database is compared with the Friday, n 5 99; Saturday, n 5 8; Sunday, n 5 5).
numbers recorded in annual reports and Porton Down This reflects the number of baseline tests that were
documents in Fig. 2. The Ministry of Defence data- carried out on the first day of each visit, Monday being
base covered the period 1941–1989. The available the most frequent day for new intakes.
Porton Down annual reports covered the periods There were 736 tests involving chemical agents:
1963–1968, 1973–1976, 1978–1981 and 1983– 102 were mustard sensitivity tests, 167 rubber mix

5. Exposures from the UK chemical warfare programme 5 of 10
Fig. 1. Example pages from the Porton Down archive. (a) Dermal exposure in the 1940s to a liquid vesicant with results
recorded on a semi-quantitative erythema and vesication scale, with the area of skin affected. (b) Exposure to a
nerve agent in 1952 with results expressed in table form. (c) Diary format, 1973.
1000 1500 2000 2500
Persons (n)
500
0
01 Jan 40 01 Jan 50 01 Jan 60 01 Jan 70 01 Jan 80 01 Jan 90
Date
MoD database Porton Down annual reports
Bramwell 1959, Bradshaw 1965 Kemp 1974
Fig. 2. Comparison of annual numbers of persons tested at Porton Down from the Ministry of Defence database
and from contemporaneous annual reports and other documents.

6. 6 of 10 T. J. Keegan et al.
Table 2. Types of exposures, dates of first and last tests and number of tests per person in the sample of 150 persons
Exposure type Persons, n Tests, n Date range of tests Number of tests
per person
First test Last test Median Range
Tests involving chemicals
Vesicant 29 103 10/12/1945 18/02/1976 2 1–16
Mustard sensitivity test 37 102 10/12/1945 16/02/1976 1 1–16
Nerve agent 32 44 17/11/1948 12/11/1979 1 1–4
Riot control 29 89 01/02/1949 13/03/1974 2 1–10
Incapacitant 10 11 21/03/1960 01/11/1972 1 1–2
Irritant 5 5 28/10/1954 19/06/1985 1 1–1
Treatment 42 118 21/08/1947 23/09/1986 2 1–17
Other chemical 19 61 01/02/1946 27/06/1988 1 1–16
Unknown chemical 13 36 18/05/1948 30/01/1985 1 1–13
Rubber mix test 49 167 30/01/1950 19/03/1984 2 1–16
Non-chemical test 66 251 16/08/1948 28/10/1988 3 1–16
Unclear exposure type 53 88 03/03/1947 04/12/1989 1 1–6
Total 150a 1075 10/12/1945 04/12/1989 3 1–33
a
Figures do not add up to 150 because individuals can appear in more than one exposure type category.
patch tests and 118 tests of treatments. The most was the most commonly tested. In this sample,
commonly tested group of chemical warfare agents 52 (51%) tests involving vesicants took place in the
in this sample was the vesicants (n 5 103) followed 1940s, and 18 (41%) tests involving nerve agents.
by riot control agents (n 5 89). There were 44 nerve In all, 65 (73%) riot control tests took place in the
agent tests, 11 with incapacitants and 5 with irritants. 1970s. Most tests involving treatments took place
No tests involved exposure to lung-damaging agents, in the 1970s (n 5 49, 42%) or 1980s (n 5 48, 41%).
blood agents, choking agents, smokes, fuels and in- Some information was found to be available on
cendiary devices or herbicides. In this sample, a rela- potential exposure or effect modifiers, depending on
tively large number of tests were classified as ‘other’ the type of agent under test and on the experimental
(n 5 61). The non-chemical tests included chest circumstances. This information included the fol-
X-rays, height and weight measurements and psy- lowing: the presence or absence of a physical barrier
chological profiling. There were 88 tests which could (such as a textile), between the exposure and expo-
not be categorized solely from the information avail- sure site; exposure site; exposure state (whether solid,
able in the books. Taken with the 36 tests involving liquid or gas/vapour); respirator use; use of other pro-
unknown chemicals, 123 (11%) of the 1075 tests tective equipment; whether the test took place in
could not be adequately categorized using only the a chamber; and the use of decontaminants, prophy-
information in the books (Table 2). lactics or antidotes. Exposure route (dermal, inhala-
The most commonly tested vesicant in this sample tion, etc.) could be imputed in many instances from
was sulphur mustard (82 of 103 vesicant tests). The the experimental circumstances.
median number of vesicant tests per person was 2
(Table 2). The commonly used method of vesicant Availability of quantitative information on exposure
testing involved placing several individual drops on The availability of quantitative information on
the skin of the anterior surface of the forearm (Fig. exposure in the tests is shown in Table 3. Overall,
1a); we noted each drop as a separate test. Sarin quantitative information on level of exposure was
was the most commonly tested nerve agent in this identified for 49% of tests, the best coverage being
sample (37 of 44 nerve agent tests). The median for vesicants (68%), treatment (67%) and nerve agents
number of nerve agent tests per person was 1. The (66%). Level of exposure was documented in a range
lachrymator dibenz(b,f)(1,4)oxazepine (CR) was of units: drops (n), drop diameter (in millimetres),
the most commonly tested riot control agent in this dilution (%), parts per billion, grams, milligrams,
sample (56 of 89 riot control agent tests). The median micrograms, cubic centimetres, milligrams per cubic
number of tests per person of riot control agents was metre and micrograms per litre. Chamber exposures
2 and many were repeat tests of one substance at dif- to nerve agents were recorded as concentration Â
ferent concentrations. The median number of tests time in units of milligrams per cubic metre minutes.
per veteran involving a treatment was 2; many in- Duration of exposure (usually in minutes) was,
volved repeat exposures to nerve agent prophylactic overall, recorded in 31% of tests, but in 66% of
tablets; pyridostigmine (41 of 118 treatment tests) vesicant tests and 77% of nerve agent tests. Mustard

7. Exposures from the UK chemical warfare programme 7 of 10
Table 3. Availability of quantitative exposure and acute toxicity data by chemical category in chemical tests identified for the
sample of 150 persons
Chemical category Number of Information available on
tests in
category Level of exposure Duration of exposure Acute toxicity
n Row % n Row % n Row %
Vesicant 103 70 68 68 66 74 72
Mustard sensitivity test 102 102 100 102 100 102 100
Nerve agent 44 29 66 34 77 25 57
Riot control agent 89 44 49 8 9 1 1
Incapacitant 11 6 55 0 0 0 0
Irritant 5 3 60 3 60 1 20
Treatment 118 79 67 5 4 1 1
Other chemical 61 12 20 3 5 6 10
Unknown chemical 36 19 53 2 6 12 33
Rubber mix test 167 n/a — n/a — n/a —
Total 736 363 49 225 31 222 30
n/a 5 not available.
sensitivity tests appeared to be standardized and it is for sarin was ‘mixed route’ (n 5 25), a description
understood that information on the test procedures we used for tests in a chamber where, as a result of
and methodology is available in Porton Down docu- the solubility and volatility of sarin, inhalation, ocu-
ments, for example in Fairley (1932). Details of indi- lar and dermal exposures were possible, although
vidual rubber mix compounds are also available. respirators were recorded as being worn in three
tests. CR exposures in this sample were principally
Availability of quantitative acute toxicity data oral, part of a series of experiments to detect the low-
Quantitative measures of acute toxicity were typi- est concentration at which food or water contami-
cally documented for vesicant (72%) and nerve agent nated with CR could be detected.
(57%) tests, but not for agents in other categories
(Table 3). The dermal effect of vesicant exposures
DISCUSSION
was recorded on a semi-quantitative scale as the
severity of erythema (reddening of the skin) and These results are based on a sample of 150 veter-
presence of vesication (blistering). Responses docu- ans derived from a representative sample of 30 from
mented in this sample were: eÀ, e, eþ, v. Our under- each decade of the study period. Because of the sam-
standing is that eÀ represents mild reddening, e pling method used, the distribution of visits to Porton
reddening, eþ pronounced reddening and v the Down will not match that of the whole cohort and no
presence of a vesicle (blister) (Kingan, 1937). Quanti- inferences about exposures in the whole cohort can
tative responses to nerve agents were documented be made from these results. The most important find-
in two ways: change in pupil size and change in level ing of this study was that it will be possible to recon-
of blood cholinesterase. The records also contained struct exposures at an individual level for the cohort
qualitative information about certain symptoms re- study from the records made available to us at Porton
ported by some of the participants. As there were Down. Despite the 50-year period of the exposure re-
no plans to make use of qualitative symptoms data cords, the data quality appeared to be reasonably
in the cohort study, none of this was abstracted for consistent over time. This exposure assessment will
the sample. be a valuable component of the cohort study of mor-
tality and cancer incidence in Porton Down veterans.
Exposures in the sample The completeness of information is an important
Information on level of exposure and duration of question in any historical cohort study (Enterline
test to three selected substances (sulphur mustard, sa- and Marsh, 1982); this study has provided evidence
rin and CR) by exposure route is shown in Table 4. that the information provided to us appeared largely
Most (85% of 82) of the mustard tests were dermal complete. It may be possible to undertake sensitivity
with between 0.28 and 4000 lg of liquid mustard analyses to estimate the effects of any missing data.
for between two and 360 min. In this sample, the test Because many veterans had multiple tests, it was
with the highest exposure concentration took place clear that it will be necessary to abstract data in
for 2 min and in the presence of a decontaminant. a test-wise format and subsequently to link all tests
The exposure route most commonly documented for each person. A database has been designed for

8. 8 of 10 T. J. Keegan et al.
Table 4. Exposures to three selected chemicals, by exposure route, in the sample of 150 persons
Chemical name Exposure route Level of exposure in test Duration of test
n n Median (range) in n Median (range)
various units in minutes
Sulphur mustard: (2,2#-dichlorodiethyl 70 Dermal 44 18 (0.28–4000) lg 50 6 (2–360)
sulphide) (n 5 82)a,b
12 Not known n/a n/a
Sarin: (O-isopropyl methyl- 3 Dermal 2 142 and 500 lg 1 30
phosphonofluoridate) (n 5 37)b 25 Mixed 16 9.6 (1–15) mg min mÀ3 24 22.5 (2–150)
9 Not known 1 0.5 mg mÀ3 3 30 (30–30)
CR: (dibenz(b,f)(1,4)oxazepine) 6 Dermal 4 0.025% (0.001–0.05%) 2 10 and 60
(n 5 56)b 32 Oral 11 0.00005% (0.00005–0.1%) n/a
1 Ocular 1 0.1% n/a
3 Mixed 2 0.1% and 3 lg n/a
14 Not known 6 0.0055% (0.001–0.01%) n/a
n/a 5 not available.
a
Does not include mustard sensitivity tests.
b
Figure denotes number of tests, not number of individuals.
the extraction of data for the exposure assessment ual exposure can only be estimated from data relating
(Fig. 3). Algorithms will be developed for data anal- to groups. Furthermore, the availability of acute tox-
ysis using a hierarchical approach. At the first level icity data for vesicants and nerve agents offers po-
above that of the cohort of veterans, it will be possi- tential for constructing a measure of absorbed dose,
ble to classify each serviceman as ever/never ex- although it is as yet unclear if this measure of dose
posed to any chemical at Porton Down. At the next is relevant to mortality and cancer incidence. One
level, it will be possible, for each serviceman and challenge for the epidemiological analysis will be
each chemical or group of chemicals, to define expo- to classify veterans’ exposure using this quantitative
sure as ever/never exposed. Mortality and cancer in- information when the unit of measurement has
cidence will be studied within major exposure changed over time, as is the case for both exposure
groupings. At the last level of complexity, it will be and acute toxicity. The analysis will also need to con-
possible, for those exposed to vesicants or nerve sider sources of variation in these quantitative mea-
agents, to define groups for analysis based on the surements, for example from difference in laboratory
level of exposure or the acute toxicity experienced. methods. For the measures of acute toxicity, such as
The exact measures of exposure and acute toxicity change in level of blood cholinesterase, inter- and
to be used will be decided after examination of all intra-individual variation in the response to exposure
the available data. are also important (Brock and Brock, 1993).
For the majority of chemical tests recorded in the The exposure and acute toxicity data for vesicants
experimental books for this sample, the name of appeared to be in a similar format to that reported
the chemical was clear and it could be assigned to in the cohort study of WWII US Navy personnel
a chemical warfare agent category derived from the exposed to sulphur mustard in human experiments
NATO Handbook (US Department of the Army, (Bullman and Kang, 2000), raising the possibility
1996). However, in .10% of the tests undergone of comparisons with this cohort.
by this sample of servicemen, the chemical could This study allows some preliminary speculation
not be categorized or the type of exposure was un- about the levels of exposure received by the veterans
clear. It may be that this percentage can be reduced who participated in these tests, although robust infer-
in the exposure assessment by enquiries in the Porton ences on the whole range of exposures over 1939–
Down library or to Porton Down staff. 1989 cannot be made from this study alone. While
In many occupational epidemiological studies the exposures have in some cases been high, leading to
only information available which bears on exposure one fatality (Ministry of Defence, 2006), indications
is job title and duration of employment and is, at are that most were not. It is likely that these sub-
best, semi-quantitative. For this cohort, in contrast, lethal exposures are comparable to the sub-lethal ex-
quantitative data on level and duration of exposure posures outside the immediate target zone in warfare
are available for vesicants and nerve agents, two of or a terrorist incident, where a decreasing gradient
the largest subgroups of chemicals in this sample. of exposure from the point of release should be ex-
Moreover, this quantitative exposure information is pected. Another similarity to warfare or a terrorist
available for individuals in contrast with many incident is that the exposures occurred as a single
occupational epidemiological studies where individ- exposure of short duration or small number of short

9. Exposures from the UK chemical warfare programme 9 of 10
Fig. 3. Diagram showing the decision path for entering data into the exposure database for the cohort study of
Porton Down veterans.
exposures. However, there are difficulties in making UK forces against nuclear, chemical or biological at-
comparisons with occupational populations, where tack, many of the exposures of test participants were
it is commonly assumed that cumulative exposure modified by, for example protective clothing, barrier
is relevant for the development of cancer and other creams or pharmaceutical prophylaxis or antidotes.
long-term outcomes. The data examined for this For example, in tests involving dermal vesicant expo-
sample suggest that the cumulative exposure of the sures, the presence or absence of clothing, protective
Porton Down veterans would appear to be lower than clothing or other barrier between the exposure and
that of exposed workforces because occupational the skin was recorded. Although it will not be possi-
exposures occur repeatedly over a working shift, for ble to transcribe every item written in the archive for
a working week, for months or years of employment. use in the epidemiological study, it was clear that
Such presumed high levels of cumulative occupa- some information on potential exposure and effect
tional exposure to sulphur mustard in a UK WWII modifiers can usefully be abstracted (Fig. 3).
munitions factory were associated with an excess of
respiratory cancers (Easton et al., 1988), although Acknowledgements—We thank our colleagues at the University
the lower cumulative exposures in experiments of Oxford for their contributions to this paper: Steven Allender,
Sageet Amlani, Terence Boyle, Kate Brockhurst, Claire
involving US Navy personnel in WWII were not Brooks, Steve Davies, Chris Shield and Sue Walker. We thank
(Bullman and Kang, 2000). If it is the case that cumu- Nick Blatchley, Dan King and Tracey Vennai from the Veterans
lative exposures were, in general, low in this cohort Policy Unit of the Ministry of Defence (formerly the Gulf Vet-
over 1939–1989, then a very large study is likely to erans’ Illness Unit) and Gradon Carter, Hugh Dyson, Rick Hall,
Marie Jones and Viv Worrall from the Defence Science and
be needed to estimate any long-term health effects, Technology Laboratory, Porton Down, who facilitated the
including rarer outcomes such as specific types of work. We also thank the members of the Medical Research
cancer. The UK cohort includes $20 000 veterans, Council (MRC) liaison group for their helpful comments on
although the numbers exposed to each specific the draft manuscript. The study was funded by a grant from
the UK Ministry of Defence administered by the MRC
chemical are lower. (G0200288). The sponsors had no role in the collection, analy-
Since much of the of the work at Porton Down ap- sis or interpretation of data or in the writing of the report. No
pears to have been related to designing protection for conflicting interests are declared. Ethical approval was granted

10. 10 of 10 T. J. Keegan et al.
by the South East Multicentre Research Ethics Committee (02/ Kingan R. (1937) Experiments to compare the skin burning
01/115) and by the Defence Medical Services Clinical Re- power of drops of viscous HS and viscous HT through cloth-
search Committee. ing. With two addendums to report. Porton Report, 1715.
Porton Down, UK.
Ministry of Defence. (2006) Historical survey of the
REFERENCES
Porton Down service volunteer programme 1939–1989.
Balali-Mood M, Hefazi M, Mahmoudi M et al. (2005) Long- London: Ministry of Defence. Available at http://www.mod.
term complications of sulphur mustard poisoning in severely uk/DefenceInternet/AboutDefence/WhatWeDo/Healthand
intoxicated Iranian veterans. Fundam Clin Pharmacol; 19: Safety/PortonDownVolunteers/HistoricalSurveyOfThePorton
713–21. DownServceVounteerProgramme19391989.htm. Accessed 30
Beebe G. (1960) Lung cancer in World War 1 veterans: possi- November 2006.
ble relation to mustard gas injury and 1918 influenza epi- Miyaki K, Nishiwaki Y, Maekawa K et al. (2005) Effects of
demic. J Natl Cancer Inst; 25: 1231–52. sarin on the nervous system of subway workers seven years
Bradshaw RP. (1965) The service volunteer observer scheme. after the Tokyo subway sarin attack. J Occup Health; 47:
Porton Note, PN 119A. Porton Down, UK. 299–304.
Bramwell ECB. (1959) History of the service volunteer observer National Research Council. (1982) Possible long term effects
scheme at CDEE. Porton Note, PN 119. Porton Down, UK. of short term exposure to chemical agents. Volume 3, Final
Brock A, Brock V. (1993) Factors affecting inter-individual report: health status of test subjects. Washington, DC:
variation in human plasma cholinesterase activity: body National Academy Press.
weight, height, sex, genetic polymorphism and age. Arch Nishiwaki Y, Maekawa K, Ogawa Y et al. (2001) Effects of
Environ Contam Toxicol; 24: 93–9. sarin on the nervous system in rescue team staff members
Bullman T, Kang H. (2000) A fifty year mortality follow-up and police officers 3 years after the Tokyo subway sarin
study of veterans exposed to low level chemical warfare attack. Environ Health Perspect; 109: 1169–73.
agent, mustard gas. Ann Epidemiol; 10: 333–8. Norman JE, Jr. (1975) Lung cancer mortality in World War I
Carter GB. (2000) Chemical and biological defence at Porton veterans with mustard-gas injury: 1919-1965. J Natl Cancer
Down, 1916-2000. London: HMSO. Inst; 54: 311–7.
Case RAM, Lea AJ. (1955) Mustard gas poisoning, chronic O’Neill MJ. (2001) The Merck Index: an encylopedia of
bronchitis and lung cancer. An investigation into the possi- chemicals, drugs, and biologicals. Whitehouse Sation, NJ:
bility that poisoning by mustard gas in the 1914–18 war Merck Research Laboratories.
might be a factor in the production of neoplasia. Br J Prev Page WF. (2003) Long-term health effects of exposure to
Soc Med; 9: 62–72. sarin and other anticholinesterase chemical warfare agents.
Chemfinder.com. (2004) ChemIndex. Cambridge, MA: Mil Med; 168: 239–45.
CambridgeSoft Corporation Available at http://chemfinder. Pechura CM and Rall DP, editors. (1993) Veterans at risk: the
cambridgesoft.com. Accessed 30 November 2006. health effects of mustard gas and lewisite. Washington,
Easton DF, Peto J, Doll R. (1988) Cancers of the respiratory DC: National Academy Press.
tract in mustard gas workers. Br J Ind Med; 45: 652–9. Porton Down. (1963–1988) 17 unpublished annual reports.
Enterline PE, Marsh GM. (1982) Missing records in occupa- Porton Down, UK.
tional disease epidemiology. J Occup Med; 24: 677–80. US Department of the Army. (1996) Field manual 8–9; NATO
Evans R. (2000) Gassed. London: House of Stratus. handbook on the medical aspects of NBC defensive opera-
Evison D, Hinsley D, Rice P. (2002) Chemical weapons. Br tions, AMedP-6(B). Washington, DC: US Department of
Med J; 324: 332–5. the Army.
Fairley A. (1932) Report on sensitivity to mustard gas in nor- Wada S, Miyanishi M, Nishimoto Y et al. (1968) Mustard gas as
mal persons. Porton Report, 999. Porton Down, UK. a cause of respiratory neoplasia in man. Lancet; 1: 1161–3.
Kemp MH. (1974) The Porton volunteer scheme. Porton Down Wiltshire Police. (2004) Operation Antler. Available at
Medical Committee Minutes, Med IT/4228/3002/73. Porton http://www.wiltshire.police.uk/antler/default.asp. Accessed
Down, UK. 30 November 2006.