Mortality of army cooks.

The possible hazard of lung cancer among cooks was studied in a cohort of 1798 cooks who had retired from the Army Catering Corps and 1310 referents retired from the Royal Army Pay Corps. During the follow-up from 1974 to 1989 the mortality of the referents was similar to that of the national population, apart from a moderate increase in lung cancer [standardized mortality ratio (SMR) 1.38]. Mortality from lung cancer among the cooks was significantly higher than among the national population [SMR 1.82, 95% confidence interval (95% CI) 1.25-2.57], as was mortality from cancer of the large intestine (SMR 3.03), ischemic heart disease (SMR 1.42), cerebrovascular disease (SMR 2.05), and digestive disease (SMR 2.27). The high rate of lung cancer among the cooks supports the hypothesis of an occupational hazard, although at least part of the excess was probably due to smoking. Possible explanations for the elevated mortality from other diseases include poor nutrition in early life, smoking, and high consumption of alcohol.


Routine analyses of occupational mortality in Great
Britain (1-3) and Denmark (4) , and of canc er incidence by occupation in California (5) , have indicated high rat es of lung canc er in cooks.These observations are supported by the result s of two case-referent studies (6,7).To explore further the risk of lung cancer associated with work as a cook , we have carried out a cohort study of retired members of the Army Catering Corps.

Subjects and methods
The study cohort comprised all retired cooks from the Arm y Catering Corp s who were on the reserve list (ie, members of the regul ar arm y reserve, the Territorial Army, or the Ulster Defence Regiment ) at any time bet ween I January 1974 and 31 December 1984.As an internal co ntro l we took retired members of the Royal Arm y Pay Corps who were on the reserve list during the sam e l l-year period.All of the subjects were men .Most of the soldiers studied would have joined the Arm y soon after leaving school, entry to a particular Corps bein g determined by personal choice and aptitude.The study and reference cohorts were identified from computerized pension record s which included surn am e, initials, date of birth, last known address, Nation al Insurance number, and date s of last ent ering and leaving regul ar service.Full first names were obtained from social security records by linkage through the National Insurance number, surname, and date of birth, and with this additional information subjects were traced through the National Health Service Central Register up to 3 1 December 1989.For those men who had died , death certificates were obtained with the und erl ying cau se of death cod ed to the ninth revi sion of the International Clas sification of Diseases.
In the anal ysis, the subjects were con sidered to be at risk from I January 1974 or their date of entry to the reserve list, whichever was later.The mortality of the study and reference cohorts was compared with that of the national population by the personyears method, the exp ected numbers of deaths being calculated for five-year age and calendar intervals, and the 95% co nfidence intervals (95% CI) for standa rdized mortality ratios (SMR) being based on the Poi sson distribution.In addition, the death rates of the two cohorts were compared directl y by Poisson regr ession anal ysis (8).
From the pen sion records we identified 1798 retired cooks and 1310 referents.At least 74.6 % of the cook s and 78.2 % of the referents had completed more than 20 years of regular service.(The exact proportion s were unknown because, for men who had left the army and then reenlisted, information was only available for the mo st recent period of service.) Table I summarizes the out come of the tracing exercise.Altogether, 92.4% of the cook s and 93.6 % of the referents were successfully traced throu gh the National Health Servi ce Central Register.The untraced subjects were excluded from further analysis, together with 33 men who had emigrated before the start of the follow-up peri od.Another 32 subjects who emigrated during follow-up were considered at risk up to the date of their migration.

Results
By the end of 1989, 227 deaths had occurred among the cooks and 150 among the referents.Table 2 shows the mortality of the cooks and referents as compare d with that of the national population.The overall death rate of the refere nce group was close to expec tat ion (SMR 1.06, 95% CI 0.89-1.24),as was mortality from most of the common causes of death.The exception was lung cancer, for which mortality was moderately elevated, although not to the point of statistica l significa nce (SMR 1.38, 95% CI 0.88-2.07).In contrast, the cooks had significantly eleva ted death rates from all causes (SMR Cooks Referents 1.48, 95% CI 1.30-1.69),all cancers (SMR 1.41, 95% CI 1.09-1.80),cancer of the large intes tine (SMR 3.03, 95% CI 1.39-5.75),lung cancer (SMR 1.82, 95% CI 1.25-2.57),ischemic heart disease (SMR 1.42, 95% CI 1.13-1.76),cere brovascular disease (SMR 2.05,95% CI 1.22-3.24),and digestive disease (SMR 2.27, 95% CI 1 .0~.30) .When the analysis was rest ricted to those men who were known to have completed at least 20 years of service, the risk esti mates were much the same.
Table 3 shows the morta lity of the cooks relative to that of the refere nts for selected causes of death .In comparison with the death rates of the refere nts, those of the cooks were significant ly elevated for all causes combined [risk rati o (RR) 1.41, 95% CI 1.14-1.73)and, specifically, for cancer of the large intestine (all nine cases occurr ed among the cooks) and circ ulatory disease (RR 1.45, 95% CI 1.07-1.96).The last-mentioned category reflected excesses of both ischemic heart disease (RR 1.43) and cerebrovascular disease (RR 2.17).Mortality from lung cancer was higher among the cooks, but the difference was not statistically significant (RR 1.29, 95% CI 0.76-2.21).A similar increase in risk was observed for other smoking-related cancers (RR 1.24), but not for respiratory disease (RR 1.04).The nine deaths from digestive disease among the cooks (RR 2.93) included four from peptic ulcer , four from cirrhosis or alcoholic liver disease, and one from intestinal obstruct ion.

Discussion
The proportion of untrace d subjects was higher in this investigatio n than in most industrial cohort stud- ies, prob ably because some of the men went to live abroad soon after leaving the regular army and never registered with a physician in the National Health Service.If so, the incompl ete tracing would not be expected to produ ce any serious bias, especially in internal comparisons with the referents from the Pay Corps.
Overall mortality in the referenc e group was similar to that in the national population.Given that entry to the study was through regi stration on the reserv e list, it is perhap s surprising that a healthy worker effect was not apparent.Chron ically disabled soldiers are excluded from entry to the reserve, but this selection does not seem to have any strong longterm impact on mortalit y.
The only common cause of death with markedly increased incidence in the reference group was lung cancer (SMR 1.38).It seems likely that this excess is attributable , at least in part, to high rates of smoking.No data are available on the prevalence of smoking in the Pay Corps spec ifically, but, in the Army as a whole, the habit has been common.In a survey of 15-to 18-year -old entrants to the Army in 1959, 76% had smoked during the previous four weeks and 53% smoked at least 20 cigarettes per week (9).Smoking habits may also have accounted for the referen ts' increased mortal ity from respiratory disease.Members of the Pay Corps are not exposed to any known occup ational causes of lung cancer.
Mortality from lung cancer among the cooks was even higher than among the referents, and it was significantly elevated in comp arison with national rates (SMR 1.82).Again, smoking may have contributed to this excess, although it would be unusual to observe such a high mortality ratio solely throu gh the confounding effect of smoking (10).Certainly, the find ing would be compatible with an occupational hazard , as suggested by earlier studies (1-7).Most of the cooking in the Army is carr ied out in permanent rather than field kitchens, mainly using gas rather than electricity, and includes a lot of frying.A hazard of lung cancer might arise from the fumes generated during cook ing, particularly when food is fried .Such fumes are known to be mutagenic (1 1, 12), and in two Chinese case-referent studies frequent frying and smokiness dur ing cooking were associated with an increased risk of lung cancer (I3, 14).
The high mort ality from other causes among the cooks, particularly from circulatory and digestive disease, is also compatible with the findings of other studies (1-3).Four of the nine observed deaths from digestive disease were ascribed to cirrhosis or alcoholic liver disease, and it seems likely that cooks drink more heavily than the average -perhaps because alcohol is more readil y available to them.
The death rates from circulatory disease were significantl y elevated not only in comparison with the nation al popul ation, but also relativ e to the internal referen ce group .Cooking fumes do not contain any known cardiotoxins which would explain this excess.Smoking may contribute, but it would be unlikely to account fully for such a large increase in cardi ovascular mortality .Another possibility is that recru its to the Catering Corps come from a different social background from tho se entering the Pay Corps.Evidence is mounting that nutrition in early life has an important influence on later risk of circulatory disease, acting through a variety of pathogenetic mechanisms (15-1 8).A cross-sectional survey of cardiovascular risk factors among curren t member s of the Caterin g Corps might help to clar ify the reasons for the high mortality from vascular disease in our study.
In contrast to the findings on other causes of death, the high mortality of the cooks from colonic cancer was unexpected .No similar increase in risk has been found in routine analyses of occup ational mortality (2,3), and no aspect of cooking is suspe cted of strongly influencing the risk of large bowel cancer.The observation may, therefore, be a chance phenomenon.If so, the excess should diminish with continued follow-up.
In summary, this study supports the hypothesis that work as a cook increases the risk of lung cancer, although a spurious association from a confound ing effect of smokin g cannot be totally ruled out.It also indicates that retired members of the Catering Corps have unusually high mortality from circulatory disease.A survey of risk factors for vascular disease among current members of the Catering Corps might help to explain this observation and allow appropriate preventi ve measures to be instituted.
They would then have received training in their chosen trade.The work of the Pay Corps is largel y clerical.While in regular servi ce, members of both the Cat ering and I MRC Environmental Epidemiology Unit, University of Southampton, Southampton General Hospital, Southampton.Reprint requests to: Dr D Coggon, MRC Environmental Epidemiology Unit, University of Southampton, Southampton General Hospital, Southampton, S09 4XY, United Kingdom.the Pay Corp s undertake physical tra inin g, and both groups are required to work irregular hours.

Table 1 .
Outcome of the tracing .

Table 2 .
Mortality of the cooks and refere nts as compared with that of the national population .(SMR : standardized mortality ratio, 95% CI : 95% confidence interval) a Code of the International Classificati on of Diseases, ninth revision, in parentheses.

Table 3 .
Mortality of the cooks relative to that of the referents, as estimated by a Poisson regression analysis .(95% CI = 95% confidence interval )