MATERIALS AND METHODS

The study population consisted of cabin crew employed at the Swedish Scandinavian Airline System (SAS), resident in Sweden and employed some time during the period 1957–94. The cabin attendants at SAS were identified using administrative company registers of employees and archive documents at the Swedish part of SAS. In all, the study population included 3202 subjects. The Swedish personal identification number and date of employment were retrieved for 2956 of the study population (92%), of whom 2324 were women and 632 men. Analyses of the completeness of the cohort indicated only a small number of cabin crew missing when compared with information about annual numbers of cabin attendants employed at SAS during the study period.

Incident cases of cancer in the study population during the period 1961–96 were identified by record linkage to the Swedish National Cancer Register using the personal identification number. Similarly, all deaths in the study population during this period were obtained from the National Cause of Death Register. From a register of migration at Statistics Sweden, all migrations into or out of Sweden by members of the study population during the years 1968–96 were recorded. Reproductive history for the female cabin attendants was obtained using the Multi-Generation Register at Statistics Sweden and the National Medical Birth Register. The Multi-Generation Register was also used for information about parity and age at birth of first child among women in Sweden in 1980.

The observation period started on 1 January 1961, at the time of employment or at the time of immigration to Sweden, whichever came last. It ended on 31 December 1996, at the time of death, or at the time of emigration out of Sweden, whichever came first. For a subject moving out of the country, the observation period was terminated at the time of emigration, irrespective of eventual later re-migration to Sweden.

Dates of employment were collected from archives at Swedish SAS. From 1979 onwards, cumulative block hours by aircraft type during employment in Swedish SAS were recorded by the company on a routine basis, yielding monthly summary reports. Block hours include the time from departure of a flight until arrival, in the air as well as on the ground. From these reports cumulative block hours by aircraft type were extracted annually by 31 December during the period 1979–95 for the cancer cases registered from 1980 to 1996 and for four controls per case. The controls were randomly selected from each case’s “risk set”, which was composed of all individuals in the cohort who matched the case by gender and age group (±5 years), and who were employed in SAS and not diagnosed with cancer at the time the case was diagnosed. In 1993 a Swedish domestic airline company (Linjeflyg) was merged with Swedish SAS. Cabin crew formerly employed in Linjeflyg could not be included in the case-control study because of a lack of information on block hours before joining Swedish SAS. Information about annual cumulative block hours for the remaining cabin crew was obtained for 48 cases and 174 controls (92%). For some cabin attendants, however, block hours were missing for one or more years during the observation period, and for these years cumulative block hours were imputed using information from adjacent years. High altitude, long distance flights were determined on the basis of the aircraft type. All block hours attributed to DC 8, DC 10, Boeing 747, and Boeing 767 were considered to concern high altitude, long distance flights.

Information on reproductive factors among the female cabin attendants and among the total Swedish female population in 1980 was used to evaluate to what extent reproductive factors may have influenced the breast cancer incidence of female cabin attendants in the present study. The method used has been recommended when evaluating confounding due to cigarette smoking in occupational studies.²¹ In this evaluation we used estimates from a meta-analysis of the relative risk of breast cancer for the Swedish female population according to reproductive factors.²²

Statistical analyses

Cancer incidence overall and for specific cancer sites for cabin crew were compared with the corresponding incidence in the general Swedish population by calculation of standardised incidence ratios (SIR), based on the ratio of the observed number of cases and the expected number of cases. The expected number of cases was computed using five-year age groups and eight different calendar periods (1961–62, 1963–67, 1968–72, 1973–77, 1978–82, 1983–87, 1988–92, 1993–96). Together with the SIR, 95% confidence intervals (95% CI) were computed, assuming that the observed number of cancer cases followed a Poisson distribution. The case-control study nested within the cohort was analysed by calculating odds ratios (OR) for cancer incidence overall, malignant melanoma, and for breast cancer among women by comparing cabin crew with at least 10 000 cumulated block hours, any high altitude, long distance flight duty, and cabin crew with at least 5000 high altitude, long distance flight hours with cabin attendants without this experience. Odds ratios were calculated, together with 95% confidence intervals, using conditional logistic regression. Analyses were carried out using Epicure (HiroSoft, Seattle, WA) and SAS for Windows, version 8.

RESULTS

The total number of person-years was 39 135 in women, and 12 774 in men. At the start of follow up, members of the cabin crew were in general below 30 years of age; a dominant part of the person-years generated in this study were below 50 years of age. Only about 5% of the person-years were above 60 years of age, which is the regular age of retirement among cabin crew in Sweden. Emigration and immigration was quite common in the cohort; 21.1% (n = 490) of the women emigrated during the study period and 5.3% (n = 124) immigrated. The corresponding figures for men were 19% (n = 120) emigrants and 8.5% (n = 54) immigrants. At the end of follow up 29 women and 48 men were dead.

A total of 109 cancer cases were observed, of which 76 were among women and 33 among men. The observed number of cases was close to that expected on the basis of the cancer incidence of the general Swedish population, taking age and calendar period into consideration (table 1). The SIR was 1.01 (95% CI 0.78 to 1.24) for women and 1.16 (95% CI 0.76 to 1.55) for men. In both genders there was an increased incidence of malignant melanoma (SIR 2.18 (95% CI 1.09 to 3.90) for women and 3.66 (95% CI 1.34 to 7.97) for men). Men also showed an increased incidence of non-melanoma skin cancer (SIR 4.42; 95% CI 1.20 to 11.32). Of the four cases of non-melanoma skin cancer in men, one was a Kaposi’s sarcoma, which, because of the very low incidence in the general population, gave a high SIR. In women, a non-significant increased incidence was found for breast cancer (SIR 1.30; 95% CI 0.85 to 1.74) and for leukaemia (SIR 3.14; 95% CI 0.86 to 8.04).

No clear associations were found between years of employment and cancer incidence overall or malignant melanoma, nor for breast cancer among women (table 2). For malignant melanoma the total number of observed cases for men and women combined was 17 compared to 6.53 expected (SIR 2.61; 95% CI 1.52 to 4.17; not shown in table). Based on these 17 cases no association with length of employment was seen. The SIR for malignant melanoma was in the order of 2.2–3.0 throughout the study period with no substantial changes. An increased incidence of leukaemia was seen among subjects with 20 or more years of employment in SAS, but this was based on only two cases (not shown in table).

Information about total flight hours and flight hours in high altitude, long distance flights was obtained for 48 cancer cases diagnosed 1980–96 and for 174 controls. Of these cases 16 had breast cancer and 10 melanoma or non-melanoma skin cancer. The numbers of total and high altitude, long distance block hours, respectively, were similar between cases and controls. There were substantial differences in block hours between men and women. The average number of total cumulative block hours in 1995 was 6440 hours for female and 13 370 hours for male cabin crew. In general, no statistically significant associations between block hours and/or experience of high altitude, long distance flights and cancer incidence could be determined. In most analyses, the number of cases was small (table 3). Female cabin crew with at least 5000 high altitude, long distance block hours showed a non-significant increased incidence for breast cancer (SIR 3.27; 95% CI 0.54 to 19.70) compared to female cabin crew with less or no experience of high altitude, long distance flights.

The proportion of the female cabin attendants who had given birth to a child was 72.5%, which was very close to the corresponding proportion among all Swedish women in 1980 (table 4). Female cabin attendants tended to have fewer pregnancies, which were later in life. The proportion of cabin attendants who gave birth before 25 years of age was only 7% compared to 51% among all Swedish women. Given these differences in reproductive history and with the relative risks for breast cancer, estimated from a Nordic meta-analysis,²² one could expect an excess incidence of breast cancer among female cabin crew in the order of 10% (SIR 1.10) compared to the Swedish female population. This is lower than, but considering random variation, consistent with, the observed SIR for breast cancer in this study.

DISCUSSION

The results of the present study showed an increased incidence of malignant melanoma in Swedish cabin crew among both men and women and of non-melanoma skin cancer in men. These increased incidences are in agreement with corresponding analyses previously reported from Iceland, Norway, and the United States^4–6; similar findings have also been suggested from studies of Nordic pilots.^15–18 The increased incidence of malignant melanoma remained stable throughout the study period. During this period the incidence of malignant melanoma increased in the population,²³ which with stable relative risks for cabin crew would mean a greater increase of the incidence among cabin attendants than in the general population. The non-significantly increased incidence of breast cancer in female cabin crew is consistent with the previously reported corresponding results from Iceland, Finland, and the United States, but at variance with the results from a Norwegian study.^1,4–6

The Swedish National Cancer Register used in this study for identification of cases is based on a compulsory reporting of diagnosed cases and has a high degree of completeness. A comparison to the National Cause of Death Register in 1978 showed a loss of cases in the Cancer Register of 4.5%.²⁴ Each case is generally reported both by the clinician first making the diagnosis and also by a pathologist or cytologist. This contributes to a comparatively low misclassification of disease in the cancer register. It is possible that the frequent and regular health examinations of the cabin attendants contributes to an earlier detection of cancer cases and thus to a higher diagnostic rate. This could result in an overestimation of the cancer incidence in cabin crew compared with the general population.

Cabin crew is an occupational group with a comparatively extensive international migration. In this study we could take immigration to and emigration from Sweden into account by using national population registers going back to 1968. In order to avoid a loss of cases due to immigration or emigration, the follow up time was started at the time of immigration and terminated at the time of first emigration. Thus, no person-time at risk was accumulated outside Sweden, where cancer cases could not be detected. The loss of person-years and cases among subjects returning to Sweden was about 10% of all potential person-years and cases in the study. By record linkage to the National Cause of Death Register, all deaths during the study period could be obtained, thereby avoiding an overestimation of the person-years at risk.

Total block hours and block hours at high altitude, long distance flights were estimated from monthly reports recorded at the Swedish SAS. This does not give a complete account of all block hours of the cohort members. In particular, block hours obtained in service for other companies would be underrepresented or completely missing. This would contribute to a misclassification of block hours, probably underestimating total exposure and exposure to high altitude, long distance flights. It is possible that this has contributed to the lack of association between total block hours and high altitude, long distance block hours, respectively, and cancer incidence in this study. This source of error cannot, however, explain the associations or lack of associations seen in analyses of all cabin crew, irrespective of classification according to block hours or aircraft type. Information about block hours was only available from 1979 and onwards. In analyses of exposure to high altitude, long distance flights the numbers of cases was often small. Since the flight duty of cabin attendants in Sweden during the 1960s and 1970s mostly concerned propeller planes flying at lower altitudes it is unlikely that inclusion of these earlier years would have substantially increased the number of cases exposed to high altitude, long distance flights.

It is likely that the observed increase of malignant melanoma of the skin in all cabin crew and of non-melanoma skin cancer in male cabin crew is real. The causes of these increased risks remain unclear, and both occupational and non-occupational causes may be involved. The melanomas often occurred on the trunk, indicating that sunburn could be of aetiological importance. Measurements of UV radiation in the aircraft during flight have shown very low levels of exposure, and that the windshield of the aircraft gives a good protection against this type of radiation.²⁵ Thus, work as a cabin attendant during flight does not seem to give an increase in exposure to UV radiation.

We observed a non-significant increased incidence of breast cancer in female cabin crew. In view of similar findings in other Nordic countries and in the United States it is likely that this reflects a real increased incidence in this occupational group. The causes of this increase are not known and may involve occupational as well as other factors, including reproductive history. We estimated that the reproductive history in female cabin crew could yield a 10% increase in breast cancer incidence compared to the general population. This could be an important contributing factor to the increased breast cancer incidence in female cabin crew, but does not seem to fully explain this excess. It has been suggested that exposure to cosmic ionising radiation and disturbances of the circadian rhythm at long distance flights may be causes of an increased risk of breast cancer in cabin crew. Our findings in analyses of block hours and high altitude, long distance flights are consistent with these hypotheses, but the numbers were small and the results therefore uncertain. Analysis of employment years in SAS indicated that the breast cancer risk was similar regardless of the number of working years. However, some cabin crew may have been working as cabin attendants for other companies during their career. Our results, together with those of previous studies, call for further investigations of breast cancer risk in female cabin crew.