Background Commercial airline crew is one of the occupational groups with the highest exposures to ionising radiation. Crew members are also exposed to other physical risk factors and subject to potential disruption of circadian rhythms.
Methods This study analyses mortality in a pooled cohort of 93 771 crew members from 10 countries. The cohort was followed for a mean of 21.7 years (2.0 million person-years), during which 5508 deaths occurred.
Results The overall mortality was strongly reduced in male cockpit (SMR 0.56) and female cabin crews (SMR 0.73). The mortality from radiation-related cancers was also reduced in male cockpit crew (SMR 0.73), but not in female or male cabin crews (SMR 1.01 and 1.00, respectively). The mortality from female breast cancer (SMR 1.06), leukaemia and brain cancer was similar to that of the general population. The mortality from malignant melanoma was elevated, and significantly so in male cockpit crew (SMR 1.57). The mortality from cardiovascular diseases was strongly reduced (SMR 0.46). On the other hand, the mortality from aircraft accidents was exceedingly high (SMR 33.9), as was that from AIDS in male cabin crew (SMR 14.0).
Conclusions This large study with highly complete follow-up shows a reduced overall mortality in male cockpit and female cabin crews, an increased mortality of aircraft accidents and an increased mortality in malignant skin melanoma in cockpit crew. Further analysis after longer follow-up is recommended.
Statistics from Altmetric.com
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.
What this paper adds
Through their occupation, commercial airline crew are exposed to ionising radiation of cosmic origin as well as other risk factors. They are subject to disruption of circadian rhythms.
The overall mortality in male cockpit and female cabin crew is strongly reduced compared with the general population; the mortality from cancer is reduced with few exceptions, notably malignant skin melanoma in cockpit crew.
The mortality from female breast cancer, leukaemia and brain cancer is similar to that of the general population.
The mortality from cardiovascular diseases is considerably reduced; the one from aircraft accidents is remarkably high, as is that of AIDS in male cabin crew.
Through their occupation, commercial airline crew are exposed to ionising radiation of cosmic origin, as well as other risk factors, including electromagnetic fields from wiring and appliances, circadian rhythm disruption, ozone and noise.1–3 Airline crew is one of the occupational groups with the highest radiation exposures, typically an excess dose of approximately 2 milliSievert (mSv) annually, roughly the dose from natural background radiation.4–8 Cumulative occupational lifetime doses of ionising radiation generally remain below 100 mSv.9
The first reports on cancer incidence and mortality of civil and military aircraft crew were published in the early 1990s.10–16 Additionally, data were combined in pooled analyses of cancer incidence and mortality.17–21 A low overall and cardiovascular mortality of aircrew was consistently observed, indicating a strong effect of selection.22 Another finding was a twofold increased incidence of skin melanoma.17 ,18 There is limited evidence that ionising radiation is a risk factor for malignant melanoma, while the risk factor with the highest attributable fraction is ultraviolet (UV) radiation.23 ,24
In pooled analyses of male cockpit crew, decreasing mortality rates for all causes and all cancers with increasing estimated doses of ionising radiation were observed.21 An increased risk for cancers of the central nervous system (CNS) has been observed in some cohorts.25 A risk increase of about 20% was found for the incidence of prostate cancer, which was associated with the number of long-haul flight hours in pooled analyses of studies from the Nordic countries (Denmark, Finland, Iceland, Norway and Sweden).17 Shift work associated with a disruption of the circadian rhythm is one potential risk factor for breast and possibly prostate cancer.2
For male cabin crew, the overall and cancer mortality is similar to that of the general population, with few exceptions. In a recent pooled analysis of cancer incidence, significant excesses in melanoma and non-melanoma skin cancer, Kaposi sarcoma and alcohol-related cancers were observed, confirming previous results on mortality.18 ,20 Additionally, increased mortality from leukaemia has been reported.19 ,20
Findings from a pooled mortality analysis of 10 national cohorts of commercial airline crew are presented here. These include six cohorts with a prolonged observation period compared to previous publications and two new cohorts. This is the first publication where European and US data are combined. With an increase of 30% in the size of the study population and 69% in the number of person-years of observation compared to previous analyses, this study has a much greater statistical power and, thereby, precision for quantifying mortality risks. The cohorts are still young and it is important to track increases in cause-specific mortality as the cohort ages, especially if preventive measures may be deduced. The present publication reports standardised mortality ratio (SMR) analyses of the pooled cohorts separately for cockpit and cabin crew.
The study population consists of cohorts of commercial airline crew from Denmark, Finland, Germany, Greece, Iceland, Italy, Norway, Sweden, UK and the USA. Individual cohort studies with similar protocols were set up so that data could be delivered in a common format for pooled analyses, following a long collaboration between eight of the study groups.19 ,20 The UK and US cohorts were included ex novo, while the follow-up of cohort members from six out of eight other countries (with the exception of Greece and Italy) was extended for these new analyses. The national cohorts were compiled from varying sources in the participating countries (see online supplementary table S6). Procedures were described previously in more detail.18 ,26–39 Briefly, airline crews were identified from personnel files of generally one airline in Finland, Germany, Greece, Iceland, Italy, Sweden and the USA. These data were supplemented by data from airline pilot associations in Finland and Iceland. In the UK, Iceland and Norway, license data of the national aviation authority were used for cockpit cohort identification and enumeration. In Denmark, all commercial cockpit crew on file in the National Clinic of Aviation Medicine in Copenhagen were included. Cabin crew had to be employed or licensed for at least 6 months (1 year in the USA) to be included in the cohort, while no such limit applied to cockpit crew. For simplicity, the pooled cohort will be termed ‘multinational cohort’ hereafter.
The data collected included sex, year of birth and employment or license dates. Information on occupational history was recorded wherever possible. The start of employment was available for each cohort member, except in the UK, where information obtained by questionnaire was available for about 50% of the cohort. The date for end of service as a crew member for retired crew members was available in the initial follow-up of the original eight cohorts. The employment status at end of follow-up for the current study was known for study subjects from Germany, Greece, Italy and the USA.
Vital status follow-up
The individual period of observation started at the first date of employment (at the date of the first valid license during the period 1 January 1989–31 December 1999 in the case of the UK cohort), at immigration, or at the country-specific start of follow-up, whichever was latest. It ended at the date of death, emigration, loss to follow-up, or at the end of the study period (2004 or later, depending on the country, except 1996 for Italy and 1997 for Greece), whichever came first. Ascertainments of vital status and causes of death were carried out according to country-specific procedures.18 ,26–39
All causes of death were coded according to the International Classification of Diseases (ICD) revision in use at the date of death (ICD 7-10), and additionally recoded to the groups of causes used in previous analyses.19 ,20 As the study focused on potential effects of exposure to ionising radiation of cosmic origin, these groups include single cancer sites as well as three combined group of neoplasms. A first group, herein termed ‘radiation-related cancers’, consists of cancers of the oesophagus, stomach, large intestine, female breast, bladder/urinary tract, thyroid gland and leukaemia excluding chronic lymphocytic leukaemia (CLL). This original categorisation was based on Boice et al40 and the 2000 United Nations Scientific Committee on the Effects of Atomic Radiation (UNSCEAR) Report. 41 Lung cancer, although known to be associated with ionising radiation, is not included due to its strong link with smoking. A second group of ‘alcohol-related cancers’ was considered, including cancers of the oral cavity, pharynx, larynx, oesophagus and liver (as in Dreyer et al 42). Finally, cancers of the lung, oral cavity, pharynx, larynx, nasal cavities or sinuses, oesophagus, pancreas, urinary bladder, renal pelvis, stomach, liver, kidney, uterine cervix and myeloid leukaemia are subsumed as ‘smoking-related cancers’.43 A list of all causes of death and corresponding ICD codes can be found in online supplementary table S7.
Mortality rates for the general population
Sex, age-group, and calendar-period-specific mortality rates for the general population of each participating country were obtained from WHO Mortality Database, or from the national statistical offices (Finland, Norway, the UK and the USA).
For the analyses, data were transferred to the Institute of Medical Biostatistics, Epidemiology and Informatics (IMBEI) at the University Medical Center, Mainz, Germany. A complete analysis plan was laid down in writing and approved by the study group prior to the statistical analysis. Persons with missing or implausible data were excluded from the dataset. Similarly, records relating to 752 female cockpit crew members were not further analysed, as the size of this subcohort was very small and only eight deaths were observed.
Person-years at risk were dynamically allocated to 5-year age and calendar-period intervals and, additionally, to categories of employment/licensing duration (>0–9.9, 10–19.9, 20–29.9, 30+ years). A latency period for cancer development after the beginning of radiation exposure (ie, start of employment) of 2 years was assumed for leukaemia, and of 10 years for solid tumours.44
SMRs and associated 95% CI were computed using the mortality rates described above. To account for missing causes of death, the method proposed by Rittgen and Becker was used.45 Following this method, the ‘corrected’ number of deaths, OC, corresponds to the ratio O/p, where O is the number of observed deaths from a known cause and p is the proportion of known causes among all causes. Similarly, a corrected SMR is computed as SMRC=SMR/p=Oc/E. Corresponding corrected confidence limits are computed by taking the variance of p into account. SMRs are reported only for single causes of death or groups with three or more observed cases. The heterogeneity between countries was analysed by χ2 tests.46
To investigate a potential effect of employment duration, SMRs stratified by employment duration were computed for single causes of death or groups with 100 or more observed cases in any of the three subcohorts of male cockpit crew and male or female cabin crew. As there is a high collinearity of employment duration with age and work-related exposures, this must be interpreted with caution and not simply as dose-response analysis. No trend analyses of mortality with increasing employment duration were done.
All analyses were performed using SAS V.9.3.47
The cohort consisted of 93 918 persons, 147 of whom were excluded from the analyses for one of the following reasons: missing birth date (1), missing date of vital status (1), death or emigration before the beginning of the study period (23), date of death before the beginning of employment (18), deceased more than a month before the end of recorded employment (16), beginning of employment at age 65 years or later (5), and end of employment at age 66 years or later (83).
The final analysis dataset included 36 816 male cockpit crew, 12 288 male, and 44 667 female cabin crew (table 1). The cohort was quite young: the mean age at the end of follow-up was 53.3 for male cockpit crew, 47.9 for male and 47.9 for female cabin crew. The mean (and median) duration of follow-up were 20.3 (17.3), 20.5 (19.7) and 23.1 (22.9) years, respectively. At the end of follow-up, most cohort members were alive, 80 290 (85.6%), 5508 (5.9%) were deceased, and 6390 (6.8%) had emigrated (table 2). Few were lost to follow-up (1.7%). Person-years tabulated by age and duration of employment are shown in online supplementary table S8.
In the male cockpit crew subcohort, most person-years of observation are contributed by individuals from the Nordic countries (37%), the UK (35%) and Germany (19%). In the male and female cabin crew subcohorts, most person-years are contributed by persons from the Nordic countries (31% and 36%), Germany (29% and 29%), and the USA (16% and 27%).
Male cockpit crew
Among male cockpit crew, a total of 2703 deaths was observed, including 135 (5.0%) from unknown causes (table 3).
The mortality from all causes (SMR=0.56, 95% CI 0.54 to 0.58), and all cancers (SMR=0.69, 95% CI 0.64 to 0.76) were significantly reduced compared to the general population. The mortality from ‘radiation-related cancers’ was also reduced (SMR=0.73, 95% CI 0.62 to 0.85); that due to non-CLL leukaemia was SMR=0.92 (95% CI 0.59 to 1.39).
The mortality from smoking-related cancers was low: it was significantly reduced for cancers of the buccal cavity and pharynx (SMR=0.27, 95% CI 0.16 to 0.44), as well as for lung cancer (SMR=0.43, 95% CI 0.36 to 0.52).
The mortality from malignant melanoma was significantly increased (SMR=1.57, 95% CI 1.06 to 2.25). Mortality from prostate cancer was elevated (SMR=1.23, 95% CI 0.98 to 1.53), albeit not significantly. The mortality rates for cancers of the brain and CNS (SMR=1.14, 95% CI 0.83 to 1.54), and thyroid/other endocrine glands (SMR=1.06, 95% CI 0.28 to 2.72) did not differ from those of the general population.
The mortality from most non-cancer causes was markedly lower than in the reference population; the observed number of cardiovascular deaths, in particular, was less than half the expected (SMR=0.42; 95% CI 0.37 to 0.46). Aircraft accidents accounted for 8.4% of all recorded deaths among cockpit crew (SMR=45.74; 95% CI 38.83 to 53.62), while mortality from other external causes, including accidents, was significantly decreased (SMR=0.62; 95% CI 0.54 to 0.72). In contrast with cabin crew (see below), the mortality due to suicide was significantly reduced in cockpit crew, with an SMR of 0.64 (95% CI 0.51 to 0.80).
In the analyses of mortality by employment duration, cockpit crew who had worked less than 10 years had the highest SMR for aircraft accidents, while the SMRs for other causes did not show strong variations according to employment duration (see online supplementary table S9).
Male cabin crew
In male cabin crew, a total of 1360 deaths were observed, including 85 (6.3%) from unknown causes (table 4).
In contrast with cockpit crew, mortality rates from all causes (SMR=1.07, 95% CI 1.02 to 1.13) were significantly higher than those of the reference population. Many (n=216, 15.9%) died of AIDS. The mortality from all cancers (SMR=0.97, 95% CI 0.84 to 1.13) and ‘radiation-related cancers’ (SMR=1.00, 95% CI 0.74 to 1.33) did not differ from those of the reference population. The mortality from malignant melanoma was non-significantly elevated (SMR=1.20, 95% CI 0.45 to 2.59), while the mortality from prostate cancer was below its expectation (SMR=0.75, 95% CI 0.40 to 1.29).
As in cockpit crew, the mortality from cancers of the brain and CNS (SMR=1.25, 95% CI 0.64 to 2.18), and that due to non-CLL leukaemia (SMR=1.33, 95% CI 0.63 to 2.47), did not differ significantly from that of the general population.
On the contrary, the mortality from AIDS (SMR=16.47, 95% CI 13.86 to 19.51), infectious diseases except AIDS (SMR=3.69, 95% CI 2.55 to 5.18), non-melanoma skin cancer (SMR=8.01, 95% CI 2.98 to 17.33), and non-Hodgkin lymphoma (SMR=2.37, 95% CI 1.41 to 3.73) was significantly elevated.
The mortality from ‘alcohol-related’ cancers (47 cases in total) was elevated, but not significantly different from that of the general population.
The SMRs from most non-cancer causes, other than those mentioned above and aircraft accidents, were below 1.
Female cabin crew
In female cabin crew, 1445 deaths were observed, including 91 (6.3%) from unknown causes (table 5).
The mortality from all causes (SMR=0.73, 95% CI 0.69 to 0.77) and all cancers (SMR=0.85, 95% CI 0.77 to 0.95) was significantly reduced. However, as in male cabin crew, the mortality from ‘radiation-related-cancer’ was close to its expectation (SMR=1.01, 95% CI 0.88 to 1.15). In contrast to male cabin crew, the mortality from alcohol-related cancers was reduced (SMR=0.68, 95% CI 0.46 to 0.98).
The mortality from breast cancer (SMR=1.06, 95% CI 0.89 to 1.27), malignant melanoma (SMR=1.17, 95% CI 0.64 to 1.97), and cancer of the thyroid/other endocrine glands (n=6, SMR=1.53, 95% CI 0.47 to 3.69) did not differ from that of the reference population.
The SMRs for most non-cancer causes, except aircraft accidents, were below 1. The SMR for suicide was 1.23 (95% CI 0.96 to 1.55).
No difference in mortality rates between strata with different duration of employment was noted (see online supplementary table S10).
This study of nearly 94 000 commercial airline crew members followed for an average of 21.7 years, including cohorts from nine European countries and the USA, is the largest study of this occupational group to date. Compared to previous pooled analyses of European aircrew and the analysis of the US cohort taken together, 28.8% more cases (20.5%, 27.9% and 48.8% more cases among male cockpit and male and female cabin crews, respectively), corresponding to 68.8% more person-years, were included.19 ,20 ,38
In male cockpit and female cabin crews, a markedly reduced overall cardiovascular mortality and decreased cancer mortality were observed, consistent with other aircrew cohorts not included here.10–13 ,48 ,49 Similar patterns are often observed in occupational cohorts. The overall mortality does not appear to increase with increasing duration of employment in this cohort. The high social status and strict health requirements may concur to explain the low overall mortality among cockpit crew. The low mortality from smoking-related cancers, especially in cockpit crew, indicates that smoking was probably less common in aircrew members than in the general population.
When comparing results of the current study with those of previous pooled analyses, including the analyses of cancer incidence in Nordic countries, it is important to keep in mind that these cohorts are part of the current analyses.17 ,18 Consistent results are hence to be expected.
The mortality experience of the US and European cohorts is very similar in general, with significant differences in SMR for all causes in female cabin crew (US SMR: 0.59, current pooled SMR: 0.73), all cancer in female cabin crew (US SMR: 0.85, current pooled SMR: 0.71).38
Compared to previous analyses of European crew only, the all-cause mortality of male pilots is significantly lower than in previous analyses (previous SMR: 0.64, current pooled SMR: 0.54), and much lower for cancer of the buccal cavity and pharynx, lungs, cardiovascular diseases and liver cirrhosis.19 The all-cause mortality of female cabin crew is lower than previously observed (previous SMR: 0.80, current pooled SMR: 0.73).20
Heterogeneity is most present in those causes of death for which confounding factors may play a major role, such as lung cancer, buccal cancer, liver cirrhosis, or causes for which national mortality rates differ substantially, such as suicide rates. No obvious pattern of higher or lower SMRs per cause emerges, as was exemplified by Zeeb et al.20
Mortality from non-cancer causes
The low mortality from cardiovascular causes and most other non-cancer causes in respect to the general population persists in the current prolonged follow-up.
Mortality from aircraft accidents is elevated in all subcohorts, and particularly high in cockpit crew where it accounts for 44% of all deaths from external causes. Young pilots seem to have the highest risk. While occupational and private accidents could not be differentiated here, a study of Norwegian pilots suggested that a large proportion of these deaths are related to leisure-time flying.50
An elevated mortality of male cabin crew from AIDS and related causes was already observed in previous analyses.
Self-perceived mental health in female aircrew was investigated earlier, and the authors found that several work-related factors, possibly implying anxiety or depressed mood, were of concern.51–54 In the current analyses, suicide rates in cabin crew were higher than in the general population, but significantly lower in cockpit crew.
An elevated incidence of, and mortality from, malignant melanoma has been reported by previous studies, including studies not included in the current analyses.12 ,15 ,55 An excess mortality was observed here for cockpit crew. The major risk factors for malignant melanoma of the skin include host factors related to skin colour and nevi, sunburn at an early age, and intermittent sun exposure, with exposure to solar UV radiation representing an overwhelming etiologic fraction in the general population.24 ,56 As aircraft windows provide effective shielding against such radiation, occupational exposures on board are unlikely.57 UV exposures among airlines crew may be recreational or occur during stopovers. Among Nordic cabin crew, a non-significant elevated incidence of skin melanoma of the trunk was observed 20 years after first employment, and a positive dose-response relationship was observed between estimated radiation exposure and melanoma incidence in Nordic pilots.17 ,18 The authors concluded that the excess incidence may be attributed to UV radiation.17 In a recent study comparing UK flight crew and air traffic controllers, no differences in skin melanoma rates were observed. The strongest risk predictors of skin melanoma in both occupations were light skin and sunbathing.58 The similar site-specific cancer risks between the two occupational groups suggest that risks are not directly related to occupational exposures.
Ionising radiation is a well-established risk factor for non-CLL leukaemia.59 Assuming a strength of association similar to that reported elsewhere, the expected SMR would not exceed 1.08.18 ,59 ,60 The SMRs for cockpit crew, male and female cabin crews, were inconclusive due to a relatively small number of cases.
Cancers of the brain and CNS
In a recent analysis of the German aircrew cohort (that contributes 14 of the 51 CNS cancer deaths in cockpit crew in the present study), a statistically significant increased mortality from CNS cancers was observed (SMR=2.13), while no increase in incidence was found in Nordic pilots or in male or female cabin crew of Nordic countries.17 ,18 ,25 In the current analyses, the SMR for brain cancers was not significantly different from the general population in any of the subcohorts.
Breast cancer is one of the cancers most consistently increased in female cabin crew when incidence is considered. No elevation was found in the present study with focus on mortality. Note that for breast cancer, the mortality estimate is a weak measure and affected by the good survival.20 ,61 ,62
A statistically significantly elevated incidence of breast cancer was observed in Nordic crew (standardised incidence ratio (SIR)=1.50), with an increased SIR at 20 years since first employment compared to shorter latency periods.18 However, in a nested case-control analysis within that cohort with adjustment for parity and age at first childbirth, an OR of 0.98 per 10 mSv cumulative effective dose was observed. No known risk factors seemed to explain the excess in breast cancer, though residual confounding remains an issue. The authors also considered screening bias unlikely.
Health effects of circadian rhythm disruptions are currently investigated by several research groups, many of which focus on breast cancer as a health outcome.63 ,64 In a recent monograph, the International Agency for Research on Cancer (IARC) classified shift-work involving circadian rhythm disruption as probably carcinogenic to humans (Group 2A).2 Human data supporting this classification are strongest for breast cancer.54 ,63–65 While there are plausible biological mechanisms for cancer development, notably the involvement of the melatonin pathway, the issue is still unresolved.
An increased mortality from cancers of the prostate was observed in cockpit crew (SMR=1.23), but not in cabin crew. In previous pooled mortality analyses, no increases were observed.19 In the Nordic studies, prostate cancer incidence was non-significantly raised in cockpit crew (SIR=1.21), with a significant trend of increasing risk with increasing cumulative effective dose in pilots below age 60 years (but not beyond this age).18 ,29
This study reports on mortality, an outcome affected by disease incidence, screening activities and treatment. For cancer, studies of disease incidence provide a better picture of risks, but such data were available only in Nordic countries.
Details of flight history (flight hours per aircraft type per year) were available for certain subcohorts only, and usually only for cockpit crew. These data were therefore not used here.
The list of cancers considered being radiation-related, and the lack of distinction between cancers with much, fair, little or no evidence for causation may be criticised. Several additional cancers were included in a list developed as part of the IARC re-evaluation of ionising radiation as a group 1 carcinogen: cancers of the salivary glands, bone, skin (basal cell carcinoma), brain and CNS, thyroid and kidney.66 As a sensitivity analysis, SMRs were computed for this extended definition. All SMRs were qualitatively unchanged.
Aircrew are exposed to a number of potential risk factors for cancer, such as ionising radiation, circadian rhythm disruptions (jet lag), pesticides, jet fuel, electromagnetic fields, and other exposures, such as leisure-time UV radiation, nutritional factors and stress. Such factors were not assessed in the current study. A good correlation between employment duration and occupational exposure to ionising radiation of aircrew was shown in Germany, although employment duration and ionising radiation exposure have not been as well correlated in US pilots.8 ,9
Finally, despite the large number of subjects, the cohort is still relatively young; only 5.9% of cohort members died during the observation period, resulting in small numbers of deaths and large risk uncertainty for many cancers. The overall mortality in male cockpit and female cabin crew is strongly reduced; the mortality from cancer was reduced with few exceptions, notably malignant skin melanoma in cockpit crew. Further analysis after a longer follow-up is recommended.
The authors thank all their collaborating partners at the respective airlines and national aviation administrations for their efforts in maintaining and updating this cohort. The UK cohort acknowledges the support of the Civil Aviation Authority (CAA) and, in particular, of Dr Saly A Evans. The authors thank Dr Ulf Tveten, who participated in previous studies of Norwegian and European crew, for his comments on the current manuscript.
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Files in this Data Supplement:
- Data supplement 1 - Online supplement
Contributors All analytical steps were laid out in the analysis plan written and approved by the study group prior to analysis. All coauthors contributed data and advice on specific caveats regarding the use of their data. GPH performed the statistical analyses and drafted the manuscript with MB and HZ. All coauthors critically reviewed the manuscript.
Funding This work was supported by the German Federal Office for Radiation Protection (grant number StSch 4558) for study coordination.
Disclaimer The findings and conclusions in this report are those of the authors and do not necessarily represent the views of the National Institute for Occupational Safety and Health.
Competing interests None.
Provenance and peer review Not commissioned; externally peer reviewed.
Ethics approval Ethical approval for national cohorts and the transfer of anonymised data were obtained by the single study groups. Ethical approval for the pooled mortality analysis was obtained locally from the Ethics Committee of the Medical Association of Rheinland-Pfalz.