Cancer incidence among Nordic airline cabin crew

Authors


Abstract

Airline cabin crew are occupationally exposed to cosmic radiation and jet lag with potential disruption of circadian rhythms. This study assesses the influence of work-related factors in cancer incidence of cabin crew members. A cohort of 8,507 female and 1,559 male airline cabin attendants from Finland, Iceland, Norway and Sweden was followed for cancer incidence for a mean follow-up time of 23.6 years through the national cancer registries. Standardized incidence ratios (SIRs) were defined as ratios of observed and expected numbers of cases. A case-control study nested in the cohort (excluding Norway) was conducted to assess the relation between the estimated cumulative cosmic radiation dose and cumulative number of flights crossing six time zones (indicator of circadian disruption) and cancer risk. Analysis of breast cancer was adjusted for parity and age at first live birth. Among female cabin crew, a significantly increased incidence was observed for breast cancer [SIR 1.50, 95% confidence interval (95% CI) 1.32–1.69], leukemia (1.89, 95% CI 1.03–3.17) and skin melanoma (1.85, 95% CI 1.41–2.38). Among men, significant excesses in skin melanoma (3.00, 95% CI 1.78–4.74), nonmelanoma skin cancer (2.47, 95% CI 1.18–4.53), Kaposi sarcoma (86.0, 95% CI 41.2–158) and alcohol-related cancers (combined SIR 3.12, 95% CI 1.95–4.72) were found. This large study with complete follow-up and comprehensive cancer incidence data shows an increased incidence of several cancers, but according to the case-control analysis, excesses appear not to be related to the cosmic radiation or circadian disruptions from crossing multiple time zones.

Airline cabin crew are occupationally exposed to ionizing radiation with doses 2–6 mSv per year.1 This is roughly twice the average annual dose from natural and medical sources received by the general population. Cosmic radiation in the common cruising altitudes (8,000–10,000 m) consists mainly of gamma and neutron radiation, with some heavy nuclei. In 1990, the International Commission on Radiological Protection recommended that in-flight radiation exposure to jet aircrew should be regarded as occupational exposure.2

Of the radiation-related cancers, only breast cancer has shown increased incidence rates among airline personnel consistently in several studies. Out of the seven cohort studies of cabin crew,3–9 all but one6 indicate an increased incidence of breast cancer. However, the excess risks seem to be higher than can be explained by the low radiation doses received,10 and several other factors may contribute to the observed excess.

Cabin crew also work in shifts including work at night and are exposed to jet lags (a temporary condition after air travel across several time zones) dependent on time, distance and direction (east-west vs. north-south) of flight routes. Such exposures may contribute to circadian disruption, including suppression of the chronobiotic neurohormone melatonin, which has anticancer properties.11–13 There is accumulated epidemiologic and biologic evidence that circadian disruption, which is characterized by desynchronization of the internal clock with the external environmental light, may contribute to the development of certain cancers, in particular breast and prostate cancer.12–14 Night shift work that involves circadian disruption was recently classified by the International Agency for Research on Cancer as probably carcinogenic to humans (category 2A), based on sufficient evidence in experimental animals and limited evidence for human breast cancer.15

It is important to determine if the risk of cancer among flight personnel is elevated due to ionizing radiation or other work-related factors and whether current occupational standards provide sufficient protection. The aim of our study was to describe the cancer incidence among airline cabin crew from four Nordic countries. In the cabin crew studies published so far, there has been a rather limited possibility for internal comparisons to characterize possible dose-response patterns related to cosmic ionizing radiation. We also evaluated the dose-dependence of cancer incidence in terms of time since first employment, cosmic radiation and number of flights crossing six time zones as a surrogate for jet lag.

Abbreviations

BCC: basal cell carcinoma; CI: confidence interval; CLL: chronic lymphatic leukemia; OR: odds ratio; SAS: Scandinavian Airlines; SIR: standardized incidence ratio; SMR: standardized mortality ratio; UVR: ultraviolet radiation

Material and Methods

Study population

National cohorts of airline cabin crew were identified from various registers in Finland, Iceland, Norway and Sweden. In Finland, all cabin crew personnel who had ever been working for Finnair and its daughter airline flight companies between 1947 and March 1993 were identified from the files of the Finnair company.3 Persons deceased before 1967 were excluded. The final cohort included 1,766 persons (1,578 women and 188 men).

The Icelandic cohort comprised 1,690 cabin crew members (1,532 females and 158 males) identified from the computerized members list from the Icelandic Cabin Crew Association and from airline companies Icelandair and Air Atlanta for the period 1947–1997.7

In Norway, the cohort was established from the files at the Personnel Licensing Section of the Civil Aviation Administration, which authorizes cabin crew members.6 The cohort included all cabin crew members who had a valid license between January 1950 and February 1994 and that were resident in Norway for some time after 1961 (and hence got the Norwegian personal identity code (PIC) used in computerized record linkages). The final number of persons included in this analysis was 3,073 women and 581 men.

The Swedish study population consisted of cabin crew employed by the Swedish part of Scandinavian Airlines (SAS) at any time during 1957–1994 and resident in Sweden.9 The cabin crew at SAS was identified using administrative company registers of employees and archival records at the Swedish part of SAS. The final number of persons included in this analysis was 2,324 women and 632 men.

The cohorts were linked to the national population registers by PIC, and the possible dates of immigration, emigration or death were obtained for every cohort member. Since the 1960s, all residents of the Nordic countries have had a unique PIC that is used in all major registers and allows automatic, accurate record linkages.

The dates of birth for live-born children among the women were obtained by linkage to the Population Register in Finland, register of the Genetical Committee of the University of Iceland and Statistics Norway. For female cabin attendants in Sweden the dates of birth of their children were obtained using the Multi-Generation Register at Statistics Sweden and the National Medical Birth Register.

Follow-up for incident cancer cases was conducted through record linkage with the national countrywide cancer registries existing in all Nordic countries.16 All these registries cover entire national populations in a nonselective way and have similar coding principles that allow, e.g., classification by subsite or morphological type of the cancer. This possibility was used in our study for skin cancers and leukemia. Basal cell carcinoma (BCC) of the skin was only registered in the Finnish and Icelandic Cancer Registries. It was analyzed as a separate category but was not included in the overall cancer rates.

Follow-up for cancer for each individual started at the date of first employment, at immigration or on the date of the beginning of cancer registration, availability of the linkage key (the PIC in Norway) or availability of computerized information of causes of death (Sweden), whichever was latest. The follow-up ended at emigration, at death or on a common closing date (the date until which cancer registration was complete at the time of the record linkage), whichever was first. For those emigrating out of the country, the observation period was terminated at the time of first emigration, irrespective of eventual later remigration (to avoid selective follow-up). In Finland, the maximal follow-up period was thus from 1953 to 2005, in Iceland from 1955 to 1997, in Norway from 1962 to 2002 and in Sweden from 1961 to 2003. During these periods a National Cause of Death Register as well as a National Cancer Register was in operation in each respective country.

The observed numbers of cases and person-years at risk were counted by gender in 5-year age groups and 5-year calendar periods. The expected numbers of cases for all cancer sites combined and for specific cancer types were calculated by multiplying the number of person–years in each stratum by the corresponding national cancer incidence rate. The specific cancer types selected a priori for the analysis included the cancer sites related to ionizing radiation or circadian disruption, cancers with a suggestion of exceptional risk levels in earlier studies and other common cancer types to understand the overall cancer profile of airline cabin crew.

To calculate the standardized incidence ratios (SIRs), the observed numbers of cases were divided by the respective expected numbers. The exact 95% confidence interval (CI) for each SIR was defined based on the assumption that the number of observed cases followed a Poisson distribution.

Case-control analysis

Case-control data for women were constructed to estimate the effect of cosmic ionizing radiation on the risk of breast cancer, skin cancers and leukemia [excluding chronic lymphatic leukemia (CLL)] and the effect of jet lags (breast cancer only). All female cabin crew members without cancer diagnosis of the same site as the case at the time of the diagnosis, born in the same year as the case and alive at the date of cancer diagnosis of the case, were used as controls. Conditional logistic regression, using control selection mentioned above, was used to assess possible relations between the factors and to estimate the statistical significance of the trends. In the models related to breast cancer, the available parity information was added to the model either as dichotomous (any vs. no children) variable or categorical factor combined from the number of children (0, 1–2 and 3+) and age at first birth (<25 and 25+). In the analysis of breast cancer, the proportion of nulliparous women was 18%, women with one or two children was 59%, women with three or more children was 23% and women given first birth at age 25 or older was 65% among cases. Corresponding figures for the controls were 22, 57, 21 and 66%.

Assessment of cosmic radiation exposure

The estimates of the annual doses per cabin crew member were based on the assumption that the crew members flew a random selection of all types of routes operated by the airline company in each year of employment. Based on the information from airline companies, we know that cabin crew flies approximately a variety of routes at any time during their work unlike pilots who have a license for a specific aircraft type and thus they have a limited range of routes at a given time. For the SAS cabin crew members in Norway exact flight histories are available (however not in a format that could be used for exposure estimation for the entire cohort). Based on a sample of these records, the Norwegian part of the study group concluded that the assumption that each cabin crew member flew proportionally their share of the routes operated per year and company was not appropriate for the Norwegian cohort. Therefore, Norway was excluded from the dose-response analysis part of our study. Information on the frequency of flights and aircraft types used on each route at 5-year intervals (from 1960 to 1995) was collected from the SAS, Finnair and Icelandair timetables.

No information was available on charter flights for SAS and Icelandair. For Finnair, the charter block hours were obtained from the Finnair archives. With the help of former Finnair pilots and the route map for the years 1984–1995, the charter hours were assigned to typical charter destinations situated in the Mediterranean area and the rest of the Europe.17 The number of cabin crew on board was estimated based on expert consultations and literature.18 The number depends mainly on the aircraft type, but also on route and time period.

The radiation dose for each flight was constructed by combining information on flight profile (time of ascent, time at cruising altitude and time of descent) and the so-called block hours (time from pulling out from the gate at departure airport until docking at arrival airport), with the dose rate based on altitude, and with some variation by calendar period reflecting heliocentric potential.17 The cosmic radiation dose for every route was calculated using the European Program Package for the Calculation of Aviation Route Doses (EPCARD) software developed for this purpose by the GSF Institute of Radiation Protection. EPCARD is based on the results of Monte Carlo radiation transport calculations, which take into account all the physical processes and affects using the most recent nuclear reaction cross-section data and the cosmic ray data of NASA.2

The collective dose from a single flight was estimated from the dose on that route based on the number of cabin crew on board during the flight. The collective dose was then multiplied by the frequency of the flights on the same route within that particular year to obtain the cumulative annual collective dose on that route. Then, the cumulative annual collective doses from all routes during that year were summed up to obtain the overall collective dose received by all cabin crew during 1 year. It was divided by the number of cabin crew during that year to estimate the average annual dose for the cabin crew, and it was assigned to each cabin crew member employed for that year.

The cumulative dose was calculated as the sum of the average annual cosmic radiation doses summed over the active work-years. For this purpose, the information on the starting and the ending year of the employment for each cabin crew member was obtained from the airline companies. Available information on maternity leaves and other work breaks was also used. Doses were estimated separately for Finnair, Icelandair and SAS Sweden (Table 1). Some of the Swedish cohort members had also worked for other airline companies than SAS. Because we did not have dose estimates for the other companies, SAS doses were applied to the years worked for other companies as well. Since the annual doses were only estimated for every fifth year, in cumulative dose calculation they were assumed constant for adjacent years as well. For example, the dose for the year 1975 was used for the period 1973–1977.

Table 1. Annual average number of flights passing six or more time zones and annual average cosmic radiation doses (mSv), for every fifth year
inline image

Circadian rhythm disruption

Circadian disruption was estimated by the average annual number of flights passing six or more time zones (Table 1). Since this information was not available on an individual level, information on flight duration and frequency was obtained from historical airline timetables for every fifth year of Finnair, Icelandair and SAS Sweden, and a similar route distribution was assumed for all cabin crew members of the airline company working at a same time period. Each one-way flight passing six time zones is counted as one “jet lag.” Alternative variables for jet lag exposure, based on thresholds of passing four and five time zones, were defined in a similar way.

Results

Cohort analyses

The combined cohort comprised 8,507 women and 1,559 men. The average length of follow-up was 23.6 years. Almost 75,000 person-years were in the follow-up category of ≥20 years since the time of first employment (Table 2). The cohort was rather young; only 10% of the person-years were above 55 years of age. At the end of follow-up, 6% of the cabin crew members had reached an estimated cumulative dose of at least 35 mSv and 40% at least 150 flights over six or more time zones (Table 2). Reproductive history was incomplete for the Finnish and Norwegian women born before 1935, altogether 3.4% of the female cohort members (Table 2).

Table 2. Numbers and percentages (%) of airline cabin crew, by study variables
inline image

During the follow-up, 577 cases of cancer were diagnosed in women; the expected number was 499.2 corresponding to an SIR of 1.16 with 95% CI of 1.06–1.25 (Table 3). Female cabin crew had a statistically significantly increased SIR for breast cancer (SIR 1.50, 95% CI 1.32–1.69), skin melanoma (1.85, 95% CI 1.41–2.38) and leukemia (1.89, 95% CI 1.03–3.17). The SIR for BCC (only registered in Finland and Iceland) was also increased among female cabin personnel (SIR 2.39, 95% CI 1.80–3.10). The SIR for breast cancer did not vary significantly between the decades of follow-up.

Table 3. Observed and expected numbers of cancer cases and standardized incidence ratios (SIR) with 95% confidence intervals (CI) among female airline cabin crew in Finland, Iceland, Norway and Sweden, by cancer site
inline image

Among men, 152 cancers were observed versus 109.7 expected (SIR 1.39, 95% CI 1.17–1.62) (Table 4). A high relative excess risk was observed for Kaposi sarcoma (SIR 86.0, 95% CI 41.2–158). The SIRs were also significantly elevated for laryngeal cancer (4.72, 95% CI 1.72–10.3), pharyngeal cancer (3.12, 95% CI 1.34–6.15), skin melanoma (3.00, 95% CI 1.78–4.74) and nonmelanoma skin cancer (2.47, 95% CI 1.18–4.53).

Table 4. Observed and expected numbers of cancer cases and standardized incidence ratios (SIR) with 95% confidence intervals (CI) among male airline cabin crew in Finland, Iceland, Norway and Sweden, by cancer site
inline image

The incidence of alcohol-related cancers (oral cavity, pharynx, esophagus, liver and larynx; as defined by Dreyer et al. 1997) combined among male cabin crew members was increased by three-fold (SIR 3.12, 95% CI 1.95–4.72). The SIR increased with age; it was 1.6 (95% CI 0.0–8.7) in ages <45, 2.8 (95% CI 1.5–5.0) in ages 45–64 and 4.1 (95% CI 1.9–7.8) in ages ≥ 65 years. The SIR of the alcohol-related cancers for women was 0.67 (95% CI 0.25–1.45).

Altogether, the SIR did not vary substantially with increasing time since first employment (Table 5). There was some (nonsignificant) tendency of a higher SIR in the follow-up category ≥20 years in breast cancer and skin melanoma of the trunk. This was also the case for alcohol-related cancers as defined above (data not shown).

Table 5. Observed (Obs) and expected (Exp) numbers of cancer cases and standardized incidence ratios (SIR) with 95% confidence intervals (CI) among airline cabin crew in Finland, Iceland, Norway and Sweden, by cancer site and time since first employment
inline image

Case-control analyses

In the conditional logistic regression case-control analyses, exposure to cosmic radiation did not have a significant dose-response association for any cancer under study (Table 6). For leukemia, excluding CLL, the odds ratio (OR) per 10 mSv increase in dose with a lag of 10 years was 1.66 (95% CI 0.77–3.55). Rather high point estimates of OR were observed in categorical analysis, but the ORs were nonsignificant because of the very small numbers of cases (Table 6).

Table 6. Odds ratios (OR) among female airline cabin crew members in Finland, Iceland and Sweden, derived from case-control analyses by conditional logistic regression model, with 95% confidence intervals (CI), for continuous and categorical estimated cumulative dose
inline image

The OR for breast cancer calculated per estimated 100 flights passing ≥6 time zones was 0.92 (95% CI 0.77–1.11), adjusted for parity (parous vs. nulliparous) when no lag time was used in the model. This result was unaffected by allowing a lag time of 10 years or by adjustment with age at first live birth (results not shown). Change of the criterion of jet lag from passing ≥6 time zones to ≥5 or ≥4 time zones did not markedly change the results.

The correlation of the estimated dose and number of flights passing ≥6 time zones was so high (0.88 with a lag time of 10 years) that mutually adjusted results were unstable.

Discussion

This Nordic study confirmed the findings of earlier reports concerning the elevated risk of skin cancers and breast cancer among airline cabin personnel. The narrow confidence intervals of the joint estimate, based on the large combined cohort, and consistent results in each of the four independent cohorts indicate that these findings cannot be attributed to chance.

Estimation of cosmic radiation

Yearly average dose estimates were constructed for the study, but flight histories for each individual cabin crew member have not been documented equally precisely as for airline pilots,19 and such information cannot be obtained in a comprehensive and objective fashion from cabin crew members themselves.20 Our exposure estimates were based on the assumption that most of the cabin crew members flew a random allocation of all types of routes operated by the airline company each year. The available flight history information indicated that this assumption was not appropriate for SAS cabin crew in Norway, and, therefore, Norway was excluded from the dose-response analyses of our study. Besides three small case-control studies on breast cancer incidence among Finnish, Swedish and Icelandic cabin attendants,9, 20, 21 this is the first cabin crew study of cancer incidence with an attempt to quantify the dose-response pattern. We acknowledge that exposure estimation method used in our case-control analyses may lead to exposure misclassification similar to studies based on job exposure matrices and may dilute risk estimates toward unity.

Estimation of circadian disruptions

Simulated chronic jet lag in mice has been shown to disrupt circadian rhythms and significantly accelerate tumor growth.22 Our way of estimating the number of flights passing six time zones is a strong simplification of the complex aspects related to circadian disruption, but this was the most detailed assessment feasible without individual flight histories and improvement compared to previous cancer incidence studies. Further, our methods might underestimate the number of flights crossing six time zones since flights containing stopovers are dealt as separate flight segments not as a single flight. However, this problem could not be defeated since only available information is from flight timetables where all flight segments are recorded separately. There are no applicable standard methods for quantification of the impact of circadian disruption by crossing multiple time zones or in the night work shift, which would be essential in the evaluation of possible carcinogenicity. Most studies on endocrine phase alterations and phase adaptation after transmeridian flights have dealt with changes over 6–10 time zones.23 Sleep disturbances are more frequent after long-haul flights than after short-haul ones,24 and lead to marked changes in several aspects of the immune system and in biological processes related to the risk of breast cancer.25 Further, disruption of menstrual cycle due to jet lag showed some association with breast cancer risk in a case-control analysis of Finnish cabin crew.20 Other previous studies showed that changes in molecular signaling pathways were already detected after a single night of partial sleep deprivation,26 and hormonal changes become manifest after a single or several nights of partial sleep deprivation.27, 28 Lack of data on work at night is a limitation of our study.

Another problem in the evaluation of the impact of shift work and transmeridian flights is circadian-infradian interactions. In experimental studies on different animal species, not all shift schedules led to harmful health effects.29 SAS in Sweden compared the impact of transmeridian flights in crews who had a rapid turnaround with crews who had a prolonged stay over and found significant differences in sleep disturbances.30

Empirically, a 6-hr time difference may be a logical limit for assuming a significant disruption of the circadian system with subjective symptomatology in the majority of subjects. If sleep deprivation plays a role as an associated outcome, as anticipated, a shorter flight associated with a 4-hr sleep deprivation may already be biologically important. In our study, change of the criterion of jet lag from ≥6 time zones to ≥5 or ≥4 did not markedly change the results.

Breast cancer

The largest number of excess cases was in breast cancer. There were 87 cases more than the 176 cases that would have been expected based on the average national cancer incidence rates. Of the main risk factors of breast cancer, we were able to adjust for age at first birth and number of children. Long-term hormonal therapy and obesity are risk factors for breast cancer among postmenopausal women, but as the majority of the crew members were younger than the normal perimenopausal age (50–55 years) these are not likely to be major confounders. Furthermore, the physical activity at work for cabin crew might be higher than in most other occupations, which should decrease the breast cancer risk.16

Breast cancer is one of the alcohol-related cancers, with a measurable risk increase starting from one daily drink.31 In a pooled analysis,32 the multivariate relative risk for a 10 g per day (one unit) in alcohol was 1.09 (95% CI 1.04–1.13). In our study of female cabin crew members, we observed a tendency of decreased relative risk of strongly alcohol-related cancers of the oral cavity, pharynx, esophagus, liver and larynx.33 This suggests that alcohol is not a positive confounder for breast cancer. Thus, none of the known risk factors seems to explain the excess risk of breast cancer.

Leukemia

Leukemia (excluding CLL) is a malignancy suitable as an indicator of health effects of ionizing radiation due to the high relative excess risk and few other risk factors. The effects of occupational exposure to ionizing radiation on developing leukemia have been studied primarily among cohorts of nuclear industry workers. A pooled analysis of mortality among nuclear workers from 15 countries demonstrated an excess rate ratio of 0.19 for non-CLL leukemia for a cumulative protracted dose of 100 mSv compared with zero dose.34 In our cohort, the estimated cumulative doses, however, only exceeded 20 mSv for 26% of the follow-up time, that is, the expected excess risk derived from the nuclear worker study would not exceed 1.05.

Cosmic radiation in flight altitudes consists of gamma and neutron radiation. Neutron radiation is more effective in inducing biological damage than gamma radiation. No human studies on carcinogenicity of neutron radiation have been published. We estimated exposure as effective doses, calculated using a radiation weighting factor (determined by radiation type and energy) of 5–20 for neutrons, depending on neutron energy, that is, presuming that the effectiveness of neutrons is 1–30 times that of gamma radiation. Accordingly, the absorbed (physical) dose from neutrons is multiplied by 1–30 to obtain the effective dose. Radiation doses in our study are so low that we would not have sufficient statistical power to detect an effect on leukemia, unless the weighting factors are too low by at least one order of magnitude. An increased frequency of chromosomal aberrations that may predict cancer risk35, 36 has also been reported among airline personnel.37, 38 Deletion or loss of chromosome 7 has been found increased among cases with myelodysplasia and acute myeloid leukemia (AML), the cases originating from cohorts of aircrews.39

The excess OR of 1.66 per 10 mSv for non-CLL leukemia in our study, based on nine cases, was not statistically significant. A Danish study in pilots40 found suggestive evidence of an increase in the risk of AML with increasing flight hours in jets, however based on only three observed cases. The SIR for AML in our study was 1.83, based on six observed cases and thus nonsignificant.

Skin cancers

This observations on the excess risk of skin cancers are in line with previously published findings. An increased incidence of cutaneous malignant melanoma among cabin crew was reported earlier in the Finnish, Norwegian, Icelandic, U.S. and Swedish studies.3, 6–9 A significantly increased incidence of squamous cell carcinoma was observed in the Norwegian, Icelandic and Swedish studies.6, 7, 9 A meta-analysis suggested a pooled meta-SIR of 2.15 (95% CI 1.56–2.88) for malignant melanoma and a meta-SIR of 1.91 (95% CI 0.71–3.73) for squamous cell carcinoma for female cabin crew.41

Exposure to ultraviolet radiation (UVR) is by default the most likely explanation for the increased risk of skin cancers, as up to 90% of all skin cancers are thought to be attributable to UVR.42 The major risk factors for malignant melanoma of the skin include intermittent sun exposure, sunburn at early age and host factors related to skin color and nevi.

There is no exposure to UVR in the aircraft cabin.43 One study reported that the aircrews spend more time in sun resorts and use more frequently sunscreen than the general population. However, there was basically no difference in frequency of nonoccupational risk factors for skin cancer including excessive exposure to sun in a study from Iceland, and nonoccupational risk factors did not seem to explain the excess risk of malignant melanoma among aircrews in our study.44 The risk in the head and neck area seems to be similar irrespective of whether the person is regularly outdoors or not, while skin melanoma of the trunk and limbs is more common among indoor workers probably owing to intermittent recreational UVR exposure combined with propensity to sunburn.16

There has been little indication of an association between ionizing radiation and malignant melanoma in earlier studies in various other settings, but the data are sparse.10 Other skin cancers have been associated with radiotherapy among children, and an excess of BCC has been found among A-bomb survivors.10

It has been suggested that night shift work may also be associated with an increased risk of melanomas,45 but a recent large prospective cohort study observed a significant decreased risk related to light-at-night.46 We conclude that despite the nonsignificantly increased melanoma risk with exposure to cosmic ionizing radiation in our case-control study, the excess risk of skin cancer may be attributable to UVR.

We observed 10 cases of the rare Kaposi sarcoma, a sentinel cancer for AIDS, among male cabin crew. In California, male cabin crew had an 8- to 9-fold increased risk of Kaposi sarcoma.8 This cancer type is not related to work exposures.

Non-Hodgkin lymphoma

The only cancer type with a significantly elevated SMR among male cabin crews in the eight-country study47 was non-Hodgkin lymphoma (9 cases, SMR 2.28 and 95% CI 1.04–4.56), while there was no excess mortality from non-Hodgkin lymphoma among the 33,000 female cabin crew members in the same study. The authors concluded that some of the deaths from non-Hodgkin lymphoma could have been related to AIDS. The SIR for non-Hodgkin lymphoma among cabin crew members in the present study was 1.89 (95% CI 0.81–3.72) in men and 1.12 (95% CI 0.30–2.88) in women. There are some data to suggest that NHL may also be related to circadian rhythm disruption but this evidence is not strong.48

Cancers of the mouth, pharynx and liver

Excess risks of cancers of the mouth, pharynx and liver have been demonstrated among persons infected with human immunodeficiency virus (HIV).49 In theory, it is possible the up to four-fold excess risk of alcohol-related cancers after retirement age in men could be associated with HIV infection.

Detection bias

The cabin crew members are subject to regular medical control surveillance, which may affect their cancer risk pattern. An increased incidence of BCC might indicate higher diagnostic activity among cabin crew members than among the average population, but the similarity of the risk of BCC and other skin cancers (for which diagnostic activity should not play an equally important role) suggests that the excess is real. Cancers of the prostate and thyroid represent other examples of cancers where active case finding increases the incidence. The incidence of these cancers among cabin crew members did not differ from the national averages. Therefore, it appears that diagnostic activity does not have a major effect on our results.

Mammography tests may have been more frequent among cabin crew than in the reference population, but the difference should have been decreasing when the organized whole-population screening programs started (in Finland 1986, Iceland 1987 and Sweden 1997). In Norway, the organized breast cancer screening started in mid-1990s in four counties (40% of population) and was stepwise introduced until national coverage in 2004. For instance, in Finland, all women in age range 50–59 years are invited to mammography screening every 2 years, and the participation rate has been close to 90%. The SIR for breast cancer has not changed over decades, which suggests that the excess risk is not an artefact due to high diagnostic activity.

Final remarks and conclusions

There are few areas outside the Nordic countries with several decades of population-based registration of cancer. Because our study cohort included most of the cabin crew ever certified in the four Nordic countries, our study can be considered as having the maximal potential world-wide to evaluate cancer incidence among cabin crew. Some of the results based on the national cohorts have been published earlier.3, 6, 7, 9 For this article, new data have been added, both in terms of additional cohort members and of increased follow-up time for those included in the national analyses. The larger material allowed analyses of more detailed classifications of exposure and subcategories of cancers than in the national settings.

Due to the accurate population registration systems in all Nordic countries, the follow-up for deaths and emigration is complete and the person-year calculations are precise. Cancer registration systems in Finland, Iceland, Norway and Sweden are also virtually complete and the computerized record linkage procedure precise.16 Therefore, the SIR estimates of our study are not affected by bias attributable to incomplete follow-up or failures in record linkages. The use of systematically registered cancer incidence data (instead of mortality data) avoids bias caused by better cancer survival between population with a relatively high educational level such as cabin crew and the reference population,50 as well as sometimes problematic definitions of the underlying cause of death. Furthermore, the use of incident cancers as outcome events instead of cancer deaths increases the study power due to a larger number of events and allows evaluation of risks for cancers that are rarely lethal, such as skin cancer.

Our study included a novel approach to compare cancer risk by levels of estimated exposure to cosmic ionizing radiation and of circadian disruption. There was a statistically nonsignificant indication of an increased risk of leukemia (excluding CLL) in the very highest dose levels of estimated radiation accumulated in cabin crews, and no other factors than radiation are evident that could explain the excess risk.

No association was observed for any metric of estimated cosmic radiation or the estimated circadian disruption and risk of breast cancer. For certain known risk factors of breast cancer, we did not find evidence to imply an explanation for our main results. These findings indicate a need of detailed studies focusing on more precise estimates of repeated jet lags, irregular night shift work and sleep deprivation, possible work-related factors involved in the increased breast cancer risk and the suggestive dose-response pattern in non-CLL. More information on the role of occupational exposure versus nonoccupational risk factors in the observed excess may potentially be obtained by collecting some data by questionnaire.

Acknowledgements

The authors thank Dr. Johnni Hansen from the Danish Cancer Society for his valuable help in updating this text related to the rapidly developing issue on circadian disruptions and cancer. The resources of the participating institutes were crucial in finalizing this last publication of the planned study series.

Ancillary