We have conducted a large and comprehensive study characterizing cancer rates among U.S. cabin crew relative to the general U.S. population, which adds to the relatively sparse literature regarding cabin crew health and has included profiling a wide range of cancers. Consistent with previous studies reporting on cancer incidence and mortality among flight attendants, we report a higher prevalence of breast, melanoma and non-melanoma skin cancers (comprising basal cell and squamous cell carcinomas) among this occupational group relative to the general population. This is striking given the low rates of overweight and smoking among flight attendants in our study population, which we take to be indicators of general health and healthy behaviors, as well as being independent risk factors for some cancers [1, 11]. We also report associations between job tenure as a flight attendant and several cancer outcomes, consistent with previous U.S. and European studies [7,8,9,10,11], though we note that our reliance on cancer prevalence rather than incidence complicates the interpretation of our findings with regard to the timing of both work exposures and cancer outcomes, and the conflation of cancer incidence and survivorship. Nevertheless, our study extends the sparse literature on this important topic, confirms previous findings, and is the first study to note an increase in non-melanoma skin cancer among U.S. cabin crew (consistent with studies of European cabin crew and pilots). Our work informs future research directions regarding the health of this understudied group of workers and highlights the question of what can be done to minimize the adverse exposures and cancers common among cabin crew.
Our finding of a greater prevalence of breast and skin cancers among flight attendants is consistent with most of the epidemiologic literature on this topic to date [7,8,9,10,11]. As noted above, our study is the first to show an increase in non-melanoma skin cancer among U.S. cabin crew relative to the general population, which replicates findings among European flight attendants and pilots [11, 24]. We also observed that job tenure as a flight attendant was associated with the prevalence of non-melanoma skin cancer, as well as breast cancer (within parity subgroups), among females. We were also able to conduct SPR analyses among crew with in-flight secondhand smoke exposure prior to 1988 and found that some associations were strengthened among this subset of participants. Interpretation of the latter results is somewhat hampered by the fact that participants’ occupational secondhand exposures ended by 1998 at the latest, and studies regarding smoking or secondhand smoking exposures and breast and skin cancers have reported mixed results [25, 26]. However, secondhand smoke has been linked to breast and skin cancers in some studies and is certainly a potential risk factor for these cancers, and unlike for cardiovascular disease, smoking-related risk of cancer never falls to baseline, even years after cessation of the exposure [25, 26].
Our results are also consistent with cabin crews’ occupational exposures to ionizing radiation [2, 6], Circadian rhythm disruption [3], historical exposures to in-flight secondhand smoke [5], and ongoing exposures to other chemical agents [2], most of which are classified as confirmed or probable carcinogens in humans [27,28,29]. Ionizing radiation is a known causal factor for non-melanoma skin cancer and breast cancer [27], whereas the studies regarding melanoma in relation to ionizing radiation are more conflicted [30]. It should be noted that cabin crew have the largest annual ionizing radiation dose of all U.S. workers (e.g. 3.07 mSv vs. 0.59 mSv for U.S. Department of Energy workers) [5]. These exposures can easily exceed guidelines released by the NCRP or the International Commission on Radiological Protection [6, 31]. Although we evaluated job tenure prior to age 45 or age 40 in relation to cancer prevalence, in part to isolate the potential effects of ionizing radiation exposure at younger ages, these restrictions generally did not meaningfully alter our results. This may be because ionizing radiation exposure is also important to cancer risk at older ages, and because it is difficult to disentangle the relevant exposure years in our study population, which has a median tenure of 19 years of employment and for which cancer diagnosis date was not recorded. One possible exception is for breast cancer, for which associations were somewhat stronger when evaluating tenure prior age 45 rather than lifetime tenure. These results, while imprecise and requiring replication in a study that estimates cosmic ionizing radiation exposure directly (rather than using tenure as a proxy), may suggest that flight-related exposures are most important to breast cancer risk when occurring at earlier ages.
We report associations between duration of employment as a flight attendant and breast cancer risk among women who had three or more children, with some evidence of an association among nulliparous women as well, though the latter association was imprecise. Nulliparity is a risk factor for breast cancer, and women who are parous may be less susceptible to the effects of carcinogenic exposures on the breast due to breast cell differentiation occurring after a first pregnancy [32, 33]. Hence, our findings of a somewhat stronger association between job tenure and breast cancer among nulliparous women is consistent with the current state of biologic and epidemiologic knowledge, though it should be noted that the few relevant previous studies among cabin crew did not show increased risk of breast cancer among nulliparous relative to parous participants [20, 21, 34]. Our findings of a stronger association between tenure as a flight attendant and breast cancer among women with three or more children is, interestingly, consistent with two other recent publications on this topic among cabin crew [20, 21]. The authors of the latter studies, which evaluated breast cancer in relation to calculated cosmic radiation exposure and Circadian rhythm disruption, hypothesized that these unexpected results may be due to Circadian rhythm changes from shift work and crossing time zones [21], especially since flight attendants report a much higher rate of sleep disorders and disturbances relative to the general population [11, 12] and these effects may be exacerbated among women with young children who have greater sleep disruptions from both their home and work lives [21].
Limitations of our study include its cross-sectional design, which precludes inferences about causality, as an observed association may reflect the effect of flight attendant work on a given condition, or the effect of an outcome on a factor related to employment as a flight attendant. Use of structured questionnaires, as in our study, aims to minimize this bias. We also note a further limitation that the date of cancer diagnosis was not recorded in the FAHS questionnaire. Hence, some reported cancers may have been diagnosed prior to employment as a flight attendant, and some flight attendant work (i.e. exposure) may have occurred following a cancer diagnosis, making the direction of the potential bias unclear. These limitations are counteracted in part by our analyses evaluating job tenure prior to age 40 and 45 years in relation to cancer prevalences, as many cancers, including of the skin and breast, occur later in life. Therefore, this restriction increases the probability that the exposure of interest occurred prior to the reported cancer outcome.
Another potential limitation of our study involves the question of whether a population of flight attendants is sufficiently comparable to the general U.S. population with regard to cancer risk factors, and whether differences in risk factors may introduce bias to the SPRs. For example, we report substantial differences in racial profile, smoking status, overweight prevalence, and number of live births between the FAHS and NHANES cohorts, all of which are related to the risk of various cancers. We have counteracted this issue in part by restricting the NHANES comparison group to currently employed adults with at least a high school degree and above a certain income to poverty ratio, and by conducting sensitivity analyses restricting to non-Hispanic white participants that showed no meaningful differences from our main results. We also note that the FAHS includes a substantially smaller percentage of current smokers and overweight participants than the NHANES population, which would be expected to decrease the risk of several cancers, whereas we consistently observed increased cancer SPRs. At the same time, we should note that FAHS participants had fewer children than NHANES participants (which elevates the risk of breast cancer), though this is in part ameliorated by the fact that we observed associations between tenure as a flight attendant and breast cancer within parity subgroups. Even with the above reported sensitivity analyses, we acknowledge that the potential for residual confounding by cancer risk factor profile differences between the two study populations (such as for race and parity) still exists.
In addition, health outcomes in our study and in the NHANES were based on self-report; validation through medical records was not possible due to the scope and cost of this endeavor. Validity of self-reported health outcomes varies by study population and the outcome of interest. Sensitivity and specificity of self-reported outcomes relative to medical records or linkage to disease registries were found to be moderate to high for common cancers (including breast cancer and melanoma), particularly among those with higher socioeconomic status, such as in our well-educated cohort [35]. However, this has not been the case for non-melanoma skin cancer. We should note that non-melanoma skin cancers are excluded from most U.S. cancer registries and may be under-reported by those that do include it [36]. This may explain why non-melanoma skin cancer assessed through the California Cancer Registry was not related to flight attendant work in a previous study [9], in contrast to many other studies conducted among cabin crew and pilots, including our own research presented here [11, 24].
A further limitation of our study is that we recruited flight attendants from a mix of company rosters, on-site airport recruitment, and an online/social media presence. Recruiting volunteer participants not recruited from employee files may have contributed to selection bias. For example, volunteer participants may differ from those recruited using a more randomized approach in terms of various factors, including their socioeconomic status, attitude toward health research, and factors related to time and ability to complete online surveys (which may also be related to health), as discussed in a recent analysis with regard to online recruitment in the Heart eHealth Study relative to NHANES [37]. However, the above analysis reported that, while selection bias was likely on a variety of factors, such as gender and marital status, it was much less likely to affect internal (rather than external) validity of exposure-outcome associations [37]. This is likely to be especially true in a relatively homogenous workforce than in a general population study recruited online. It is also important to note that an online recruitment strategy has many advantages in terms of efficiency, reliability of data collection and coding, and the ability to reach a wider range of potential study participants [37].
Our study may have attracted a disproportionate number of flight attendants with cancer, leading to detection bias, as flight attendants with worse health are likely to be more motivated to participate in an epidemiological study of flight attendant health, are likely to attend regular medical check-ups (this is true for flight attendants in general), and the question of cancer risk in relation to flight exposures is well known within the aviation community. However, it is reassuring that our results are consistent with previous studies that recruited participants from employee rosters [8,9,10].
Additional limitations of our study include reliance on job tenure as a surrogate for occupational exposures, lack of correction for multiple testing, insufficient power to evaluate less common cancers, and insufficient information on confounders for some cancers. In particular, we were not able to control for leisure-time UV exposure when evaluating skin cancer risk, though it should be noted that a large study found no difference in sunbathing habits between flight attendants and the general population [38]. We plan to evaluate specific exposures in future individual exposure-outcome analyses. Finally, we note that our reliance on prevalence rather than incidence of cancer confuses the issues of cancer risk and survivorship in interpreting our results. This is in part ameliorated by the fact that breast and skin cancers have relatively low mortality rates (especially for basal cell carcinoma, which is not considered fatal or disabling), and that we are comparing to prevalence rates in NHANES as well. Nevertheless, the limitation remains, and it is also important to note that flight attendants may differ from the general population of U.S. workers with regard to health insurance access, paid leave policies, and other benefits that could affect survivorship, and may be more likely to have access to an urban center with better quality health care for cancer treatment. It is reassuring that our results are consistent with previous studies that relied on cancer incidence [9, 10].
Strengths of our study include access to the resources of a large cohort of cabin crew with information on a range of cancer outcomes, work experiences, and potential confounders. In addition, online questionnaires are an increasingly popular option in epidemiologic research, including high profile studies such as the Millennium Cohort and the Nurses’ Health Study 3 [39]. This mode of data collection allows for validation checks, reduced data entry and coding errors, personalized question administration, convenience to participants, equal or better validity compared to hard copy questionnaires, and the collection of metadata, such as date, time, and time to completion, which can be used for quality control and sensitivity analyses [39].
Our study findings contribute to the sparse literature on flight attendant health, which may also be applicable to passengers, especially frequent flyers. Conducting high quality studies within this group of workers is important given that U.S. cabin crew are subject to fewer protections than most workers in this country and relative to flight attendants working in the European Union (EU). For example, the EU requires airlines to monitor radiation dose, organize schedules to reduce radiation exposure, and inform workers of current studies [40].
