IJC

International Journal of Cancer

Cancer incidence in professional ight crew and air trafc control ofcers: Disentangling the effect of occupational versus lifestyle exposures
Isabel dos Santos Silva1, Bianca De Stavola1, Costanza Pizzi1, Anthony D. Evans2 and Sally A. Evans3
1

Departments of Non-Communicable Disease Epidemiology and Medical Statistics, Faculty of Epidemiology and Population Health, London School of Hygiene and Tropical Medicine, London, United Kingdom 2 International Civil Aviation Organization, Montreal, QC, Canada 3 Medical Department, UK Civil Aviation Authority, Gatwick Airport South, West Sussex, United Kingdom

Flight crew are occupationally exposed to several potentially carcinogenic hazards; however, previous investigations have been hampered by lack of information on lifestyle exposures. The authors identified, through the United Kingdom Civil Aviation Authority medical records, a cohort of 16,329 flight crew and 3,165 air traffic control officers (ATCOs) and assembled data on their occupational and lifestyle exposures. Standardised incidence ratios (SIRs) were estimated to compare cancer incidence in each occupation to that of the general population; internal analyses were conducted by fitting Cox regression models. Allcancer incidence was 2029% lower in each occupation than in the general population, mainly due to a lower incidence of smoking-related cancers [SIR (95% CI) 5 0.33 (0.270.38) and 0.42 (0.280.60) for flight crew and ATCOs, respectively], consistent with their much lower prevalence of smoking. Skin melanoma rates were increased in both flight crew (SIR 5 1.87; 95% CI 5 1.452.38) and ATCOs (2.66; 1.554.25), with rates among the former increasing with increasing number of flight hours (p-trend 5 0.02). However, internal analyses revealed no differences in skin melanoma rates between flight crew and ATCOs (hazard ratio: 0.78, 95% CI 5 0.371.66) and identified skin that burns easily when exposed to sunlight (p 5 0.001) and sunbathing to get a tan (p 5 0.07) as the strongest risk predictors of skin melanoma in both occupations. The similar sitespecific cancer risks between the two occupational groups argue against risks among flight crew being driven by occupationspecific exposures. The skin melanoma excess reflects sun-related behaviour rather than cosmic radiation exposure.

Epidemiology

There are concerns that professional ight crew may have raised cancer risks because of their exposure to several occupational hazards known, or suspected, to be carcinogenic. In particular, professional ight and cabin crew are classied as occupationally exposed to ionising radiation1,2 because they sustain an annual ionising radiation dose of 26 mSv,3 which
Key words: cancer, skin cancer, malignant melanoma, radiation, ight crew, air trafc control ofcers Abbreviations: ATPL: airline transport pilot; BMI: body mass index; ATCOs: air trafc control ofcers; CAA: Civil Aviation Authority; CI: condence interval; HR: hazard ratio; ICD: International Classication of Diseases; MRS: Medical Records System; OR: odds ratio; SIR: standardised incidence ratio; UV: ultraviolet radiation Additional Supporting Information may be found in the online version of this article. Grant sponsor: Medical Department, UK Civil Aviation Authority (CAA); Grant number: 491/SRG/R&AD DOI: 10.1002/ijc.27612 History: Received 16 Nov 2011; Accepted 13 Apr 2012; Online 24 Apr 2012 Correspondence to: Isabel dos Santos Silva, Faculty of Epidemiology and Population Health, London School of Hygiene and Tropical Medicine, Keppel Street, London WC1E 7HT, UK, Fax: 44-20-7436-4230, E-mail: isabel.silva@lshtm.ac.uk

is in addition to the background radiation of the general population. This dose is regarded as low and within the limits for occupational exposure to radiation for non-pregnant adults; however, there are concerns about the health effect of high-energy radiation from neutrons.2 Previous studies of ight and cabin crew have reported increased risks of melanoma and non-melanoma skin cancers,46 acute myeloid leukaemia5,7 and cancers of the breast in women8 and prostate in males7; however, these incidence studies tended to be relatively small. A large pooled analysis of mortality data from several European countries conrmed the increased risks for skin melanoma, but did not observe increases for any other cancer sites.911 Flight and cabin crew members have a complicated exposure history because their occupation leads to specic lifestyle characteristics, and these may act as possible confounders to the health effects of occupational hazards. Such lifestyle characteristics include recreational sunlight exposure during rest periods in hot places overseas or leisure activities. Exposure of ight crew to solar ultraviolet (UV) radiation entering through the cockpit windows is unlikely to explain their increased skin cancer risk.12 We examined cancer risks in a large United Kingdom population-based cohort of ight crew and attempted to disentangle for the rst time the effects of occupational

C Int. J. Cancer: 132, 374384 (2013) V 2012 UICC

dos Santos Silva et al.

375

exposures from those of lifestyle characteristics. This study benets from several unique features. First, to minimise the impact of the healthy worker effect and other health-related selection bias, we included a cohort of air trafc control ofcers (ATCOs) to act as a comparison group. ATCOs share a similar socioeconomic background and undergo regular medical surveillance as do professional ight crew. Second, we had access to data on occupational exposures as well as lifestyle characteristics, including smoking habits and UV-related exposures, from medical records and postal questionnaires. Third, cancer incidence data were obtained through linkage to the UK national cancer registries. Many previous studies have relied on mortality data, but because of high survival rates for many cancers, mortality studies are much less informative.

in England, Scotland and Wales or through the Central Services Agency if resident in Northern Ireland. These are virtually complete population registers for their respective countries. Follow-up information included details of cancer registrations, death certications and migration between and outside the UK countries. Cancer sites and underlying causes of death were coded in accordance with the International Classication of Diseases (ICD), revisions 9 and 10.16,17 Grouping by main cancer sites, according to the ICD-10 main chapters, was used in the analyses. Ethical approval was obtained from the Defence Medical Services Clinical Research Ethics Committee, the London School of Hygiene and Tropical Medicine Ethics Committee and the National Health Service South East Multi-Centre Research Ethics Committee. The study was conducted according to Section 23 (Disclosure of Information) of the UK Civil Aviation Act, as advised by the Queens Counsel.
Statistical analysis

Material and Methods

Study design

The study design and data sources have been described in detail elsewhere.1315 In summary, the study population was identied from the Medical Records System (MRS) of the UK Civil Aviation Authority (CAA), which holds data gathered from routine surveillance examinations of holders of UK licences, that is, holders of a UK professional ight crew licence or ATCO licence. Flight crew licences comprise licenses for airline transport pilot (ATPL), senior commercial pilot, commercial pilot, basic commercial pilot, ight engineer and ight navigator. CAA medical examinations are undertaken every 6 or 12 months by ight crew and every year or alternate year by ATCOs (both depending on age). Flight crew and ATCOs who held a valid professional licence at any time during the period from January 1, 1989 (when the MRS was computerised) to December 31, 1999 were eligible for entry into the study. During 20002001, all eligible individuals were mailed a letter of invitation to complete a postal questionnaire and consent for researchers to access their CAA medical records and follow them up passively through UK health population registers. All eligible UK resident subjects were mailed twice (the second time using registered post) to ensure that they were given the opportunity to refuse participation. The CAA MRS is completed at the time of each medical examination when a standardised form is administered to aircrews and ATCOs. This form collects data on a limited number of demographic, occupational (e.g., type of licences) and lifestyle variables (e.g., smoking habits) as well as data on selected physical characteristics (e.g., current height and weight, hair and eye colour). The postal questionnaire obtained more detailed information on demographic (e.g., country of birth and residence) and lifestyle (e.g., reproductive history and UV-related exposures) variables as well as a full occupational history. Study subjects were followed up to the end of 2008 through the National Health Service Central Registers (recently re-named as NHS Information Centre) if resident
C Int. J. Cancer: 132, 374384 (2013) V 2012 UICC

Follow-up time was dened from date of entry (January 1, 1989 or the date of issue of the rst professional licence, if later) to the date of exit (date of cancer registration, date of death, date of emigration, 90th birthday, or December 31, 2008, whichever occurred rst). All-cancer (excluding nonmelanoma skin cancer because of its acknowledged underascertainment and under-registration; see below) and sitespecic cancer incidence rates among ight crew and ATCOs were compared to those of the UK general population by calculating standardised incidence ratios (SIRs), that is, the ratio of observed to expected numbers of cancers for a set of individuals with a given length of follow-up, where expectations were calculated according to the age-, sex-, calendar yearand country-specic incidence rates of the reference population using Poisson regression models.18 We used respectively cancer incidence rates of England and Wales for the English and Welsh participants, of Scotland for the Scottish participants and of the United Kingdom for the Northern Irish ones (no reliable estimates were available for Northern Ireland). Initial comparisons of cancer incidence between the two occupational groups were based on the ratio of their SIRs. SIRs were also stratied by occupational and lifestyle exposures of the cohort members using Poisson regression,18 separately by occupational group, to identify potential explanatory factors for the observed reduced (or raised) SIRs. Internal analyses were also conducted by tting Cox regression models (with the time scale dened by age) to estimate hazard ratios (HR) as a measure of the association of occupational and lifestyle variables with all sites, site specic and cancer rates, while adjusting for sex, calendar period and (implicitly) age.19 Heterogeneity and linear trends were assessed using the Wald test.19 Analyses of cancer risks among ight crew and ATCOs relative to the UK general population excluded non-melanoma skin cancers because of concerns of differential under-

Epidemiology

376

Cancer in flight crew and air traffic controllers

Figure 1. Flowchart illustrating how the participants were selected for the study. [Color gure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]

Epidemiology

diagnosis and under-registration as ight crew and ATCOs are from a higher socioeconomic group and undergo more intensive medical surveillance than the general population. However, to exploit all available information, both melanoma and non-melanoma skin cancers were included in internal analyses as the degree of under-diagnosis/registration for the latter is likely to have been comparable for the two occupational groups. When appropriate, further analyses were conducted among ight crew by type of their licence (categorised as ATPL/not-ATPL, as ATPL holders tend to y large planes over World routes); cumulative ying hours at baseline and cumulatively up to year 2000-1 (both categorised using the tertiles of their respective distributions); route own (World/ Europe/UK) and type of aircraft (large/medium/single pilot). Similarly, further analyses were conducted among ATCOs by cumulative number of hours of radar duties and cumulative number of night shifts (both categorised using the tertiles of their respective distributions). All P-values are two sided.

slightly higher percentage of ight crew than ATCOs reported drinking alcohol regularly with about 90% in each group reporting exercising regularly at least once a week (Table 1). On average, female ight crew had a lower BMI at entry than female ATCOs [median (inter-quartile range): 22.0 (20.6 23.7) and 23.0 (21.525.1) kg/m2], whereas males had similar values [24.4 (22.726.3) and 24.2 (22.326.3) kg/m2]. The skin type distribution and the proportion of those reporting having ever been sunburnt were similar between the two occupational groups and sexes; however, females consistently reported a higher use of sunscreens and sunbeds (Table 1). At entry, 69% ight crew had an ATPL, about one-third had accumulated more than 5,500 ight hours (Table 1), and 23% had own world routes (data not shown). As expected, the cumulative number of hours own were substantially higher in 20002001 (Table 1), when the questionnaire was administered, than when the study commenced. Among ATCOs who completed the questionnaire, the large majority were radar qualied and worked night shifts (Table 1). There was no difference in the distribution of baseline variables between participants who completed the questionnaire and those who did not (Supporting Information Table S1).
Cancer incidence in each occupational group relative to the general population

Results
Characteristics of the study subjects

A total of 27,392 ight crew and ATCOs satised the criteria for entry into the study. Of these, 7,903 were excluded for reasons given in Figure 1. Thus, the analyses were based on 19,494 participants, 50% of whom completed a postal questionnaire. Flight crew were slightly older at entry than ATCOs [median age (in years) of males: 37 for ight crew and 35 for ATCOs; of females: 29 and 25]. They were also less likely to be current smokers than ATCOs at entry, but not at the time of the questionnaire (10 years later) when the prevalence in both groups was much lower (Table 1). Among those who completed the questionnaire, and within each gender, a

A total of 773 incident neoplasms (excluding non-melanoma skin cancers) occurred among ight crew and 151 among ATCOs during 285,259 and 54,045 person-years at risk (median follow-up time: 19 years for each group). Flight crew and ATCOs had a 29% (SIR 0.71, 95% CI 0.660.76) and 20% (SIR 0.80, 95% CI 0.680.94], respectively, lower all-cancer incidence than the UK general population. Rates were lower among ight crew and ATCOs than among the general population for most cancer sites, the only exceptions being the similar rates for cancers of the prostate and female breast and the much higher rates for malignant melanoma of the skin, in each occupational group (Fig. 2; Supporting Information Table S2). The low all-neoplasm incidence in each occupational group was largely accounted for by a very low incidence of smoking-related cancers, with ight crew having only 33% and ATCOs only 42%, of the rates found in the general population (Table 2). There was a clear trend in the relative incidence of smoking-related cancers with smoking status categories, with current smokers at entry experiencing similar rates to the general population (especially if ATCOs). In contrast, there was no clear variation in the incidence of smokingrelated or non-smoking-related cancers across categories of BMI, height, regular alcohol consumption and regular physical exercise (Table 2), and no obvious trend with amount of alcohol intake [p for linear trend (p-trend) 0.83 in ight crew and p-trend 0.26 in ATCOs] or amount of regular exercise (p-trend 0.76 and p-trend 0.95). (Data not shown for these two analyses).
C Int. J. Cancer: 132, 374384 (2013) V 2012 UICC

Table 1. Selected lifestyle, host and occupational characteristics of ight crew and ATCOs Flight crew (N 16,329) Males Characteristic All (from MRS): Current smoker at entry2 Blonde/fair/ginger hair type
2

ATCOs (N 3,165) Males %1 100 12 36 N 2,706 479 316 %1 100 18 12 N 459 63 152 Females %1 100 14 33

Females %1 100 15 14 N 462 56 164

N 15,867 2,349 2,254

Type of licence held at entry3 ATPL Not ATPL Cumulative ying hours (baseline) <400 4005,499 5,500 With questionnaire data
5 4

10,958 4,909

69 31

200 262

43 57

5,066 5,277 5,482 7,656 450 7,217 6,750

32 33 35 100 6 95 89

296 144 22 222 10 197 204

64 31 5 100 5 92 93 1,562 95 1,425 1,350 100 6 92 87 260 22 234 243 100 8 90 94

Current smoker in 20012002 Regular alcohol drinker in 20012002 Regularly exercises in 20012002 Skin type Burns easily Burns occasionally Ever sunburnt Regular sunscreen use Days per year spent sunbathing 0 115 1630 >30 Times per year used sunbed 0 15 >5 Cumulative ight hours (up to 2000) <5,000 5,00011,699 11,700 Radar qualied Cumulative radar days (19902000)4 <360 360727 >727 Night shift work Cumulative night shifts (overall; days)4 <384 384999
1

3,489 4,100 7,057 5,411

46 54 94 72

126 93 201 184

57 43 95 84

716 832 1,377 1,101

46 53 91 71

153 107 241 239

59 41 94 92

357 2,032 2,039 1,230

6 36 36 22

11 56 75 37

6 31 42 21

64 418 371 234

6 39 34 21

5 83 93 36

2 38 43 17

6,998 268 182

4

94 4 2

157 31 17

77 15 8

1,414 71 43

92 5 3

176 41 32

71 17 13

2,196 2,392 2,410

31 34 34

142 42 2

76 23 1 1,369 89 190 76

516 363 382 1,430

41 29 30 93

124 56 35 212

58 26 16 85

491 466
2

33 31

141 67

61 29

Percentages were calculated out of total without missing values. Missing values for ight crew and ATCOs: 42 and 17 for current smoking, and 26 and 5 for hair colour. 3The Not ATPL category includes senior commercial pilots license (SCPL), commercial pilots license (CPL), basic commercial pilots license (BCPL), ight engineers license (FE) and ight navigator license (FN). 4Categorisations are based on tertiles of the respective distributions. 5Information obtained from the postal questionnaire administered in 20012002. Missing values for ight crew and ATCOs: 58 and 14 for current smoking; 81 and 14 for regular alcohol drinker; 36 and 4 for regular exercise; 70 and 14 for skin type; 155 and 52 for ever sunburnt; 114 and 18 for regular sunscreens use; 2041 and 518 for days spent sunbathing and 223 and 45 for frequency of sunbeds use. Abbreviations: ATCOs: air trafc control ofcers; ATPL: airline transport pilots license; MRS: Civil Aviation Medical Records System.

378

Cancer in flight crew and air traffic controllers

Figure 2. Cancer incidence in ight crew and ATCOs relative to the general population and internal comparisons of ight crew versus ATCOs for all cancers (excluding non-melanoma skin cancer) and selected main sites. [Color gure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]

Epidemiology

Site-specic cancer incidence in relation to occupational exposures

There were no clear differences in rates of any cancer site between the two occupational groups, before (Fig. 2) or after further adjustment for smoking habits and BMI at entry (Supporting Information Table S2). The incidence of skin melanoma among ight crew increased with increasing cumulative number of hours own, as recorded at entry into the study (ptrend 0.02) or at the time of questionnaire administration (p-trend 0.07; Table 3); however, there were no clear associations in the risk of this cancer with licence type (Table 3), route own or type of aircraft (data not shown). The incidence of digestive cancers was higher among ight crew who held licences other than ATPL (p-trend 0.02); however, there was no association with cumulative number of hours own (Table 3). The incidence of male genital cancers was higher among ight crew who held ATPL licences (p-trend 0.07), with rates increasing with cumulative number of ight hours at entry (p-trend 0.04) (Table 3). Similar associations with licence type were present when the analyses were restricted to colorectal cancer (p 0.05), which accounted for

63% of all digestive cancers, and prostatic cancer (p 0.09), which accounted for 90% of all male genital cancers, with a positive trend in rates with ight hours at entry also present for the latter (p-trend 0.06). There were no associations between cumulative number of hours of radar duties, or cumulative number of night shifts, and site-specic cancer rates among ATCOs (Supporting Information Table S3).
Skin cancer risks in relation to host characteristics and occupational and sun-related exposures

Internal comparisons by occupational group focussed on skin cancer rates because of the raised incidence of this cancer relative to the general population. Both melanoma and nonmelanoma skin cancers were included in these analyses to exploit all available information (Table 4). Skin melanoma and non-melanoma rates were similar between ight crew and ATCOs [HR of ight crew versus ATCOs adjusted for age, sex and calendar period: 0.72 (95% CI 0.421.24) and 0.59 (95% CI 0.271.29)]. Host characteristics and exposure to recreational UV-related radiation were therefore examined controlling for occupational group.
C Int. J. Cancer: 132, 374384 (2013) V 2012 UICC

dos Santos Silva et al.

379

Table 2. Incidence of smoking related and non-smoking-related cancers among ight crew and ATCOs relative to the UK general population by lifestyle characteristics Smoking related cancers1 Flight crew Stratifying variable From MRS (N 19,494) All 142 0.33 0.270.38 30 0.42 0.280.60 631 0.94 0.871.02 120 0.98 0.811.17 D SIR 95% CI D ATCOs SIR 95% CI D Non-smoking-related cancers2 Flight crew SIR 95% CI D ATCOs SIR 95% CI

Smoking status (at baseline) Never Ex Current Linear trend (p-value) BMI (kg/m2) (at baseline)3 25 >25 Heterogeneity (p-value) Height (at baseline)3 <Sex-specic median Sex-specic median Heterogeneity (p-value) 75 67 0.34 0.31 0.53 0.270.43 0.240.39 14 16 0.32 0.58 0.09 0.190.53 0.260.95 314 317 0.94 0.94 0.99 0.841.05 0.841.05 65 55 0.87 1.15 0.14 0.691.12 0.881.49 63 65 0.31 0.32 0.95 0.250.40 0.250.40 15 13 0.44 0.36 0.60 0.270.73 0.210.62 291 295 0.89 0.97 0.29 0.791.00 0.871.09 62 52 0.99 0.92 0.71 0.771.27 0.701.21 73 18 51 0.22 0.49 0.76 <0.001 0.180.28 0.310.77 0.570.99 13 3 14 0.26 0.42 0.95 <0.001 0.150.45 0.141.30 0.561.61 473 53 105 0.92 0.98 1.02 0.30 0.841.01 0.751.29 0.841.24 83 11 26 0.95 1.02 1.09 0.52 0.761.17 0.561.83 0.741.61

With questionnaire data (prospective analyses from questionnaire date; N 9,413)4 All Smoking status Never Ex Current Linear trend (p-value) Regular alcohol drinking3 No Yes Heterogeneity (p-value) Regular exercise3 No Yes Heterogeneity (p-value)
1

36
3

0.30

0.220.42

14

0.57

0.340.95

211

1.08

0.941.24

47

1.08

0.791.43

3 25 8

0.06 0.44 1.02 <0.001

0.020.17 0.300.65 0.512.04

3 9 2

0.30 0.70 1.19 0.10

0.100.93 0.361.34 0.304.74

95 102 12

1.06 1.12 0.95 0.99

0.871.30 0.921.36 0.541.67

22 23 2

1.15 1.09 0.66 0.54

0.761.75 0.721.63 0.172.64

1 35

0.17 0.31 0.55

0.021.21 0.220.44

2 12

1.16 0.53 0.30

0.294.63 0.300.93

14 194

1.43 1.06 0.28

0.852.42 0.921.22

2 45

0.64 1.12 0.44

0.162.55 0.841.50

7 29

0.39 0.29 0.46

0.190.82 0.200.41

1 13

0.24 0.63 0.35

0.031.70 0.371.09

30 180

1.06 1.09 0.88

0.741.51 0.941.26

6 41

0.87 1.12 0.56

0.391.93 0.821.52

ICD-10 codes for smoking-related cancers: C01.006.9, C09.010.9, C12.014.9, C15.015.9, C16.016.9, C25.025.9, C32.032.9, C33.034.9, C38.4, C45.045.9 and C67.068.9. 2ICD-10 codes for non-smoking-related cancers are all sites not classied as smoking related but excluding non-melanoma skin cancers. 3Missing values for MRS data were for ight crew and ATCOs: 629 and 64 for BMI; none for height. From questionnaire data, missing values were for ight crew and ATCOs: 55 and 13 for smoking status; 77 and 14 for regular alcohol drinking and 36 and 3 for regular exercise. 4N is slightly smaller than the total number of participants who completed the questionnaire (i.e., 9,700 as shown in Fig. 1), because only those who were still at risk at the time of completion of their questionnaires were included in these prospective analyses. Abbreviations: ATCOs: air trafc control ofcers; BMI: body mass index; CI: condence interval; D: number of cancers; MRS Civil Aviation Medical Records System; SIR: country-sex-age-calendar year standardised incidence ratio.

Occupational exposure to cosmic radiation was examined in terms of cumulative ying hours, with analyses conducted initially on all participants and subsequently on the subset with complete data to allow assessment of whether results from the latter were affected by selection bias. Minimally adjusted analyses (i.e., adjusted for age, sex, calendar period and occupational group) showed statistically

signicant raised melanoma rates for participants who reported their skin burning easily when exposed to hot sunlight without using sunscreens (p 0.001) and among those who reported having blonde/fair/ginger hair (p 0.03), as well as non-signicant raised rates among those who reported using sunscreens regularly and having ever been sunburnt or having ever sunbathed to get a tan (Table 4),

C Int. J. Cancer: 132, 374384 (2013) V 2012 UICC

Epidemiology

Epidemiology

380

Table 3. Incidence of all-cancers, and of selected cancer sites, among ight crew, by their occupational exposures

All cancers 95% CI D D D D D SIR 95% CI SIR 95% CI SIR 95% CI SIR 95% CI

Digestive system SIR

Respiratory system

Skin melanoma

Male genital system

Lymphatic and haematopoietic system 95% CI

SIR

From MRS (N 16,329) 0.660.76 164 0.56 0.470.65 58 0.25 0.190.33 67 1.87 1.452.38 232 1.07 0.941.22 70 0.65 0.500.81

All

773

0.71

Type of licence2 0.650.77 0.620.80 0.02 0.53 0.79 0.07 61 0.72 0.560.92 19 0.29 0.180.45 17 1.77 1.122.81 57 0.88 103 0.49 0.410.60 39 0.24 0.180.33 49 1.91 1.452.53 175 1.15 0.991.34 0.681.14 48 22 0.62 0.71 0.61 0.470.82 0.471.07

ATPL

548

0.71

Not ATPL

225

0.70

Heterogeneity (p-value)

0.94

Cumulative ight hours (at entry)3 0.570.83 0.530.73 0.670.79 0.63 0.11 0.02 109 0.56 0.470.68 42 0.27 0.200.36 43 2.47 31 0.49 0.350.70 11 0.24 0.130.43 13 1.23 0.712.12 1.833.33 24 0.69 0.461.02 2 0.08 0.020.32 10 1.30 0.702.42 20 43 162 0.70 0.99 1.13 0.04 0.451.08 0.731.33 0.971.32 16 15 39 0.85 0.54 0.64 0.49 0.521.38 0.330.89 0.470.87

<400

103

0.69

4005,499

152

0.62

5,500

505

0.73

Linear trend (p-value)

0.28

With questionnaire data (prospective analyses from questionnaire date; N 7,645)4 0.640.84 44 0.53 0.390.71 13 0.21 0.120.35 18 2.04 1.283.23 88 1.19 0.971.47 14 0.51 0.300.86

All

228

0.73

Cumulative ight hours in 2001 (from questionnaire)5 0.521.13 0.600.99 0.540.78 0.32 0.23 21 0.46 0.300.71 5 0.14 12 0.55 0.310.97 4 0.25 6 0.73 0.331.63 2 0.36 0.091.42 0.090.66 0.060.34 0 5 9 0 1.90 2.43 0.07 0.794.57 1.264.67 9 26 42 1.35 1.47 1.00 0.17 0.702.59 1.002.16 0.741.35 4 2 7 1.02 0.26 0.52 0.46 0.382.73 0.071.05 0.251.09

<5,000

26

0.77

5,00011,699

62

0.77

11,700

107

0.65

Linear trend (p-value)

0.28

Cancer in flight crew and air traffic controllers

C Int. J. Cancer: 132, 374384 (2013) V 2012 UICC

>Excluding non-melanoma skin cancers. ICD codes as in Supporting Information Table S2. 2The Not ATPL category includes senior commercial pilots license (SCPL), commercial pilots license (CPL), basic commercial pilots license (BCPL), ight engineers license (FE) and ight navigator license (FN). 3Categorisations are based on tertiles of the distribution among UK professional ight crew. 4N is slightly smaller than the total number of ight crew who completed the questionnaire (i.e., 7,878 as shown in Fig. 1), because only those who were still at risk at the time of completion of their questionnaires were included in these prospective analyses. 5Categorisations are based on tertiles of the distribution among UK professional ight crew with questionnaire data. Missing values for questionnaire cumulative ying hours: 663 (of which 33 events). Abbreviations: ATCOs: air trafc control ofcers; ATPL: airline transport pilots license; CI: condence interval; D: number of cancers; SIR: country-sex-age-calendar year standardised rate ratio.

Table 4. Minimally and mutually adjusted hazard ratios for skin melanomas and non-melanomas among ight crew and ATCOs Melanoma skin cancers1 All (D 82/48)2 Subset with complete data (D 44) Minimally adjusted4 Mutually adjusted5 HR 95% CI HR 95% CI HR 95% CI HR HR 95% CI Minimally adjusted4 Minimally adjusted4 Mutually adjusted5 95% CI Subset with complete data (D 24) Minimally adjusted4 HR 95% CI All (D 38/28)3 Non-melanoma skin cancer1

dos Santos Silva et al.

Risk factor

Hair colour 1 1.83 0.03 0.10 0.28 0.009 0.005 1.063.16 1.89 0.894.00 1.52 0.723.21 2.77 1.295.94 3.65 1 1 1 1 1.498.91 1 3.35 0.01 1.348.35

Dark

Blonde/fair/ginger

Heterogeneity (p-value)

Type of skin 1 2.75 0.001 <0.001 0.001 0.48 1.495.08 3.62 1.737.05 3.49 1.737.05 1.31 0.622.75 1 1 1 1 1.44 0.38 0.643.21 1 1.22 0.65 0.522.87

C Int. J. Cancer: 132, 374384 (2013) V 2012 UICC

Burns occasionally

Burns easily

Heterogeneity (p-value)

Sunscreens use 1 1.72 0.12 0.17 0.78 0.883.37 1.64 0.813.31 1.11 0.542.28 1.73 0.20 1 1 1 0.754.01 1 1.81 0.22 0.714.62 1 1.51 0.40 0.574.00

No

Yes

Heterogeneity (p-value)

Ever been sunburnt 1 4.48 0.14 0.17 0.38 0.6232.64 4.05 0.5629.6 2.46 1 1 0.3318.26 1 1.24 0.77 0.295.28 1 1.13 0.87 0.264.84 1 0.85 0.83 0.193.76

No

Yes

Heterogeneity (p-value)

Ever sunbathed to get a tan 1 1.78 0.12 0.13 0.863.70 1.81 1 0.843.92 1 2.08 0.07 0.954.56 1 1.68 0.26 0.684.16 1 1.63 0.33 0.614.39 1 1.59 0.37 0.584.39

No (baseline)

Yes

Heterogeneity (p-value)

Cumulative ying hours, risk stratied by anatomical site

Whole body 1 0.95 1.84 0.07 0.412.20 0.844.05 1 1.02 1.65 0.30 0.323.32 0.544.98 1 0.99 1.49 0.39 0.313.23 0.494.51 1 1.17 2.37 0.14 0.216.46 0.5310.68 1 1.82 2.87 0.26 1 0.1917.90 1.82 0.1917.92 0.3523.48 2.60 0.3221.16 0.33

<4006

4005,499

5,500

Linear trend (p-value)

Head, neck and upper limbs7

381

Epidemiology

Epidemiology

382

Table 4. Minimally and mutually adjusted hazard ratios for skin melanomas and non-melanomas among ight crew and ATCOs (Continued) Melanoma skin cancers1 All (D 82/48) Minimally adjusted4 Minimally adjusted4 Mutually adjusted5 HR 0.405.75 0.344.97 95% CI HR 95% CI HR HR 95% CI Minimally adjusted4 HR 1 1.51 1.30 0.82 95% CI
2

Non-melanoma skin cancer1 All (D 38/28)3 Subset with complete data (D 24) Minimally adjusted4 95% CI HR Mutually adjusted5 95% CI

Subset with complete data (D 44)

Risk factor

<4006

4005,499

5,500

Trunk and lower limbs7 1 1.19 2.84 0.10 0.7311.01 0.304.69

<4006

4005499

5500

ICD-10 codes for malignant melanoma of the skin: C43; for non-melanoma skin cancer: C44. 2Analyses based on 82 melanoma cases for variables derived from the MRS (i.e., hair colour and cumulative ying hours); analyses based on 48 cases for the remaining exposure variables in the table (all derived from questionnaire data). 3Analyses based on 38 non-melanoma cases for variables derived from the MRS (i.e., hair colour and cumulative ying hours); analyses based on 28 cases for the remaining exposure variables in the table (all derived from questionnaire data). 4 Minimally adjusted analysis: adjusted for sex, age, calendar time and occupational group. 5Mutually adjusted analyses: additionally adjusted for all other variables in the table; analyses restricted to participants without missing values for all exposure variables shown in the table. 6Categories of ight hours dened by tertiles of the distribution in participating ight crew. 7Codes used to dene anatomical site: head and neck: ICD9: 172.0, 172.1, 172.2, 172.3, 172.4; ICD10: C43.0, C43.1, C43.2, C43.3, C43.4; trunk: ICD9: 172.5; ICD10: C43.5; limbs: ICD9: 172.6, 172.7; ICD10: C43.6, C43.7. Data on anatomical location of the cancer were available only for a subset of melanoma cases (D 32 of the head, neck and upper limbs; D 30 of the trunk and lower limbs). There were 2 and 0 non-melanomas in these subgroups, and therefore, no analyses were possible. Mutually adjusted analyses were not possible for melanomas because of the small number of cases. Abbreviation: ATCOs: air trafc control ofcers; CI: condence intervals; D: number of cancers; HR: hazard ratio referring to the follow-up from entry into the study (i.e., assuming the information collected by questionnaire was valid retrospectively; see Results section).

Cancer in flight crew and air traffic controllers

C Int. J. Cancer: 132, 374384 (2013) V 2012 UICC

dos Santos Silva et al.

383

but no trend in risk with days/year spent sunbathing (p-trend 0.88). Similarly, non-melanoma rates were signicantly raised among participants who reported having blonde/fair/ ginger hair and non-signicantly raised among those who reported their skin burning easily when exposed to hot sunlight, using sunscreens regularly, and having ever sunbathed to get a tan (Table 4), but no trend in risk with days/year spent sunbathing (p-trend 0.81). The effect of these host and recreational sun exposures on skin cancer did not differ between ight crew and ATCOs, except for borderline evidence that an association with sunscreen use was stronger among ight crew (p for effect modication 0.05). There was an indication of a linear trend in the rates of both skin melanomas and non-melanomas with increasing ight hours (p-trend 0.07 and p-trend 0.14; Table 4) in line with the observed trend in SIRs for melanomas (Table 3). When analyses were repeated distinguishing between melanomas in normally covered (i.e., trunk and lower limbs) from those in uncovered areas of the body (i.e., head and neck and upper limbs), we found that the trend was strongeralbeit not signicantin the latter (Table 4). Mutual adjustment for occupational and lifestyle exposuresrestricted to subjects with complete datarevealed that skin that burns easily when exposed to sunlight (p 0.001) and sunbathing to get a tan (p 0.07) were the strongest predictors of melanoma risk, whereas blonde/fair/ ginger hair (p 0.01) was the strongest predictor of nonmelanoma risk (Table 4). Notably, there was no evidence of differential melanoma and non-melanoma rates between ATCOs and ight crew after adjustment for these host and recreational exposures (ight crew versus ATCOs, melanomas: HR 0.78, 95% CI 0.371.66; non-melanomas: HR 0.66, 95% CI 0.25, 1.77). Furthermore, there was no evidence that cumulative ying hours had a signicant effect when adjusted for environmental exposures (p-trend 0.39 for melanomas; p-trend 0.33 for non-melanomas; Table 4). To maximise power, the analyses reported above were carried out treating the questionnaire data as valid retrospectively. However, prospective analyses from the date of the questionnaire administration (based on 21 melanomas and 22 non-melanomas diagnosed subsequently) showed a similar pattern [e.g., mutually adjusted analyses identied skin that burned easily (p 0.02) and blonde/fair/ginger skin (p 0.02) as the main predictors of risk for melanomas and nonmelanomas, respectively].

Discussion
By necessity of their profession, ight crew are healthier individuals than the general population. They also undergo strict medical surveillance, leading to even greater health advantages. Indeed, we found that UK professional ight crew had a markedly lower cancer incidence than the general population. However, such comparison may hide increases in specic site-specic cancer risks experienced by this occupational
C Int. J. Cancer: 132, 374384 (2013) V 2012 UICC

group. To be able to identify such increases, if present, we took UK ATCOs, which have a similar socioeconomic background and type of medical surveillance as ight crew,13 as a comparison group. We found that ight crew had similar cancer risks to ATCOs, arguing against their risks being driven mainly by occupation-specic exposures. The reduced all-cancer incidence in both occupational groups relative to the general population was mainly due to a markedly low incidence rate of smoking-related cancers, in line with their much lower prevalence of smoking.13 We have previously shown that the prevalence of current smoking at the time of questionnaire administration was much lower in the two occupational groups than in the general population (e.g., 7% vs. 27%, respectively, for males),13 with ever smokers having started to smoke later than the general population.13 The slightly lower incidence of smoking-related cancers among ight crew than ATCOs is also consistent with the observation that ight crew were less likely to be current smokers at entry (see Results section). Consistently with previous reports,4,5,9,11 our study found that professional ight crew had a marked excess of skin melanoma relative to the general population. The reasons for this excess had not been properly investigated in the past. The increases in risk relative to the general population with increasing cumulative number of ight hours observed in this study would be consistent with a putative effect of exposure to cosmic radiation on the ight deck. However, no signicant trend in the risk of skin melanoma or non-melanoma with ight hours was observed in internal analyses for which differential distributions in host characteristics and recreational sun exposures were adjusted for. In fact, there was no indication that rates were higher among ight crew than ATCOs, even in the highest ight hours category [5,500 cumulative ying hours versus ATCO, HR 0.91 (95% CI 0.402.10) for melanomas and 0.69 (95% CI 0.241.94) for non-melanomas]. Circadian disruptions, leading to melatonin disturbances, have also been proposed as putative risk factors for the excess skin cancer among ight crew.4 We did not have direct information on night shifts for ight crew; however, interestingly, night shift work was not associated with skin cancer in ATCOs. Furthermore, a recent study found a protective effect of night shift work on skin cancer.20 Alternatively, it has been postulated that the high incidence of skin melanoma among ight crew may be due to recreational sun exposure, including exposure during stopovers in sunny places. The ndings from this study are consistent with this interpretation. The fact that the trend in risk with ight hours appeared to be stronger for cancers in areas of the body that are usually covered (i.e., the trunk and lower limbs) suggests that number of ight hours is possibly a correlate of intermittent and intense sunbathing during stopovers in sunny places. The high skin melanoma rates in ATCOs are also likely to be related to their easy access to holidays in sunny places, as ATCOs in the

Epidemiology

384

Cancer in flight crew and air traffic controllers

United Kingdom have historically had access for discounted air travel. We found a positive association between risk of skin melanoma and having ever sunbathed to get a tan, albeit of borderline statistical signicance, but no association of risk with days spent sunbathing. However, accurate recall of past sun exposures is notoriously difcult. Furthermore, UV exposure is likely to be affected by an individuals host characteristics as those who tend to burn rather than tan are more likely to avoid high exposure to sunlight and use sunscreens regularly. Thus, unfavourable host characteristics may have acted as negative confounders on the association of UV-related exposures with skin cancer. In contrast to previous reports,5,7 we found no evidence that ight crew had an increased risk of leukaemias or lymphomas; however, the number of cases was too small to be conclusive. Furthermore, chronic lymphocytic leukaemia is not related to ionising radiation, and they represented a larger proportion of the all-leukaemia cases (7 of 13 among ight crew and 1 of 3 among ATCOs). Some previous studies reported excesses in prostatic cancer among ight crews.7 The overall incidence of prostatic cancer among ight crew in the our study was similar to those in the general population and among ATCOs; however, there was a trend in the risk of this cancer with increasing number of hours own. We also observed an association between type of licence and cancer of digestive organs, particularly colorectal cancer. The aetiological signicance of these ndings is, however, unclear. This study is the largest population-based incident study of ight crew conducted so far. Furthermore, access to a relevant occupational comparison group allows addressing the

potential biases inherent in investigations of extremely healthy occupational groups. Other strengths include the availability of data on both occupational and lifestyle exposures, which allowed the examination of their independent effects on cancer risks. However, there were some limitations, namely, the lack of cosmic radiation dose estimates for ight crew, the limited power to detect small effects, the small number of female participants (which precluded examination of risks for female-specic cancers while adjusting for differences in reproductive history) and the possibility that some statistically signicant results may have arisen by chance given the large number of comparisons performed. In short, by using ATCOs as a comparison group, we were able to rule out any major cancer risks associated with occupational exposures specically associated with being ight crew. The excess of skin melanoma relative to the general population, which was seen for both ight crew and ATCOs, is likely to be due to recreational sunlight exposure in hot places overseas rather than exposure to cosmic radiation. Organisations with an interest in the health of ight crew and ATCOs should emphasise the importance of minimising exposure to potentially harmful sunlight and the importance of early diagnosis of malignant melanoma and nonmelanoma skin cancers.

Acknowledgements
This work was supported by the Medical Department of the UK Civil Aviation Authority (CAA) (to I.S.S. and B.D.S.). The authors thank Mrs. Roberta North (CAA), Mrs. Emma Forrest (CAA) and Ms. Jocelyn Hawkins (LSHTM) for their clerical assistance, and Mr. David Mayer (LSHTM), the late Mr. John Adams (CAA) and Mr. Chris Barrow (Steria CAA) for their computing and data management support.

References

Epidemiology

1.

2. 3.

4.

5.

6.

7.

International Commission on Radiobiological Protection. 1990 Recommendations of the international commission on radiobiological protection. Oxford: Pergamon Press, 1991. Lim MK. Cosmic rays: are air crew at risk? Occup Environ Med 2002;59:42832. Bagshaw M, Irvine D, Davies DM. Exposure to cosmic radiation of British Airways ying crew on ultralonghaul routes. Occup Environ Med 1996;53:4958. Rafnsson V, Hrafnkelsson J, Tulinius H. Incidence of cancer among commercial airline pilots. Occup Environ Med 2000;57:1759. Gundestrup M, Storm HH. Radiation-induced acute myeloid leukaemia and other cancers in commercial jet cockpit crew: a population-based cohort study. Lancet 1999;354: 202931. Nicholas JS, Swearingen CJ, Kilmer JB. Predictors of skin cancer in commercial airline pilots. Occup Med (Lond) 2009;59:4346. Band PR, Le ND, Fang R, et al. Cohort study of Air Canada pilots: mortality, cancer incidence, and leukemia risk. Am J Epidemiol 1996;143: 13743.

8.

9.

10.

11.

12.

13.

14.

Reynolds P, Cone J, Layefsky M, et al. Cancer incidence in California ight attendants (United States). Cancer Causes Control 2002;13: 31724. Zeeb H, Blettner M, Langner I, et al. Mortality from cancer and other causes among airline cabin attendants in Europe: a collaborative cohort study in eight countries. Am J Epidemiol 2003;158:3546. Langner I, Blettner M, Gundestrup M, et al. Cosmic radiation and cancer mortality among airline pilots: results from a European cohort study (ESCAPE). Radiat Environ Biophys 2004;42: 24756. Blettner M, Zeeb H, Auvinen A, et al. Mortality from cancer and other causes among male airline cockpit crew in Europe. Int J Cancer 2003;106: 94652. Diffey BL, Roscoe AH. Exposure to solar ultraviolet radiation in ight. Aviat Space Environ Med 1990;61:10325. Pizzi C, Evans SA, De Stavola BL, et al. Lifestyle of UK commercial aircrews relative to air trafc controllers and the general population. Aviat Space Environ Med 2008;79:96474. dos Santos Silva I, Pizzi C, De Stavola B, et al. Reproductive history and adverse reproductive

15.

16.

17.

18.

19.

20.

outcomes in UK commercial aircrews and air trafc controllers. J Occup Environ Med 2009;51: 1298305. De Stavola BL, Pizzi C, Clemens F, et al. Causespecic mortality in professional aircrew and air trafc control ofcers: ndings from a UK population-based cohort of over 20,000 subjects. Int Arch Occup Environ Health 2012;85:28393. World Health Organisation (WHO).International classication of diseases, 9th revision.Geneva: WHO, 1977. World Health Organisation (WHO).International statistical classication of diseases and related health problems, 10th revision.Geneva:WHO, 1992. Breslow NE, Day NE.Statistical methods in cancer research, vol. II: The design and analysis of cohort studies. Scientic Publications No. 82.Lyon:International Agency for Research on Cancer (IARC), 1987. Clayton D, Hills, M.! Statistical models in epidemiology. Oxford: Oxford University Press, 1993. Schernhammer ES, Razavi P, Li TY, et al. Rotating night shifts and risk of skin cancer in the nurses health study. J Natl Cancer Inst 2011; 103:18.

C Int. J. Cancer: 132, 374384 (2013) V 2012 UICC