- Email: [email protected]
Cancer incidence in Dutch Balkan veterans
Cancer Epidemiology 37 (2013) 550–555
Contents lists available at SciVerse ScienceDirect
Cancer Epidemiology The International Journal of Cancer Epidemiology, Detection, and Prevention journal homepage: www.cancerepidemiology.net
Cancer incidence in Dutch Balkan veterans Rik P. Bogers a,*, Flora E. van Leeuwen b, Linda Grievink a, Leo J. Schouten c, Lambertus A.L.M. Kiemeney d, Dieneke Schram-Bijkerk a a
National Institute for Public Health and the Environment, PO Box 1, 3720 BA Bilthoven, The Netherlands Department of Epidemiology, The Netherlands Cancer Institute, PO Box 90203, 1006 BE Amsterdam, The Netherlands c Department of Epidemiology, GROW – School for Oncology and Developmental Biology, Maastricht University, PO Box 616, 6200 MD Maastricht, The Netherlands d Radboud University Nijmegen Medical Centre, PO Box 9101, 6500 HB Nijmegen, The Netherlands b
A R T I C L E I N F O
A B S T R A C T
Article history: Received 10 January 2013 Received in revised form 17 April 2013 Accepted 18 April 2013 Available online 22 May 2013
Suspicion has been raised about an increased cancer risk among Balkan veterans because of alleged exposure to depleted uranium. The authors conducted a historical cohort study to examine cancer incidence among Dutch Balkan veterans. Male military personnel (n = 18,175, median follow-up 11 years) of the Army and Military Police who had been deployed to the Balkan region (1993–2001) was compared with their peers not deployed to the Balkans (n = 135,355, median follow-up 15 years) and with the general Dutch population of comparable age and sex. The incidence of all cancers and 4 main cancer subgroups was studied in the period 1993–2008. The cancer incidence rate among Balkan deployed military men was 17% lower than among non-Balkan deployed military men (hazard ratio 0.83 (95% confidence interval 0.69, 1.00)). For the 4 main cancer subgroups, hazard ratios were statistically non-significantly below 1. Also compared to the general population cancer rates were lower in Balkan deployed personnel (standardised incidence rate ratio (SIR) 0.85 (0.73, 0.99). The SIR for leukaemia was 0.63 (0.20, 1.46). The authors conclude that earlier suggestions of increased cancer risks among veterans are not supported by empirical data. The lower risk of cancer might be explained by the ‘healthy warrior effect’. ß 2013 Elsevier Ltd. All rights reserved.
Keywords: Cohort studies Historical Incidence Military personnel Neoplasms Occupational exposure
1. Introduction In 2000 the international lay-press reported on cases of leukaemia among soldiers or peacekeepers who had been deployed to the Balkans, i.e. former Yugoslavia. A suspected cause was alleged exposure to depleted uranium (DU), a radioactive and chemically toxic material originating from DU-containing ammunition used by NATO (North Atlantic Treaty Organisation) Forces. The media reports prompted questions in various NATO countries about the occurrence of leukaemia and its relationship with Balkan deployment and DU-exposure in particular. It was therefore recommended by the Committee of the Chiefs of Military Medical Services in NATO (COMEDS) that the NATO countries would study cancer risk among Balkan veterans [1]. Studies from Italy [2], Sweden [3], and Denmark [4] did not demonstrate a higher than expected risk of all cancers including
Abbreviations: 95% CI, 95% confidence interval; DU, depleted uranium; HR, hazard ratio; NATO, North Atlantic Treaty Organisation; SIR, standardised incidence rate ratio; n, number. * Corresponding author. Tel.: +31 30 2742150; fax: +31 30 2744451. E-mail address: [email protected] (R.P. Bogers). 1877-7821/$ – see front matter ß 2013 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.canep.2013.04.005
leukaemia among Balkan veterans compared to the general population. However, the follow-up periods in these studies of about 10 years were still short considering the latency time of cancer development by ionising radiation. In addition, since lifestyle factors between the military and general population differ [5], the results of these studies were not convincing and suspicions remained. The present study includes an additional comparison group of non-deployed military, has a large study population, and uses a longer observation period of maximal 15 years. Others have studied cancer in Gulf War veterans, which might be of particular relevance here because questions also arose on health risks of exposure to DU from ammunition deployed during the Gulf War. A study on cancer incidence among UK Gulf veterans (Gulf cohort) up to 11 years since the end of the war showed no increased overall or site-specific cancer incidence as compared with service personnel not deployed to the Gulf War back then [6]. A report of the US Committee on Gulf War Veterans’ Illnesses [7] concluded that depleted uranium was not likely to have caused Gulf War Illness. The aim of the present study was to describe cancer incidence including haematological malignancies up to 2008 among Dutch military personnel deployed to the Balkans between 1993 and
R.P. Bogers et al. / Cancer Epidemiology 37 (2013) 550–555
2001. This epidemiological study does not address any specific causative factors, such as exposure to DU, but provides evidence intended to evaluate public health and policy concerns. Comparisons were made with a cohort of military personnel who were in service during the same time period but who were not deployed to the Balkans. We also compared cancer incidence in Balkan deployed military personnel with the general population of comparable age and sex. 2. Materials and methods 2.1. Study design The study was a historical cohort study, i.e. deployment of military personnel was identified from information recorded in the past. Cancer incidence rates in the period 1993–2008 were obtained through record linkage with the national cancer registry, and rates in Balkan-deployed personnel were compared with rates in personnel not deployed to the Balkans and the general Dutch population of comparable sex and age. 2.2. Study population The cohort was restricted to men who were in service any time between January 1, 1993 and March 1, 2001, who worked in the Army or Military Police. Personnel from the Navy and Air Force could not be included because complete historical employee records were not available. However, since personnel of the Air Force and most Navy personnel were only minimally involved in ground missions, exposure to DU was unlikely. Civil personnel was excluded. The (ex-)military personnel could refuse to participate by sending back a no-consent form. Ninety-seven percent of all military personnel could be traced, and 2% of them (3261 out of 160 737) returned their no-consent form. Women (n = 3946) were excluded because the number of women was too small to perform meaningful analyses. The final cohort included 153 530 men. All men ever been deployed for at least 28 days to the Balkans between 1993 and March 2001 were included in the ‘Balkan deployed’ group (n = 18,175). Those deployed for less than 28 days were included in the reference population in order to exclude short working visits. Those who died during Balkan deployment were excluded, because the study question concerns post-deployment cancer incidence. The first deployments to the Balkans started in 1991, but employee records were only available from 1993 onwards. However, there is no information showing that DUcontaining ammunition was used in the Balkans before 1993, and given the type of the early missions (monitoring), exposure to DU is unlikely. Moreover, the number of deployed personnel before 1993 is considered to be small. 2.3. Data sources Available data included dates of entrance into service and termination of service, branch of military service, rank, appointment, deployments (name of mission, start- and end date), sex, year of birth, and personal details (for record linkages only – see below). Cancer incidence data were collected for the years 1993– 2008 from the Netherlands Cancer Registry (NCR) held by the Association of Comprehensive Cancer Centres (IKNL&IKZ). 2.4. Record linkage with cancer incidence registry Addresses, place of birth and vital status by January 1, 2008 and unique, anonymous, Dutch personal ID numbers given to all Dutch citizens (municipal ‘A-numbers’) were obtained through record linkage with the Computerised Municipal Population Registry
551
using name, birth date and, if necessary, (old) addresses as identifiers. Using sex, name and date of birth, record linkage with the NCR (http://www.kankerregistratie.nl) was performed. Addresses and place of birth, and if applicable, date of death, were used to check the matches. The registry contains data on all new cancer cases in the Netherlands; the sensitivity of record linkage has been reported to be 98% [8]. Observed cancer cases were extracted from the registry and recoded from the International Classification of Diseases for Oncology (ICD-O) into the general International Classification of Diseases (ICD-10). Only invasive cancers were included, as well as non-invasive bladder cancers. Only the first primary cancers were included for incidence determination, because subsequent cancers can be caused by treatment of the first primary. Because modest numbers of cancer cases were expected, cancer types were categorised into large cancer subgroups. The following main cancer subgroups were studied (ICD-10 codes in parentheses): -
digestive organs (C15–C26); respiratory system and intrathoracic organs (C30–C39); bronchus and lung (C34); urinary tract and genital organs (C60–C68); haematolymphopoetic system (including leukaemia; C81–C85, C88, C91–C96); - leukaemia (C91–C96). The classification we used is a standard one when not enough cases of individual cancer sites are observed to examine cancer risks for all individual organs and tissues separately. We categorised the sites according to tracts (digestive tract, respiratory tract, urogenital tract) or tissues (haematolymphopoetic system), which is often done in epidemiologic studies because the etiologic factors for sites within a tract are assumed to be more similar. Lung cancer was considered separately because of specific concerns about lung cancer in relation to inhaled DU. 2.5. Data analysis All statistical analyses were performed using SPSS for Windows version 14.0. 2.5.1. Differences between Balkan deployed military and their peers Differences in characteristics between Balkan- and non-Balkan deployed personnel were tested with a Chi-square or Wilcoxon test, as appropriate. To study differences in cancer incidence between Balkan- and non-Balkan deployed military, proportional hazard models were used with age defined as a follow-up variable. Time under study (follow-up) was included as a time-dependent variable. In this way, differences in follow-up time between groups were taken into account. Time under study started at the beginning of the study (January 1, 1993) or at entrance into duty (if > January 1, 1993) and ended at death, emigration, diagnosis of cancer or the end of the study (January 1, 2008). Deployment to the Balkans was included as a separate time-dependent variable, depending on end date of first deployment. Thus, person-time and cases between entrance in the cohort and end of first Balkan deployment contributed to the non-Balkan deployed group (see Fig. 1). Analyses were adjusted for branch of military service (Army or Military Police), rank (commissioned or non-commissioned officers) and appointment (permanent, temporary or conscription). Branch of military service was included as a (crude) proxy of type of work during deployment (and thereby possibly occupational exposures). Military rank and type of appointment were used as crude measures of socioeconomic status in the proportional hazard models. Those with missing information on any of
R.P. Bogers et al. / Cancer Epidemiology 37 (2013) 550–555
552
Time at risk
Control time
1, 1, 1993 1st deployment
2nd deployment
1, 1, 2008
Entrance into duty Fig. 1. Calculation of time at risk for a virtual Balkan deployed soldier who was deployed to the Balkans twice.
these variables (n = 18) were excluded from the proportional hazards analysis.
3. Results 3.1. Characteristics of the study population
2.5.2. Comparison with Dutch population Expected numbers of cases of cancer in the study population were estimated according to the person-year method, based on cancer incidence rates for the general population in the Netherlands, stratified by age (five year classes), sex and time period (two or three year classes). Standardised incidence rate ratios (SIRs) were computed by dividing the observed numbers by the expected numbers, using a statistical programme written in SPSS developed by the Netherlands Cancer Institute [9]. Confidence intervals were calculated using the Poisson distribution. Cancer incidence rates in the general population, from 1996 to 2007, were obtained from the Netherlands Cancer Registry. Cancer incidence rates between 1993 and 1996 were available, but coded according to ICD-9, which does not completely match with coding of some specific tumours in ICD-10 after 1996. Therefore, rates in the Dutch population between 1993 and 1996 were not used. For those cancers which occurred between 1993 and 1996 in the study population (ca. 10%), rates in the Dutch population in 1996 were used as a reference instead. 2.5.3. Sensitivity analyses All analyses were also conducted without conscripts, because conscription was abolished in the Netherlands in 1996. Also, because conscripts are relatively young, rank may be a less stable indicator for socioeconomic status in this group. Proportional hazard models were also performed without cases of cancer that occurred within 18 months from the end date of the mission, as most cancers have a longer latency period. SIRs were also calculated for three 5-year periods separately to study possible trends over time.
The study population included 18,175 men deployed to the Balkans between 1993 and 2001, and 135,355 non-Balkan deployed men, with median follow-up times of 10.9 years and 14.8 years, respectively. As follow-up of Balkan deployed personnel started later than January 1, 1993, i.e. after the date of first Balkan deployment, 60% could be followed for at least 10 years, but by definition, no one could be followed for the complete follow-up period of 15 years. Five percent and 1% of the Balkan and nonBalkan deployed personnel had been deployed to other areas between 1993 and 2001. Table 1 presents general characteristics of Balkan and nonBalkan deployed military personnel. Most of them worked in the Army. Whereas the majority of the non-Balkan deployed personnel were conscripts, Balkan deployed personnel consisted mostly of commissioned officers and only 15% conscripts. The median age of both groups was 21 years in 1993. At the end of the study period (1 January 2008), 13% and 5% of Balkan and non-Balkan deployed personnel, respectively, was older than 50. The majority (77%) of the Balkan deployed military personnel was deployed only once (Table 1). Almost all deployments lasted approximately 6 months. 3.2. Cancer incidence 3.2.1. Comparison between Balkan deployed military and their peers In Balkan deployed and non-Balkan deployed men, 175 and 1337 new cases of cancer, respectively, were observed during 15 years of follow-up (until 2008). The cases among Balkan deployed personnel that were not included in the cancer subgroups were
Table 1 General characteristics of male military personnel deployed and non-deployed to the Balkans. Balkan deployeda (n = 18,175) b
Median age at January 1, 1993 (p25–p75 ) Rank (% commissioned officers)c Branch (%) Army Military Police Appointment (%) Permanent Temporary Conscription Other (chaplains)d Duration of 1st deployment (median number of days (p25-p752))e Number of Balkan deployments 1 (%) 2 (%) 3 or more (%)
Non-Balkan deployed (n = 135,355)
21 (18–28) 13.1
21 (19–23) 5.3
94.0 6.0
94.9 5.1
37.2 47.1 15.3 0.3 181 (166–185) 77.1 19.1 3.8
8.9 10.3 80.8 0.05 – – – –
a All characteristics were significantly different between Balkan vs. non-Balkan deployed military at p < 0.01. This was tested by using Chi-Square (sex, rank, appointment, branch of military service) and Wilcoxon statistics (mean age)). b 25th and 75th percentiles. c Rank was available for n = 18,157 Balkan deployed and 128,077 non-Balkan deployed men. d Categorised in the analyses as permanent appointment. e Start date of deployment not available for n = 20.
R.P. Bogers et al. / Cancer Epidemiology 37 (2013) 550–555 Table 2 Hazard ratios of total and site-specific incidence of cancer in Dutch male military personnel deployed to the Balkans (n = 18,175) a compared with non-Balkan deployed personnel (n = 135,355). Cancer site
ICD-10 codes
HR (95% CI)b
All cancers Digestive organs Respiratory system and intrathoracic organs Bronchus and lung Urinary tract and genital organs Haematolymphopoetic system
C00–C96c C15–C26 C30–C39
0.83 (0.69–1.00) 0.85 (0.54–1.33) 0.88 (0.44–1.76)
C34 C60–C68 (C81–C85, C88, C91–C96)
0.91 (0.45–1.90) 0.90 (0.65–1.25) 0.76 (0.47–1.23)
a 18 Balkan deployed men were excluded because of missing information on covariates. b Hazard ratios (95% confidence intervals) adjusted for rank, appointment and branch of military service. c C44 (skin, other/NOS) included; C44 basal cell carcinoma excluded.
mainly skin cancers. The total cancer incidence rate among Balkan deployed personnel was 17% lower than among non-Balkan deployed personnel (hazard ratio (HR) = 0.83, 95% confidence interval (CI) 0.69–1.00, p = 0.05; see Table 2). For the 4 main cancer subgroups and lung cancer, rates were lower among Balkan deployed men, although the differences were not statistically significant. Leukaemia was included in the subgroup haematological cancers, but the number of Balkan and non-Balkan-deployed men with leukaemia (n = 5 and n = 56, respectively) was too low to calculate a meaningful HR for leukaemia separately. Results without adjustment for rank, appointment and branch of military service were essentially similar (not shown). The number of deployments did not seem to affect the number of cancer cases in the Balkan-deployed group. Sensitivity analyses without conscripts (excluding 2788 persons) or cases of cancer that occurred
553
within 18 months after the mission (excluding 20 cases) yielded similar results (see supplement). 3.2.2. Comparison with the Dutch population The observed number of new cancer cases (n = 175) among Balkan deployed military personnel was 15% lower than the expected number of cases (nexp = 205) in the age- and sex-matched Dutch population (SIR = 0.85 (0.73–0.99); see Table 3). The SIR for all cancers combined was close to unity in the first 5 years of follow-up and then declined to 0.65 (0.43–0.95) in the last 5 years of follow-up. For the 4 main cancer subgroups and lung cancer, observed numbers of cancer cases among Balkan deployed men were lower than the expected number. These differences were statistically significant for tumours of the respiratory system and intrathoracic organs, and lung cancer. The observed number of leukaemia cases was lower than expected but not statistically significantly (SIR = 0.63 (0.20–1.46)). Sensitivity analyses without conscripts yielded similar results (not shown). 4. Discussion This study showed that the cancer incidence in Balkan deployed men was 17% lower than in non-Balkan deployed men in the follow-up period of 15 years, and 15% lower than expected from incidence rates in the general Dutch population of comparable age and sex. Cancer risks were also somewhat lower for the main cancer site subgroups separately. There was no indication of an increased risk of leukaemia among Balkan deployed personnel. Although we do not have any data on DU exposure during Balkan deployment, our results do not substantiate concerns about health effects of DU. This is in line with a desk top study about the fate of and potential exposure to DU in the 1999 Kosovo War, which concluded that health effects of DU exposure of Balkan deployed Dutch soldiers are unlikely [10].
Table 3 Cancer incidence among Dutch military men deployed to the Balkans (n = 18,175) as compared to the age- and sex-matched Dutch population, overall and for three follow-up intervals separately. Cancer site and follow-up period All cancers Total period 0–5 y 5–10 y 10–15 y Digestive organs Total period 0–5 y 5–10 y 10–15 y Respiratory system and intrathoracic organs Total period 0–5 y 5–10 y 10–15 y Bronchus and lung total periodc Urinary tract and genital organs Total period 0–5 y 5–10 y 10–15 y Haematolymphopoetic system Total period 0–5 y 5–10 y 10–15 y Leukaemia Total periodc a b c
ICD-10 codes
Observed
Expected
SIR (95% CI)a
175 74 74 27
205.3 70.7 93.3 41.3
0.85 1.05 0.79 0.65
(0.73–0.99) (0.82–1.31) (0.62–1.00) (0.43–0.95)
30 7 16 7
36.5 11.1 17.1 8.2
0.82 0.63 0.93 0.85
(0.56–1.17) (0.25–1.30) (0.53–1.52) (0.34–1.76)
12 8 3 1 10
25.0 8.0 11.6 5.5 20.8
0.48 1.00 0.26 0.18 0.48
(0.25–0.84) (0.43–1.97) (0.05–0.76) (0.00–1.01) (0.23–0.89)
55 22 27 6
54.1 17.9 25.1 11.2
1.02 1.23 1.08 0.54
(0.77–1.32) (0.77–1.86) (0.71–1.57) (0.20–1.17)
28 13 8 7
29.3 11.8 12.6 4.9
0.95 1.10 0.63 1.43
(0.63–1.38) (0.59–1.88) (0.27–1.25) (0.58–2.96)
5
8.0
b
C00–C96
C15–C26
C30–C39
C34 C60–C68
C81–C96 (C81–C85, C88, C91–C96)
C91–C96
Standardised incidence rate ratio. C44 (Skin, other/NOS) included; C44 basal cell carcinoma excluded. Total number too low to stratify analyses by follow-up period (0–5, 5–10, 10–15 years).
0.63 (0.20–1.46)
R.P. Bogers et al. / Cancer Epidemiology 37 (2013) 550–555
554
Table 4 Previously reported studies on cancer incidence among male UN Balkan veterans. Study design
Study population
Follow-up time (years)
Outcome
Obs
Exp
SIR
95% CI
Netherlands (1993–2008)
Population-based historical cohort study
11 (median)
Cancer incidence (registry-based)
175
205.3
0.85
0.73–0.99
Denmark (1992–2001) (4)
Population-based historical cohort study
11 (maximum)
Cancer incidence (registry-based)
84
n.r.
0.9
0.7–1.1
Sweden (1989–1999) (3)
Population-based historical cohort study
10 (maximum)
Cancer incidence (registry-based)
26
21.8
1.2
0.8–1.8
Italy (1996–2007) (2)
Historical cohort study using number of persons instead of person-years
Military personnel deployed between 1993 and March 2001 (n = 18,175 men) Military personnel deployed between 1992 and 2001 (n = 13,552 men) Military personnel deployed between 1989 and 1999 (n = 8347 men) Military personnel deployed between 1996 and 2007 (n = 49,862 men)
11 (maximum)
Cancer incidence based on preliminary army cancer surveillance activity
344.9
0.48
164
0.39–0.57
n.r.: not reported.
The results of this study are in line with previous studies on cancer incidence in Balkan deployed military personnel. Studies in Italy [2], Sweden [3], and Denmark [4] did not demonstrate a higher than expected risk of all cancers including leukaemia among Balkan veterans compared to the general population (see Table 4). However, the follow-up periods in these studies of about 10 years were still short considering the latency time of cancer development by ionising radiation. The study in Italy was hampered by the fact that the geographical distribution of cancer registries is not homogeneous across Italy and the coverage of the resident population is low [2]. However, new estimates taking the low completeness into account showed similar results [11]. Previous studies did not include analyses comparing Balkan deployed with non-Balkan deployed military personnel. Such analyses are important because there are differences in lifestyle factors between the military and general population [5]. Our study included non-deployed personnel, had a relatively large study population and a long follow-up period. The number of personyears in our study is more than twice of those in the Danish study and almost five times those of the Swedish study. Therefore, our study was able to produce more precise and less biased estimates of the relative risks of cancer and allows for longer latency times than previous studies. The lower cancer incidence among Balkan veterans in the present study can be explained by the ‘healthy warrior effect’, an analogy of the ‘healthy worker effect’ or ‘healthy soldier effect’. These terms refer to the observation that the working population is healthier than the general population on a whole. The healthy worker effect was probably first described by William Ogle at the end of the 19th century [12]. This effect is relatively strong for military personnel, due to health examinations at the time of enlistment and the physical demands of the profession, which reduce the risk of some site-specific cancers [13]. The healthy warrior effect results from selection at the time of deployment, as those who are ill are less likely to be deployed [14]. Nevertheless, the lower risk of developing tumours in respiratory organs in Balkan deployed military personnel (and non-Balkan deployed personnel; SIR = 0.64 (95% CI 0.51–0.79)) as compared to the Dutch population was against our expectations. Data on smoking habits were not available for this study population, and we do not know the cause of the lower risk. It was generally assumed that smoking (an important risk factor for respiratory tumours) is more common in the military population [5] than in the general population. Latency time should be taken into account when considering the
risk of developing lung cancer, but because it is likely that most smokers in the study population started smoking many years before the onset of the study, an effect of smoking on the development of lung cancer should have been detected. However, our cohort was relatively young, and lung cancer in men is most observed among 65–80 year olds. Another explanation might be the relatively high amount of physical activity among military personnel [5]. However, despite strong evidence for a protective role of physical activity on cancers of the breast and endometrium, for lung cancer the evidence is weaker [15]. Almost the complete cohort could be retrieved (97%), and highly reliable registration data were available on cancer incidence. Bias by missing persons in the cohort because of the no-consent procedure is unlikely because the 2% personnel (all non-Balkan deployed) that returned their no-consent card had similar characteristics (rank, appointment, age) to those included in the cohort. A limitation of the study is that no data were available on (individual) exposure to DU in the Balkans and on individual risk factors for the development of cancer (such as smoking, alcohol consumption, and job history after deployment to the Balkans). Also, late effects of Balkan deployment on the development of cancer cannot be excluded, as many cancers do not become apparent until 10–20 years after an initiating event and the study population was relatively young. For acute leukaemia, however, which has frequently been associated with radiation, the latency time is relatively short [16] and our results should be more conclusive. Finally, in theory an adverse effect of deployment could remain unnoticed because of the generally better health of deployed soldiers. Ideally, epidemiological studies should be planned before military action is contemplated to avoid difficulties in assessing exposures and health effects afterwards [17]. It then would be feasible to collect additional data on e.g., potentially hazardous exposures during deployment and lifestyle factors such as smoking habits and physical activity. In conclusion, the present study provides the best evidence against an increased risk of cancer following Balkan deployment to date. Role of the funding source This work was supported by the Dutch Ministry of Defence. The funder contributed to the collection of data from registers of the Dutch Armed Forces. All authors were independent of the funder.
R.P. Bogers et al. / Cancer Epidemiology 37 (2013) 550–555
Conflict of interest The authors declare that they have no conflict of interest. Acknowledgements No ethics approval was required for this study. Handling of personal data complied with the Personal Data Protection Act and the Municipal Database Act. Operating procedures for confidential and careful handling of data were described in detail in a privacy agreement between the Ministry of Defence, the Netherlands Cancer Registry (NCR) and the research institute (RIVM). The authors are grateful to Marc Ruijten, Jeanne van Loon, and Diane Houweling, who coordinated the project during different time periods, Bas Verhage, Jeroen de Hartog and Leontien Korteweg, who contributed to the data cleaning and analysis. From RIVM, also Ellis Franssen, Ingrid van Kuilenburg, Adriaan van Kessel, Bennie Bloemberg, Ben Bom, Dirk-Jan Griffioen, Wil Lokhorst, Caroline Ameling and Hendriek Boshuizen contributed to this study. Otto Visser reviewed the study as a member of the initial Scientific Advisory Board. The Netherlands Cancer Institute provided the computer programme for indirect standardisation. We would also like to thank the persons at the Ministry of Defence who contributed to this study; in particular Ad de Koning, Angelique Hardij, Tjalling Leenstra and Peter Krab, who was the manager of the Veteran Registration System. Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/j.canep.2013.04.005. References [1] COMEDS Meeting on Health Concerns Related to the Balkan Deployments, Brussels [updated 13.02.01, cited April 2013]; available from: http://www.nato.int/du/docu/d010115a.htm.
555
[2] Peragallo MS, Lista F, Sarnicola G, Marmo F, Vecchione A. Cancer surveillance in Italian army peacekeeping troops deployed in Bosnia and Kosovo, 1996–2007: preliminary results. Cancer Epidemiol 2010;34(1):47–54. [3] Gustavsson P, Talback M, Lundin A, Lagercrantz B, Gyllestad PE, Fornell L. Incidence of cancer among Swedish military and civil personnel involved in UN missions in the Balkans 1989–1999. Occup Environ Med 2004;61(2):171–3. [4] Storm HH, Jorgensen HO, Kejs AM, Engholm G. Depleted uranium and cancer in Danish Balkan veterans deployed 1992–2001. Eur J Cancer 2006;42(14):2355–8. [5] Bray RM, Hourani LL, Olmsted KLR, Witt M, Brown JM, Pemberton MR, et al. Department of defense survey of health related behaviors among active duty military personnel. A component of the defense lifestyle assessment program (DLAP). Research Triangle Park, NC: RTI International2006, 2005, Contract no.: RTI/7841/106-FR. [6] Macfarlane GJ, Biggs AM, Maconochie N, Hotopf M, Doyle P, Lunt M. Incidence of cancer among UK Gulf war veterans: cohort study. Br Med J 2003;327(7428): 1373. [7] Research Advisory Committee on Gulf War Veterans’ Illnesses. Gulf War illness and the health of Gulf War veterans: scientific findings and recommendations. Washington, DC: U.S. Department of Veterans Affairs, 2008. [8] van den Brandt PA, Schouten LJ, Goldbohm RA, Dorant ELIS, Hunen PMH. Development of a record linkage protocol for use in the Dutch Cancer Registry for epidemiological research. Int J Epidemiol 1990;19(3):553–8. [9] van Leeuwen FE, Klokman WJ, Hagenbeek A, Noyon R, van den Belt-Dusebout AW, van Kerkhoff EH, et al. Second cancer risk following Hodgkin’s disease: a 20-year follow-up study. J Clin Oncol 1994;12(2):312–25. [10] Busker RW, van Ham NHA, van’t Hof G, Langeberg JP, van der Meer BJ, Nieuwenhuizen MS. Fate of depleted uranium in the 1999 Kosovo war: unlikely relation with health effects of Dutch military personnel. Rijswijk: Netherlands Organisation for Applied Scientific Research (TNO), 2005, Report No.: PML 2002-A16. [11] Peragallo MS, Urbano F, Lista F, Sarnicola G, Vecchione A. Evaluation of cancer surveillance completeness among the Italian army personnel, by capture– recapture methodology. Cancer Epidemiol 2011;35(2):132–8. [12] Ogle W. Letter to the Registrar-General on the mortality in the registration districts of England and Wales during the ten years 1871–80. Supplement to the 45th annual report of the Registrar General of Births, Deaths, and Marriages, in England.; 1885. [13] Lee I, Oguma Y. Physical activity. In: Schottenfeld D, Fraumeni JF, eds. Cancer epidemiology and prevention. 3rd ed., Oxford: Oxford University Press, 2006: 449–68. [14] Haley RW. Point: bias from the ‘‘healthy-warrior effect’’ and unequal followup in three government studies of health effects of the Gulf War. Am J Epidemiol 1998;148(4):315–23. [15] Friedenreich CM, Neilson HK, Lynch BM. State of the epidemiological evidence on physical activity and cancer prevention. Eur J Cancer 2010;46(14):2593– 604. [16] Boice Jr JD. Ionizing radiation. Cancer epidemiology and prevention. 3rd ed. Oxford: Oxford University Press, 2006. 259–93. [17] Hotopf M, Wessely S. Can epidemiology clear the fog of war? Lessons from the 1990–91 Gulf War. Int J Epidemiol 2005;34(4):791–800.
Contents lists available at SciVerse ScienceDirect
Cancer Epidemiology The International Journal of Cancer Epidemiology, Detection, and Prevention journal homepage: www.cancerepidemiology.net
Cancer incidence in Dutch Balkan veterans Rik P. Bogers a,*, Flora E. van Leeuwen b, Linda Grievink a, Leo J. Schouten c, Lambertus A.L.M. Kiemeney d, Dieneke Schram-Bijkerk a a
National Institute for Public Health and the Environment, PO Box 1, 3720 BA Bilthoven, The Netherlands Department of Epidemiology, The Netherlands Cancer Institute, PO Box 90203, 1006 BE Amsterdam, The Netherlands c Department of Epidemiology, GROW – School for Oncology and Developmental Biology, Maastricht University, PO Box 616, 6200 MD Maastricht, The Netherlands d Radboud University Nijmegen Medical Centre, PO Box 9101, 6500 HB Nijmegen, The Netherlands b
A R T I C L E I N F O
A B S T R A C T
Article history: Received 10 January 2013 Received in revised form 17 April 2013 Accepted 18 April 2013 Available online 22 May 2013
Suspicion has been raised about an increased cancer risk among Balkan veterans because of alleged exposure to depleted uranium. The authors conducted a historical cohort study to examine cancer incidence among Dutch Balkan veterans. Male military personnel (n = 18,175, median follow-up 11 years) of the Army and Military Police who had been deployed to the Balkan region (1993–2001) was compared with their peers not deployed to the Balkans (n = 135,355, median follow-up 15 years) and with the general Dutch population of comparable age and sex. The incidence of all cancers and 4 main cancer subgroups was studied in the period 1993–2008. The cancer incidence rate among Balkan deployed military men was 17% lower than among non-Balkan deployed military men (hazard ratio 0.83 (95% confidence interval 0.69, 1.00)). For the 4 main cancer subgroups, hazard ratios were statistically non-significantly below 1. Also compared to the general population cancer rates were lower in Balkan deployed personnel (standardised incidence rate ratio (SIR) 0.85 (0.73, 0.99). The SIR for leukaemia was 0.63 (0.20, 1.46). The authors conclude that earlier suggestions of increased cancer risks among veterans are not supported by empirical data. The lower risk of cancer might be explained by the ‘healthy warrior effect’. ß 2013 Elsevier Ltd. All rights reserved.
Keywords: Cohort studies Historical Incidence Military personnel Neoplasms Occupational exposure
1. Introduction In 2000 the international lay-press reported on cases of leukaemia among soldiers or peacekeepers who had been deployed to the Balkans, i.e. former Yugoslavia. A suspected cause was alleged exposure to depleted uranium (DU), a radioactive and chemically toxic material originating from DU-containing ammunition used by NATO (North Atlantic Treaty Organisation) Forces. The media reports prompted questions in various NATO countries about the occurrence of leukaemia and its relationship with Balkan deployment and DU-exposure in particular. It was therefore recommended by the Committee of the Chiefs of Military Medical Services in NATO (COMEDS) that the NATO countries would study cancer risk among Balkan veterans [1]. Studies from Italy [2], Sweden [3], and Denmark [4] did not demonstrate a higher than expected risk of all cancers including
Abbreviations: 95% CI, 95% confidence interval; DU, depleted uranium; HR, hazard ratio; NATO, North Atlantic Treaty Organisation; SIR, standardised incidence rate ratio; n, number. * Corresponding author. Tel.: +31 30 2742150; fax: +31 30 2744451. E-mail address: [email protected] (R.P. Bogers). 1877-7821/$ – see front matter ß 2013 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.canep.2013.04.005
leukaemia among Balkan veterans compared to the general population. However, the follow-up periods in these studies of about 10 years were still short considering the latency time of cancer development by ionising radiation. In addition, since lifestyle factors between the military and general population differ [5], the results of these studies were not convincing and suspicions remained. The present study includes an additional comparison group of non-deployed military, has a large study population, and uses a longer observation period of maximal 15 years. Others have studied cancer in Gulf War veterans, which might be of particular relevance here because questions also arose on health risks of exposure to DU from ammunition deployed during the Gulf War. A study on cancer incidence among UK Gulf veterans (Gulf cohort) up to 11 years since the end of the war showed no increased overall or site-specific cancer incidence as compared with service personnel not deployed to the Gulf War back then [6]. A report of the US Committee on Gulf War Veterans’ Illnesses [7] concluded that depleted uranium was not likely to have caused Gulf War Illness. The aim of the present study was to describe cancer incidence including haematological malignancies up to 2008 among Dutch military personnel deployed to the Balkans between 1993 and
R.P. Bogers et al. / Cancer Epidemiology 37 (2013) 550–555
2001. This epidemiological study does not address any specific causative factors, such as exposure to DU, but provides evidence intended to evaluate public health and policy concerns. Comparisons were made with a cohort of military personnel who were in service during the same time period but who were not deployed to the Balkans. We also compared cancer incidence in Balkan deployed military personnel with the general population of comparable age and sex. 2. Materials and methods 2.1. Study design The study was a historical cohort study, i.e. deployment of military personnel was identified from information recorded in the past. Cancer incidence rates in the period 1993–2008 were obtained through record linkage with the national cancer registry, and rates in Balkan-deployed personnel were compared with rates in personnel not deployed to the Balkans and the general Dutch population of comparable sex and age. 2.2. Study population The cohort was restricted to men who were in service any time between January 1, 1993 and March 1, 2001, who worked in the Army or Military Police. Personnel from the Navy and Air Force could not be included because complete historical employee records were not available. However, since personnel of the Air Force and most Navy personnel were only minimally involved in ground missions, exposure to DU was unlikely. Civil personnel was excluded. The (ex-)military personnel could refuse to participate by sending back a no-consent form. Ninety-seven percent of all military personnel could be traced, and 2% of them (3261 out of 160 737) returned their no-consent form. Women (n = 3946) were excluded because the number of women was too small to perform meaningful analyses. The final cohort included 153 530 men. All men ever been deployed for at least 28 days to the Balkans between 1993 and March 2001 were included in the ‘Balkan deployed’ group (n = 18,175). Those deployed for less than 28 days were included in the reference population in order to exclude short working visits. Those who died during Balkan deployment were excluded, because the study question concerns post-deployment cancer incidence. The first deployments to the Balkans started in 1991, but employee records were only available from 1993 onwards. However, there is no information showing that DUcontaining ammunition was used in the Balkans before 1993, and given the type of the early missions (monitoring), exposure to DU is unlikely. Moreover, the number of deployed personnel before 1993 is considered to be small. 2.3. Data sources Available data included dates of entrance into service and termination of service, branch of military service, rank, appointment, deployments (name of mission, start- and end date), sex, year of birth, and personal details (for record linkages only – see below). Cancer incidence data were collected for the years 1993– 2008 from the Netherlands Cancer Registry (NCR) held by the Association of Comprehensive Cancer Centres (IKNL&IKZ). 2.4. Record linkage with cancer incidence registry Addresses, place of birth and vital status by January 1, 2008 and unique, anonymous, Dutch personal ID numbers given to all Dutch citizens (municipal ‘A-numbers’) were obtained through record linkage with the Computerised Municipal Population Registry
551
using name, birth date and, if necessary, (old) addresses as identifiers. Using sex, name and date of birth, record linkage with the NCR (http://www.kankerregistratie.nl) was performed. Addresses and place of birth, and if applicable, date of death, were used to check the matches. The registry contains data on all new cancer cases in the Netherlands; the sensitivity of record linkage has been reported to be 98% [8]. Observed cancer cases were extracted from the registry and recoded from the International Classification of Diseases for Oncology (ICD-O) into the general International Classification of Diseases (ICD-10). Only invasive cancers were included, as well as non-invasive bladder cancers. Only the first primary cancers were included for incidence determination, because subsequent cancers can be caused by treatment of the first primary. Because modest numbers of cancer cases were expected, cancer types were categorised into large cancer subgroups. The following main cancer subgroups were studied (ICD-10 codes in parentheses): -
digestive organs (C15–C26); respiratory system and intrathoracic organs (C30–C39); bronchus and lung (C34); urinary tract and genital organs (C60–C68); haematolymphopoetic system (including leukaemia; C81–C85, C88, C91–C96); - leukaemia (C91–C96). The classification we used is a standard one when not enough cases of individual cancer sites are observed to examine cancer risks for all individual organs and tissues separately. We categorised the sites according to tracts (digestive tract, respiratory tract, urogenital tract) or tissues (haematolymphopoetic system), which is often done in epidemiologic studies because the etiologic factors for sites within a tract are assumed to be more similar. Lung cancer was considered separately because of specific concerns about lung cancer in relation to inhaled DU. 2.5. Data analysis All statistical analyses were performed using SPSS for Windows version 14.0. 2.5.1. Differences between Balkan deployed military and their peers Differences in characteristics between Balkan- and non-Balkan deployed personnel were tested with a Chi-square or Wilcoxon test, as appropriate. To study differences in cancer incidence between Balkan- and non-Balkan deployed military, proportional hazard models were used with age defined as a follow-up variable. Time under study (follow-up) was included as a time-dependent variable. In this way, differences in follow-up time between groups were taken into account. Time under study started at the beginning of the study (January 1, 1993) or at entrance into duty (if > January 1, 1993) and ended at death, emigration, diagnosis of cancer or the end of the study (January 1, 2008). Deployment to the Balkans was included as a separate time-dependent variable, depending on end date of first deployment. Thus, person-time and cases between entrance in the cohort and end of first Balkan deployment contributed to the non-Balkan deployed group (see Fig. 1). Analyses were adjusted for branch of military service (Army or Military Police), rank (commissioned or non-commissioned officers) and appointment (permanent, temporary or conscription). Branch of military service was included as a (crude) proxy of type of work during deployment (and thereby possibly occupational exposures). Military rank and type of appointment were used as crude measures of socioeconomic status in the proportional hazard models. Those with missing information on any of
R.P. Bogers et al. / Cancer Epidemiology 37 (2013) 550–555
552
Time at risk
Control time
1, 1, 1993 1st deployment
2nd deployment
1, 1, 2008
Entrance into duty Fig. 1. Calculation of time at risk for a virtual Balkan deployed soldier who was deployed to the Balkans twice.
these variables (n = 18) were excluded from the proportional hazards analysis.
3. Results 3.1. Characteristics of the study population
2.5.2. Comparison with Dutch population Expected numbers of cases of cancer in the study population were estimated according to the person-year method, based on cancer incidence rates for the general population in the Netherlands, stratified by age (five year classes), sex and time period (two or three year classes). Standardised incidence rate ratios (SIRs) were computed by dividing the observed numbers by the expected numbers, using a statistical programme written in SPSS developed by the Netherlands Cancer Institute [9]. Confidence intervals were calculated using the Poisson distribution. Cancer incidence rates in the general population, from 1996 to 2007, were obtained from the Netherlands Cancer Registry. Cancer incidence rates between 1993 and 1996 were available, but coded according to ICD-9, which does not completely match with coding of some specific tumours in ICD-10 after 1996. Therefore, rates in the Dutch population between 1993 and 1996 were not used. For those cancers which occurred between 1993 and 1996 in the study population (ca. 10%), rates in the Dutch population in 1996 were used as a reference instead. 2.5.3. Sensitivity analyses All analyses were also conducted without conscripts, because conscription was abolished in the Netherlands in 1996. Also, because conscripts are relatively young, rank may be a less stable indicator for socioeconomic status in this group. Proportional hazard models were also performed without cases of cancer that occurred within 18 months from the end date of the mission, as most cancers have a longer latency period. SIRs were also calculated for three 5-year periods separately to study possible trends over time.
The study population included 18,175 men deployed to the Balkans between 1993 and 2001, and 135,355 non-Balkan deployed men, with median follow-up times of 10.9 years and 14.8 years, respectively. As follow-up of Balkan deployed personnel started later than January 1, 1993, i.e. after the date of first Balkan deployment, 60% could be followed for at least 10 years, but by definition, no one could be followed for the complete follow-up period of 15 years. Five percent and 1% of the Balkan and nonBalkan deployed personnel had been deployed to other areas between 1993 and 2001. Table 1 presents general characteristics of Balkan and nonBalkan deployed military personnel. Most of them worked in the Army. Whereas the majority of the non-Balkan deployed personnel were conscripts, Balkan deployed personnel consisted mostly of commissioned officers and only 15% conscripts. The median age of both groups was 21 years in 1993. At the end of the study period (1 January 2008), 13% and 5% of Balkan and non-Balkan deployed personnel, respectively, was older than 50. The majority (77%) of the Balkan deployed military personnel was deployed only once (Table 1). Almost all deployments lasted approximately 6 months. 3.2. Cancer incidence 3.2.1. Comparison between Balkan deployed military and their peers In Balkan deployed and non-Balkan deployed men, 175 and 1337 new cases of cancer, respectively, were observed during 15 years of follow-up (until 2008). The cases among Balkan deployed personnel that were not included in the cancer subgroups were
Table 1 General characteristics of male military personnel deployed and non-deployed to the Balkans. Balkan deployeda (n = 18,175) b
Median age at January 1, 1993 (p25–p75 ) Rank (% commissioned officers)c Branch (%) Army Military Police Appointment (%) Permanent Temporary Conscription Other (chaplains)d Duration of 1st deployment (median number of days (p25-p752))e Number of Balkan deployments 1 (%) 2 (%) 3 or more (%)
Non-Balkan deployed (n = 135,355)
21 (18–28) 13.1
21 (19–23) 5.3
94.0 6.0
94.9 5.1
37.2 47.1 15.3 0.3 181 (166–185) 77.1 19.1 3.8
8.9 10.3 80.8 0.05 – – – –
a All characteristics were significantly different between Balkan vs. non-Balkan deployed military at p < 0.01. This was tested by using Chi-Square (sex, rank, appointment, branch of military service) and Wilcoxon statistics (mean age)). b 25th and 75th percentiles. c Rank was available for n = 18,157 Balkan deployed and 128,077 non-Balkan deployed men. d Categorised in the analyses as permanent appointment. e Start date of deployment not available for n = 20.
R.P. Bogers et al. / Cancer Epidemiology 37 (2013) 550–555 Table 2 Hazard ratios of total and site-specific incidence of cancer in Dutch male military personnel deployed to the Balkans (n = 18,175) a compared with non-Balkan deployed personnel (n = 135,355). Cancer site
ICD-10 codes
HR (95% CI)b
All cancers Digestive organs Respiratory system and intrathoracic organs Bronchus and lung Urinary tract and genital organs Haematolymphopoetic system
C00–C96c C15–C26 C30–C39
0.83 (0.69–1.00) 0.85 (0.54–1.33) 0.88 (0.44–1.76)
C34 C60–C68 (C81–C85, C88, C91–C96)
0.91 (0.45–1.90) 0.90 (0.65–1.25) 0.76 (0.47–1.23)
a 18 Balkan deployed men were excluded because of missing information on covariates. b Hazard ratios (95% confidence intervals) adjusted for rank, appointment and branch of military service. c C44 (skin, other/NOS) included; C44 basal cell carcinoma excluded.
mainly skin cancers. The total cancer incidence rate among Balkan deployed personnel was 17% lower than among non-Balkan deployed personnel (hazard ratio (HR) = 0.83, 95% confidence interval (CI) 0.69–1.00, p = 0.05; see Table 2). For the 4 main cancer subgroups and lung cancer, rates were lower among Balkan deployed men, although the differences were not statistically significant. Leukaemia was included in the subgroup haematological cancers, but the number of Balkan and non-Balkan-deployed men with leukaemia (n = 5 and n = 56, respectively) was too low to calculate a meaningful HR for leukaemia separately. Results without adjustment for rank, appointment and branch of military service were essentially similar (not shown). The number of deployments did not seem to affect the number of cancer cases in the Balkan-deployed group. Sensitivity analyses without conscripts (excluding 2788 persons) or cases of cancer that occurred
553
within 18 months after the mission (excluding 20 cases) yielded similar results (see supplement). 3.2.2. Comparison with the Dutch population The observed number of new cancer cases (n = 175) among Balkan deployed military personnel was 15% lower than the expected number of cases (nexp = 205) in the age- and sex-matched Dutch population (SIR = 0.85 (0.73–0.99); see Table 3). The SIR for all cancers combined was close to unity in the first 5 years of follow-up and then declined to 0.65 (0.43–0.95) in the last 5 years of follow-up. For the 4 main cancer subgroups and lung cancer, observed numbers of cancer cases among Balkan deployed men were lower than the expected number. These differences were statistically significant for tumours of the respiratory system and intrathoracic organs, and lung cancer. The observed number of leukaemia cases was lower than expected but not statistically significantly (SIR = 0.63 (0.20–1.46)). Sensitivity analyses without conscripts yielded similar results (not shown). 4. Discussion This study showed that the cancer incidence in Balkan deployed men was 17% lower than in non-Balkan deployed men in the follow-up period of 15 years, and 15% lower than expected from incidence rates in the general Dutch population of comparable age and sex. Cancer risks were also somewhat lower for the main cancer site subgroups separately. There was no indication of an increased risk of leukaemia among Balkan deployed personnel. Although we do not have any data on DU exposure during Balkan deployment, our results do not substantiate concerns about health effects of DU. This is in line with a desk top study about the fate of and potential exposure to DU in the 1999 Kosovo War, which concluded that health effects of DU exposure of Balkan deployed Dutch soldiers are unlikely [10].
Table 3 Cancer incidence among Dutch military men deployed to the Balkans (n = 18,175) as compared to the age- and sex-matched Dutch population, overall and for three follow-up intervals separately. Cancer site and follow-up period All cancers Total period 0–5 y 5–10 y 10–15 y Digestive organs Total period 0–5 y 5–10 y 10–15 y Respiratory system and intrathoracic organs Total period 0–5 y 5–10 y 10–15 y Bronchus and lung total periodc Urinary tract and genital organs Total period 0–5 y 5–10 y 10–15 y Haematolymphopoetic system Total period 0–5 y 5–10 y 10–15 y Leukaemia Total periodc a b c
ICD-10 codes
Observed
Expected
SIR (95% CI)a
175 74 74 27
205.3 70.7 93.3 41.3
0.85 1.05 0.79 0.65
(0.73–0.99) (0.82–1.31) (0.62–1.00) (0.43–0.95)
30 7 16 7
36.5 11.1 17.1 8.2
0.82 0.63 0.93 0.85
(0.56–1.17) (0.25–1.30) (0.53–1.52) (0.34–1.76)
12 8 3 1 10
25.0 8.0 11.6 5.5 20.8
0.48 1.00 0.26 0.18 0.48
(0.25–0.84) (0.43–1.97) (0.05–0.76) (0.00–1.01) (0.23–0.89)
55 22 27 6
54.1 17.9 25.1 11.2
1.02 1.23 1.08 0.54
(0.77–1.32) (0.77–1.86) (0.71–1.57) (0.20–1.17)
28 13 8 7
29.3 11.8 12.6 4.9
0.95 1.10 0.63 1.43
(0.63–1.38) (0.59–1.88) (0.27–1.25) (0.58–2.96)
5
8.0
b
C00–C96
C15–C26
C30–C39
C34 C60–C68
C81–C96 (C81–C85, C88, C91–C96)
C91–C96
Standardised incidence rate ratio. C44 (Skin, other/NOS) included; C44 basal cell carcinoma excluded. Total number too low to stratify analyses by follow-up period (0–5, 5–10, 10–15 years).
0.63 (0.20–1.46)
R.P. Bogers et al. / Cancer Epidemiology 37 (2013) 550–555
554
Table 4 Previously reported studies on cancer incidence among male UN Balkan veterans. Study design
Study population
Follow-up time (years)
Outcome
Obs
Exp
SIR
95% CI
Netherlands (1993–2008)
Population-based historical cohort study
11 (median)
Cancer incidence (registry-based)
175
205.3
0.85
0.73–0.99
Denmark (1992–2001) (4)
Population-based historical cohort study
11 (maximum)
Cancer incidence (registry-based)
84
n.r.
0.9
0.7–1.1
Sweden (1989–1999) (3)
Population-based historical cohort study
10 (maximum)
Cancer incidence (registry-based)
26
21.8
1.2
0.8–1.8
Italy (1996–2007) (2)
Historical cohort study using number of persons instead of person-years
Military personnel deployed between 1993 and March 2001 (n = 18,175 men) Military personnel deployed between 1992 and 2001 (n = 13,552 men) Military personnel deployed between 1989 and 1999 (n = 8347 men) Military personnel deployed between 1996 and 2007 (n = 49,862 men)
11 (maximum)
Cancer incidence based on preliminary army cancer surveillance activity
344.9
0.48
164
0.39–0.57
n.r.: not reported.
The results of this study are in line with previous studies on cancer incidence in Balkan deployed military personnel. Studies in Italy [2], Sweden [3], and Denmark [4] did not demonstrate a higher than expected risk of all cancers including leukaemia among Balkan veterans compared to the general population (see Table 4). However, the follow-up periods in these studies of about 10 years were still short considering the latency time of cancer development by ionising radiation. The study in Italy was hampered by the fact that the geographical distribution of cancer registries is not homogeneous across Italy and the coverage of the resident population is low [2]. However, new estimates taking the low completeness into account showed similar results [11]. Previous studies did not include analyses comparing Balkan deployed with non-Balkan deployed military personnel. Such analyses are important because there are differences in lifestyle factors between the military and general population [5]. Our study included non-deployed personnel, had a relatively large study population and a long follow-up period. The number of personyears in our study is more than twice of those in the Danish study and almost five times those of the Swedish study. Therefore, our study was able to produce more precise and less biased estimates of the relative risks of cancer and allows for longer latency times than previous studies. The lower cancer incidence among Balkan veterans in the present study can be explained by the ‘healthy warrior effect’, an analogy of the ‘healthy worker effect’ or ‘healthy soldier effect’. These terms refer to the observation that the working population is healthier than the general population on a whole. The healthy worker effect was probably first described by William Ogle at the end of the 19th century [12]. This effect is relatively strong for military personnel, due to health examinations at the time of enlistment and the physical demands of the profession, which reduce the risk of some site-specific cancers [13]. The healthy warrior effect results from selection at the time of deployment, as those who are ill are less likely to be deployed [14]. Nevertheless, the lower risk of developing tumours in respiratory organs in Balkan deployed military personnel (and non-Balkan deployed personnel; SIR = 0.64 (95% CI 0.51–0.79)) as compared to the Dutch population was against our expectations. Data on smoking habits were not available for this study population, and we do not know the cause of the lower risk. It was generally assumed that smoking (an important risk factor for respiratory tumours) is more common in the military population [5] than in the general population. Latency time should be taken into account when considering the
risk of developing lung cancer, but because it is likely that most smokers in the study population started smoking many years before the onset of the study, an effect of smoking on the development of lung cancer should have been detected. However, our cohort was relatively young, and lung cancer in men is most observed among 65–80 year olds. Another explanation might be the relatively high amount of physical activity among military personnel [5]. However, despite strong evidence for a protective role of physical activity on cancers of the breast and endometrium, for lung cancer the evidence is weaker [15]. Almost the complete cohort could be retrieved (97%), and highly reliable registration data were available on cancer incidence. Bias by missing persons in the cohort because of the no-consent procedure is unlikely because the 2% personnel (all non-Balkan deployed) that returned their no-consent card had similar characteristics (rank, appointment, age) to those included in the cohort. A limitation of the study is that no data were available on (individual) exposure to DU in the Balkans and on individual risk factors for the development of cancer (such as smoking, alcohol consumption, and job history after deployment to the Balkans). Also, late effects of Balkan deployment on the development of cancer cannot be excluded, as many cancers do not become apparent until 10–20 years after an initiating event and the study population was relatively young. For acute leukaemia, however, which has frequently been associated with radiation, the latency time is relatively short [16] and our results should be more conclusive. Finally, in theory an adverse effect of deployment could remain unnoticed because of the generally better health of deployed soldiers. Ideally, epidemiological studies should be planned before military action is contemplated to avoid difficulties in assessing exposures and health effects afterwards [17]. It then would be feasible to collect additional data on e.g., potentially hazardous exposures during deployment and lifestyle factors such as smoking habits and physical activity. In conclusion, the present study provides the best evidence against an increased risk of cancer following Balkan deployment to date. Role of the funding source This work was supported by the Dutch Ministry of Defence. The funder contributed to the collection of data from registers of the Dutch Armed Forces. All authors were independent of the funder.
R.P. Bogers et al. / Cancer Epidemiology 37 (2013) 550–555
Conflict of interest The authors declare that they have no conflict of interest. Acknowledgements No ethics approval was required for this study. Handling of personal data complied with the Personal Data Protection Act and the Municipal Database Act. Operating procedures for confidential and careful handling of data were described in detail in a privacy agreement between the Ministry of Defence, the Netherlands Cancer Registry (NCR) and the research institute (RIVM). The authors are grateful to Marc Ruijten, Jeanne van Loon, and Diane Houweling, who coordinated the project during different time periods, Bas Verhage, Jeroen de Hartog and Leontien Korteweg, who contributed to the data cleaning and analysis. From RIVM, also Ellis Franssen, Ingrid van Kuilenburg, Adriaan van Kessel, Bennie Bloemberg, Ben Bom, Dirk-Jan Griffioen, Wil Lokhorst, Caroline Ameling and Hendriek Boshuizen contributed to this study. Otto Visser reviewed the study as a member of the initial Scientific Advisory Board. The Netherlands Cancer Institute provided the computer programme for indirect standardisation. We would also like to thank the persons at the Ministry of Defence who contributed to this study; in particular Ad de Koning, Angelique Hardij, Tjalling Leenstra and Peter Krab, who was the manager of the Veteran Registration System. Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/j.canep.2013.04.005. References [1] COMEDS Meeting on Health Concerns Related to the Balkan Deployments, Brussels [updated 13.02.01, cited April 2013]; available from: http://www.nato.int/du/docu/d010115a.htm.
555
[2] Peragallo MS, Lista F, Sarnicola G, Marmo F, Vecchione A. Cancer surveillance in Italian army peacekeeping troops deployed in Bosnia and Kosovo, 1996–2007: preliminary results. Cancer Epidemiol 2010;34(1):47–54. [3] Gustavsson P, Talback M, Lundin A, Lagercrantz B, Gyllestad PE, Fornell L. Incidence of cancer among Swedish military and civil personnel involved in UN missions in the Balkans 1989–1999. Occup Environ Med 2004;61(2):171–3. [4] Storm HH, Jorgensen HO, Kejs AM, Engholm G. Depleted uranium and cancer in Danish Balkan veterans deployed 1992–2001. Eur J Cancer 2006;42(14):2355–8. [5] Bray RM, Hourani LL, Olmsted KLR, Witt M, Brown JM, Pemberton MR, et al. Department of defense survey of health related behaviors among active duty military personnel. A component of the defense lifestyle assessment program (DLAP). Research Triangle Park, NC: RTI International2006, 2005, Contract no.: RTI/7841/106-FR. [6] Macfarlane GJ, Biggs AM, Maconochie N, Hotopf M, Doyle P, Lunt M. Incidence of cancer among UK Gulf war veterans: cohort study. Br Med J 2003;327(7428): 1373. [7] Research Advisory Committee on Gulf War Veterans’ Illnesses. Gulf War illness and the health of Gulf War veterans: scientific findings and recommendations. Washington, DC: U.S. Department of Veterans Affairs, 2008. [8] van den Brandt PA, Schouten LJ, Goldbohm RA, Dorant ELIS, Hunen PMH. Development of a record linkage protocol for use in the Dutch Cancer Registry for epidemiological research. Int J Epidemiol 1990;19(3):553–8. [9] van Leeuwen FE, Klokman WJ, Hagenbeek A, Noyon R, van den Belt-Dusebout AW, van Kerkhoff EH, et al. Second cancer risk following Hodgkin’s disease: a 20-year follow-up study. J Clin Oncol 1994;12(2):312–25. [10] Busker RW, van Ham NHA, van’t Hof G, Langeberg JP, van der Meer BJ, Nieuwenhuizen MS. Fate of depleted uranium in the 1999 Kosovo war: unlikely relation with health effects of Dutch military personnel. Rijswijk: Netherlands Organisation for Applied Scientific Research (TNO), 2005, Report No.: PML 2002-A16. [11] Peragallo MS, Urbano F, Lista F, Sarnicola G, Vecchione A. Evaluation of cancer surveillance completeness among the Italian army personnel, by capture– recapture methodology. Cancer Epidemiol 2011;35(2):132–8. [12] Ogle W. Letter to the Registrar-General on the mortality in the registration districts of England and Wales during the ten years 1871–80. Supplement to the 45th annual report of the Registrar General of Births, Deaths, and Marriages, in England.; 1885. [13] Lee I, Oguma Y. Physical activity. In: Schottenfeld D, Fraumeni JF, eds. Cancer epidemiology and prevention. 3rd ed., Oxford: Oxford University Press, 2006: 449–68. [14] Haley RW. Point: bias from the ‘‘healthy-warrior effect’’ and unequal followup in three government studies of health effects of the Gulf War. Am J Epidemiol 1998;148(4):315–23. [15] Friedenreich CM, Neilson HK, Lynch BM. State of the epidemiological evidence on physical activity and cancer prevention. Eur J Cancer 2010;46(14):2593– 604. [16] Boice Jr JD. Ionizing radiation. Cancer epidemiology and prevention. 3rd ed. Oxford: Oxford University Press, 2006. 259–93. [17] Hotopf M, Wessely S. Can epidemiology clear the fog of war? Lessons from the 1990–91 Gulf War. Int J Epidemiol 2005;34(4):791–800.