Article Text
Abstract
Objectives Until recently, enrichment of uranium for civil and military purposes in France was carried out by gaseous diffusion using rapidly soluble uranium compounds. We analysed the relationship between exposure to soluble uranium compounds and exposure to external γ-radiation and mortality in a cohort of 4688 French uranium enrichment workers who were employed between 1964 and 2006.
Methods Data on individual annual exposure to radiological and non-radiological hazards were collected for workers of the AREVA NC, CEA and Eurodif uranium enrichment plants from job-exposure matrixes and external dosimetry records, differentiating between natural, enriched and depleted uranium. Cause-specific mortality was compared with the French general population via standardised mortality ratios (SMR), and was analysed via Poisson regression using log-linear and linear excess relative risk models.
Results Over the period of follow-up, 131 161 person-years at risk were accrued and 21% of the subjects had died. A strong healthy worker effect was observed: all causes SMR=0.69, 95% CI 0.65 to 0.74. SMR for pleural cancer was significantly increased (2.3, 95% CI 1.06 to 4.4), but was only based on nine cases. Internal uranium and external γ-radiation exposures were not significantly associated with any cause of mortality.
Conclusions This is the first study of French uranium enrichment workers. Although limited in statistical power, further follow-up of this cohort, estimation of internal uranium doses and pooling with similar cohorts should elucidate potential risks associated with exposure to soluble uranium compounds.
- uranium
- enrichment
- solubility
- mortality
- cohort study
Statistics from Altmetric.com
What this paper adds
During the uranium enrichment step of the nuclear fuel cycle, a protracted exposure to rapidly soluble uranium compounds (under three isotopic forms: natural, enriched and depleted uranium) may occur.
We analysed mortality in a historical cohort of 4688 French uranium enrichment workers employed for at least 6 months between 1964 and 2006, with a median follow-up of 30 years.
As in other studies of nuclear workers, a strong healthy worker effect was observed when comparing this worker cohort with the general population; no cause of mortality was significantly associated either with exposure to rapidly soluble uranium compounds (assessed via job-exposure matrixes) or external γ-radiation (individual doses).
Although our study did not find strong evidence for an association between exposure to rapidly soluble uranium compounds and cause-specific mortality, a reanalysis based on extended follow-up and incorporating estimated internal uranium doses is needed.
Introduction
In recent years, research on the health of workers involved in the uranium fuel cycle has focused on relationships between internal exposures to radionuclides and health effects, including cancer and non-cancer outcomes.1–4 When inhaled or ingested, these compounds are distributed to various organs depending on biokinetic processes specific to their physicochemical form. Internal exposure to airborne uranium compounds is of concern because they may exhibit both radiological (α-emitters) and chemical toxicity.5 Even though a substantial amount of animal and human toxicological data exists for the radiological and chemical health effects of uranium,6 ,7 data on the health impact of chronic inhalation of industrial uranium compounds on human health are limited.7 ,8
Inhalation is the main route of exposure to uranium in nuclear fuel cycle workers in the course of their work. French nuclear fuel cycle workers are subject to uranium exposure of various physicochemical forms and other chemical and physical hazards9 during the following steps in the fuel cycle: ore milling and refining to produce a uranium concentrate powder known as ‘yellow cake’, conversion to uranium tetrafluoride (UF4) and uranium hexafluoride (UF6), enrichment of isotope 235U, uranium fuel fabrication and spent fuel reprocessing and disposal. Since exposures to these workers are regularly monitored for the purposes of radiation protection, they represent one of the most pertinent groups for studying health risks associated with chronic radiation exposure.10
A recent literature review suggested that further studies of subgroups of nuclear fuel cycle workers with homogeneous exposure to soluble or insoluble uranium are needed to examine whether they experience an increased risk of mortality from cancerous and non-cancerous diseases.8
The Tricastin nuclear site, situated in south-eastern France, is the only French nuclear site where uranium undergoes enrichment at three plants operated by AREVA NC, CEA and Eurodif. Although the main industrial enrichment technology during the period 1964 and 2008 was gaseous diffusion, some experimental work was also performed on laser enrichment in the 1970s and 1980s. During uranium enrichment by gaseous diffusion, uranium is in the form of highly soluble UF6, containing molecules of 234U, 235U and 238U, which are separated by mass. Enriched uranium (high weight per cent of 235U) is produced in industrial quantities after many repetitions of this process. Depleted uranium, which contains a low weight per cent of 235U, is a by-product of the enrichment process. Both enriched and depleted uranium are used in civil and military applications. Uranium enrichment workers constitute a specific subgroup which has protracted exposure to soluble uranium compounds of various isotopic compositions which can be categorised as natural, enriched and depleted uranium. External γ-radiation exposure in uranium enrichment workers is of lower magnitude compared with other groups of nuclear fuel cycle workers11 In contrast to insoluble uranium compounds, which are retained in the lungs, soluble uranium compounds rapidly enter the systemic circulation where part of the uranium can be taken up by the skeleton, kidneys, liver and other tissues, and the remaining amount excreted within the following day via urine. Although previous studies in uranium enrichment workers have reported excesses in mortality from lymphohaematopoietic, bladder and stomach cancers,12–14 until now no epidemiological study has explored long-term health effects of inhalation of different isotopic forms of uranium (natural, enriched and depleted). However, in vitro studies have shown that enriched and depleted uranium may have different toxicological profiles.15
Our study aimed to examine mortality risks due to cancerous and non-cancerous diseases in a national cohort of French uranium enrichment workers who were employed in enrichment of uranium by gaseous diffusion at three uranium enrichment plants, and were exposed to both radiological and non-radiological hazards. Associations with exposures to soluble uranium compounds of various isotopic compositions and external γ-radiation were examined.
Materials and methods
Cohort construction and follow-up
A roster of 5070 uranium workers involved in enrichment activities was identified from the French TRACY U (TRAvailleurs du CYcle du combustible potentiellement exposés à l'Uranium) cohort of 12 739 nuclear fuel cycle workers.16 Inclusion in the cohort required that the workers should have worked for at least 6 months between 1964 and 2006 in the AREVA NC, CEA and Eurodif uranium enrichment plants. Membership of the ‘uranium enrichment subcohort’ (AREVA NC, CEA and Eurodif) was based on the longest employment period at these plants (see online supplementary table S1). Workers with a previous history of employment in uranium mining (n=31) were excluded. The final data set used in the statistical analyses included 4688 eligible uranium enrichment workers.
Each worker contributed person-years at risk from either the date of first employment at the uranium enrichment plant plus 6 months or 1 January 1968 (whichever occurred later), up to the date of death, last date known to be alive or 31 December 2008 (whichever occurred earlier). Sixteen deaths occurred before 1968, but the follow-up in our study began on 1 January 1968 because data on individual causes of deaths are not available in France before this date. Follow-up ended in 2008 because completeness of the death registry could not be guaranteed for more recent years at the time of the collection of individual causes of death.
Occupational radiation exposure assessment
The main exposures of interest in our study were internal radiation exposure from inhalation of uranium and external γ-radiation exposure. Ingestion of uranium in drinking water and food was considered negligible.
Estimates of annual internal exposure to uranium were reconstructed using two plant-specific job-exposure matrices (JEM) for the AREVA NC and Eurodif plants. The construction of these JEMs has been described in detail elsewhere.17 ,18 The AREVA NC JEM was validated against individual bioassay data with 64% sensitivity and 80% specificity.19 The two JEMs were constructed using the same strategy. The Eurodif JEM had additional information on current occupational exposure limits, which served to validate the intensity and frequency of exposure. The AREVA NC JEM was extrapolated to the CEA plant because of the identical nature of the work. The JEM was used to assign annual (1964–2008) levels of frequency and intensity of exposure on a four-level scale for each hazard. A multiplicative product of frequency, intensity and duration of employment (years) allowed deriving an individual exposure score which was used for epidemiological analyses.20 Exposure to soluble uranium compounds (UF6 and UO2F2) was defined as exposure to type F (rapidly soluble) uranium compounds according to the classification of the International Commission on Radiological Protection (ICRP).21 For the Eurodif subcohort, it was possible to further distinguish between isotopic forms of uranium (natural, enriched and depleted). Exposure scores were cumulated for any worker with a history of working at uranium enrichment plants.
External γ-radiation exposure was monitored individually on either a monthly (workers susceptible to receiving between 6 and 20 mSv) or quarterly (those susceptible to receiving between 1 and 6 mSv) basis, and reported as an annual whole-body dose in mGy. External radiation dosimetry records were extracted from the plant monitoring files and the electronic SISERI system (French national database of occupational external exposure to ionising radiation).22
Assessment of other occupational hazards
Information on occupational exposure to trichloroethylene (TCE), heat and noise was considered because of their possible influence on cancerous23 and circulatory diseases.24 These were also selected due to their high prevalence and availability of monitoring data from the industrial hygiene services at uranium enrichment facilities.12 ,17 Similarly to uranium exposure, exposure scores to TCE, heat and noise were estimated using JEMs. Noise was classified as a binary time-dependent variable (never exposed vs ever exposed to sound pressure of ≥80 dB(A)). Annual exposure to noise was available for the Eurodif subcohort.
Mortality ascertainment
Individual vital status and underlying causes of death were identified from the French national mortality registries by deterministic linkage via name, gender and date, and place of birth. Contributing causes of death listed on death certificates were not included as mortality events in this study. Causes of death were coded according to the eighth revision of the International Classification of Diseases (ICD-8) from 1968 to 1977, the ninth revision (ICD-9) from 1978 to 1999 and the 10th revision (ICD-10) for the period 2000–2008.
Statistical analysis
We calculated standardised mortality ratios (SMR) for selected health outcomes using the French general population as a reference. Expected numbers of deaths for each cause were calculated using French sex-specific and age-specific mortality rates grouped in 5-year intervals from 1968 to 2008.
In addition, we performed within-cohort analyses via Poisson regression on grouped data25 for all solid (n=406), lung (n=100) and lymphohaematopoietic (n=28) cancers, as well as circulatory (n=281), ischaemic heart (n=95) and cerebrovascular (n=71) diseases. In these analyses, person-years were cross-classified by sex, age (15–19, 20–24…80–84, 85 and over), calendar period (1968–1972, 1973–1977…1998–2003, 2004–2008), socioprofessional status at hire (managerial/professional, clerical, skilled technical, unskilled), subcohort (AREVA NC, CEA and Eurodif) and 5-year lagged cumulative exposures to soluble uranium, external γ-radiation, TCE, heat and noise. Time-dependent exposure levels were categorised (unexposed, low exposed, medium exposed and highly exposed) using quartiles of each cumulative exposure score weighted by the person-years. Cut-points for external γ-radiation were 0, 0.01, 0.13, 0.9 and 10 and more mGy so as to obtain a balanced number of deaths in each dose category. Log-linear risk models were used to obtain relative risk (RR) and corresponding 95% CI. In addition, linear excess relative risk (ERR) models were used to estimate ERR per 100 mGy and 95% CI associated with external γ-radiation dose. Models were stratified on sex, attained age, calendar period, socioeconomic status at hire and subcohort. We assessed confounding by TCE for cancer outcomes, and confounding by heat and noise for circulatory diseases. We examined the impact of the isotopic forms of rapidly soluble uranium compounds (enriched and depleted) within the Eurodif subcohort for solid cancers (n=85), lung cancer (n=23) and circulatory diseases (n=45).
A correlation between uranium compounds and external γ-radiation exposures were examined by Pearson's partial correlation coefficients controlling for the individual component effect.
All analyses were performed using Stata (StataCorp, College Station, Texas, USA) and EPICURE (HiroSoft International Corporation, Seattle, Washington, USA) statistical software.
Results
Cohort description
Male workers constituted more than 90% of the study population (table 1). The median duration of follow-up was 30.2 years, and, as a whole, the cohort cumulated 136 161 person-years. Causes of death were ascertained for 99% of decedents (between 1968 and 2008). Less than 1% of the workers (n=37) were lost to follow-up. At the end of follow-up, 21% (n=1010) of the cohort had died, and 25% (n=1164) of the workers were still employed in the French nuclear industry. Almost 30% (n=1312) of the workers had been employed at more than two nuclear facilities. Seventy per cent (n=3295) of workers were potentially exposed to soluble uranium and 90% (n=4253) were monitored for external γ-radiation. Median external γ-radiation among exposed monitored workers was 0.75 mGy (minimum=0.03, maximum=230.2; table 1). More than 60% of the workers were exposed to several occupational hazards, but only 34% of the workers were exposed to both soluble uranium and external γ-radiation (data not shown). There was no correlation between exposure to rapidly soluble uranium compounds and external γ-radiation (Pearson's r=0.1). Within the Eurodif subcohort, exposure to enriched uranium was moderately correlated with depleted uranium (Pearson's r=0.7; data not shown).
Comparison of the cohort mortality with the general population
Mortality rates for all causes of death (SMR=0.69, 95% CI 0.65 to 0.74) and all cancers (SMR=0.79, 95% CI 0.72 to 0.87) were substantially below expectation based on national rates (table 2). An excess in mortality was observed for pleural cancer (SMR=2.3, 95% CI 1.06 to 4.4; based on nine deaths). A somewhat smaller mortality risk, albeit non-statistically significant, was also observed for kidney cancer (SMR=1.1, 95% CI 0.60 to 1.9), pancreatic cancer (SMR=1.3, 95% CI 0.87 to 1.8), biliary system cancer (SMR=1.5, 95% CI 0.50 to 3.6), malignant neoplasms of the central nervous system (SMR=1.6, 95% CI 0.94 to 2.6), malignant melanoma (SMR=1.9, 95% CI 0.83 to 3.8) and breast cancer (SMR=1.5, 95% CI 0.63 to 2.9) in females. Notable deficits were observed for smoking-related cancers (SMR=0.73, 95% CI 0.64 to 0.83), including lung cancer (SMR=0.74, 95% CI 0.60 to 0.90), non-malignant respiratory diseases (SMR=0.64, 95% CI 0.47 to 0.84), circulatory diseases (SMR=0.79, 95% CI 0.70 to 0.89) and deaths due to external causes (SMR=0.53, 95% CI 0.42 to 0.66; table 2).
Within-cohort exposure–response analyses
Associations between cumulative exposures to rapidly soluble uranium compounds and external γ-radiation, and mortality outcomes are presented in table 3 and in table 4. Exposure to natural soluble uranium compounds was not significantly associated with any cause of mortality, and a monotonic decreasing trend from low exposed to highly exposed was observed for lung and lymphohaematopoietic cancers. A highly imprecise positive trend across exposure to natural soluble uranium compounds (RR=0.85, 95% CI 0.56 to 1.3, low exposed vs never exposed; RR=0.98, 95% CI 0.71 to 1.3, moderately exposed vs never exposed; RR=1.2, 95% CI 0.85 to 1.6, highly exposed vs never exposed) was observed for circulatory diseases (table 3). A positive non-significant association was found between external γ-radiation dose and mortality due to circulatory (ERR/100 mGy=0.38, 95% CI <0 to 2.3) and ischaemic heart diseases (ERR/100 mGy=0.91, 95% CI <0 to 5.1; table 4). Additional adjustments for non-radiological occupational hazards (TCE, heat and noise) did not substantially change RR, ERR or improve the model fit (data not shown). Cause-specific RRs associated with exposures to enriched and depleted uranium were of comparable magnitude (table 5). Associations of mortality with non-radiological hazards are presented in online supplementary table S2.
Discussion
In our study, we analysed mortality in a national cohort of French uranium enrichment workers exposed to soluble uranium compounds, external γ-radiation and other non-radiological occupational hazards. Overall, this workforce exhibits a favourable mortality pattern (healthy worker effect), with the exception of a significantly elevated mortality risk for pleural cancer. We did not find an association between exposure to soluble uranium compounds and external γ-radiation and cause-specific mortality. There was an imprecise trend of increased risk of mortality due to circulatory diseases across increasing exposure to natural soluble uranium compounds.
Study strengths and limitations
A unique strength of our study is exposure reconstruction of both radiological and non-radiological (chemical and physical) occupational hazards and distinguishing isotopic forms of soluble uranium compounds (natural, enriched and depleted). Together with ionising radiation, uranium enrichment workers are known to be exposed to numerous non-radiological hazards.9 ,17 ,26 While these chemical and physical hazards are present in nuclear fuel cycle activities, they are rarely considered in epidemiological studies. Also, partly owing to historical and regulatory reasons, employers and employees in the French nuclear industry might have been more concerned with radiation protection compared to other non-radiological hazards.27 Although exposure data were collected on more than 20 hazards, we only considered those three non-radiological risk factors (TCE, heat and noise) that raise concerns among occupational physicians and workers and that are most prevalent at French uranium enrichment plants. We did not consider other established carcinogens (eg, chromium and asbestos), because of the limited number of exposed workers. For example, TCE, as a chlorinated solvent, is a known carcinogen of group 1 according to the International Agency for Research on Cancer (IARC).23 This exposure was not statistically significantly associated with excess lung and lymphohaematopoietic cancer as observed in our previous study.9 Finally, models were not adjusted for non-radiological exposures, because their simultaneous inclusion in the models produced unstable risk estimates.
Uranium enrichment in France started in the beginning of the 1960s, which is late compared to the USA where the first uranium enrichment facilities were opened during the Manhattan Project in the 1940s. Hence, the proportion of workers alive at the end of follow-up is still high and the statistical power of our study will be improved by continuing the follow-up, as well as by conducting combined analyses with similar cohorts of nuclear fuel cycle workers. Most medical files, radiological bioassays and industrial hygiene data are hard copy and not adapted for immediate use in large-scale epidemiological studies. Even though the use of a JEM can cause non-differential misclassification of exposure, its use is particularly advantageous in this situation. The JEM exposure score used in this study was calculated individually as a product of frequency, intensity and duration of exposure, allowing for quantitative exposure–response analyses in the absence of internal uranium doses. In the short term, the lack of information on potential lifestyle confounders may be overcome by conducting nested case–control studies.28 After the available bioassay data have been collected for this cohort, internal dose estimation will be possible. The harmonised approach developed in the European Commission-funded Concerted Uranium Research in Europe (CURE) project for the computation of internal doses in European cohorts of uranium workers will be used for that purpose.29 Use of the JEM to assign solubility and isotopic composition of uranium compounds will improve the accuracy of internal dosimetry.30 Therefore, a major limitation of our study is the absence of individual uranium dose estimates.
Excluding an unknown number of construction workers employed by subcontractor companies that may have been highly exposed to ionising radiation during maintenance and construction work from this study may have affected the strength of the tested associations. As in other studies,31 ,32 the obstacles to including this workforce are the difficulty of locating payroll rosters and the impracticality of collecting occupational health monitoring files due to the frequent structural changes of subcontractor companies.
In addition, other limitations of our study are its limited statistical power and the lack of information on smoking and other lifestyle factors.
Comparison with the general population
Nuclear workers are subject to selection at the time of hiring on the basis of initial health status, and regular surveillance by occupational health services, which leads to selection of healthy workers. Decreased mortality in comparison with the general population—or healthy worker effect (HWE)—is common in occupational studies. As in other occupational cohorts,33 an HWE was evident in our study for many causes of death (including cancer and circulatory diseases), indicative of selection bias. An excess risk typically becomes apparent when workers are exposed to an occupational hazard associated with a high risk of disease. Although it was possible to find by chance a significant association in the SMR analysis due to the large number of tests performed, the significant result for pleural cancer might be linked with previous exposure to asbestos. The magnitude of latency for pleural mesothelioma is 40–50 years after first asbestos exposure, depending on the occupation and the intensity of exposure.34 This increased pleural cancer mortality (mostly represented by pleural mesothelioma) is a common finding in studies of nuclear workers exposed to low-level radiation, and a critical role of unmeasured confounding by asbestos has been emphasised.35 The excess for pleural cancer, albeit based on nine cases, may be a true finding due to the fact that many French nuclear workers started their career at naval shipyards where exposure to asbestos and external γ-radiation was quite substantial. Exposure to asbestos at uranium enrichment plants was of lower magnitude (P. Collomb, personal communication). Nine workers who died from pleural cancer in our study had a higher mean γ-radiation dose (13.3 mGy), compared to the cohort average (2.81 mGy) and started their employment in uranium enrichment at the age of 37.6 years, on average. Thus, the increased mortality due to pleural cancer may be attributed to exposures received before the work in uranium enrichment. Continuing monitoring of mortality due to pleural cancer is necessary in this study; however, detailed exposure–response analyses are not feasible at this stage due to the limited number of cases.
Associations with soluble uranium and external γ-radiation
An absence of significant associations between exposure to soluble uranium compounds and cause-specific mortality is noticeable. This may be due to a low influence of rapidly soluble UF6 on studied causes of mortality. In fact, the products of the UF6 hydrolysis (HF, UO2F2) dissolve in the upper airways by forming solid UO2F2 aggregates and not entering deeply into the lungs. Knowledge gained from several accidental exposures of UF6 has shown that 73% of the uranium was excreted during the first 24 h.36 Thus, the acute toxic effects of HF (skin damages and lung oedema) may prevail over the long-term health effects of UO2F2.
An increase in mortality due to lymphohaematopoietic cancer was reported in a recent study of the US Paducah gaseous diffusion plant workers.12 This may be explained by the use of reprocessed uranium at this plant, which may have been contaminated with other radionuclides such as 99Tc, 237Np and 239Pu,12 having a shorter half-life period. In addition, as recently suggested by one case of accidental exposure to UF6, its biokinetics may be modified by the lung oedema and lead to prolonged material retention in the lungs and lymphatic nodes.37 While leukaemias are known to originate in haematopoietic stem cells of the red bone marrow, some lymphomas (non-Hodgkin's lymphoma and Hodgkin's disease) originate in the mature lymphoid cells situated in the lymphatic nodes.38 ,39 In our study, an additional analysis excluding 246 workers with potential exposure to insoluble uranium compounds did not produce different risk estimates for lymphohaematopoietic cancer (results not shown).
The only suggestive non-significant trend across exposure categories of exposure to soluble uranium compounds was noted for circulatory diseases. A recent review of toxic effects of chronic uranium ingestion in animals has reported heterogeneous tissue sensitivity to uranium.40 It seems that toxic effects of uranium exposure are not directly correlated with the amount of uranium accumulated in an organ.40 While the studies reviewed by Dublineau et al40 were not focused on cancer or circulatory diseases, there are numerous mechanistic theories of the relationship between circulatory diseases and low-dose radiation, such as induction of atherosclerosis, microvascular damage to the heart, kidney and lung and direct damage to the heart.41 Owing to the lack of statistical significance of our observations and the lack of radiobiological studies on the effect of chronic uranium inhalation on the circulatory system, our findings should be considered very cautiously. A positive but non-significant association was also observed between circulatory diseases and external γ-radiation, which was comparable with other studies of French nuclear workers.42 ,43
Differences in the magnitude of mortality risks associated with exposures to natural, enriched and depleted uranium were indistinguishable in our study. Natural, enriched and depleted uranium share the same chemical toxicity, but the radiological toxicity of these three types of isotope mixtures varies, from lowest for depleted, intermediate for natural and highest for enriched uranium. Although enriched uranium, having strong α-emission potential, is more likely to produce double-strand breaks in DNA, a recent study showed that depleted uranium caused the same kind of DNA damage in bronchoalveolar cells of rats.44 It should be noted, however, that an analysis stratified by the isotopic form of soluble uranium compounds was only possible within the Eurodif subcohort. This subcohort is the youngest of three subcohorts included in this study, with only 9% of workers having died at the end of follow-up. In time, it will therefore be necessary to include more workers exposed to enriched and depleted uranium to allow for more powerful analyses. At this stage, the most appropriate risk estimates of soluble uranium compounds are those obtained in the analysis of the total cohort of French uranium enrichment workers presented in this paper.
Conclusion
In summary, the first mortality analysis of the cohort of French uranium enrichment workers has not shown affirmative associations between exposure to soluble uranium compounds and cause-specific mortality. The findings obtained in this study should be revisited after continuing follow-up of this cohort, carrying out further analyses using individual-level internal uranium doses, and ultimately combining the data with those of similar cohorts of nuclear fuel cycle workers to increase statistical power. Opportunities to conduct such analyses in Europe were recently demonstrated by the European Commission-funded CURE project.29
Acknowledgments
The authors would like to thank Dr Eve Bourgkard from the Institut national de recherche et de sécurité pour la prévention des accidents du travail et des maladies professionnelles (INRS) for her help in the use of national mortality statistics, Dr Pascale Scanff and Dr Hervé Roy from the Institut de Radioprotection et de Sûreté Nucléaire (IRSN) for providing access to external dosimetry data, and Dr Estelle Rage, Dr Klervi Leuraud, Ms Lucy Fournier and Professor Dale Preston for their help in using EPICURE software. Finally, the authors thank Professor David Richardson for his critical review of the manuscript and helpful suggestions.
References
Supplementary materials
Supplementary Data
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
- Data supplement 1 - Online table 1
- Data supplement 2 - Online table 2
Footnotes
Contributors SZ designed the study, conducted statistical analyses and had the lead role in writing the manuscript. DL, IGC, LBZ, ES, OL and JG participated in developing the study analytical strategy and in revising the manuscript. PC helped in the acquisition of occupational records.
Funding This research was funded by a research grant from the ‘Health, Environment, and Toxicology’ scientific programme of the Région Île-de-France. This article was prepared within the framework of the CURE project, supported by the EU FP7 DoReMi Network of Excellence (grant agreement: 249689). The construction of the TRACY U cohort was partly supported by AREVA in the framework of a bilateral IRSN-AREVA research agreement. The views of the authors do not necessarily reflect the views of the funding agencies.
Competing interests PC was employed by AREVA and had no control over the study design or statistical analysis.
Ethics approval The study has been approved by the French Data Protection Authority (CNIL) (declaration No. DR-2012–611).
Provenance and peer review Not commissioned; externally peer reviewed.