Article Text
PDF
Abstract
Objective This follow-up study of uranium processing workers at the Fernald Feed Materials Production Center examines the relationship between radiation exposure and cancer and non-cancer mortality among 6403 workers employed for at least 30 days between 1951 and 1985.
Methods We estimated cumulative, individual, annualised doses to 15 organs/tissues from external, internal and radon exposures. Vital status and cause of death were ascertained in 2017. The analysis employed standardised mortality ratios, Cox proportional hazards and Poisson regression models. Competing risk analysis was conducted for cardiovascular disease (CVD) mortality risk given several assumptions about risk independent of competing outcomes. Emphysema was examined to assess the potential for confounding by smoking.
Results Vital status was confirmed for 98.1% of workers, with 65.1% deceased. All-cause mortality was less than expected in salaried but not hourly workers when compared with the US population. A statistically significant dose response was observed between external (but not total or internal) lung dose and lung cancer mortality (HR at 100 mGy adjusted for internal dose=1.45; 95% CI=1.05 to 2.01). Significantly increased HRs at 100 mGy dose to heart were observed for CVD (1.27; 95% CI=1.07 to 1.50) and ischaemic heart disease (1.30; 95% CI=1.07 to 1.58). CVD risk remained elevated regardless of competing risk assumptions. Both external and internal radiation were associated with emphysema.
Conclusions Lung cancer was associated with external dose, though positive dose responses for emphysema imply residual confounding by smoking. Novel use of competing risk analysis for CVD demonstrates leveraging retrospective data for future risk prediction.
- Radiation
- Epidemiology
- Radiation, Ionizing
- Occupational Health
Data availability statement
Data are available in a public, open access repository. The datasets generated during and/or analysed during the current study are available in the Comprehensive Epidemiologic Data Resource (CEDR) repository, https://oriseapps.orau.gov/CEDR/search_results.aspx?DataSet=FMM23A01.
Statistics from Altmetric.com
WHAT IS ALREADY KNOWN ON THIS TOPIC
Although ionising radiation is an established carcinogen, uncertainties remain surrounding the effects of internal emitters, radiation received at low-dose rates and induction of cardiovascular disease at doses <500 mGy. The exposure profile of uranium processing workers to low-dose rates of external and internal radiation makes them an informative population for interrogating these uncertainties.
WHAT THIS STUDY ADDS
In a follow-up assessment of 6403 previously studied uranium processing workers at the Fernald Feed Materials Production Center, we observed radiation dose-dependent increases in cardiovascular disease, emphysema and lung cancer, though lung cancer risk was only associated with external gamma dose and confounding by tobacco use could not be excluded.
HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY
We used novel competing risk analysis to provide cardiovascular disease risk estimates as population cancer rates decrease while cardiovascular disease rates increase. Our use of emphysema as a negative control to assess whether residual confounding by smoking is present is translatable to other settings where occupational characteristics such as pay type are used as a smoking surrogate.
Introduction
Ionising radiation is an established carcinogen across a broad range of doses received at low-dose and high-dose rates. However, important uncertainties remain, including the difference between acute and chronically received doses,1 effects of densely ionising radiation from internal emitters1 and induction of cardiovascular disease (CVD) at doses below 500 mGy.2 Non-significantly elevated risks of kidney cancer and CVD from low doses of chronically received internal alpha radiation have been reported in uranium-exposed gaseous diffusion workers,3 4 and with the exception of alpha exposures from radon, there is uncertainty regarding a lung cancer radiation association following exposures to uranium at protracted dose rates.5 Gaseous diffusion worker studies and meta-analyses provide evidence of a radiation dose response for CVD below 500 mGy, which has been difficult to detect in past studies.2 3 The exposure profile of uranium processing workers to chronic alpha and gamma radiation makes them an informative population for assessing these uncertainties.6 The Fernald Feed Materials Production Center (‘Fernald’) comprises a cohort of 6403 such workers.6 7
Fernald was a uranium processing facility operated by National Lead of Ohio (NLO) between 1951 and 1985,7 of which several follow-up studies have been previously conducted. Cragle et al collected the initial data on white males8; findings from this study and two subsequent analyses of the same data by Ritz revealed a healthy worker hire effect and a dose response for lung cancer and non-malignant respiratory disease (NMRD).9 10 Silver et al expanded the dataset to include females and people of colour and increased follow-up through 2004. Unlike Cragle et al,8 they found evidence of the healthy worker hire effect among salaried but not hourly workers and no association between external or internal radiation dose and either lung cancer or NMRD.7
As part of the Million Person Study (MPS), this study expands follow-up of the Silver et al Fernald cohort by an additional 13 years and includes additional outcomes in dose-response analyses.
Materials and methods
Cohort definition
The Oak Ridge Institute for Science and Education collected employment data on all identifiable Fernald employees though Cragle et al only assessed 4014 white males.8 Silver et al expanded 6409 existing records to include additional workplace history variables.7 Through a data sharing agreement with the National Institute for Occupational Safety and Health (NIOSH), 6403 workers from the Cragle et al cohort were matched on either full name and birthdate or social security number to the Silver et al cohort.7 NIOSH provided work history information including radiation organ doses, pay category (ie, salaried/hourly) and other workplace exposures. The study population, therefore, consisted of 6403 NLO employees who worked ≥30 days at Fernald from 1951 to 1985 (data reference11).
Data abstracted from Fernald work history records, the NLO Employee Database and the site’s Health Physics Information System (HIS20) included date of birth, sex, race, pay status, date of hire and date of employment termination. Personal information was updated when available from mortality records. Because most employees were white, we assumed employees were white when otherwise unverifiable.
Vital status and outcome determination
Cragle et al ascertained vital status through 1989.8 Follow-up was extended through 2017 and expanded to include females and people of colour using MPS methods.12 Briefly, workers were matched to records from the Social Security Administration, Pension Benefit Information and National Death Index (NDI). The LinkPlus programme was used to match the study roster against the Social Security Administration Death Master File and 34 state mortality files to ascertain vital status where not already available.13 14 Underlying cause of death (UCOD), the outcome of interest for all analyses, was determined using the International Classification of Diseases revision at time of death. UCOD was available from the NDI beginning in 1979; for earlier mortality, a trained nosologist coded UCOD from death certificates. 37 employees who were confirmed dead but had no cause of death were flagged as having died and included in the ‘all causes of death’ category with no specific causes assigned. Outcomes of interest were selected a priori based on observations from previous assessments of Fernald, Mallinckrodt Chemical Works and the Life Span Study of atomic bomb survivors (online supplemental table 1).1 7 14
Supplemental material
Radiation dose reconstruction
The comprehensive radiation dosimetry analysis of Fernald performed by Anderson et al formed the basis of our dosimetry.15 Briefly, Anderson et al abstracted exposures from Fernald site records and the site’s HIS20, supplemented with doses reported in the Department of Energy Radiation Exposure Monitoring System for work outside Fernald. Long-lived radionuclides, external exposures and radon were evaluated.
Long-lived radionuclides
Uranium intakes were evaluated for workers using urine samples with reported positive uranium concentrations based on the assumption of chronic inhalation.15 We extended the Anderson et al analysis to an additional 21 organs/tissues by reevaluating the uranium dosimetry model based on daily chronic intake over the reported duration using the Dose and Risk Calculation Software platform.16 We calculated yearly organ dose estimates from date of first uranium intake projected through 2017.
External exposures
Anderson et al reported external doses based on film badge and thermoluminescent dosimeter measurements as unadjusted penetrating whole-body dose estimates, that is, personal dose equivalents.15 We used zero doses as recorded and did not estimate potential missed dose. We assumed workers spent half their exposure time in the anterior–posterior orientation to the external radiation source and half the time in the rotational orientation and used whole-body dose-based HT/H×10 factors specific to the orientations and organs to estimate dose to 15 organs.17 18 We added external doses from exposures received at other workplaces using data from Landauer; the US Air Force and the US Navy.
Radon
Anderson et al estimated radon exposure in working level months (WLMs) using a validated Gaussian dispersion model which estimated air concentrations from the K-65 silos built to store radioactive materials described by Anderson et al and Hornung et al.15 19 We estimated annual lung dose using an 8.19 mGy/WLM conversion coefficient.6 20
Total dose
The total organ absorbed dose was calculated for all organs of interest, lagged 10 years (2 years for red bone marrow) (online supplemental table 3) by summing external photon dose, dose from long-lived radionuclides and dose from radon, using a weighting factor of 1 for each source of exposure.
Other occupational exposures
Division, department, plant/building and job title information were used in combination with site process records, industrial hygiene monitoring records and institutional knowledge to assign qualitative (ie, ever/never) exposure flags for asbestos, coal dust, miscellaneous dusts, uranium dust, routine laboratory chemicals use and emergency conditions exposure (ie, exposed to potentially high-dose rates during a facility emergency such as episodic releases of UF6 gas).7 15
Statistical methods
Person-time began accruing 30 days after date of hire and ended at date of death, age 95, date lost to follow-up, or 31 December 2017, whichever occurred first.
Standardised mortality ratios (SMRs) were calculated as the ratio of observed to expected number of deaths in the average US population based on US population rates for persons of the same age (5-year categories), sex, race and calendar year (5-year categories). Outcomes were selected a priori from knowledge of uranium deposition in tissue and their potential relation to radiation (online supplemental table 4). Sensitivity analyses assessed SMRs stratified by sex and pay type.
Cox and Poisson regression models were used for dose-response analyses. We fit Cox regression models with age as the time scale across both continuous total organ dose and categorical dose, using time-dependent, annualised, total cumulative absorbed dose (table 1). We used a single, total organ dose for our main analyses because of the very small internal dose component to most organs except the lung and to enable direct comparisons with other MPS cohorts.14 21 Because radon exposure contributed to high internal lung doses, we assessed differential effects of external gamma and internal alpha radiation on all lung outcomes, with and without adjustment for internal and external exposures, respectively. Cox regression and SMRs were conducted using Stata V.17 (StataCorp).
HRs for selected causes of death for categorical and continuous organ-specific doses for 6403 Fernald employees employed for at least 30 days, 1951–1985 and followed through 2017
We conducted linear Poisson regression analyses using the AMFIT programme in Epicure.22 We used the DATAB programme to create person-tables tabulated over sex, pay type (hourly/salaried), organ dose (23 categories with cut points at 10/20/30/40/50/60/70/80/90/100/110/120/130/140/150/200/250/300/350/400/450/500 mGy), age (14 categories with cut points in 5-year intervals from 20 to 80), age-at-hire (8 categories with cut points in 5-year intervals from 20 to 50), binary non-radiological exposures of interest (all dusts, laboratory chemicals and asbestos exposures), binary emergency conditions exposure and follow-up time (14 categories with cut-points in 5-year intervals from 1955 to 2015). The excess relative risk (ERR) for radiation was specified as:
We did not report a lower bound when the lower bound dose parameter fell below the parameter space boundary (−1/maximum dose).
Common endpoints such as CVD mortality may have distorted effect estimates in the presence of competing outcomes.23 24 We conducted competing risk analysis to interrogate this possibility for CVD and ischaemic heart disease (IHD), the most common causes of death in this cohort. We tested two extreme assumptions: the Fine and Gray model, for which all employees who died of competing causes were not censored until end of follow-up (‘maximum CVD survival assumption’),25 and a model for which we assumed all employees who died of competing causes would have died of CVD/IHD on the same day otherwise (‘minimum CVD survival assumption’). We conducted these analyses in Stata V.17, using graphs described by Lunt.26
We used directed acyclic graphs (DAGs) to select potential confounders27 based on exposures known to impact the outcome of interest, or where the outcome of interest occurred in a target organ for that exposure,7 28 shown in figure 1. DAG-selected confounders included pay code (at time of hiring), birth year, emergency conditions exposure and DAG-specific concurrent exposures. Poisson models additionally adjusted for log(age/50) because age was not implicit in person-time, which was defined using calendar year.
DAGs representing the effect of radiation exposure on outcomes of interest in Fernald workers. Variables at the base of solid lines represent measured or partially measured factors in each cohort; variables at the base of dashed lines represent unmeasured factors. DAG (A) presents cardiovascular outcomes and DAG (B) presents all other outcomes. The only difference between these DAGs is the node for concurrent exposures in DAG (B). Outcome-specific concurrent exposures were non-radiation exposures that increased with duration of work and were either (1) known to impact the outcome of interest or (2) where the outcome of interest occurred in a target organ for that exposure.7 28 Concurrent exposures refer to asbestos and dust exposures for lung outcomes; dust exposures for colorectal and stomach outcomes; dust and chemical exposures for kidney outcomes and chemical exposures for brain and liver outcomes. DAGs, directed acyclic graphs.
Because sex is a complex construct representing both biological and societal factors,29 we stratified the baseline hazard rates on sex to reflect sex-specific differences without implicit confounding assumptions. The competing risk analysis adjusted for sex in the model because Fine and Gray model implementation did not allow baseline hazard stratification, which did not greatly impact results.
When proportional hazards assumptions were not met for a particular variable, the variable was simplified if appropriate (eg, categorical, dichotomous), otherwise we stratified baseline hazards by the variable.
Results
Table 2 displays demographics for 6403 Fernald workers contributing 268 579 person-years, and online supplemental figure 1 shows vital status tracing. We did not receive race information from NIOSH; therefore, the racial distribution differs slightly from Silver et al.7 Nonetheless, most workers were white males (5273 (82.3%)). Most workers were hired in the 1950s (4360 (68.1%)). Median length of employment was 5.7 years; longest length of employment was 53.7 years. Median length of follow-up was 42.4 years (range=3 days to 66.9 years). Most workers were paid hourly (3814 (59.6%)). Vital status was known for 6283 (98%) workers, and 4134 (>99%) deceased workers had a confirmed cause of death.
Demographic and occupational characteristics for 6403 Fernald Feed Materials Production Center workers employed for at least 30 days, 1951–1985 and followed through 2017
Online supplemental table 2 shows monitoring status for radiation and non-radiation exposures. All employees had some measured or estimated dose. Most Fernald employees had measured external photon exposure (4080 (63.7%)) and internal uranium intakes (5996 (93.6%)), and all had radon estimates. Nearly half the cohort had exposures to dusts (2798 (43.7%)), including 1378 (21.5% of the total cohort) exposed to uranium dust (not shown), 91 (1.4%) exposed to coal dust and 2082 (32.5%) exposed to miscellaneous dusts.
Organ doses for individuals exposed to radiation are shown in online supplemental table 3. Total lung dose mean (212.6 mGy) and median (28.0 mGy) were moderate. High radon exposures were not uncommon; 435 employees had estimated radon-associated doses >500 mGy, contributing to a highly right-skewed internal lung dose distribution. Doses were lower and distributions less skewed for other organs; for example, mean and median heart dose were 12.0 and 1.0 mGy, respectively. Females consistently had lower doses than males.
Online supplemental table 4 displays overall SMR results for a priori selected outcomes with ≥5 cases (for confidentiality). Online supplemental tables 5 and 6 show SMRs for males and females, respectively; online supplemental tables 7 and 8 show SMRs for salaried and hourly workers, respectively. Overall mortality was significantly decreased for all workers together (SMR=0.94; 95% CI=0.91 to 0.97), though mortality among hourly workers was not different from expectation (SMR=1.01; 95% CI=0.98 to 1.05). Only cancers of the pleura, peritoneum and mesothelioma had a statistically significant SMR (SMR for all workers=2.73; 95% CI=1.49 to 4.59). Results were similar when stratified by sex, despite fewer outcomes with ≥5 cases among females. Outcomes were generally attenuated among salaried workers and increased among hourly workers.
Table 1 displays results from categorical and continuous Cox regression analyses over total cumulative absorbed organ dose. Lung cancer had a flat dose response (HR at 100 mGy=1.00; 95% CI=0.98 to 1.01), as did NMRD and lung cancer combined with NMRD, though categorical results indicated elevated NMRD above 250 mGy (HR=1.47; 95% CI=1.00 to 2.16). Emphysema had elevated categorical estimates though the p value for trend was 0.08. Elevated responses were also detected for CVD (HR at 100 mGy=1.27; 95% CI=1.07 to 1.50) and IHD (HR at 100 mGy=1.30; 95% CI=1.07 to 1.58). Other outcomes including kidney, liver and stomach cancer; non-chronic lymphocytic leukaemia (‘non-CLL leukaemia’); non-Hodgkin’s lymphoma and dementia, Alzheimer’s, Parkinson’s, and motor neuron diseases had non-significantly elevated estimates with wide CIs. Both colorectal cancer and non-malignant kidney disease (NMKD) had non-significantly decreasing HRs, though corresponding categorical analyses had both elevated and attenuated estimates depending on dose category.
Online supplemental table 9 presents Poisson regression results. Maximum likelihood estimate CIs were wide for all outcomes, and lower bounds were frequently outside the parameter space boundary. Nonetheless, Poisson models were generally consistent with Cox models.
Table 3 displays the results of lung outcome analyses after regressing on total dose, external dose and internal dose. All lung outcomes had a positive dose response with increasing external dose; for instance, the lung cancer HR at 100 mGy adjusted for internal dose was 1.45 (95% CI=1.05 to 2.01). Emphysema had a non-significantly increased risk with increasing internal dose (HR at 100 mGy adjusted for external dose=1.11; 95% CI=0.78 to 1.58).
HRs for lung outcomes over continuous total, external and internal lung dose for 6403 Fernald employees employed for at least 30 days, 1951–1985 and followed through 2017
Table 4 shows competing risk analysis results, with HRs for the main analysis and the equivalent subdistribution hazard functions for the competing risk analyses. The HR was attenuated for both CVD and IHD under the minimum CVD survival assumption (HR at 100 mGy for both outcomes=1.08; 95% CI=0.97 to 1.20). Under the maximum CVD survival assumption, the HR was higher (HR at 100 mGy for CVD=1.43; 95% CI=1.21 to 1.69; IHD=1.39; 95% CI=1.14 to 1.69). Online supplemental figure 2 illustrates the difference in CVD survival by heart dose categories using a Kaplan-Meier curve unadjusted for any covariates and competing risks-corrected curves for minimum and maximum CVD survival assumptions. All curves indicate an association between heart dose and CVD, but the maximum CVD survival curve has a steadily increasing gap between the 50–100 mGy and ≥100 mGy categories while survival is similar by dose in the minimum CVD survival curve.
HRs and subdistribution hazard functions for cardiovascular outcomes in 6403 Fernald employees employed for at least 30 days, 1951–1985 and followed through 2017
Discussion
Similar to the Silver et al follow-up of the Fernald cohort,7 we observed greater mortality in hourly than salaried workers. Despite considerably different dose-response models, our Cox and Poisson effect estimates were also similar to those of the previous study. For example, Silver et al reported an HR at 100 mGy external dose (adjusted for internal dose) for intestinal cancer of 0.57 (95% CI=0.18 to 1.4), which is comparable to our HR at 100 mGy for colorectal cancer from all sources combined (HR=0.51; 95% CI=0.22 to 1.20). This similarity is reassuring, as it shows the results were robust to differences in both the definition of colorectal cancer and in the dose-response models. Estimates for stomach cancer, non-CLL leukaemia and non-Hodgkin’s lymphoma were likewise non-significantly elevated in both reports.
Also in agreement with the previous study were the positive associations we observed for lung cancer, NMRD and emphysema over increasing external but not uranium or radon dose.7 In both studies, the outcomes chronic obstructive pulmonary disease (COPD) and emphysema can be used as negative controls because they are typically considered smoking—but not radiation induced. The positive association between external and internal radiation and emphysema in this study (and COPD in the previous study) implies uncontrolled confounding for smoking.5 30 Although results may be unstable because only 34 workers died of emphysema, their similarity to COPD from the previous study is reassuring. However, emphysema may have uncontrolled confounding from non-smoking workplace exposures.
Our study expanded analyses to include kidney, liver and brain and CNS cancer, as well as NMKD; however, none of these outcomes were significantly associated with radiation exposure.
Although other occupational cohorts have reported lung cancer associations with low dose-rate radiation, analytical and exposure circumstances were frequently different. Cohorts of nuclear workers and medical radiation workers reporting radiation-lung cancer associations have not adjusted for smoking.5 31 However, a separate large study of radiologic technologists that adjusted for smoking did not detect increased lung cancer mortality with increasing external dose.32 In this and the previous study, the lack of a lung cancer association with radon, an established lung carcinogen from mining and residential exposure studies,1 5 implies bias such as residual smoking confounding. Non-differential radon exposure misclassification is also possible because radon exposure was estimated entirely using a model (independent of outcome status), which may have biased estimates toward the null.
The previous Fernald follow-up did not examine the relationship between radiation and CVD. In this follow-up, we observed an increasing hazard for CVD and IHD mortality with increasing dose. Increased IHD risk at low-dose rates of radiation has not been detected in previous large MPS cohorts21; however, systematic reviews, meta-analyses and several studies of uranium-exposed gaseous diffusion workers have reported positive associations between low-dose rate radiation and CVD.2 3 Other analyses of uranium processing workers have also reported a slightly increased risk of IHD with increasing dose.14 33 34
To our knowledge, our use of competing risk assessment to further investigate these relationships is novel in occupational radiation epidemiology. Most competing risk analyses in radiation epidemiology have been in medical contexts as a decision-to-treat tool.35–38 In occupational settings, competing risks-adjusted estimates may help clarify the potential magnitude of a hazard on a specific outcome of interest, particularly if competing outcomes are expected to decrease over time. CVD is the most common cause of death both in this cohort and the US population,39 and cancer mortality (a competing risk) is in decline.40 Therefore, radiation-induced CVD may soon become more relevant. Competing-risks adjusted estimates in Fernald and other retrospective cohorts may predict radiation dose responses in future populations by showing the extreme boundaries between which a dose response may be anticipated in a future with less cancer mortality. For CVD, this range would be an HR at 100 mGy between 1.08 and 1.43, with an estimate from standard Cox regression of 1.25. Online supplemental figure 2 further illustrates this application.
Additional strengths of this study include the detailed dosimetry with organ/tissue-specific doses for all outcomes of interest and using extensive urinalysis data,15 a long period of follow-up with over 98% confirmation of mortality and over 99% known causes of death, and use of DAGs to enable an a priori analytical plan accounting for worker characteristics and non-radiological exposures.
Although we assessed lung outcomes separately by external and internal dose, internal dose distributions comprised a very small portion of dose for all other organs. Because of uncertainty in internal dose calculations, we were concerned about the validity of assessing a dose response at smaller increments than an increase in 100 mGy, which would have resulted in tighter confidence bounds because of the small unit size. Dosimetric uncertainties included simplifying assumptions such as aerosol size, lack of concomitant radionuclide and transuranic estimations, and inadequate dose-rate representation. Additionally, residual confounding may be present from dichotomisation of non-radiological exposures. Smoking has been previously implicated as an uncontrolled confounder for a positive relationship seen between radiation and CVD,41 though the thorough competing risk assessment lends credence to our main findings. Due to small sample size and limited power from a fairly narrow dose distribution, we did not assess temporal effect modification. Because of the low number of deaths among specific causes in females (online supplemental table 6) combined with the even narrower dose distribution than the total population (online supplemental table 3), we did not assess effect modification by sex. However, our inclusion of females and people of colour in this analysis improved the power of this study.
Workers at Fernald exposed to uranium, external radiation, radon and various non-radiological toxins were followed up from 1951 to 2017. A new finding in this study was an association between both CVD and IHD and increasing radiation dose, which was further explored in novel competing risk analyses. Although lung cancer was not associated with total radiation dose, it was associated with external dose, though positive emphysema dose responses imply potential residual confounding by smoking. Next steps will be to pool Fernald with other uranium processing cohorts6 and obtain individual smoking histories from linkages to Centers for Medicare and Medicaid records.21
Data availability statement
Data are available in a public, open access repository. The datasets generated during and/or analysed during the current study are available in the Comprehensive Epidemiologic Data Resource (CEDR) repository, https://oriseapps.orau.gov/CEDR/search_results.aspx?DataSet=FMM23A01.
Ethics statements
Patient consent for publication
Ethics approval
This study involves human participants and human subjects research approval was received from both the Vanderbilt University Institutional Review Board and the Department of Energy Central Institutional Review Board, under HSRD VANU-12-VANU, ID 329174. Data were only obtained through retrospective records and linkages (ie, no contact with participants), and the research involves no more than minimal risk.
Acknowledgments
We would like to recognise and thank everyone who has contributed to this work through participation and contribution to the Million Person Study. We deeply thank Dr Jeri Anderson, who published the previous Fernald follow-up and provided support and guidance throughout the data preparation, analysis, and writing phase of this effort. We recognise and thank the Fernald employees, whose contributions are helping the world better understand radiation health effects. We would also like to thank Dr Melinda Aldrich and Dr Loren Lipworth, who provided insights and expertise on this paper as dissertation committee members for CMM, Mr David Girardi, who contributed considerably to data preparation and Dr Dale Preston, who worked regularly with CMM to help him better understand Epicure.
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.
Footnotes
Contributors All authors contributed to study design and analytical plan. MTM completed vital status and cause of death determination. MBB conducted the dosimetric update for the present follow-up. CMM completed data cleaning in consultation with SCH, EDE, APG and SSC. CMM conducted statistical analyses in consultation with BF, LBZ and JDB. CMM and MBB composed the first draft of this paper. CMM acted as guarantor for for the overall content of the paper. All authors edited the manuscript and provided feedback prior to manuscript finalisation.
Funding This work was supported in part by grants awarded to the National Council on Radiation Protection and Measurements from the US Department of Energy (grant numbers DE-AU0000042 and DE-AU0000046). Contract support was additionally received by Oak Ridge National Laboratory from the Office of Radiation and Indoor Air, US Environmental Protection Agency through Interagency Agreement DOE No. 1824 S581-A1, under contract number DE-AC05-00OR22725 with UT-Battelle. Contract support was also received by Oak Ridge Associated Universities from the US Department of Energy through contract number DE-SC0014664. LBZ’s work was funded by National Institute for Occupational Safety and Health and National Institutes of Health award (5R21OH011452) (principal investigator: LBZ).
Competing interests None declared.
Provenance and peer review Not commissioned; externally peer reviewed.
Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.