Mortality among Tennessee Eastman Corporation (TEC) uranium processing workers, 1943–2019

Abstract Background There are few occupational studies of women exposed to ionizing radiation. During World War II, the Tennessee Eastman Corporation (TEC) operated an electromagnetic field separation facility of 1152 calutrons to obtain enriched uranium (235U) used for the Hiroshima atomic bomb. Thousands of women were involved in these operations. Materials and methods A new study was conducted of 13,951 women and 12,699 men employed at TEC between 1943 and 1947 for at least 90 days. Comprehensive dose reconstruction techniques were used to estimate lung doses from the inhalation of uranium dust based on airborne measurements. Vital status through 2018/2019 was obtained from the National Death Index, Social Security Death Index, Tennessee death records and online public record databases. Analyses included standardized mortality ratios (SMRs) and Cox proportional hazards models. Results Most workers were hourly (77.7%), white (95.6%), born before 1920 (58.3%), worked in dusty environments (57.0%), and had died (94.9%). Vital status was confirmed for 97.4% of the workers. Women were younger than men when first employed: mean ages 25.0 years and 33.0 years, respectively. The estimated mean absorbed dose to the lung was 32.7 mGy (max 1048 mGy) for women and 18.9 mGy (max 501 mGy) for men. The mean dose to thoracic lymph nodes (TLNs) was 127 mGy. Statistically significant SMRs were observed for lung cancer (SMR 1.25; 95% CI 1.19, 1.31; n = 1654), nonmalignant respiratory diseases (NMRDs) (1.23; 95% CI 1.19, 1.28; n = 2585), and cerebrovascular disease (CeVD) (1.13; 95% CI 1.08, 1.18; n = 1945). For lung cancer, the excess relative rate (ERR) at 100 mGy (95% CI) was 0.01 (–0.10, 0.12; n = 652) among women, and −0.15 (–0.38, 0.07; n = 1002) among men based on a preferred model for men with lung doses <300 mGy. NMRD and non-Hodgkin lymphoma were not associated with estimated absorbed dose to the lung or TLN. Conclusions There was little evidence that radiation increased the risk of lung cancer, suggesting that inhalation of uranium dust and the associated high-LET alpha particle exposure to lung tissue experienced over a few years is less effective in causing lung cancer than other types of exposures. There was no statistically significant difference in the lung cancer risk estimates between men and women. The elevation of certain causes of death such as CeVD is unexplained and will require additional scrutiny of workplace or lifestyle factors given that radiation is an unlikely contributor since only the lung and lymph nodes received appreciable dose.


Introduction
From June 1943 to May 1947, the Tennessee Eastman Corporation (TEC) operated a uranium processing facility that employed nearly 40,000 workers in Oak Ridge, TN. Y-12 was the wartime code name given to the facility. Enriched uranium ( 235 U) was obtained by electromagnetic processing and was then sent to Los Alamos National Laboratory (LANL) and used to develop atomic weapons, including the Hiroshima atomic bomb (Polednak and Frome 1981). The mass spectrometer separation process resulted in many TEC workers being exposed to varying amounts of 238 U, 235 U and their progeny. These radionuclides are primarily alpha-particle emitters, and the primary health concern was from the inhalation of uranium dust. The exposure to external radiation was minimal (Polednak et al. 1982;Beck et al. 1983;Dupree et al. 1995). There are few occupational studies of women exposed to ionizing radiation and even fewer investigations that have evaluated intakes of inhaled radionuclides or uranium dust (ATSDR 2013;UNSCEAR 2017). Thus, the TEC operation offers a unique opportunity to study a sizable number of women who have never been studied before. The women employed at this facility were popularized in the book 'The Girls of Atomic City' (Kiernan 2013).
The electromagnetic field separation of the uranium isotopes was done by 1152 calutrons, named for the California University cyclotron and designed by Ernest O. Lawrence. The calutrons were housed in nine buildings. Five buildings were designated as 'alpha' buildings and four as 'beta' buildings. The designations were not related to radioactive emissions and were just building names.
TEC operated the electromagnetic separation process from 1943 to 1947 and employed the largest number of workers of all the early uranium processing plants involved in the Manhattan Project. The plant converted uranium oxide (UO 3 ) received from the Mallinckrodt Chemical Works in Missouri to uranium chloride (UCl 4 ). The UCl 4 was then enriched, i.e. the percentage of 235 U was increased by the calutron (mass spectrometer) electromagnetic separation process. The separation process involved two stages ('alpha' and 'beta'), each consisting of multiple calutron separators. The alpha stage was discontinued in late 1945, when enriched uranium fluoride (UF 6 ) was received from the Oak Ridge gaseous diffusion plant (K-25), converted to UF 4 , enriched further by the beta calutrons, and shipped to LANL for development of nuclear weapons.
White male workers at TEC were previously studied through 1973 (Polednak and Frome 1981;Cookfair 1982;Cookfair et al. 1983), and as part of a four-site case-control study of lung cancer, including female workers, based on follow-up through 1982 (Dupree et al. 1995). The initial studies suggested an association between lung cancer and lung dose (Polednak et al. 1982;Cookfair et al. 1983), but this was not confirmed in a later study that also accounted for tobacco use (Dupree et al. 1995). The current investigation builds upon these previous studies, extends the followup of male workers through 2018 (for an additional 45 years), improves upon the estimation of lung dose from uranium dust inhalation, enhances the evaluation of tobacco use, includes a female TEC cohort for the first time, and includes a separate analysis of black women. As occupational exposure is from inhalation of airborne uranium dust with essentially no external radiation, the focus is on lung cancer, NMRD and NHL as possible late-occurring health outcomes. Herein, we report lifetime risks for all causes of death and sex-specific radiation risk estimates for these outcomes among workers employed between 1943 and 1947 for more than 90 days at the TEC uranium conversion and enrichment plant. The TEC cohort is a component of the Million Person Study (MPS) of U.S. Radiation Workers and Veterans evaluating health effects following low-level chronic (low-dose rate) exposures (Boice, Cohen, Mumma, & Ellis 2022; Boice, Quinn, Al-Nabulsi, Ansari, et al. 2022).

Materials and methods
Human subjects research approval was received from the Oak Ridge Site-Wide Institutional Review Board and the Vanderbilt University Institutional Review Board.

Study population
The previous TEC studies restricted the cohort to white males who were employed only at the TEC facility and not subsequently at any other Oak Ridge facility (Polednak 1980;Polednak and Frome 1981;Polednak et al. 1982). The current study extended the follow-up of this white male population; and includes all women, white and black, who were employed at the TEC facility. The study does not exclude women who transferred to the newer Y-12 plant operated by the Union Carbide Corporation or who transferred to any other Oak Ridge facility such as Oak Ridge National Laboratory (X-10) or the gaseous diffusion plant (K-25). Overall, there were 20,434 women first employed at TEC between 1943 and 1947. Women employed for fewer than 90 days (n ¼ 6483) were excluded, leaving an eligible population of 13,951 female workers. The initial male population consisted of 18,869 workers. After excluding those who worked fewer than 90 days (n ¼ 6170), 12,699 male workers remained eligible for study. In total, the study population comprised 26,650 workers.

Vital status and outcome determination
The procedures described in Mumma et al. (2022) were followed to determine vital status as of 31 December 2019 for the TEC women and 31 December 2018 for the TEC men. Probabilistic matching programs, such as LinkPlus (Campbell et al. 2008), were used to obtain date of death and cause of death from state and national sources of vital status information, including the National Death Index (NDI), the California Death Statistical Master File, and the Social Security Administration (SSA) Death Master File. Alive status was confirmed from the SSA Service for Epidemiological Researchers. To confirm or correct identifying information used in the matching procedures (e.g. name, social security number and date of birth), TLO, a subscription service of TransUnion (www.tlo.com) was used. Individual one-on-one searches also were employed to locate study subjects using resources such as Ancestry.com and Google obituary search. Unique databases in Tennessee were also searched on a one-on-one basis because of the difficulty in tracing young unmarried women who worked in the 1940s.

Radiation dose reconstruction
The radiation dose reconstruction approach followed the guidance set forth in NCRP Report 178 (2018) for deriving organ doses and their uncertainty specifically for the MPS. The manual nature of the operations at the TEC plant resulted in many workers being exposed to varying amounts of uranium compounds of 234 U, 235 U, 238 U and their progeny. Since these uranium isotopes are chiefly alpha-radiation emitters and inhalation was the primary mode of exposure, internal doses to the lung and thoracic lymph nodes (TLNs) were most important when considering health risks. External radiation was not expected to be consequential due to relatively low energies of photons emitted by natural uranium isotopes, and no film badges or other measurements were used. The internal radiation dosimetry for TEC workers built upon the foundational work of previous dose reconstructions (Polednak and Frome 1981;Cookfair 1982;Polednak et al. 1982;Beck et al. 1983;Cookfair et al. 1983;Dupree et al. 1995). The present dose estimations were based on air monitoring data for locations near primary personnel work areas at the two processing facilities (i.e. the alpha and beta buildings), and accounted for the likely chemical and physical forms of uranium. The highest average concentrations of airborne uranium dust (300 and 250 lg/m 3 ) were measured in the chemistry department in the alpha building and were associated with the sublimation (i.e. the transition from a solid to a gaseous state) and bottle filling areas, respectively. Concentrations of airborne uranium in several beta chemistry areas averaged 50 l g/m 3 (Polednak and Frome 1981).
An estimate of annual radiation absorbed doses to the lung from inhaled uranium was obtained for each worker. First, to obtain estimates of the uranium intakes, and recognizing that many TEC employees worked in more than one department (and some changed jobs within departments), an employee's total work experience was divided into groups according to specific job factors: job title, job code, department, location, and calendar date range. Second, these work experiences were linked to appropriate average airborne uranium dust concentration measurements (Dupree et al. 1995) accounting for the changes in enrichment percentages that occurred over time (e.g. for beta areas the uranium enrichment changed from 13% before August 1945 to 17.8% by October 1945, and ultimately to 25% thereafter). Third, the job-specific exposure concentrations were multiplied by days worked in that specific job to obtain daily integrated inhaled activity of the uranium isotopic mixture. Finally, these activities were converted to a lung absorbed dose in accordance with current ICRP internal dosimetry methodology for 234 U, 235 U, and 238 U (ICRP 2007(ICRP , 2017 and were based on daily intake estimates and job start and end times. The analyses assumed an aerosol size with an activity median aerodynamic diameter (AMAD) of 5 lm, a breathing rate of 27.3 m 3 /d for females and 28.8 m 3 /d for males based on ICRP Publication 66 (ICRP 1994), absorption type M/S, and 6h/d of exposed time. Thus, the female and male daily air intakes while at work would be 10.4 and 12.6 m 3 , respectively. Similar procedures were used to estimate individual doses to TLNs.
AMAD is the particle size in an aerosol where 50% of the activity is associated with particles of aerodynamic diameter greater than the AMAD. Clearance of the deposited chemical form of the inhaled material is characterized by the absorption type parameter. Compounds that are readily absorbed are referred to as type F compounds. No uranium compounds were designated as F at any of the TEC facilities. Type M compounds have intermediate rates of absorption into blood, and type S compounds are relatively insoluble and slowly absorbed. This analyses utilized an absorption type M/S in accord with ICRP Publication 137 that notes that 238 U and U dioxide compounds have absorption rates between those of default type M and S. Sensitivity analyses were conducted to consider the impact of these assumptions (i.e. considering alternate assumptions of AMAD of 1 or 10lm; breathing rate of 14.5 or 51.1 m 3 /d; absorption type S or M, and exposure times of 4 h or 8 h per day).

Career doses
Dosimetry records documenting radiation exposure received before or after employment at TEC (1943TEC ( -1947 were obtained following the procedures outlined in  and Ellis et al. (2018). There were only 315 women and 333 men who received occupational exposures elsewhere, and the additional contribution to lung dose, primarily from gamma radiation, was small.

Smoking history
Tobacco use for individual workers is often unavailable or too expensive to obtain in retrospective epidemiologic studies. Smoking data, however, were available in the TEC occupational medical records and were abstracted and considered in two previous overlapping studies that included TEC workers Dupree et al. 1995).
Smoking histories were sought for 641 women who died of lung cancer and for a comparison group of 707 women who did not die of lung cancer (selected to be 10% larger than the case group to ensure adequate data collection for missing medical records). The level of detail of smoking information was 'tobacco use Y/N.' During the abstraction process, information on chest X-ray examinations also was collected.
A similar evaluation of tobacco use was conducted among 1002 men who died of lung cancer and an approximate 10% sample of the 12,699 men who did not die of lung cancer as comparison subjects (n ¼ 1107). The number of male lung cancer deaths was larger than in previous studies because our follow-up was longer by up to 35 years through 2018 compared with 1983 (Dupree et al. 1995), and was nearly to the end of life as 98.6% of the men had died. Information on chest X-ray examinations was not collected because the results from the women indicated only one chest X-ray on average.

Statistical analysis
The start of follow-up began 90 days after date of first hire at TEC (1943TEC ( -1947, and ended at the date of death, age 95, date lost to follow-up, or 31 December 2019 for women and 31 December 2018 for men, whichever came first. Standardized mortality ratios (SMRs) for over 50 causes of death were computed as follows. The observed numbers of cause-specific deaths were compared with the expected numbers based on U.S. mortality rates accounting for age and calendar year in five year categories, and sex. SMRs were further computed within pay category (hourly/salary) work area (alpha, beta, other), age at hire, race and years of follow-up. Exact 95% Poisson confidence intervals (CIs) for the SMRs were computed (Rothman and Boice 1979).
Cox proportional hazards models were used in the doseresponse analyses to compute organ-specific radiation risks (Cox 1972). Such internal cohort analyses are conducted to address the 'healthy worker effect' often seen in occupational studies (Monson 1986). Cox models were conducted using SAS/STAT software (SAS/STAT software, Version 9.4 of the SAS System for Windows, SAS Institute Inc., Cary, NC). Excess relative rates (ERRs) at 100 mGy were estimated using the Peanuts program in Epicure (Preston et al. 2015). Year of birth and a measure of socioeconomic status (SES), i.e. pay category (hourly/salary), were included in all models as was race which was available for women. Pay category was selected as the measure of SES to indirectly account for smoking and other life choices. Previous studies of TEC workers found that pay category was correlated with tobacco use (Dupree et al. 1995). Questionnaire interviews at the time of employment were available in medical records to validate the use of pay category as a reasonable surrogate for tobacco use. Internal dose-response analyses were conducted for lung cancer among the 1168 black women who worked at TEC. For female workers, both race and pay category were included as adjustment factors in all models. Conceivably, collinearity might exist which could distort results. To address this possibility, a sensitivity analysis was conducted only adjusting for pay category and then only adjusting for race. Results were not meaningfully different so both factors were kept as adjustments in the dose-response analyses.
For the internal cohort analyses, radiation dose was treated as a categorical measure, with categories based on the distribution of dose to lung. The dose category was treated as a time-dependent measure, allowing workers to be assigned to increasingly higher dose categories over time as their individual organ-specific radiation doses accrued. To allow for a possible latent period between radiation exposure and any consequent effect, doses were lagged, i.e. excluded if they occurred during some assumed interval prior to the event of interest, 10 years for lung cancer and also for nonmalignant respiratory disease (NMRD) excluding flu and pneumonia (NMRD), and for NHL. Parameter estimates and standard errors for the dose categories in the Cox timedependent regression models were used to obtain hazard ratios (HRs) and 95% CIs for death due to lung cancer, NMRD and NHL compared with those in the referent group taken as the workers with low cumulative radiation dose. Trend tests treated the radiation dose as a single continuous time-dependent measure, and two-sided p values are presented. Dose weighting factors (DWFs) of 1, 10 and 20 were applied to the alpha-particle absorbed doses as possible indicators of biological effectiveness. ERRs were estimated using the same model specifications as the Cox time-dependent regression model. Sensitivity analyses addressed changing the minimum duration of employment required for inclusion; excluding workers with lung dose ¼ 0 and using >0 to <5 mGy as the referent category; restricting the analysis of male workers to those with lung doses <300 mGy; examining different dose-lag intervals; and varying the assumptions with regard to the uranium dust parameters used in the absorbed lung dose calculations. Malignant and nonmalignant lung disease and NHL were the primary outcomes of interest because of the substantial contribution of airborne uranium dust to lung dose and TLN dose. Due to the nature of the uranium conversion and enrichment processes relying on mass spectrometry, there was little external radiation exposure, and no other organs received more than a tiny dose from intakes of uranium (Polednak and Frome 1981).

Results
Descriptive characteristics of the TEC workers are shown in Table 1. There were 26,650 workers employed 1943-1947 at TEC for at least 90 days. Follow-up was through 2019 for the 13,951 female workers (52.3%) and through 2018 for the 12,699 male workers (47.7%). The total person-years of observation were 1,199,423 for an average of 45.0 years. The majority of workers were hourly workers (77.7%), white (95.6%), born between 1910 and 1929 (73.5%), under age 30 when first hired (60.4%), worked for more than one year (55.5%), and had died (94.9%). Fifty-seven percent of workers were employed in buildings with dusty environments (alpha or beta processing, or alpha chemistry). Through the end of follow-up, 672 workers (2.5%) were confirmed alive, and cause of death was determined for all but 473 (1.9%) of the 25,288 workers who had died. Only 690 (2.6%) of the workers were lost to follow-up ( Figure 1). Table 2 presents the cumulative lung doses and the TLN doses among the female and male workers. These dose estimates were used in the main analysis of dose response for lung disease and non-Hodgkin lymphoma (NHL). The preferred analyses assumed a particle size of 5 lm AMAD, breathing rate of 27.3 m 3 /d for females and 28.8 m 3 /d for males (ICRP 1994), solubility class of M/S, and 6 h/d of exposed time. For a DWF of 1, the mean dose to lung was 32.7 mGy and the maximum was 1.05 Gy for women. Assuming a DWF of 20, the mean is 591 weighted-mGy lung dose with a maximum of 18.5 weighted-Gy lung dose. For men, the mean absorbed dose to lung was 18.9 mGy (maximum 0.50 Gy). For a DWF of 20, the mean is 352 weighted-mGy lung dose with a maximum of 8.91 weighted-Gy dose to lung. The mean dose for DWF ¼ 20 is a bit lower than what would be computed by multiplying the mean dose for DWF ¼ 1 by 20 because the total dose computations also incorporate photon doses which do not change when the uranium dose is multiplied by any DWF. Thus, the mean dose for DWF ¼ 20 will always be lower than the product of 20 times the mean dose with DWF ¼ 1 given concomitant photon contributions to dose. The mean dose to TLN was 127 mGy, and the patterns between women and men were similar to those for lung dose, except that the TLN measures of dose were about five times that of the lung doses. Women had higher estimates of cumulative dose in large part because they worked longer in dusty environments than their male counterparts. Sensitivity analyses were conducted using different assumptions for the four dose parameters used in the dose estimation procedure. Realistic low-sided and high-sided estimates for the four parameters were applied for particle size, breathing rate, duration of daily exposure, and the solubility of the inhaled uranium (Supplemental Table 1). For example, compared with the AMAD particle size of 5 lm used in the main analysis, the dose to lung would be two times higher for a 1 lm assumption and 50% lower for a 10 lm assumption.
The SMRs for 59 causes of death among the 13,951 TEC women and the 12,699 men are shown in SMR analyses by pay category (hourly/salary) and sex were also conducted (Supplemental Table 2). The 20,696  IHD was significantly higher among hourly workers compared with salaried workers as were deaths from accidents. For smoking-related cancers, the salaried male workers had a significantly low SMR (0.80; n ¼ 251) in contrast to significantly high SMRs among hourly male workers (1.27; n ¼ 1123), hourly female workers (1.14; n ¼ 722), and salaried female workers (1.17; n ¼ 253). This pattern was particularly noticeable for lung cancer, smoking-related cancers, and NMRD. Hourly female workers had a significantly low risk of breast cancer (0.89; n ¼ 303) and a significantly high risk of cervical cancer (1.32; n ¼ 68). These comparisons indicate the importance of adjusting for pay category in the internal dose-response analyses. Table 4 presents the internal dose-response analysis for lung cancer over categories of radiation dose to the lung by sex for three different DWFs of 1, 10, and 20. The estimates of risk, ERR per 100 weighted-mGy dose to lung, did not change when the DWF was increased from 1 to 20. However, for DWF ¼ 20, the number of workers in the high-dose category (>250 weighted-mGy lung dose) did increase appreciably by 9137 and the number of lung cancer deaths by 595. There was no evidence for an increased risk of lung cancer among either women or men. For a DWF of 1, the ERR per 100 mGy lung dose (95% CI) was 0.01 (-0.10; 0.12) for females, À0.22 (-0.43; À0.01) for males, and À0.09 (-0.19, 0.02) for all workers. Graphical representation of the dose response for all workers is presented in Figure 2, for women in Figure 3, and for men in Figure 4 for DWFs 1 and 20. Figure 4(c) presents an analysis of male workers with lung absorbed doses <300 mGy, i.e. it excludes 25 workers (0.2% of the male population) with very high lung absorbed doses, >300 mGy. This was done in part because of concern that bias in the dust to dose conversion process for a few workers in the highest dose categories might have overestimated the lung doses and disproportionately influence the dose response which was significantly negative. Excluding these few workers (none who died of lung cancer) increased the estimated ERR per 100 mGy from À0.22 (95% CI À0.43; À0.01; n ¼ 1002; p for trend ¼.042) to À0.15 (95% CI À0.38, 0.07; n ¼ 1002; p for trend ¼.18) and the trend was no longer significantly negative.
Three previous studies of TEC workers reported a radiation-related increase of lung cancer among workers first hired over the age of 45 years, although not statistically significant (Polednak et al. 1982;Cookfair et al. 1983;Dupree et al. 1995). The current study included of 2139 workers over age 45 years when first hired; 286 were women and 1853 were men (Table 1). The SMR for lung cancer among all workers hired at age 45 years or older was 1.16 (95% CI 0.94, 1.43) based on 90 lung cancer deaths. For workers hired at ages 15-24 years (n ¼ 11,448) and 25-44 years (n ¼ 10,565), lung cancer SMRs were 1.28 (95% CI 1.14, 1.32; n ¼ 737) and 1.28 (95% CI 1.20, 1.37; n ¼ 827), respectively (data not shown). An internal dose-response analysis for male workers first hired after age 45 years found the ERR (95% CI) at 100 mGy to be 0.30 (95% CI À0.43, 1.02). The ERRs (95% CI) at 100 mGy for male workers hired at ages 15-24 years and 25-44 years were À0.41 (95% CI À0.85, 0.04) and À0.19 (95% CI À0.45, 0.06), respectively (Supplemental Table 3). The number of female workers hired after age 45 years was only 286 and too small for any meaningful analyses. Table 5 presents the internal dose-response analysis for NMRD, excluding flu and pneumonia, over categories of radiation dose to the lung by sex for three different DWFs of 1, 10, and 20. Similar to the estimates of radiation risk for lung cancer, the estimates for NMRD did not vary by DWF. There was no evidence for a radiation-related increased risk of NMRD among either women or men. For all workers, the sex-adjusted NMRD ERR per 100 mGy lung dose (95% CI) for a DWF of 1 was À0.07 (-0.16, 0.02; n ¼ 1798); it was À0.05 (-0.15, 0.05; n ¼ 881) for females and À0.02 (-0.20, 0.17; n ¼ 917) for males.
In the past, there were few occupational studies that included women and even fewer that included nonwhite women. This was true in large part because there were relatively few women in radiation jobs, the likelihood of receiving meaningful radiation exposures was limited, and women were more challenging to trace than men. The number of nonwhite workers was a small component of the TEC female radiation workforce. Given the limited numbers of studies including black women, a dose-response evaluation Cumulative organ doses include the sum of the internal doses received at TEC and any external doses received at other facilities. For DWF ¼ 1, the lung dose is in units of absorbed dose, mGy or Gy. For DWF ¼ 10 and 20, the lung dose is in units of weighted-lung dose, weighted-mGy or weighted-Gy. Doses are unlagged. There were 7 TEC workers with no available dose records who are not included in these tabulations. --Ã SMR is significantly different from 1.0 (p<.05). a ICD: International Classification of Disease. Revision 9 codes are shown. b Mesothelioma did not have an explicit code in ICD9 but did in ICD10 as denoted. Table 4. Internal cohort dose-response analyses for lung cancer over categories of lung dose among 13,951 women and 12,699 men employed for more than 90 days and followed through 2019 (women) or 2018 (men for the 1168 black women who worked at TEC was conducted, recognizing the small numbers and statistical uncertainties (Supplemental Table 4). There were 34 women who died of lung cancer and their ERR (95% CI) at 100 weighted-mGy lung dose was estimated as 0.13 (95% CI À0.03, 0.28). A DWF of 10 was applied because a DWF ¼ 1 resulted in a narrow dose range that precluded model convergence. Future analyses will combine black workers from other nuclear facilities (Boice, Quinn, Al-Nabulsi, Ansari, et al. 2022), such as the Savannah River site (Wartenberg et al. 2001). The highest dose from the inhalation of uranium dust was to the TLNs: the mean dose was 127 mGy and the maximum dose was 5.1 Gy (Table 2). An internal dose-response analysis was conducted of the 207 workers who died of NHL. For a DWF ¼ 1, the ERR (95% CI) at 100 mGy for NHL was estimated as 0.03 (95% CI À0.02, 0.07) ( Table 6). Table 7 presents a sensitivity analysis on the influence of the assumptions made to estimate lung dose on the lung cancer dose response estimates of risk. The four key parameters to estimate lung dose from inhaled uranium dust included the particle size of the dust inhaled, the breathing rate of the worker, the daily exposure during time on the job, and the solubility factor of the uranium whether it was absorbed into the blood at a slow rate (S) or a moderate rate (M). The low and high level parameter assumptions were judged to be realistic, though somewhat extreme, possibilities. Dose response evaluations were made for best estimate (Table 4), realistic low estimate, and realistic high estimate (Table 7). The Hazard ratio minus one (HR-1) is presented as an estimate of ERR. For female workers, the HR-1 estimates at 100 mGy lung dose ranged between 0.00 and 0.03 and all were consistent with the results from the main analysis (ERR per 100 mGy 0.01; 95% CI À0.10, 0.12).
For male workers, however, there was a broader range in the estimates of HR-1 at 100 mGy of À0.59 to À0.09, which were not as consistent with the main analysis (ERR per 100 mGy estimate of À0.22; 95% CI À0.43, À0.01). The solubility assumptions (totally S or totally M) for male workers result in the most inconsistent risk estimates which might be expected because the inhaled uranium dust was known to be of various solubilities and not totally S or totally M. Table 8 presents additional sensitivity analyses of lung cancer risk considering employment duration, dose lagging intervals, a different referent category, and a modified high dose category. For the 14,778 workers employed for more than 1 year, the ERR estimate was essentially the same as the analysis based on all workers: ERR per 100 mGy of À0.09 (95% CI À0.19; 0.03; n ¼ 871) compared with À0.09 (95% CI À0.19; 0.02; n ¼ 1654). Similarly, the lagging intervals made little difference: for the 5 years lag and 20 years lag intervals the ERRs per 100 mGy were À0.09 (95% CI À0.19; 0.01) and À0.06 (95% CI À0.17; 0.04), respectively. Related in part to the unexpected statistically significant negative dose response for lung cancer among the 12,699 male workers (ERR per 100 mGy of À0.22 (95% CI À0.43, À0.01; n ¼ 1002); p for trend ¼.042) (Table 4), and the possible disproportional influence of a few very high dose workers on the dose response, two additional analyses were conducted: one excluding the 1256 male workers with 0 lung dose, and the other excluding the 25 male workers with estimated lung dose >300 mGy. For both analyses, the ERR estimate increased, and the negative dose response was no longer statistically significant. Due to the apparent influence of a handful of high dose workers perhaps related to bias in the dust to dose conversion for the dustiest environments, the preferred ERR per 100 mGy was taken as -0.14 (95% CI -0.32, 0.08; n ¼ 1002; p for trend ¼.18) for male workers (Figure 4(c)).

Tobacco use
Smoking information was sought from TEC medical records on all female lung cancer cases (n ¼ 641) and a sample of comparison women who did not die from lung cancer (n ¼ 707). Medical records were found for 1235 (91.6%) of these 1348 women: 224 (18.1%) women had used tobacco, 374 (30.3%) had not used tobacco, and 637 (51.6%) had no information on tobacco use. For the 598 women for whom tobacco use was recorded, either yes or no, a higher proportion of tobacco use was seen among the 294 lung cancer cases compared with the 304 comparison women who did not die of lung cancer. Of the 294 women who died of lung cancer, 146 (49.7%) used tobacco, and 78 (25.7%) of the 304 comparison women who did not die of lung cancer used tobacco. These percentages are similar to those reported in an earlier study (Dupree et al. 1995). A dose-response analysis directly adjusting for tobacco use (Supplemental Table  5) did not modify the ERR estimate of lung cancer risk compared with the estimate adjusting for pay category as a surrogate for smoking (Table 4). A small number of black women (n ¼ 16) had information on tobacco use, and their percentage who used tobacco (31.3%) was similar to the percentage for the 582 white women (37.6%).
Smoking information also was sought for all male lung cancer cases (n ¼ 1002) and a comparison sample of men who did not die from lung cancer (n ¼ 1107). Medical records were found for 1905 (94.3%) of these 2109 men: 782 (41.0%) men had used tobacco, 181 (9.5%) had not used tobacco, and 942 (49.4%) had no information on tobacco use. For the 963 men for whom tobacco use was recorded, a higher proportion of tobacco use was seen among the 489 lung cancer cases compared with the 474 comparison men who did not die of lung cancer. Of the 489 men who died of lung cancer, 437 (89.4%) used tobacco, and 345 (72.8%) of the 474 comparison men who did not die of lung cancer used tobacco. A dose-response analysis directly adjusting for tobacco use did not appreciably modify the ERR estimate of lung cancer risk compared with the estimate adjusting for pay category as a surrogate for smoking (Table 4), although the negative trend was no longer statistically significant (ERR per 100 mGy À0.13 (95% CI À0.45, 0.19; n ¼ 489; p trend ¼.41)) (Supplemental Table 5).

Medical X-rays
Chest X-ray information was sought for the 1234 women included in the tobacco use medical record abstraction study. Chest X-ray information was found for 78% of the women, and there was no difference between the number of X-rays among lung cancer cases and the comparison women. Among the 585 women who died from lung cancer, 270 (46.1%) received only one X-ray and only 13 women received more than three X-rays. The numbers and proportions of X-rays received by the comparison women were similar. The dose to the lung from a chest X-ray was very low and of the order of a mGy, similar to other cohorts ). An evaluation of low-dose chest X-rays for men was not deemed necessary so was not included in the medical record abstraction.

Discussion
The study of 26,650 workers employed 1943-1947 at the TEC-Y12 electromagnetic field uranium processing facility is unique because of the large number of women who received higher exposures than the male workers, the follow-up for mortality was up to 75 years and 45 years on average, and practically all workers had died (94.9%). There was little exposure to external radiation in contrast to relatively large Table 5. Internal cohort dose-response analyses for nonmalignant respiratory disease (ICD9 codes 460-478, 490-519) excluding flu and pneumonia over categories of lung doses among 13,951 women and 12,699 men employed 1943-1947 for more than 90 days at TEC, and followed through 2019 (women) or 2018 (men).   doses to lung from inhalation of airborne uranium dust, i.e. the mean absorbed dose to lung for all workers was 26.3 mGy and the maximum was 1.05 Gy. The high doses and large numbers of deaths from lung cancer (n ¼ 1654) resulted in high statistical power to reveal any radiationrelated lung cancer risks and to distinguish any differences in risk estimates between men and women had they occurred. The data, however, did not support an association between radiation dose to the lung from inhaled uranium dust and lung cancer or NMRD. There was no difference in the lung cancer risk by sex. There was no significant radiation association for NHL for doses to the TLNs that were higher than the lung doses. The unexpected significant negative dose response for male lung cancer was addressed by conducting several sensitivity analyses (Table 8). First, male workers who had no lung dose were removed as the referent group, after recognizing that they were not comparable in job categories, potential for dust exposure, or tobacco use as other workers. This was reflected in the SMR analyses where salaried male workers had significantly low risks of lung cancer whereas the hourly males, hourly females and salaried females all had significantly high lung cancer SMRs. The dose response was no longer significantly negative after these zero-dose workers were excluded.
Second, concern over possible measurement assessments that may have been exceptionally high among a few high dose workers, was confirmed in an analysis that excluded all male workers with cumulative absorbed lung doses >300 mGy (n ¼ 25 and no lung cancer deaths). The ERR  [1943][1944][1945][1946][1947] for at least 90 days and followed through 2019 (women) or 2018 (men). Death is from underlying cause only. Model adjusted for year of birth (categories), race (white/black) for females, and pay category (hourly/salary). Underlying timescale is age. a Cumulative lung doses include the sum of the internal doses received at TEC and any external doses received at other facilities. b Rounded to two significant digits. The negative sign indicates that the HR-1 was between 0.000 and -0.005.
(95% CI) per 100 mGy became À0.14 (95% CI À0.32, 0.08; n ¼ 1002; p for trend ¼.18). Removal of this small number of apparent outliers resulted in a dose response that was no longer statistically significant and the ERR estimate was compatible with the TEC female estimate and with other MPS studies with intakes of radionuclides, i.e. workers at LANL ) and at the Mallinckrodt Chemical Workers . Additional analyses were conducted removing 16 male workers with lung doses >350 mGy (ERR¼ À0.17) and 11 workers >400 mGy (ERR¼ À0.18) and neither dose response trend was statistically significant. Although Table 8. Sensitivity analyses for lung cancer considering (1) only the 14,778 TEC workers employed for more than 1 year; (2) applying different lag periods of 5 years and 20 years for the 26,650 workers employed 1943-1947 for more than 90 days; (3) analyses of male workers excluding those with lung dose ¼ 0 mGy (primarily salaried workers) and using >0 to <5 mGy as the referent category; (4) and analyses of male workers with lung doses <300 mGy (excluding the 25 workers with lung dose >300 mGy).
(1) All workers employed >1  .18 (-) HR-1 (95% CI) at 100 mGy À0.14 -0.32; 0.08 ERR (95% CI) at 100 mGy À0.15 -0.38; 0.07 HR: hazard ratio; CI: confidence interval; ERR: excess relative rate; DWF: dose weighting factor. Death is from underlying cause only. Model adjusted for year of birth (categories), sex (for all worker analyses), and pay category (hourly/salary). Underlying timescale is age. DWF ¼ 1 for all analyses. a Cumulative lung doses include the sum of the internal doses received at TEC and any external doses received at other facilities. speculative as to why the dose assessment may have been biased, conceivably it might be related to unknown use of respirators in the dustier environments. Earlier assessments of the TEC male cohort noted a possible limitation if respiratory protection was used during certain operations involving high uranium air concentrations (Polednak et al. 1982). Unfortunately, we and Polednak et al. (1982) did not identify any specific documents on the use of respirators. If, however, respirators were worn and were fully operational in the dustier workplaces, i.e. worn and properly cleaned, actual exposures would probably be only 25% of the measured average concentration (Lippmann 1959). Accordingly, the analysis of workers with lung doses <300 mGy (i.e. excluding 25 workers with doses >300 mGy and no lung cancer deaths) was taken as the preferred ERR estimate for male workers (Table 8, Figure 4(c)).

Comparison with previous TEC studies
In a subgroup analysis of men who were 45 years of age or older when first hired, Cookfair et al. (1983) reported a significant smoking-adjusted lung cancer risk among those with a cumulative lung dose of greater than 200 mGy. There were 330 lung cancer deaths evaluated by Cookfair et al. (1983) and follow-up was to 1 July 1973. The earlier study of white males at TEC also noted an elevation of lung cancer (n ¼ 324 total) for those over age 45 years when hired (lung cancers, n ¼ 97) and among a subset of 'chemical' workers (lung cancers, n ¼ 7) who had the highest potential for uranium dust exposure (Polednak and Frome 1981). The larger four-facility study of 787 lung cancer cases (567 among workers at TEC), included contributing causes of death and follow-up was through 1982 (Dupree et al. 1995). The findings resembled those of Cookfair et al. (1983) but did not reach a level of statistical significance after adjusting for tobacco use (Dupree et al. 1995). Among 1853 male workers over age 45 years when first hired in our study, the ERR (95% CI) at 100 mGy was increased, 0.30 (95% CI À0.43, 1.02), but not at the level of statistical significance. In contrast, for workers 15-24 years and 25-44 years at first hire, the ERRs were all negative, but also not statistically significant. This pattern of decreasing ERRs with increasing age at hire might partially reflect the increase background rates for lung cancer with age at hire. Future combined cohorts of MPS cohorts will evaluate further the possible influence of age at exposure and subsequent lung cancer risk.

Comparison with other uranium processing cohorts
A comprehensive study of uranium processing workers who inhaled uranium dust was conducted of 2514 white males who worked at the Mallinckrodt Chemical Works Ellis et al. 2018;Golden et al. 2022). The exposure assessment was extensive and included 39,451 uranium urine bioassays, 2341 breath radon measurements, and 15,518 occupational medical X-rays in addition to film badge measurements of external gamma radiation exposures.
The estimated mean lung absorbed dose from all sources of exposure was 69.9 mGy (maximum 0.89 Gy). The mean lung dose from uranium inhalation alone was 27.1 mGy (max 0.42 Gy) which is similar to our mean lung dose of 26.3 mGy (max 1.0 Gy) for all workers. Similar to our findings that are based only on airborne uranium measurements, there was no evidence of a dose response for either lung cancer (n ¼ 157) or NMRD excluding flu and pneumonia (n ¼ 238). It remains equivocal whether exposure to uranium is a cause of lung cancer (UNSCEAR 2017), and occupational guidelines are based in large part on the chemical toxicity of uranium, notably to the kidney (ATSDR 2013). An extensive review of the literature in 2013 concluded that 'no definitive evidence has been found in epidemiologic studies that links human deaths to uranium exposure' (ATSDR 2013).

Smoking considerations
For the 1561 workers (598 women and 963 men) with available information on tobacco use obtained from medical questionnaires completed during employment at TEC, internal cohort dose-response analyses were conducted for lung cancer. The estimates of radiation risk (ERRs) with direct adjustment for tobacco use were comparable to those computed for the full cohort adjusting for pay category (salary/hourly) (Supplemental Table 5). For all workers, the lung cancer ERR per 100 mGy with direct adjustment for tobacco use was À0.03 (95% CI À0.20, 0.13; n ¼ 784; p trend .70). The lung cancer ERR per 100 mGy with adjustment for pay category was À0.09 (95% CI À0.19, 0.02; n ¼ 1654; p trend .10) and statistically indistinguishable. For males, the ERR per 100 mGy approached the null, increasing from À0.22 to À0.13, and the dose response was no longer statistically significant. These analyses suggest that the absence of a lung cancer radiation effect among TEC workers is unlikely to be related to confounding by cigarette smoking. Further, that adjusting for pay category can be assumed to be closely equivalent to directly adjusting for tobacco use in this population. To support this conclusion that pay category is an adequate surrogate for cigarette smoking, we evaluated the risk of lung cancer among hourly workers (SMR ¼ 1.32; 95% CI 1.25-1.40; n ¼ 1299) and salaried workers (SMR ¼ 1.01; 95% CI 0.90-1.12; n ¼ 344) (Supplemental Table 2). The ratio of SMRs, 1.31, was statistically significant (95% CI 1.16, 1.47), as was the SMR for lung cancer among the hourly workers (but not salaried workers).

Non-Hodgkin lymphoma
Death from NHL was increased overall based on comparison with the general population (SMR 1.11;95% CI 0.96,1.27;n ¼ 207). The TLNs received a higher mean dose (127 mGy) than the lung (26.3 mGy) ( Table 2). An internal analysis revealed little evidence for a radiation association with NHL (ERR at 100 mGy 0.03; 95% CI À0.02, 0.07) ( Table 6). This finding is consistent with previous MPS studies of workers with intakes of radionuclides at LANL in New Mexico ) and at the Mallinckrodt Chemical Works in Missouri . No radiation association for NHL was observed following external radiation among nuclear power plant workers in the MPS (Boice, Cohen, Mumma, Hagemeyer, et al. 2022). In contrast, a recent report from the UK National Registry of Radiation Workers found a significant dose response for NHL which appeared concentrated among workers who received cumulative external doses >0.5 Sv (Hunter and Haylock 2021). Polednak and Frome (1981) commented that the dose to the skin from beta-emitting progeny of 238 U might be a potential hazard but provided no dosimetric details. Melanoma of the skin (SMR 1.26; 95% CI 0.98, 1.59; n ¼ 90) and cancer of the eye (SMR 2.77; 95% CI 1.26, 5.25; n ¼ 9) were elevated which prompted us to look carefully at the possible range of doses to the skin from the uranium dust. We found that the skin dose was miniscule (mean dose <0.05 mGy; max $1.0 mGy) and was lower than the lung dose by a factor of 1000.

Parkinson's disease
A study of Mayak plutonium workers in Russia recently reported a dose-response relationship between low-LET radiation and Parkinson's disease (Azizova et al. 2020). Similar positive associations have been reported among MPS cohorts (Boice and Dauer 2021;Boice, Cohen, Mumma, Hagemeyer et al. 2022;Boice, Quinn, Al-Nabulsi, Ansari, et al. 2022). The SMR for Parkinson's disease among TEC workers was slightly elevated at 1.13 (95% CI 0.96, 1.32; n ¼ 153). As workers had little radiation dose to brain, dose-response analyses were precluded. However, we were able to evaluate whether smoking cigarettes was related to a protective effect of Parkinson's disease as seen in epidemiologic studies of smokers, and most recently in the 65-year follow-up of British doctors (Mappin-Kasirer et al. 2020). Assuming that pay category is an adequate surrogate for tobacco use, we evaluated the SMRs of Parkinson's disease among hourly workers (SMR ¼ 1.03; 95% CI 0.85, 1.26; n ¼ 97) and among salaried workers (SMR ¼ 1.30; 95% CI 0.97, 1.71; n ¼ 52) (Supplemental Table 2). These results are as predicted and consistent with a protective effect of cigarette smoking for Parkinson's disease, i.e. hourly workers had a lower SMR than salaried workers. The ratio of SMRs, 1.26, was close to statistical significance (95% CI 0.90, 1.76), as was the increased SMR among the small number of salaried workers.

Interpretation of SMR analysis and future evaluations
Interpretation of SMR analysis must be done cautiously for a number of reasons, including non-comparability of the general population with a working population, and the ability to make hundreds of comparisons when evaluating many causes of death by many categories. There were nearly 400 SMRs computed for the male and female TEC workers which resulted in many significantly high and significantly low estimates (Table 3, Supplemental Table 2). By chance alone about 20 statistically significant estimates might be expected. Thus, specific hypotheses are important beforehand to discourage, in part, attempts to explain every significant observation or subgroup analysis that may have been serendipitous and not of a priori interest. In our study of uranium processing the highest radiation doses were anticipated for the lung and TLNs, so that focusing on malignant and nonmalignant lung disease and perhaps NHL and kidney disease seemed appropriate. A number of causes of death had statistically significant SMRs that are not easily interpreted with respect to radiation exposure because the organ-specific doses were very small. For example, there was a significant excess of deaths from CeVD in contrast to a nearly significant deficit of deaths from IHD. Similarly, a significantly high excess of CLL was seen whereas there was no increase in deaths from leukemia excluding CLL; the latter is frequently found in excess following external (although not internal) radiation exposures whereas the former is not. SMR elevations were seen for cancer of the eye, melanoma, accidents, and other sites suggesting that the comparison population, the general population of the United States, is not ideal for radiation risk indications. Perhaps equally important, the TEC WWII worker population may have had unique lifestyle and demographic features that are not readily characterized. The population smoked heavily, for example, as seen by the significant excesses from smokingrelated cancers and the significant deficits for nonsmoking related cancers.
The WWII population of men and women working in Oak Ridge on a secret Manhattan Project program could very well differ in many ways from the general population of the United States. Accordingly, we computed SMRs based on population rates of the state of Tennessee. The overall SMR for all causes of death decreased to 1.01 (95% CI 1.00, 1.03), and there were other changes observed such as the deficit of IHD became significantly low (SMR 0.92;95% CI 0.89,0.94). These examples point to the challenges in interpreting SMRs even when best attempts are made to obtain an adequate general population. Thus, internal dose-response relationships considering only the workers themselves and making comparisons over categories of radiation dose with adjustment for potential confounders is optimum. Such an approach addresses the so-called healthy worker effect where occupational groups are healthier and less likely to die during the years after first hire than segments of the general population that are unable to work because of illnesses or other infirmities. However, in circumstances such as TEC there are few causes of death for which radiation dose-response evaluations can be conducted because only the lung and lymph nodes received appreciable radiation dose, and other occupational exposures are either unknown or difficult to estimate. Future evaluations should consider, if possible, any chemical exposures (e.g. phosgene, CI 2 , HNO, NO X ); and any medical conditions such as tuberculosis, other lung diseases, or other conditions that may be available in medical records of the TEC population. For TEC workers alive in 1999 and after, we will obtain detailed information on disease occurrence, such as cancer incidence, and risk factors, such as smoking, with linkages to Medicare claims data and other extensive datasets available from the Centers for Medicare and Medicaid Services (CMS) (Boice, Quinn, Al-Nabulsi, Ansari, et al. 2022). There were 9995 TEC workers alive in 1999, 7624 women and 2371 men, who are eligible for the CMS linkages.
Quality of death information for those dying after age 85 years Analyses of MPS cohorts commonly end the follow-up of workers not known to have died when they reach the age of 95 years. This is done because of perceived concern that the causes of death recorded on death certificates for the very old would be nonspecific and noninformative, for example 'old age', and that the rates of mortality for cause-specific deaths for ages greater than 95 years accordingly would be uncertain.
We conducted a sensitivity analysis (Supplemental Table  6) where we ended the follow-up when a worker reached the age of 85 and compared the resulting SMRs with those who died at older ages (Table 3). There were essentially no differences in the SMRs in relation to the ages when followup ended. For all causes of death, the SMRs using an 85 years cutoff was 1.06 (95% CI 1.05, 1.08; n ¼ 18,327) compared with 1.07 (95% CI 1.06, 1.08; n ¼ 25,282) for using the later cutoff. There were 6,955 additional deaths that occurred after age 85, and the force of mortality was not seen to change, i.e. the women did not die at a higher or lower rate as compared with the general population, nor did the SMRs change. A key issue, however, is whether the distribution of the causes of death changed over these advanced ages of follow-up. In brief, they did not. Based on large numbers of deaths from all cancers, the SMRs at cutoff age 85 and at ages >85 years were the same at 1.01 with an additional 750 cancer deaths observed after age 85. For lung cancer, the SMRs at cutoff age 85 and at ages >85 years were also the same at 1.24 with an additional 164 lung cancer deaths observed after age 85. Similar patterns are seen for the other major causes of death including breast cancer, prostate cancer, leukemia, mental and behavioral disorders, all heart disease including IHD, CeVD, NMRD, nephritis and nephrosis, and all external causes of death. There were some differences seen for non-Hodgkin lymphoma and diabetes, but based on small numbers and, conceivably, might indicate better diagnoses of these causes of death than in the general population of elderly persons.
Medical diagnoses and treatment for the elderly and disabled in the United States improved markedly since 1965 with the expansion of Medicare programs that ensured access to and payment for health care (CMS 2015). These programs were accompanied with an increased accuracy of disease diagnoses and cause of death determination compared with years past, e.g. the top five causes of death among centenarians in 2014 were heart disease, Alzheimer's disease, stroke, cancer, and influenza and pneumonia and similar to death distributions for those younger but considered elderly (Xu 2016). We conclude that using an age at follow-up cutoff of 95 years is reasonable, and does not bias or dilute the analyses. Future analyses will consider the ramifications of extending the follow-up without using any age cutoff and recording deaths until the end of life, e.g. up to age 105 years.
High-LET dose to lung Table 9 presents lung cancer risk estimates for MPS cohorts with intakes of radionuclides . The workers in these cohorts had intakes of uranium, polonium, plutonium, or other radionuclides, and exhibited little evidence for any increase in lung cancer risk. The over 35,000 workers with intakes of radionuclides were employed at Rocketdyne (Leggett et al. 2005;Boice et al. 2011), Mound (Boice et al. 2014), Mallinckrodt Chemical Works , LANL , and Rocky Flats (Boice, Quinn, Al-Nabulsi, Ansari, et al. When a different metric was used to estimate risk in a particular study, an approximation to an estimated ERR is presented. The mGy doses are unweighted absorbed lung doses (Dose Weighting Factor of 1) and include external doses. a Rocketdyne (Boice et al. 2011); Mound (Boice et al. 2014); Mallinckrodt ; Los Alamos ). Summaries also found in Boice, Quinn, Al-Nabulsi, Ansari, et al. (2022). b Male workers with lung doses <300 mGy. c These values were converted to an approximate ERR per 100 mGy. 2022). Other than the male TEC estimate, the ERRs per 100 mGy lung dose ranged from À0.06 to 0.01, which can be compared with the estimate of 0.02 (95% CI À0.03, 0.07) for low-LET exposure in the combined cohorts of nuclear power plant workers, industrial radiographers and medical radiation workers (n ¼ 367,722) Boice, Cohen, Mumma, Hagemeyer, et al. 2022;Boice, Ellis, Golden, Zablotska, et al. 2022).
Our study of high-LET alpha particle exposure to lung from inhaled uranium dust is relevant to concerns for astronauts embarking on long missions outside of Earth orbit. The possible health effects of galactic cosmic rays (GCRs), i.e. the very high-LET heavy ions traveling near the speed of light, are an area of intense investigation (Boice 2017(Boice , 2022NASA 2021;NASEM 2021;NCRP 2021aNCRP , 2021bBoice, Ellis, Golden, Zablotska, et al. 2022). There are no circumstances on Earth that can represent GCR so that animal studies and GCR simulations are used to estimate the risk of cancer and nonmalignant diseases, including cognitive function. The studies of the Japanese atomic bomb survivors exposed acutely to radiation in 1945 indicate that women are at 2-3 times higher risk than men on a relative scale for lung cancer (NCRP 2014). Until recently (NASA 2022) based its radiation protection guidance on individual lifetime risk estimates and relies heavily upon the Japanese data. Accordingly, women were limited in their flight hours compared with men because of the differences in radiationrelated lung cancer mortality risk. Although an imperfect analog to GCR, the alpha particle exposure to lung from intakes of radionuclides is from high-LET particles and of high dose from protracted exposure over years (Boice, Quinn, Al-Nabulsi, Ansari, et al. 2022). The dose distributions, however, are not similar within the lung, the track structure differs, and there is the possible effect of inhomogeneous distributions of the heavy metal uranium within the lung as well as any effect of metal toxicity. Nonetheless, it provides another set of human data that could be considered for radiation protection guidance, including the combination with other informative data sets (Simonsen and Slaba 2021).

Strengths and limitations
The study of the TEC workers employed 1943-1947, has a number of epidemiologic strengths as mentioned: a large number of workers, including female workers, near complete lifetime follow-up for mortality, information on individual use of tobacco from medical questionnaires, and a comprehensive dose-reconstruction process. Further, a wide range of sensitivity analyses were conducted to help interpret observations such as varying the input parameters required to estimate lung dose from uranium air monitoring data, evaluating different dose lagging procedures, examining the effect of different inclusion and exclusion criteria for conducting dose response (such as excluding 25 highest dose male workers and limiting the analyses to worker with <300 mGy cumulative lung dose). The number of women was larger than the number of men and their lung doses were larger; this is unusual in occupational studies and a strength when conducting sex-specific comparisons of lung cancer risks. The reason for the higher doses among the women is that they remained in the 'dusty environments' for longer employment periods than the men.
A major limitation in most occupational studies of lung cancer is the limited information available on smoking histories for individual workers. During WWII, smoking was not uncommon among men or women. Pay category (hourly/salary) was used as the measure of SES and as a surrogate for tobacco use. Medical record abstraction on the use of tobacco among TEC men and women confirmed that pay category was directly correlated with tobacco use. Nearly, 50% of the workers who died of lung cancer had information on tobacco use available in medical questionnaires. Analyses adjusting directly for tobacco use among those workers with information available in their medical records (both lung cancer cases and comparison workers who did not die of lung cancer) were similar to analyses of the full cohort adjusting for pay category as a surrogate for tobacco use, although the negative dose response for male workers was no longer statistically significant. The current study is generally consistent with previous MPS studies evaluating lung cancer risk following intakes of radionuclides (Table 9).
Another limitation is the possibility of inaccuracies within the dose reconstruction process since the lung doses are based almost entirely on air monitoring measurements in the various calutron (alpha and beta processing) areas and other TEC buildings, and no urine samples or other bioassays measurements were taken. Also, it was not known whether or when respiratory protections may have been used by individual workers which could bias the estimates of inhaled airborne dust and overestimate the lung doses. Conceivably, this bias might have been possible for workers engaged in very dusty environments where respirators might have been available and recommended. Separately, inaccuracies or misclassifications could be in the assignment of exposure for job-specific categories, the duration of exposure, uranium solubility assumptions, and particle size assumptions. While a comprehensive effort was made to convert air measurements to worker dose based on likely airborne dust exposure for specific job locations (alpha building, beta building, chemistry lab, other), this process was nonetheless based on a series of assumptions. The sensitivity analysis associated with varying the parameters needed for dose reconstruction did not indicate any appreciable difference in risk estimates for female workers, but the male workers were more affected by changes in the parameter values (Table 7).
Our previous studies found that air monitoring measurements alone are subject to uncertainties that are difficult to address Golden et al. 2022).
Though not as optimal as lung-dose estimates based on periodic individual bioassay measurements over time, air sampling measurements for particulate radioactivity can help to determine the intake of airborne radioactivity for individual workers or groups (Marshall and Stevens 1980;Polednak and Frome 1981;Keane and Polednak 1983). In addition, sampling at TEC included collection of dust settled on surfaces (Polednak and Frome 1981), and later studies of Oak Ridge workers exposed to enriched airborne compounds demonstrated positive correlations between surface contamination levels and employee uranium excretion rates (Schultz and Becher 1963). Although based on only a few postmortem examinations of TEC male workers employed for several years in environments with high airborne uranium dust levels, the fact that high uranium retention in the lungs and TLNs occurred is not questioned (Keane and Polednak 1983). It is the uncertainty surrounding accurate estimates of lung dose that is a limitation, particularly for individual workers. Static (area) air monitoring may provide reasonable estimates of average airborne activity in a work area but less accurate estimates of intake by workers at individual workstations within that area. In general, estimates of intake based on air monitoring alone are prone to sizable errors for individual workers even if personal air samplers are worn (ICRP 2015).
State of the art approaches were used in the conversion processes to estimate absorbed dose to the lung, and account was taken of the number of environmental samples and the estimates of uranium dust over the years of TEC operation, including the changing enrichment factors and solubility factors. The absence of a dose response in previous studies at four uranium processing plants, including TEC, suggests that with the possible exception of male workers errors in measurements are likely to be small (Dupree et al. 1995) as does the consistency with previous MPS studies of lung cancer following intakes of uranium (Boice et al. 2011;Golden et al. 2022) and other radionuclides (Boice et al. 2014; (Table 9). Finally, a standard approach was followed for estimating organ doses after intakes of radionuclides for epidemiologic studies, specific to the MPS investigations, and described in detail in NCRP Report 178 (NCRP 2018) and in other publications (Bouville et al. 2015;Ellis et al. 2018;Dauer et al. 2022).

Conclusions
The study of 26,650 men and women who inhaled uranium dust during the years 1943-1947 found no evidence that high-LET alpha particle exposures increased the risk of lung cancer, NHL, or NMRD in either men or women. The absorbed doses to lung and to TLN were broad and reached over 1 Gy. The large population of female workers at this uranium conversion and enrichment plant had not been studied before, their exposures were greater than those for their male counterparts, and sex-specific comparisons indicated no significant difference in the lung cancer risk estimates between male and female TEC workers.