The Colorado Plateau cohort is based on records from 4,137 underground uranium miners who were employed for at least 1 month and participated in at least 1 medical screening between 1/1/1950 and 12/13/1960. Person time began either at the start of mortality rates (1/1/1960) or the date of first medical screening, whichever was later. Person time ended at date of death, date lost to follow-up, or the end of follow-up, 12/31/2016. Mortality ascertainment is described in detail in a previous study.⁴ Briefly, mortality follow-up was conducted by linking name, Social Security number, and date of birth to the US National Death Index. The US Social Security Administration Death Master File and Internal Revenue Service records were also used to confirm or obtain vital status when needed.

Cumulative radon progeny exposure is measured in Working Level Months (WLM). WLM calculations are described elsewhere,^2–4,6 briefly, WLM in this cohort is based on interpolation of dates when miners reached 1 of 9 categories of cumulative radon exposure.^2–4 Since the last published update of the Colorado Plateau cohort, WLM estimates have been recalculated to include more accurate estimates of annual cumulative exposure and exposure rates by reviewing original records to verify employment and exposure data. Additionally, cumulative radon exposure from previous non-uranium hard rock mining was estimated. Updated estimates were calculated by linking each miner back to original employment records by mine location and year, combined with available radon progeny measurement data.

Cigarette smoking is quantified in pack-years. Smoking information has also been improved since the last update of this cohort. Previously, miners’ smoking status and pack-years was based on surveys last conducted in the 1970s or, for white miners, in 1985. In this update, smoking status, pack-years, and the last smoking date among ever-smokers were imputed from the date of the last survey to the end of follow-up. Smoking imputations were based on age and 5-year birth cohort and the cessation probabilities for U.S. men using U.S. National Health Interview Survey data.²⁰ Silica dust exposure was not quantified in the Colorado Plateau cohort.

We selected NMRD subtypes of a priori interest: asthma, COPD, idiopathic pulmonary fibrosis, pneumonia, silicosis, tuberculosis, other pneumoconiosis, and other respiratory diseases. Underlying causes of death are coded by the ICD revision in effect at time of death. Appendix table I reports the ICD codes for revisions 7 through 10 for each subtype. For temporal and ERR analyses, we focus on subtypes with more than 50 deaths.

Overall Standardized Mortality Ratios (SMRs) were calculated using SAS statistical software (SAS 9.4; SAS Institute Cary, NC). 95% confidence intervals were calculated using the Byar approximation of the Poisson distribution.²¹ SMRs were adjusted for 5-year calendar period, 5-year age groups, and racialization. Expected numbers of deaths were calculated by multiplying the person-time at risk by regional population rates. The regional standard population rates for white miners are the age- and calendar period-specific underlying mortality rates for white non-Hispanic males in Colorado, Utah, New Mexico, and Arizona. The regional standard population rates for American Indian miners are the age- and calendar period-specific underlying mortality rates for non-white non-Hispanic males in New Mexico and Arizona. We also calculated SMRs by duration of employment, calendar period, and period of hire.

We performed regression analyses of associations between cumulative radon progeny exposure and mortality using Cox regression²² with attained age as the time scale. Risk sets were further matched on year of birth (+/− 5 years of index case), and duration of employment in decades. The relationships between cumulative radon progeny exposure and NMRD deaths were modeled as linear excess relative rates and 95% profile likelihood-based confidence intervals using the general model form rate=h0(1+β1d) where β1 is an additive term representing the excess relative rate per WLM cumulative radon exposure d ²³ and h0 represents the baseline/unexposed rate by the matched variables of age and birth date. Smoking was assessed as a potential modifier by performing analyses stratified by smoking pack-years (≥ or <10 pack-years).

Several sensitivity analyses were conducted: (1) In addition to using a 10 year lag in the main analyses, we explore the induction and latency periods⁵ for different NMRD types. The latent period is the time between disease onset and detection (in this case NMRD death), and the induction period is the time between exposure and disease onset. Latency and induction were estimated through an empirical evaluation of model fit based on likelihood ratio tests. (2) We fit additional models of the associations between radon and silicosis and IPF, restricted to person time and events accrued at <1000 WLM and modeled as a restricted cubic spline with three evenly spaced knots across the lower 95% of exposures.in the period prior to ICD-10, (3) We conducted a sensitivity analysis that restricted idiopathic pulmonary fibrosis to deaths coded in the 10^th revision in order to assess potential bias due to outcome misclassification in prior revisions. We also analyzed silicosis, other pneumoconioses, and IPF together as a group to assess the impact of outcome misclassification. (4) We assessed the potential impact of unmeasured silica as a potential confounder in several ways. We restricted analyses to workers who had hard rock mining experience prior to uranium mining employment to assess the impact of prior hard rock mining experience on overall results. We also indirectly assessed potential confounding of the association between cumulative radon progeny exposure and silicosis by unmeasured silica exposure, using COPD as a negative control outcome. And, we indirectly assessed potential confounding of the association between cumulative radon progeny exposure and IPF by unmeasured silica exposure, using silicosis as a negative control outcome. (5) We assessed the impact of adjusting for duration of employment, since duration of exposure could be correlated with unmeasured silica exposures. We also assessed the impact of adjustment for racialization because of different working conditions and different health care systems between whites and American Indians.²⁴ (6) Finally, we conducted a sensitivity analysis restricting ERR estimates to person time and events among workers exposed to <1000 cumulative working level months to assess potential impacts of radon progeny exposure misclassification, which is higher for miners who worked in the early years of mining operations.

Large and persistent excess SMRs for NMRD types indicate that the uranium mining experience results in excess NMRD deaths compared to the regional population. We observed elevated SMRs for silicosis that increased by calendar period and duration of employment and decreased by period of hire. IPF SMRs also increased with advancing calendar period.

We observed differences in SMRs by racialization, which has been observed in prior analyses.⁴ We found elevated SMRs for COPD among whites, but American Indian workers had an SMR below the null for COPD. This is likely driven by the large differences in smoking rates between whites and American Indian workers.⁴ Smoking is a strong risk factor for COPD,²⁵ and SMRs for ≥10 pack-years smoking were much higher than <10 pack-years smoking for COPD. Higher silicosis and other pneumoconioses SMRs among white miners may be due to more dust exposure from other mining activities before or after uranium mining employment. These differences likely reflect several other factors including differences in the workplace safety and social experiences,²⁴ differences in healthcare access and diagnostic patterns, and disparities in the burden of pollution from uranium mining activities, which disproportionately impact American Indians.

The NMRD SMRs among this cohort are higher than in other uranium miner studies. In the largest pooled study of uranium miners, the Pooled Uranium Miners Analysis (PUMA), investigators analyzed SMRs for silicosis and COPD using data from 7 cohorts, including Colorado Plateau.²⁶ The PUMA SMR for COPD was 0.98 (95% CI 0.93, 1.02), and 13.56 (95% CI 12.64, 14.52) for silicosis.²⁶ Unlike PUMA, we observed a large excess COPD mortality, although this excess was driven entirely by smokers. We also observed an excess mortality rate of silicosis that was three times higher than PUMA. This is likely because the Colorado Plateau cohort is composed of workers from an early calendar period of hire when ventilation was poor, with high average cumulative radon exposure, and likely high dust exposures.^26,27

The SMR analysis of the Colorado Plateau cohort indicates that uranium miners in the US continue to die of NMRDs at a higher rate than the comparable general population. The uranium industry employed tens of thousands of workers,²⁸ including 3,000 to 5,000 American Indian miners,²⁹ so the Colorado Plateau cohort represents only a small proportion of the US uranium worker population. Based on the SMRs by calendar year, it is expected that this occupational group will continue to exhibit high NMRD mortality. The Radiation Exposure Compensation Act (RECA), which provides benefits to uranium miners who developed certain diseases from uranium mining work between 1942 and 1971, is scheduled to expire in 2024.²⁸ This analysis indicates there will be uranium miners who will continue to develop RECA-eligible NMRDs after RECA ends.

We observed no exposure-response associations between radon and pneumonia or other pneumoconioses despite a large SMR for other pneumoconioses. We also found no evidence of an association between radon and COPD death in this cohort.

Silicosis death was associated with radon exposure, and this association was modified by smoking, with a stronger association among heavier ≥10 pack-years smoking and no association among <10. Analyses restricted to workers with prior hard rock mining experience had a stronger association between radon and silicosis, suggesting that the associations observed in the full cohort are driven by workers exposed to silica via prior hard rock mining experience.

There was also a small positive association between radon and IPF that did not appear to be modified by smoking. As with silicosis, IPF analyses restricted to workers with prior hard rock mining experience had a stronger association. A prior analysis of this cohort found positive associations between radon and silicosis and pulmonary fibrosis.⁴ The authors suspected that this was a result of outcome misclassification of death certificates, or due to unmeasured exposure to dust which may have been correlated with duration of employment, or from.⁴

As in prior analyses of this cohort, unmeasured silica dust exposure is a limitation. In this analysis we assessed for confounding by silica exposure using negative control outcome logic. COPD and silicosis are strong candidates for negative control outcomes since there is little evidence that they caused by radon progeny, but could be caused by the unmeasured confounder of interest (silica dust exposure). We observed no association between radon and COPD, which suggests that radon may not be associated with silica and would not confound other radon-NMRD associations. But we did observe a strong association between radon and silicosis, which suggests that radon could be associated with silica, and could confound other radon-NMRD associations.

A key assumption for a valid negative control outcome is that the exposure of interest is not a cause of the negative control outcome. Although our analysis supports a large German uranium miner study that also found no association between radon and COPD,³⁰ a cohort study with an ecological metric of residential radon exposure reported an association between county-level radon and COPD mortality.¹⁵ Given the null association between radon and COPD in the Colorado Plateau cohort it is unlikely that violation of this assumption is a major limitation of the current negative control outcome evaluation.

Results were also sensitive to restriction to person time and events <1000 WLM suggesting issues with exposure misclassification among higher exposures, which mostly occurred in early periods. As shown in figure 1, the RRs are lower in the highest exposure category, and the exclusion of the high exposure group leads to elevated estimates. Additionally, results were sensitive to categorization of exposure. Relative rates differed between quintiles and continuous estimates because the referent category of the quintiles included deaths among miners with radon exposures (figure 1).

Another source of uncertainty is that NMRD mortality based on death certificates is subject to outcome misclassification. Because of the known risks of silicosis, miners may be more likely to be diagnosed with silicosis than other radiographically similar NMRDs. Among the 52 silicosis deaths listed as underlying deaths in the cohort, 11 had a multiple cause of death listed as idiopathic pulmonary fibrosis, other pneumoconioses, or both. ERR analyses of silicosis, pneumoconioses, and IPF combined were lower than silicosis alone. Therefore, outcome misclassification and confounding are still substantial concerns in this analysis.

This is the first analysis to report NMRD risk per unit exposure in the Colorado Plateau uranium miner cohort. Other than in the German cohort, there have been few in-depth studies of NMRD subtypes among uranium miners. This analysis updates mortality follow-up 11 years since the last update, for a total of 56-years of mortality follow-up. This extended follow-up allows for the investigation of diseases with long latency and induction periods. Other strengths of this study include the improved cumulative radon exposure estimates, which reduces bias from exposure misclassification, improved smoking pack-year estimates, and minimal loss to follow-up. This study is also strengthened by the smoking data, which was based on repeated questionnaires.

This study is limited by small numbers of deaths for several NMRD subtypes and limited information on occupational co-exposures. Silica dust and diesel fuel exhaust exposures may have been higher prior to the introduction of ventilation in the mid-1960s.⁴ The exposure-response analyses are substantially limited by the absence of silica exposure estimates. Silica is known to cause silicosis and other NMRDs, including COPD independently of silicosis.³¹ In estimating ERR/100WLM, we adjusted for duration of employment and temporal factors, which may be a proxy measure of silica exposure. We conducted a sensitivity analysis to assess the impact of adjustment for duration of employment but only saw slight changes in some estimates. (Appendix table II). We also used negative control outcome methods and restriction to evaluate the role of silica as a confounder of the radon-silicosis association.

Finally, it is important to note healthy worker survivor bias (HWSB) affects estimates with internal comparisons. We adjusted for duration of employment, but it is unlikely that this adjustment fully corrects for HWSB.³² A prior study of lung cancer using data from this cohort indicated that HWSB caused substantial underestimation of radon-lung cancer associations.³³

Despite these limitations, research on NMRD risks among uranium miners and other radon-exposed worker populations should continue for both epidemiological studies and public health surveillance. Epidemiologic studies of uranium miners are the basis for occupational radon standards.³⁴ The results of the ERR analysis suggest that radon may play a role in the development of some types NMRDs, even in the absence of silica exposure, although confounding and outcome misclassification remain concerns. Currently, the International Commission on Radiological Protection (ICRP) only considers radon risks from lung cancer in their recommendations for radiological protection systems.³⁵ There is increasingly more evidence that cancers other than lung, cerebrovascular disease, and some NMRDs are associated with occupational radon exposure,^26,36,37 which could influence the assessment of risks of radon exposure for contemporary workers and the general population.

We observed excess NMRD mortality for several subtypes of interest, indicating that the occupational hazards of uranium mining result in elevated mortality rates compared to regional standard population rates. In ERR analyses, we observed small positive associations between radon and silicosis, and radon and IPF. However, there are several limitations: data for outcomes were sparse, measurements for important potential confounders such as silica were not fully accounted for, there are concerns with outcome misclassification. and associations were driven by workers with prior hard rock mining experience. While our analysis provides only suggestive evidence for an association between radon and select NMRDs, we demonstrated a large excess mortality compared to the standard population, and we observed that these elevated rates continue to persist among surviving cohort members. The legacy of uranium mining continues to impact former miners, and the disease burden among this aging population requires ongoing attention.