| dc.description.abstract | Long-term health effects following exposures to mixed high- and low-linear energy transfer (LET) ionizing radiation received at low dose-rates are unclear, and uranium processing workers represent a population of relatively healthy workers with this exposure profile. While a variety of health effects were of interest, I focused particularly on cancer and noncancer diseases of the lung and noncancerous heart diseases. Initially, I conducted a systematic review of non-malignant respiratory disease (NMRD) following uranium and photon exposures in working populations to determine whether past evidence supported a causal relationship. There was minimal evidence for such a relationship, particularly after accounting for the quality of the published literature to assess causality. I then conducted detailed analyses of three individual uranium processing cohorts: Middlesex Sampling Plant, Mallinckrodt Chemical Works, and Fernald Feed Materials Production Center. In each cohort and in a pooled cohort with individualized data, I estimated standardized mortality ratios compared with a sex, race, age, and birth cohort-matched US population. Additionally, I used Cox and Poisson regression to estimate dose-response relationships between radiation and a priori selected outcomes, with adjustment factors selected using directed acyclic graphs (DAGs). Standardized mortality ratios indicated a difference in health outcomes between salaried and hourly workers. None of the individual cohorts or the pooled cohort showed evidence of a dose-response between radiation and either lung cancer or NMRD; however, all had indications of a cardiovascular disease (CVD) and ischemic heart diseases (IHD) dose-response. The CVD and IHD hazard ratios were statistically significant in Mallinckrodt, Fernald, and the pooled cohort. In the Fernald and pooled cohorts, I used Fine and Gray methods to estimate competing risks-adjusted CVD and IHD dose-responses, creating a pseudo-confidence interval for the potential effect of radiation on heart disease in a population with less mortality from competing causes. In both Fernald and the pooled cohort, CVD and IHD dose responses remained statistically significantly elevated in competing risks analyses. However, results may be biased due to lack of smoking adjustment in all models. Future work should pool additional cohorts, adjust for smoking status, and conduct effect modification analyses. | |
