Risk of leukemia mortality from exposure to ionizing radiation in U.S. nuclear workers

Stephen Bertke, NIOSH 
*Robert Douglas Daniels, NIOSH 
Mary K Schubauer-Berigan, NIOSH 
Kathleen M Waters, NIOSH 

Keywords: Ionizing radiation, leukemia, nuclear workers, epidemiology, dose-response modeling

We examined the relation between protracted low-dose ionizing radiation exposure and leukemia mortality in a pooled cohort of U.S. nuclear workers (n=105,245) followed through 2005. We used a nested design to select four controls for each leukemia decedent (n=369) from risk sets matched on attained age of the case. Conditional logistic regression analyses were conducted using general relative risk models to estimate the excess relative risk (ERR) of leukemia, leukemia excluding the chronic lymphocytic (CLL) subtype, acute myeloid leukemia (AML), chronic myeloid leukemia (CML), and CLL. Models were adjusted for sex, race, benzene exposure, social economic status, and either hire date or birth date. Sensitivity analyses were used to examine temporal effects by varying exposure lags and time windows of exposure. Linear models exhibited positive but imprecise estimates of non-CLL leukemia risk (ERR per 10 mGy = 0.009; 95% CI:-0.014, 0.051). Risk attenuation in the low dose (<10 mGy) and high dose (>100 mGy) regions of the dose response contributed to the imprecision in full models; however, was not evident in time windows-based models. Leukemia risk appeared wave-like and differed by subtype, where peak risks were observed from exposures occurring 5-10 years prior to attained age for AML cases (ERR per 10 mGy = 0.78; 95% CI: 0.051, 2.7) and from 10-15 years for CML (ERR per 10 mGy = 0.44; 95% CI: <0, 2.4). The non-linearity in the low-dose region may be the result of a short latency period and increased follow-up over periods of diminished exposure. This study suggests that continued analysis of risk temporality may be critical in future studies examining risk factors for leukemia.