Health effects of plutonium exposure

Determination of radiation protection guidelines for persons working with plutonium has been complicated by limited human data on the biological behavior and subsequent health effects from internally deposited plutonium. One solution has been the use of animal models to predict likely health effects in humans. The relationship of plutonium inhalation and lung fibrosis and lung cancer was explored with extended Cox proportional hazards methods in data from life-span studies of beagle. The lung cancer hazard was dependent on cumulative dose, regardless of isotope, while the lung fibrosis hazard depended on both cumulative dose and isotope. The extended Cox methods used to estimate the hazard of lung, liver and bone cancer were unable to distinguish between a lagged cumulative dose or a cumulative dose adjusted for a threshold. However, multistage modeling techniques allow restriction of the dose to that accumulated during specific time periods and can assist in determining the timing of radiosensitivity. Models generated with multistage methods were an improvement over extended Cox models as measured by the deviance.;Studies of plutonium-induced health effects in animal models can complement human studies by providing more specific data than is possible in human observational studies. The Mayak Production Association (MPA) worker registry has shown evidence of plutonium-induced health effects in humans. Lung, liver and bone cancer mortality rate ratios in the MPA worker cohort were compared to those seen in beagle dogs and models of the excess relative risk of lung, liver and bone cancer mortality from the MPA worker cohort were applied to data from life-span studies of beagle dogs. The lung cancer mortality rate ratios in beagle dogs are similar to those seen in the MPA worker cohort as are the liver cancer mortality rate ratios at cumulative doses less than 3 Gy. Models of lung cancer mortality based on MPA worker data with additional covariates adequately described the beagle dog data while the liver and bone cancer models were less successful. Approaches to analysis that combine observational human data with experimental animal data may allow better estimates of risk.