Assessing Nonlinearity in Harderian Gland Tumor Induction Using Three Combined HZE-irradiated Mouse Datasets

Recently reported studies considering nonlinearity in the effects of low-dose space radiation have assumed a nontargeted mechanism. To date, few analyses have been performed to assess whether a nontargeted term is supported by the available data. The Harderian gland data from Alpen et al. (published in 1993 and 1994), and Chang et al. (2016) provide the most diversity of ions and energies in a tumor induction model, including multiple high-energy and charge particles. These data can be used to investigate various nonlinearity assumptions against a linear model, including nontargeted effects in the low-dose region or cell sterilization at high doses. In this work, generalized linear models were used with the log complement link function to analyze the binomial data from the studies independently and combined. While there was some evidence of nonlinearity that was best described by a cell-sterilization model, the linear model was adequate to describe the data. The current data do not support the addition of a nontargeted effects term in any model. While adequate data are available in the low-dose region (<0.5 Gy) to support a nontargeted effects term if valid, additional data in the 1-2 Gy region are necessary to achieve power for cell-sterilization analysis validation. The current analysis demonstrates that the Harderian gland tumor data do not support the use of a nontargeted effects term in human cancer risk models.