Estimating dose and exposure fraction from radiation biomarkers in the presence of overdispersion

Abstract

It is typically assumed that the total -H2AX foci produced in a sample of blood cells is Poisson distributed, whose expected yield can be represented by a linear function of the absorbed dose. However, in practice, because of unobserved heterogeneity in the cell population, the standard Poisson assumption of equidispersion will most likely be contravened which will cause the variance of the foci counts to be larger than their mean. In both whole and partial body exposure this phenomenon is perceptible, unlike in the context of the dicentric assay in which overdispersion is usually considered only to be linked to partial exposure. For such situations, and as we will demonstrate, it is suitable to utilise a model that can handle overdispersion such as the quasi-Poisson or negative binomial regression.

The scenarios of most radiation accidents result in partial-body exposures or non-uniform dose distribution, leading to a differential exposure of lymphocytes in the body. Subsequently for the exposed individuals, their blood will contain a mixture of cells showing no radiation impact at all and cells featuring a distribution of counts according to dose of exposure. For such exposure scenarios, it remains that there are no statistical procedures to follow for the -H2AX assay. Part of this work will focus on updating the contaminated Poisson method, traditionally used in conjunction with cytogenetic biomarkers, to enable an estimate of the radiation dose and irradiated fraction to be found in the presence of both zero-inflation and overdispersion. As an extension, we discuss and compare how to measure the uncertainty associated with a given dose estimate via the delta, Merkle and ISO methods. We illustrate their applications firstly via simulated zero-inflated Poisson and NB1 data, with the non-inflated part being generated using an external -H2AX whole-body calibration curve, before applying the methodology to practical data.