An Analysis of Uncertainty for Dose Estimation through the γ-H2AX Assay

Abstract

Currently, the primary biomarker for radiation biodosimetry is the dicentric chromosome. However, the γ-H2AX histone is an alternative assay that is less labour-intensive. Blood samples can be taken more quickly than for the dicentric biomarker, and a larger number of samples can be handled within a given time frame. In this thesis, we discuss several statistical techniques for how to handle scored γ-H2AX foci data. We then apply these techniques to two datasets from Public Health England, using one to demonstrate techniques, and the second to check that the dose-response curve calculation and dose estimation techniques work. Throughout we choose to fit quasi-Poisson models instead of Poisson ones to account for overdispersion present within the foci count data. After fitting both linear and quadratic dose-response curves we create controls to validate the curves, using a reference sampling ratio to scale them if necessary. By calculating the uncertainty we show why linear fits are preferable to quadratic ones. We finally compare our linear dose-response curves from both datasets for multiple timepoints with pre-existing ones from the literature to see how they compare and what conclusions can be drawn about dose-response curves for this assay in general.