RENEB Inter-Laboratory comparison 2017: limits and pitfalls of ILCs

In case of a mass-casualty radiological event, there would be a need for networking to overcome surge limitations and to quickly obtain homogeneous results (reported aberration frequencies or estimated doses) among biodosimetry laboratories. These results must be consistent within such network. Inter-laboratory comparisons (ILCs) are widely accepted to achieve this homogeneity. At the European level, a great effort has been made to harmonize biological dosimetry laboratories, notably during the MULTIBIODOSE and RENEB projects. In order to continue the harmonization efforts, the RENEB consortium launched this intercomparison which is larger than the RENEB network, as it involves 38 laboratories from 21 countries. In this ILC all steps of the process were monitored, from blood shipment to dose estimation. This exercise also aimed to evaluate the statistical tools used to compare laboratory performance.

Blood samples were irradiated at three different doses, 1.8, 0.4 and 0 Gy (samples A, C and B) with 4-MV X-rays at 0.5 Gy min^-1, and sent to the participant laboratories. Each laboratory was requested to blindly analyze 500 cells per sample and to report the observed frequency of dicentric chromosomes per metaphase and the corresponding estimated dose.

This ILC demonstrates that blood samples can be successfully distributed among laboratories worldwide to perform biological dosimetry in case of a mass casualty event. Having achieved a substantial harmonization in multiple areas among the RENEB laboratories issues were identified with the available statistical tools, which are not capable to advantageously exploit the richness of results of a large ILCs. Even though Z- and U-tests are accepted methods for biodosimetry ILCs, setting the number of analyzed metaphases to 500 and establishing a tests' common threshold for all studied doses is inappropriate for evaluating laboratory performance. Another problem highlighted by this ILC is the issue of the dose-effect curve diversity. It clearly appears that, despite the initial advantage of including the scoring specificities of each laboratory, the lack of defined criteria for assessing the robustness of each laboratory's curve is a disadvantage for the 'one curve per laboratory' model.

Based on our study, it seems relevant to develop tools better adapted to the collection and processing of results produced by the participant laboratories. We are confident that, after an initial harmonization phase reached by the RENEB laboratories, a new step toward a better optimization of the laboratory networks in biological dosimetry and associated ILC is on the way.