oa Statistical Issues for Instrument Validation and Inter-laboratory Comparison of Automated Systems

Statistical treatment of data for human biomonitoring has greatly improved within the last decade, and the most advanced techniques have been translated from the analysis of classic epidemiologic studies to molecular epidemiology. The use of more sophisticated techniques has improved the precision of estimates in human population studies, increasing the reliability of study results. In parallel, the increased popularity of pooled analyses, created the opportunity for a deeper insight into the sources of variability. Large collaborative studies published over the last few decades have revealed that the inter-laboratory and especially the inter-scorer variation are the most important source of variability, setting this heterogeneity as a priority field to address. The recent development of automated systems for chromosome damage scoring is going to dramatically change the level of reliability of these biomarkers. Before introducing these methods, robust standardization studies have to be started, aimed at comparing automated systems in different setting and the overlapping between different company systems. During the presentation we will discuss the list of priorities for systems standardization and the most suitable study design and the statistical analyses to be implemented for addressing these priorities.