Mucilair 3D bronchial airway models, cultured at an air-liquid interface, were exposed to aerosols of copper oxide (CuO) nanoparticles in Vitrocell air exposure modules. Four cell donors, four exposure modules and four exposure concentrations were varied within four different exposure sessions using a statistical experimental design called a hyper-Graeco-Latin square. Analysis of variance techniques were used to investigate the effects of these factors on release and RNA expression of inflammation markers monocyte chemoattractant protein-1 (MCP-1) interleukines 6 and 8 (IL-6 and IL-8) an cytotoxicity marker lactate dehydrogenase (LDH) determined 24 h after exposure. The same techniques were also used to conduct a global analysis on RNA expressions of 10,000 genes. There were no major signs of cytotoxicity. Release of IL-6 and MCP-1 was affected by CuO concentration, and, for MCP-1, by donor variation. IL-8 release was not affected by these factors. However, gene expression of all three inflammation markers was strongly affected by CuO concentration but not by the other factors. Further, among the 10,000 genes involved in the global analysis of RNA expression, 1736 were affected by CuO concentration, 704 by donor variation and 269 by variation among exposure sessions. The statistical design permitted the assessment of the effect of CuO nanoparticles on 3D airway models independently of technical or experimental sources of variation. We recommend using such a design to address all potential sources of variation. This is especially recommended if test materials are expected to be less toxic than CuO, because the variation among the concentration levels could then be close to the variation among donors or exposure sessions.