The aim of the study was to investigate the ability of human reconstructed epidermis EpiSkin(LM) to identify the phototoxic potency of topically or systemically applied chemicals (EPA: EpiSkin phototoxicity assay). Three classes, according to their available human phototoxic potential, were evaluated: systemic phototoxic compounds, topical phototoxic chemicals and non-phototoxic compounds. Non-cytotoxic concentrations of chemicals were applied topically or directly added to the underlying culture medium in order to mimic a systemic-like administration. Following treatment, tissues were exposed to non-cytotoxic dose of UVA (50Jcm(-2)). Cell viability and pro-inflammatory mediators (IL-1alpha) were investigated 22h after UVA exposure. Our results show that the phototoxic potential of chemicals can be determined using cell viability combined with inflammatory mediator measurements (cytokine IL-1alpha) in a 3-D epidermis model. Moreover, the EPA was able to discriminate efficiently between phototoxic and non-phototoxic products. Furthermore, the EPA is sensitive to the administration route in the prediction of the phototoxic potency of the tested chemical. Differences observed between the two routes of administration (topical or systemic-like) may be linked in part to chemicals bioavailability which depends on specific penetration potential, epidermis barrier function and also on keratinocytes absorption/metabolization processes. Results were very promising and showed a very good sensitivity (92.3%) and an excellent specificity (100%) with an overall accuracy of 94.1%. The performances of the EPA showed that the EpiSkin(LM) model is an interesting tool able to integrate decision-making processes to address the question of phototoxicity linked to the application site.