The aim of this study was to determine the reproducibility of data obtained from in vitro irritation testing using three industrial reconstructed human epidermis models, EpiDermTM, EpiskinTM and SkinEthicR, and one in-house model developed at Wella/Cosmital. A common protocol was established based on the measurement of cytotoxicity in the 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyl tetrazolium bromide (MTT) assay and of extracellular release of proinflammatory mediators and cytosolic enzymes after a range of exposure times to sodium lauryl sulfate (SLS). This time course protocol was applied to 6 different batches of each skin model using triplicate tissue cultures per test condition. The parameters analyzed for intra- and inter-batch reproducibility were the cell viability determined as MTT reduction capacity and the ET-50 values in the 6 batches, as well as the release of the cytokine IL-1α and of the cellular enzymes LDH and GOT in 3 batches only. The MTT viability results showed that EpiDerm was the most resistant to the SLS treatment and at the same time the most reproducible model, SkinEthic was the most sensitive to SLS and the least reproducible, and Episkin and the Cosmital model were intermediate. Measurements of IL-1α release showed a relatively high intra- and inter-batch variability in all the skin models. It was not possible to detect the extracellular release of the enzymes LDH and GOT in the Episkin assay medium. With the 3 other models, the release of LDH and GOT varied in about the same range as that of IL-1α. For all the parameters in this study, the inter-batch variability was generally greater than the intra-batch variability. A possible reduction in the number of batches and replicates for future applications in routine irritancy testing is discussed on the basis of the results obtained using 6 batches in triplicate.