Cosmetics Europe collected, de-coded and evaluated the respective results. As a minimum, test developers were asked to complete a checklist including the results but also e.g. information on timing or protocol adherence. If provided or available, further supplementary information including the test protocol, publications or raw test data were collected. Information on 15 of the 16 test methods was compiled systematically to enable evaluation on the basis of criteria that were defined by the Cosmetics Europe

Skin Sensitisation Task Force. The PPRA is not included in this compilation because its standardisation was finalised only after evaluation had commenced. Twenty evaluation Trametinib criteria addressing various aspects of interest were considered. For clarity, these were grouped under the headings ‘General points’, ‘Standard Operation Procedure (SOP) and prediction model’, ‘data’, ‘ease of transfer’ and ‘throughput’ (Table 1). Each test method was also mapped onto the skin sensitisation Idelalisib in vitro AOP (Fig. 1). The data analysis focused on the test results for the ten substances. These were available for all 16 methods. The completeness of results and their concordance with the pre-defined reference results based on LLNA EC3 values (and human data for SLS) was evaluated. If data on other substances were available,

overall sensitivity, specificity and concordance were calculated. For the 15 test methods differentiating non-sensitising and sensitising substances, the reference results were derived from both LLNA EC3 values distinguishing five potency categories and in parallel from human data using six classes (Basketter et al., 2014). Both result in the same potential,

for nine substances (no EC3 value: non-sensitiser; EC3 value: sensitiser; human potency classes 5 and 6: non-sensitiser; human potency classes 1–4: sensitiser). As SLS is false positive in the LLNA compared to Methisazone human, it was considered as a non-sensitiser. The seven test methods attempting to predict skin sensitisation potency results used method-specific potency categories that did not necessarily correspond to those of the reference results. Therefore, the potency prediction results are described only, without detailed predictivity analysis. The focus of the method evaluation exercise was to establish a harmonised knowledge base for each of the test methods in order to prioritise methods for further consideration. This evaluation was carried out in close collaboration with the test method developers, whose review concluded the evaluation process. The method developers were invited to a two-day workshop with the Cosmetics Europe Skin Tolerance Task Force held in Brussels on December 3rd and 4th 2012 to discuss their methods and results, the requirements of the cosmetics industry and the strategy of the task force to meet these needs.