Performance validation of the KeraSkin™ reconstructed human epidermis (RhE) model according to OECD document no. 219 and exploratory optimization of its predictive capacity for in vitro skin corrosion testing

초록

This study evaluated the KeraSkin™ reconstructed human epidermis (RhE) model for in vitro skin corrosion testing per OECD Series on Testing and Assessment No. 219. A validation study across three laboratories assessed reliability and predictive performance using OECD reference chemicals, followed by an expanded evaluation of 79 chemicals.In accordance with Document No. 219, KeraSkin™ demonstrated overall compliance with mandatory requirements. For binary classification (n = 24), the model showed within-laboratory reproducibility (WLR) ≥95.8% and between-laboratory reproducibility (BLR) of 100.0%, satisfying predictive benchmarks (n = 20) with 100% sensitivity and 85.6% accuracy. For UN GHS subcategorization (n = 30), the model met acceptance criteria with WLR ≥90.0% and BLR ≥80.0% using the original prediction model (PM), achieving 100% sensitivity for Category 1A. While one laboratory's three-category accuracy (65.6%) fell below the indicative target, the aggregated multi-laboratory results satisfied all mandatory requirements of Document No. 219.In the expanded dataset (n = 79), the optimized PMvar2 effectively addressed laboratory variability, increasing overall accuracy to 73.5% while maintaining strict control over Category 1A misclassification. Whereas the original PM ensures current regulatory compliance, PMvar2 is presented as a promising strategy for enhancing predictive stability, with its adoption intended for further discussion during the formal OECD evaluation. Together, KeraSkin™ is a robust, regulatory-compliant RhE model fulfilling all OECD TG 431 inclusion criteria for non-animal skin corrosion hazard classification. Highlights •KeraSkin™ RhE validated according to OECD Series on Testing and Assessment No. 219.•100% sensitivity for UN GHS Category 1A chemicals.•No false-negatives in safety-critical subcategorization.•Optimized prediction model improved classification accuracy.

키워드