Abstract
Photosafety assessments are a key requirement for the safe development of pharmaceuticals, cosmetics, and agrochemicals. Although in vitro methods are widely used for phototoxicity and photoallergy testing, their limited applicability and predictive power often necessitate supplemental in vivo studies. To address this, we developed the PhotoChem Reference Chemical Database, comprising 251 reference compounds with curated data from in vitro, in vivo, and human studies. Using this database, we evaluated the predictive capacity of three OECD in vitro test guidelines-TG 432 (3T3 NRU), TG 495 (ROS assay), and TG 498 (reconstructed human epidermis)-by comparing the results against human and animal data. Against human reference data, all three test methods showed high sensitivity (≥82.6%) and strong overall accuracy: TG 432 (accuracy: 94.2% (49/52)), TG 495 (100% (27/27)), and TG 498 (86.7% (26/30)). In comparison with animal data, sensitivity remained high for all tests (≥92.0%), while specificity varied: TG 432 (54.3% (19/35)), TG 495 (63.6% (7/11)), and TG 498 (90.5% (19/21)). TG 498 demonstrated the most balanced performance in both sensitivity and specificity across datasets. We also analyzed 106 drug approvals from major regulatory agencies to assess real-world application of photosafety testing. Since the mid-2000s, the use of in vitro phototoxicity assays has steadily increased in Korea, particularly following the 2021 revision of the MFDS regulations. Test method preferences varied by region, with 3T3 NRU and ROS assays most widely used to evaluate phototoxicity, while photo-LLNA and guinea pig tests were frequently employed for photoallergy assay. Collectively, this study provides a valuable reference for optimizing test method selection and supports the broader adoption of validated, human-relevant non-animal photosafety assessment strategies.