Abstract
PURPOSE: Given the subjective nature of clinical activity score (CAS) and variability between observers, this study evaluated the inter-observer variability of soft tissue signs (STS) in the CAS using periocular photographs and compared the diagnostic performance of photographic CAS (P-CAS) with clinical assessment for detecting active thyroid eye disease (TED). METHODS: This retrospective cohort study included 754 TED patients who underwent periocular photography at Seoul National University Bundang Hospital between 2006 and 2021. Five oculoplastic specialists independently evaluated five STS of CAS using periocular photographs. Inter-observer agreement and concordance between photographically (P-STS) and clinically (C-STS) assessed STS, assessment consistency, were analyzed using Kappa statistics. Diagnostic accuracy of P-CAS was evaluated via receiver operating characteristic (ROC) curves with additional analysis incorporating demographic variables such as sex, age, and smoking dose. RESULTS: Assessment consistency was moderately good for redness of eyelid and conjunctiva (Kappa = 0.418, 0.499). The diagnostic performance of P-CAS (area under the curve, AUC = 0.774) was lower than that of clinical CAS (C-CAS) (AUC = 0.818, p = 0.007), However, it significantly improved when demographic factors were added (AUC = 0.851). CONCLUSION: Periocular photography provides a reproducible and clinically valuable adjunct for assessing TED activity. The redness-related items showed higher agreement than swelling-related ones, and integrating demographic variables enhances diagnostic performance. This approach may contribute to standardizing TED assessment and serve as a foundation for future automated or image-based evaluation tools.