Ocular Signs and Testing Most Compatible with Sarcoidosis-Associated Uveitis: A Latent Class Analysis.

ABSTRACT

Purpose: This study aims to explore the potential subgroups of sarcoidosis-associated uveitis (SAU) within a multicenter cohort of uveitis participants. Design: Cross-sectional study. Participants: A cohort of 826 uveitis patients from a uveitis registry from 19 clinical centers in 12 countries between January 2011 and April 2015. Methods: We employed a latent class analysis (LCA) incorporating recommended tests and clinical signs from the revised International Workshop on Ocular Sarcoidosis (IWOS) to identify potential SAU subgroups within the multicenter uveitis cohort. Additionally, we assessed the performance of the individual tests and clinical signs in classifying the potential subclasses. Main Outcome Measures: Latent subtypes of SAU. Results: Among 826 participants included in this analysis, the 2-class LCA model provided a best fit, with the lowest Bayesian information criteria of 7218.7 and an entropy of 0.715. One class, consisting of 548 participants, represented the non-SAU, whereas the second class, comprised of 278 participants, was most representative of SAU. Snowballs/string of pearls vitreous opacities had the best test performance for classification, followed by bilaterality and bilateral hilar lymphadenopathy (BHL). The combination of 4 tests with the highest classification importance, including snowballs/string of pearls vitreous opacities, periphlebitis and/or macroaneurysm, bilaterality, and BHL, demonstrated a sensitivity of 84.8% and a specificity of 95.4% in classifying the SAU subtypes. In the exploratory analysis of the 3-class LCA model, which had comparable fit indices as the 2-class model, we identified a candidate non-SAU subtype, candidate SAU subtype with pulmonary involvement, and a candidate SAU with less pulmonary involvement. Conclusions: Latent class modeling, incorporating tests and clinical signs from the revised IWOS criteria, effectively identified a subset of participants with clinical features indicative of SAU. Though the sensitivity of individual ocular signs or tests was not perfect, using a combination of tests provided a satisfactory performance in classifying the SAU subclasses identified by the 2-class LCA model. Notably, the classes identified by the 3-class LCA model, including a non-SAU subtype, an SAU subtype with pulmonary involvement, and an SAU subtype with less pulmonary involvement, may have potential implication for clinical practice, and hence should be validated in further research. Financial Disclosures Proprietary or commercial disclosure may be found in the Footnotes and Disclosures at the end of this article.