Choroidal measurements in decision making for retinopathy of prematurity: a decision tree analysis

BACKGROUND: To compare the choroidal thickness and vascular profile of premature infants with ROP (retinopathy of prematurity) using a handheld SD-OCT device. METHODS: We performed horizontal SD-OCT scans through the fovea in 115 eyes of 66 premature infants. Premature infants included 2 groups infants with ROP requiring treatment (as treatment group) vs. infants without ROP or with ROP not- requiring treatment (as no-treatment group) Choroidal thicknesses (CT) were measured at 5 points, including the fovea, 250 µm, and 500 µm mm nasal and temporal to the fovea. The choroidal vascularity index (CVI) and choroidal stromal index (CSI) were also calculated. The classification and regression tree (CRT) algorithm was used to predict the need for treatment based on all OCT characteristics. RESULTS: Mean CT was higher in 500 µm nasal to the fovea compared to temporal CT (275.8 ± 64.8 and 257.1 ± 57.07, P value 0.05, GEE). The decision tree analysis showed that in infants with ROP, the most important predictor for the need for treatment is CSI. CONCLUSION: This study delineated the possible effectiveness of choroidal measurements as an additive to decision-making for ROP. We also demonstrated that choroidal involution is associated with the presence of plus disease, not with the stage of ROP. We demonstrated that choroidal measurements are very sensitive but not specific tools for assessing the need for treatment in ROP patients.