Predictive models for the incidence and progression of myopia in children

Purpose : Myopia related axial elongation causes structural changes in the optic disc and retinal features (ODRFs). We aimed to determine if changes in the ODRFs can be used to predict the incidence and progression of myopia in children. Methods : We retrospectively analyzed 1-year longitudinal data from a 2-year school-based randomized controlled study conducted by Shanghai Eye Disease Prevention and Treatment Centre, China. Of the 12-24 months’ data analyzed, 12 months’ data was considered as baseline. The right eyes of 2851 children (age: 6-9 years) were analyzed. A swept-source optical coherence tomography was used to acquire 3D wide scans and fundus images. ODRFs analyzed were optic disc tilt, rotation, and ovality index, fovea-to-disc distance (FDD), the presence of tessellations, extent of temporal crescent, retinal nerve fiber layer thickness (RNFLT) and superior- and inferior- temporal retinal artery (STRAA and ITRAA) and vein angles. Fundus images were analyzed using custom-coded MATLAB software. Binary logistic regression and receiver operating characteristic (ROC) curves were used to determine the association of ODRFs with a) 12-month incidence of myopia and b) 12-month progression of axial length (AL) > 0.5mm. Level of significance was set at 5%. Results : The baseline prevalence of myopia in the 2851 participants (mean age 7.4 ± 0.6) was 15.2%. The mean change in spherical equivalent and AL (baseline to 12 months) were -0.50 ± 0.47 DS and 0.30 ± 0.18 mm. Annual incidence of myopia was 12.8% (n=365) and AL progression >0.5 mm was 14.5% (n=414). The significant ODRFs in ‘myopia incidence model’ were reduced FDD (p<0.001), RNFLT– temporal > nasal (p<0.000), smaller optic disc short axis (p=0.001) increased total tessellations (p=0.027), an increased STRAA (p=0.028), and increased optic disc tilt (p=0.039). The significant ODRFs in ‘AL progression model’ were myopia at baseline (p<0.001), increased STRAA and ITRAA (p<0.001), increased optic disc tilt (p=0.002), reduced FDD (p=0.003) and RNFLT– temporal > nasal (p=0.008). For both models area under the ROC showed good predictability (0.80 & 0.82 respectively). Conclusions : Multiple ODRFs contributed to the ‘myopia incidence’ and ‘AL progression’ models evidencing superior-nasal changes. AL progression was also associated with inferior changes. Following a planned validation study, these models could aid in predicting the incidence and progression of myopia in children. This is a 2020 ARVO Annual Meeting abstract.