Predictive models for myopia in children : Detection, Incidence, and Progression

Download files
Access & Terms of Use
open access
Embargoed until 2021-10-13
Copyright: Ashari Kandy, Divya Jagadeesh
Aim: To explore the association of optic disc and retinal parameters with myopia, and to determine if models incorporating these parameters can discriminate between myopic versus non-myopic eyes as well as aid in predicting future onset/incidence and increased progression. Methods: Optic nerve head and retinal parameters were systematically analysed using retinal images, spherical equivalent refraction, and axial length data obtained from two pilot studies (n=58 adults; n=56 myopic children, respectively). Thereafter, an association of optic disc and retinal parameters with myopic versus non-myopic eyes were analysed using data from a large population study of Asian children (n= 2995; 6 to 9 years). Multiple logistic regression analysis was performed and receiver operating characteristic (ROC) curves used to determine the ability of optic disc/retinal features to a) differentiate myopes and non-myopes at baseline; b) predict 12-month myopia incidence c) predict 12-month spherical equivalent and axial length change of ≥-1.00D and ≥0.50 mm respectively. The models were validated in a small, independent sample of 380 Asian children aged 5 to 14 years. Results: Optic disc and retinal parameters such as disc rotation, tilt from Vertical, ovality, reduced short axis of the disc, temporal crescent, tessellations, temporal > nasal pRNFL thickness, reduced fovea to disc distance as well a thinner macular thickness were associated with myopia and/or incidence and/or progression. Models incorporating one or more of these parameters had sensitivity/specificity of 77%/83% to discriminate between myopic versus non-myopic eye; 69%/68% to predict the incidence of myopia, and 58%/ 82% and 66%/79% to predict spherical equivalent and axial length change of ≥ 1D and ≥ 0.50 mm respectively. In an independent sample, the sensitivity/specificity of the models to discriminate between myopic versus non-myopic eyes was 90%/79%; to predict the incidence of myopia was 83%/ 50%; whereas models to predict progression had low sensitivity but high specificity. Conclusion: This body of work adds to the knowledge on the optic disc and retinal features associated with myopia and demonstrates that models incorporating such parameters can successfully discriminate between myopic versus non-myopic eyes. These features are also associated with future incidence and/or progression; however, further work is needed to improve the accuracy of models to predict incidence and/or progression.
Persistent link to this record
Link to Publisher Version
Link to Open Access Version
Additional Link
Ashari Kandy, Divya Jagadeesh
Sankaridurg, Padmaja
Jong, Monica
Fedtke, Cathleen
Conference Proceedings Editor(s)
Other Contributor(s)
Corporate/Industry Contributor(s)
Publication Year
Resource Type
Degree Type
PhD Doctorate
UNSW Faculty
download public version.pdf 8.76 MB Adobe Portable Document Format
Related dataset(s)