Purchase this article with an account.
Jing Xie, Yanxian Chen, Mingguang He; Prediction of Future Spherical Equivalent Refraction in Children using the Longitudinal Data from the Guangzhou Twins Eye. Invest. Ophthalmol. Vis. Sci. 201657(12):.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
The prediction of Spherical Equivalent (SE) Refraction development plays an important role in myopic prevention. In this paper, we characterize individual and sample average growth curves based on longitudinal data and further determine the best set of predictors for Spherical Equivalent (SE) in school-aged children effect from Zhongshan Ophthalmic Center, Sun Yat-sen University.
The first-born twins (n=1221) in Guangzhou Twin Eye Study with 10-year annual visit data (baseline age 7-15 years) were used to develop a mixed effect model. SE was calculated as the sum of sphere and 1/2 cylinder. We included baseline SE, age, indoor reading time, outdoor activity, and 39 candidate SNPS identified from the Consortium for Refractive Error and Myopia (CREAM) to develop the best model to predict the SE at 15 years old (endpoint SE). Root Mean Squared Error (RMSE) was used to assess the performance by comparing the difference between the true SE and the predicted endpoint SE. Bootstrapping method was used for the internal validation of the prediction models.
Of the factors evaluated, SE at baseline, age, age2, 4 SNPs were statistically associated with the risk for endpoint SE (all P < 0.05). Indoor reading time and outdoor activity were not associated. The mixed effects quadratic polynomial regression model with SE at baseline, age and age2 had better predictive ability in terms of SE absolute bias (-8.10e-11), RMSE (0.35), AIC (12708.97), BIC (12763.7) than the models with additional four SNPs (SE absolute bias (-2.48e-10), RMSE (0.63), AIC (1768.07), BIC (1736.48).
The proposed simple model using age, age2 and baseline SE can produce highly accurate prediction on the SE at adulthood whereas susceptible SNPs are statistically significantly associated but do not provide further improvement on the prediction.
This is an abstract that was submitted for the 2016 ARVO Annual Meeting, held in Seattle, Wash., May 1-5, 2016.
This PDF is available to Subscribers Only