Purchase this article with an account.
Jaclyn Hernandez, Loraine T Sinnott, Noel A Brennan, Xu Cheng, Karla Zadnik, Donald O Mutti; Analysis of CLEERE data to test the feasibility of identifying future fast myopic progressors. Invest. Ophthalmol. Vis. Sci. 2018;59(9):3388.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Faster myopia progression and early onset are more likely to lead to higher amounts of myopia, but it is unclear how to identify children at risk of fast progression other than by age. To assess how well the use of the present and the previous year’s axial length and spherical equivalent data predict the rate of progression over the next year, we analyzed data from the Collaborative Longitudinal Evaluation of Ethnicity and Refractive Error (CLEERE) Study.
CLEERE data were restricted to visits from subjects with measurements of axial length (AL) and spherical equivalent refractive error (SER) for three consecutive years. Data were from 916 subjects with SER between -0.75 and -5.00 D (inclusive) and between 7 and 14.5 years of age at the second visit, resulting in 2,231 observations. Subjects were randomly assigned to either the model training set or the model testing set, using a 55%/45% split. Models predicting AL and SER at the third visit were fit using the training set. Two optimal models were selected for each outcome, one incorporating AL and SER from the prior two years as predictors (the “History” Model) and the other allowing AL and SER taken in only the prior year (the “No History” Model). Race/ethnicity, gender, age, parental history of myopia, and reading and outdoor activity levels were also examined as possible predictors. Model fitting aimed to classify subjects as fast or slow progressors. A range of definitions of fast progression was examined. Optimal definitions were used for model evaluation.
For both the History and No History models of future AL, 0.222 mm/year was the optimal choice for the definition of fast progression. For the History and No History Models of future SER, -0.375 D/year was the optimal choice. The table provides the test set evaluation findings for these optimal change levels.
The models provide modest prediction of future refractive progression, but AL and SER history add little to prediction in this extensive data set, decreasing misclassification only about 2%. Additional input parameters may be required for better prediction of refractive error progression.
This is an abstract that was submitted for the 2018 ARVO Annual Meeting, held in Honolulu, Hawaii, April 29 - May 3, 2018.
This PDF is available to Subscribers Only