|Title||MYOPIA ONSET PREDICTION IN SCHOOLCHILDREN|
|Author, Co-Author||Karla Zadnik, Loraine Sinnott, Donald Mutti|
|Abstract|| PURPOSE: To develop a model for juvenile myopia onset prediction.
METHODS: Data from 4,512 children enrolled in the CLEERE Study were analyzed, excluding children who entered the study myopic. Four hundred fourteen children became myopic between ages 8 and 14, with annual myopia incidences ranging from 3.7 to 5.4%. Candidate predictors included baseline refractive error (SPHEQ; cycloplegic autorefraction), parental myopia, diopter hours; hours per week outdoors; axial length (AL), lens thickness, corneal power (CP), Gullstrand lens power (GLP), AC/A ratio, relative peripheral refraction, accommodative lag, and astigmatism (J0, J45). Using discrete time survival analysis, models were created and areas under the curve (AUCs) computed. Cross-validation was used to select the best model with varying numbers of predictors.
RESULTS: This data-driven method generated optimal models for baseline ages 6-12 with a total of 8 unique variables (SPHEQ, parental myopia, AL, AC/A, CP, GLP, J0, and J45), from which we created 5 models to be used for all ages, ranging from a model with all 8 variables to one with only SPHEQ. AUCs in this model ranged from 0.593 to 0.962; only the models trying to predict myopia at age 14 from age 6 data yielded unacceptably low AUCs. Because of the similarity of the AUCs, we concentrated on a model using only SPHEQ at age 7 to predict myopia at ages 8 through 14. Separately, it was determined that non-cycloplegic SPHEQ data yielded unacceptably poor AUCs. The SPHEQ cutpoint that predicted myopia across ages 8 through 14 based on age 7 data ranged from +0.41 to +0.58 D and yielded sensitivities ranging from 0.716 to 0.905 and specificities from 0.622 to 0.762.
CONCLUSIONS: Data from 7-year-old children yielded a single best predictor of myopia onset: spherical equivalent refractive error as measured by cycloplegic autorefraction, and mild hyperopia at age 7 was the optimum cutpoint for predicting myopia onset between ages 8 and 14. The usefulness of this predictor depends on the safety and efficacy of any proposed prevention of myopia.
ADDITIONAL COMMENTS: NIH/NEI Grant U10-EY08893
|Affiliation of Co-Authors||The Ohio State University, College of Optometry, The Ohio State University, College of Optometry|