RESUMO
PURPOSE: To evaluate the agreement between non-cycloplegic autorefraction (NCAR) and cycloplegic autorefraction (CAR) in an ethnically diverse population of preschool-aged children and the validity of the screening criteria used to refer for further evaluation. METHODS: This study included data from 7,073 preschoolers who underwent NCAR and CAR, which enabled refractive error classification based on the American Association for Pediatric Ophthalmology and Strabismus (AAPOS) 2013 criteria. Right eye data of sphere and cylinder were used to compare NCAR to CAR via paired t-testing and vector analyses, and left eye data for an analysis on anisometropia. The sensitivity and specificity of screening referral criteria for refractive error were calculated. RESULTS: Mean values of sphere differed between NCAR and CAR by 1.95 ± 1.45 D (p < .05) with 95% limits of agreement (LoA) of -0.94 to 4.85 D, with less discrepancy found in myopic eyes. The mean values of cylinder differed by -0.08 ± 0.43 D (p < .05) with 95% LoA of -0.93 to 0.77 D. Power vector results reflected a similar lack of agreement. The sensitivity and specificity of our screening referral criteria were, respectively, 66% and 84% for myopia, 66% and 98% for hyperopia, and 98% and 58% for astigmatism. CONCLUSION: NCAR is insufficient in preschoolers for spherical refractive error referrals. Level of agreements was lower for spherical (15.5% within 0.5D) and higher for cylindrical refractive errors (89.6%) compared to CAR. In the absence of cycloplegic examination, screening programs using NCAR should utilize low referral thresholds for spherical refractive error.
