Clinical Validation of Apollo™, A Novel Web-Based Tool for Refractive Error Self-Assessment.

Refractive errors are a leading cause of visual impairment, with substantial economic impact. Digital refraction systems may help improve access to eye care, especially for underserved populations. Considering their relevance, this study evaluated the Apollo™, a novel web-based system for self-assessed refractive error measurement, in estimating spherical equivalent (SEQ) compared to manifest refraction (also known as dry refraction or dynamic refraction) performed by experienced ophthalmologists. A prospective, single-center, single-blind, non-randomized study was conducted with 63 eyes of 64 healthy adults aged 18-40 years. Each participant underwent both manifest subjective refraction and digital refraction using Apollo™. Exclusion criteria included astigmatism exceeding 3.00 diopters (D) in absolute value and SEQ beyond - 4.00 D to + 4.00 D. Primary outcome was the comparison of SEQ between methods. No significant difference was observed between SEQ from Apollo™ (mean - 1.20 ± 1.11 D) and manifest refraction (mean - 1.17 ± 1.14 D; p = 0.67). Bland-Altman analysis showed a bias of 0.03 ± 0.90 D, with 95% limits of agreement (LOA) from - 1.73 to 1.78 D. LOA was narrower in myopic SEQ stratum (-1.48 to 1.35D). The ICC for the total sample SEQ was 0.69 (95% CI: 0.52-0.79), and for the myopic SEQ subgroup was 0.74 (95% CI: 0.58-0.84), indicating moderate to good reliability. The accuracy for detecting the sign of SEQ was 82.5% in the overall cohort and 96.2% in the myopic subgroup. Regression analysis revealed a proportional bias, with the system tending to underestimate refractive error as the magnitude of myopia increases (β = -0.27, p = 0.005). In conclusion, Apollo™ demonstrates moderate-to-good reliability and high diagnostic accuracy (96%) in categorizing refractive status among myopes. These results highlight its potential as a valuable digital screening tool, warranting further large-scale studies to refine its predictive algorithm.