Diagnostic Performance of Prespecified OCT Rules for Glaucomatous Optic Neuropathy in Nonpathologic Myopia.

Abstract

Diagnosing glaucoma in myopic eyes is challenging due to overlapping structural features, such as optic disc tilt and retinal nerve fiber layer (RNFL) bundle shifts. Lacking standardized criteria for differentiating glaucomatous optic neuropathy (GON) from nonpathologic myopia, current commercial databases often flag healthy myopic eyes as abnormal. To evaluate the diagnostic performance of prespecified optical coherence tomography (OCT) rules for detecting GON in myopic eyes across diverse international populations and devices. This multicenter diagnostic study used a sequential 2-phase design consisting of a modified Delphi process to prespecify diagnostic rules and a cross-sectional diagnostic validation. Participants included adults with nonpathologic moderate or high myopia recruited from an internal validation cohort (Zhongshan Ophthalmic Center, China) and an international external validation cohort (centers in Hong Kong, Taiwan, the US, and India). Eyes with pathologic myopia (staphyloma or myopic maculopathy category ≥2) were excluded. Data were collected from January 2019 through December 2024 and analyzed from December 2024 through June 2025. OCT imaging of the peripapillary RNFL and macular ganglion cell-inner plexiform layer (mGC-IPL), with application of 5 prespecified diagnostic rules to detect GON. Sensitivity and specificity of 5 prespecified OCT rules were evaluated against a clinical reference diagnosis established by masked experts. The rules were as follows: rule A, temporal-superior-inferior-nasal-temporal (TSNIT) curve dip or depression; rule B, inferior peripapillary RNFL (pRNFL) thinning; rule C, inferotemporal mGC-IPL thinning; rule D, rule B or C; and rule E, rule B and C. Participants included 943 adults (1525 eyes) with nonpathologic moderate or high myopia recruited from an internal validation cohort. Of 1525 eligible eyes (mean [SD] participant age, 45.9 [17.2] years; 665 eyes [43.6%] from 402 female participants), 814 (53.4%) had confirmed GON. Rule A demonstrated the highest diagnostic utility. In the internal cohort (n = 841), sensitivity was 0.96 (95% CI, 0.94-0.98) and specificity was 0.95 (95% CI, 0.92-0.97). In the multiethnic external cohort (n = 684 eyes), rule A maintained a sensitivity of 0.93 (95% CI, 0.90-0.95) and specificity of 0.93 (95% CI, 0.90-0.96). Rule D also performed robustly, with an external sensitivity of 0.90 (95% CI, 0.87-0.93) and specificity of 0.93 (95% CI, 0.89-0.97). In this multicenter diagnostic study, morphological assessment of TSNIT curves and combinatorial analysis of inferior pRNFL and inferotemporal mGC-IPL thinning demonstrated high diagnostic accuracy for GON in nonpathologic myopic eyes. These expert-derived rules offer objective, generalizable decision support for reducing diagnostic uncertainty in the myopic population.