Topographic Agreement of Retinal Nerve Fiber and Ganglion Cell Loss Improves Incipient Glaucoma Detection.
Adam Z Xu, Giacinto Triolo, Maria J Chaves-Samaniego, Maximilian G Konarzewski, Steven J Gedde, Jean Claude Mwanza, Donald L Budenz, Felipe A Medeiros, Luis E Vazquez
Summary
Combining corresponding retinal nerve fiber and ganglion cell layer parameters, factoring in their correlation, significantly improves detection of incipient glaucoma, offering clinicians a powerful diagnostic tool.
Abstract
PURPOSE
To evaluate combinations of geographically corresponding retinal nerve fiber layer (RNFL) parameters and ganglion cell inner plexiform layer (GCIPL) parameters for detection of incipient glaucoma.
DESIGN
Cross-sectional study.
SUBJECTS
The early glaucoma exploratory cohort consisted of 156 diseased subjects and 199 age-matched controls. The incipient glaucoma validation cohort consisted of 75 diseased subjects and 247 age-matched controls. Only one eye from each subject was used.
METHODS
RNFL and GCIPL scans were obtained with Cirrus HD-OCT. Linear regression was used to determine the degree of correlation between corresponding RNFL and GCIPL sector thinning in early glaucoma. Logistic regression models of parameter combinations were fitted on the early glaucoma cohort, then used to generate receiver operating characteristic (ROC) curves. Fitted models were validated on the incipient glaucoma dataset. Diagnostic power was assessed by calculating area under the ROC curve (AUC).
MAIN OUTCOME MEASURES
Area under the ROC curve (AUC)
RESULTS
In early glaucoma patients, inferotemporal GCIPL loss is moderately correlated to inferior RNFL loss (R= 0.38), and supertemporal GCIPL loss is moderately correlated to superior RNFL loss (R= 0.32). A logistic regression model of all these parameters combined improved significantly when these correlations were accounted for with interaction terms (ΔAIC = -17.4, ΔBIC = -9.7). The combined model performed strongly in diagnosing both early glaucoma (AUC = 0.969) and incipient glaucoma (AUC = 0.902) and outperformed all of the individual parameters in both early and incipient glaucoma (unpaired bootstrapped AUC comparisons, p<0.0001 for all).
CONCLUSIONS
Combining corresponding sectors of the GCIPL and RNFL, and factoring in their correlation, produces a model with strong diagnostic accuracy both in early and incipient glaucoma. Attention to corresponding GCIPL and RNFL thickness loss in these sectors can help clinicians detect incipient glaucoma.
Top Research in Diagnosis & Screening
Browse all →Efficacy of a Deep Learning System for Detecting Glaucomatous Optic Neuropathy Based on Color Fundus Photographs.
Dry eye disease and oxidative stress.
Central Corneal Thickness in the Ocular Hypertension Treatment Study (OHTS).
Discussion
Comments and discussion will appear here in a future update.