Color-Code classification versus continuous data by Spectral Domain-Optical Coherence Tomography for detecting glaucoma: Multicenter Italian Glaucoma Imaging Study (MIGIS).

Abstract

Purpose : To evaluate whether the diagnostic accuracy of OCT parameter as estimated by either ROC curve (from continuous data) and Sensitivity/Specificity approaches(from color code classification), is consistent. And to provide clinicians, researchers and manifacturer with practical suggestion on how to make the best use of OCT analysis for detecting glaucoma.

Methods : Healthy controls and primary open angle glaucoma were enrolled consecutively from the Multicentric Italian Glaucoma Imaging Study. All patients underwent a full eye examination, standard automated perimetry (Humphrey Field Analyzer) and imaging with Spectralis SD-OCT (Heidelberg Engineering, Optic Nerve Head protocol). For both peripapillary retinal nerve fibe rlayer (pRNFL) and minimum rim width (MRW), color code classification as well as continuous data were considered. ROC curve, sensitivity and specificity (sp) were the outcome measures. For continuous data sensitivity at categorical matched specificity was extracted from the ROC curve. Borderline results were considered either as normal or glaucoma. P value <0.05 was considered statistically significant.

Results : One hundred and ninety-seven glaucoma and 83 controls were included in this study. For categorical data, when borderline was considered as glaucoma, the sensitivities significantly increase with minimal decrease in specificities. For pRNFL parameters, continuous data provided similiar sensitivity (range 47.21-84.26) to that of categorical classification (range 42.64-84.77). For MRW, continuous parameters provided significantly higher sensitivities (range 67.01-89.85) to that of categorical classification (range 48.57-78.68).

Conclusions : Categorical and continuous data offered similar diagnostic performance for pRNFL parameters but significantly different for MRW analysis. Diagnostic accuracy studies' results are is affected by which outcome measures are considered, and clinicians must be careful when translating the results from diagnostic accuracy study in their clinical daily practice.

This abstract was presented at the 2019 ARVO Annual Meeting, held in Vancouver, Canada, April 28 - May 2, 2019.