Agreement between clinical evaluation, Heidelberg-Retina-Tomograph (HRT) and Nerve Fiber Analyzer (GDx) in glaucoma diagnosis]

Background: We have made a comparison of clinical evaluation, Heidelberg-Retina-Tomograph (HRT) and Nerve Fiber Analyzer (GDx) in diagnosing glaucoma. Materials and Methods: One eye of 38 glaucoma patients and 26 non-glaucoma patients was examined with HRT and GDx. Assignment of the patients into the two groups by clinical evaluation was performed on the basis of visual field and optic disc results. As classification criterion for glaucoma/non-glaucoma by means of the instruments we employed the statistical classification of the HRT and "the number" of the GDx. Statistical analysis was performed with Cohen's kappa and McNemar tests. Additionally, ROC curves for evaluation of the sensitivity and specificity of the glaucoma classification are demonstrated for HRT and GDx. Results: Agreement between clinical evaluation and HRT was moderate (Cohen's kappa = 0.43), there was no significant over- or underdiagnosis by HRT (p = 0.48). Agreement between clinical evaluation and GDx (diagnosis of glaucoma when "the number" > 50) was bad (Cohen's kappa = 0.27) with a significant overdiagnosis by GDx compared to clinical evaluation (p < 0.05). Agreement between clinical evaluation and GDx (diagnosis of glaucoma when "the number" > 40) was better (Cohen's kappa = 0.54), there was no significant over- or underdiagnosis by GDx (p = 0.79). ROC curves for glaucoma classification showed no difference between HRT and GDx (area under the curve: HRT = 0.8, GDx = 0.78). Conclusions: Because of the only moderate agreement between clinical evaluation on the one hand and HRT and GDx classification on the other hand for the discrimination between glaucoma and non-glaucoma, the clinician should not completely rely on the instrument-derived glaucoma classification.