COMPARING STRUCTURAL AND FUNCTIONAL TECHNIQUES FOR DETECTING GLAUCOMA

L. M. Zangwill, C. Bowd, C. Berry, E. Z. Blumenthal, C. Vasile, P. A. Sample and R. N. Weinreb

Department of Ophthalmology, University of California, San Diego, USA

Purpose: To compare the abilities of confocal scanning laser ophthalmoscopy (HRT), scanning laser polarimetry (GDx), optical coherence tomography (OCT), short-wavelength automated perimetry (SWAP) and frequency doubling technology perimetry (FDT) to detect glaucoma classified in two ways, (1) based on optic disc appearance and, (2) based on standard automated perimetry (SAP).  Methods: One eye from 94 subjects was examined. Glaucoma by disc appearance (n=51) was based on stereophotograph evaluation. Glaucoma by SAP result (n=42) required a repeatable abnormal result. Healthy eyes (n=38) had both normal appearing optic discs and SAP results. Areas under the ROC curve, sensitivity, and specificity values were calculated for each diagnosis. Results: Several summary measures from each instrument had similar areas under the ROC curve.  The largest ROC curve area was found for OCT inferior quadrant thickness (0.91 and 0.89, for diagnosis based on SAP and diagnosis based on disc appearance, respectively), followed by FDT number of total deviation plot points < 5% (0.88 and 0.87), HRT RNFL thickness temporal/inferior region (0.85 and 0.82), SWAP number of points < 1% (0.83 and 0.81), and GDx discriminant function (0.80 and 0.82). ROC curve areas were not significantly different among the best parameter from each instrument for diagnosis based on disc appearance. For diagnosis based on SAP, ROC curve area was significantly larger for the best parameter from OCT than GDx. Agreement for classifying eyes as glaucomatous was limited.  Conclusion: Limited diagnostic agreement among instruments suggests that combining techniques may improve diagnostic precision.