Combining structural and functional assessment to detect glaucoma

COMBINING STRUCTURAL AND FUNCTIONAL

ASSESSMENT TO DETECT GLAUCOMA*

LINDA M. ZANGWILL,1 CHRISTOPHER BOWD,1 CHARLES C. BERRY,2 SANAZ FARID,1 PAMELA A. SAMPLE1 and ROBERT N. WEINREB1

1Hamilton Glaucoma Center, Department of Ophthalmology, and 2Department of Family and Preventive Medicine, University of California, San Diego, La Jolla, CA, USA

Abstract

Purpose

To assess whether combining results of scanning laser polarimetry (GDx Nerve Fiber Analyzer), optical coherence tomography (OCT), confocal scanning laser ophthalmoscopy (HRT), short wavelength automated perimetry (SWAP), and frequency doubling technology (FDT), can improve the detection of glaucoma.

Methods

Seventy-two healthy eyes (mean age, 62.4 ± 10.4 years) and 66 eyes with early to moderate glaucoma (mean age, 64.4 ± 1.3 years) were included. Two definitions of glaucoma were used; glaucoma based on repeatable abnormal results on standard automated perimetry (SAP) (n = 43; SAP MD; -6.9 ± 3.4 dB) and glaucoma based on optic disc appearance assessed by masked assessment of stereoscopic optic disc photographs (n = 64; SAP MD, -4.8 ± 6.0 dB).

Parameters from different structural and functional instruments were combined using flexible discriminant analysis and k-fold cross validation techniques (to estimate ROC curve areas without resubstitution bias). The area under the ROC curve for discriminating between healthy and glaucomatous eyes (based on the two criteria above) was estimated for each instrument separately and in combination with additional instruments. In addition, sensitivity was assessed with eyes classified as glaucoma if either one structural or functional parameter was outside the normal range (at ≥ 80% specificity).

Results

In general, the area under the ROC curve was larger when the diagnosis was based on SAP compared to a diagnosis based on disc. Several parameters from each instrument had similar areas under the ROC curve. Of the structural measures, HRT and OCT had larger areas under the ROC curve than GDx for a diagnosis based on disc (0.91, 0.85, and 0.76, respectively) and on field (0.89, 0.89, and 0.81, respectively). Of the functional measures, FDT had a larger area under the ROC curve than SWAP for a diagnosis based on disc (0.86 and 0.76, respectively) and field (0.96 and 0.90, respectively). Combinations of structural and

* The full article will be published elsewhere.

Address for correspondence: Linda M. Zangwill, PhD, Glaucoma Center and Diagnostic Imaging Laboratory, Department of Ophthalmology, University of California, San Diego, 9500 Gilman Drive, La Jolla, CA 92093-0946, USA. Email: Zangwill@eyecenter.ucsd.edu

Perimetry Update 2002/2003, pp. 187–188

Proceedings of the XVth International Perimetric Society Meeting, Stratford-upon-Avon, England, June 26–29, 2002

edited by David B. Henson and Michael Wall

© 2004 Kugler Publications, The Hague, The Netherlands

functional measures (via flexible discriminant analysis) did not improve on the overall discriminating ability of OCT, HRT, and FDT measures alone (area under the ROC curve of 0.92 for a diagnosis based on disc and 0.94 for a diagnosis based on field). However, at a specificity of ≥ 80%, the sensitivities for detecting glaucoma improved for some instruments when a structural parameter was combined with a functional parameter.

Conclusions

Given the amount of data available, combining structural and functional parameters did not improve on the overall ability of HRT, OCT, and FDT parameters alone to differentiate between healthy and glaucoma eyes. At a fixed specificity of > 80%, combining results from some structural and functional instruments may improve the sensitivity of detecting glaucoma.

Acknowledgment

This study was supported in part by National Eye Institute grant EY11008 (LMZ) and EY08208 (PAS).