Abstract
Purpose :
To investigate the performance of visual field indices based on right–left eye asymmetries at detecting early loss in Primary Open Angle Glaucoma (POAG) and to establish whether their performance is superior to uniocular hemifield indices.
Methods :
206 control and 108 POAG patients (clinical diagnosis) were selected from a visual field database of patients attending Manchester Royal Eye Hospital. All POAG eyes had early glaucoma (GSS2 stage<=1). Age related defect asymmetries were calculated for each of the 44 test point pairs used in the Humphrey Glaucoma Hemifield Test (GHT). Two new indices were investigated: 1) Right/Left Standard Deviation (R/L SD) of differences; 2) Number of test Pairs (R/L NP) outside 95% CL of controls. Both indices were derived for whole field data (44 test pairs) and for hemi field data (2x22 test pairs). ROC curves were generated for the R-L indices along with those for 3 uniocular worse eye hemifield indices: 1) GHT; 2) Hemifield Standard Deviation of differences; 3) Hemifield Number of Pairs outside 95% CL.
Results :
The control and POAG samples median MD was 0.29dB (IQR=1.07) and -2.4dB (IQR 1.55) while the median absolute MD differences between the two eyes were 0.47dB (IQR 0.61) and 0.99dB (IQR 1.21). Figure 1 gives the ROC curves for all indices. Uniocular worse eye hemifield indices performed better than the new R/L eye asymmetry indices. were R/L MD=0.870 (95% Confidence Interval (CI) 0.828-0.911), R/L SD=0.925 (95% CI 0.894-0.957), R/L NP=0.899 (95% CI 0.862-0.936) and GHT=0.794 (95% CI 0.7.43–0.845) and hemifield analysis, area under the ROC for HMD, HNP and HSD were: 0.899 (95% CI 0.855-0.924), 0.945 (95% CI 0.925-0.966) and 0.963 (95% CI 0.948-0.978) respectively. .
Conclusions :
R/L asymmetry analysis has good discriminatory power but did not perform better than worse eye uniocular hemifield asymmetry indices. The performance of GHT was similar to that of combined uniocular indices in differentiating between normal and POAG eyes.
This is an abstract that was submitted for the 2016 ARVO Annual Meeting, held in Seattle, Wash., May 1-5, 2016.