Abstract
Purpose :
To investigate how well clinicians of variable experience can predict presence and location of 10-2 visual field (VF) loss in early glaucoma using 24-2 VF and spectral-domain optical coherence tomography (SD-OCT) macular and retinal nerve fiber layer (RNFL) data.
Methods :
After 6 clinicians of varying clinical experience reviewed a training set that demonstrated structure/function relationships between SD-OCT RNFL thickness, SD-OCT macular ganglion cell thickness, 24-2 VF and 10-2 VF tests, each clinician was asked to predict presence and associated hemispheric location of 10-2 VF loss using SD-OCT (RNFL, macular) and 24-2 data within a test set of 369 eyes from 190 subjects. The test set was comprised of subjects participating in a longitudinal glaucoma research study at the Albuquerque VA Medical Center whose 24-2 mean deviation (MD) was better than -6 dB and who were diagnosed primary open-angle glaucoma (POAG: glaucomatous optic neuropathy with repeatable VF loss on 10-2 and/or 24-2 testing) or glaucoma suspect (GS: ocular hypertension and/or optic nerve appearances that were suspicious for glaucoma but without repeatable VF loss). Predictive accuracy, false positives, and false negatives were recorded, and agreement was compared by ROC analysis and by weighted kappa-statistic (k).
Results :
Among 369 test eyes, median (IQR) MD for 24-2 and 10-2 VF tests was -0.79 (-2.35, 0.14) dB and -0.71 (-2.65, 0.29) dB respectively. Among all 6 clinicians, 10-2 VF loss predictive accuracy was 85-89% (mean: 88%). False-negative rates ranged from 12-29% with a mean of 20%, and false positive rates ranged from 4-13% with a mean of 9%. Hemispheric location of 10-2 VF loss (n=130 total defects) was accurately predicted 67% of the time (range 62-74% across raters) with most errors involving under-call of second hemisphere involvement. Agreement between clinicians was good to very good (k=0.62 to 0.82).
Conclusions :
While presence/absence of 10-2 VF loss was generally well predicted by all clinicians using routinely obtained clinical diagnostic data, there was a subset of 10-2 VF loss eyes (20-30%) in which presence and/or location of 10-2 VF loss was under-called. Considering that central glaucomatous visual field loss appears to be an important consideration within clinical risk stratification, our results suggest that 10-2 VF testing may have a valuable adjunctive role in early glaucoma evaluation.
This is a 2020 ARVO Annual Meeting abstract.