Prediction of 10-2 Visual Field Loss Using Optical Coherence Tomography and 24-2 Visual Field Data

Using standard glaucoma structural and functional tests, clinicians accurately predicted the presence/absence of 10-2 glaucomatous visual field (VF) loss in 90% of the eyes in this study.

To investigate how well clinicians with variable experience can predict the presence and location of 10-2 VF loss using structural and functional data that are routinely obtained for glaucoma assessment.

Within a test set of 416 eyes (210 subjects) who were diagnosed glaucoma suspect or primary open-angle glaucoma (with most eyes having mild disease), 6 clinicians were asked to predict the presence and hemispheric location of 10-2 VF loss using 24-2 VF and spectral-domain optical coherence tomography structural data. Prediction accuracies were calculated for each clinician and compared using the weighted κ-statistic. Receiver operating characteristic analyses were used to evaluate models for predicting 10-2 VF loss.

Among the 6 clinicians, mean (range) accuracy, false negatives, and false positives for predicting presence/absence of 10-2 VF loss were 90% (87% to 92%), 4.7% (2.4% to 7.0%), and 5.4% (1.7% to 7.5%) respectively. The mean (range) weighted κ-statistic was 0.75 (0.64 to 0.83), suggesting good or very good inter-rater agreement between examiners. Mean accuracy for correctly predicting hemispheric location was 73% (range, 65% to 82%) with the most common error occurring in eyes with both superior and inferior 10-2 VF defects in which one hemisphere was correctly identified but the other missed.

In this study, the presence/absence of 10-2 glaucomatous VF loss was highly predictable using standard functional and structural clinical metrics. These findings suggest that 10-2 VF testing is not needed to reliably recognize and confirm central VF involvement in most eyes with glaucoma. Whether error related to identifying second hemisphere involvement in 10-2 VF loss is important requires further study.