Performance of iPad-based threshold perimetry in glaucoma and controls.

Abstract

Independent validation of iPad visual field testing software Melbourne Rapid Fields (MRF). To examine the functionality of MRF and compare its performance with Humphrey SITA 24-2 (HVF). Prospective, cross-sectional validation study. Sixty glaucomas mean deviation (MD:-5.08±5.22); 17 pre-perimetric, 43 HVF field defects and 25 controls. The MRF was compared with HVF for scotoma detection, global indices, regional mean threshold values and sensitivity/specificity. Long-term test-retest variability was assessed after 6 months. Linear regression and Bland Altman analyses of global indices sensitivity/specificity using (ROC) curves, intraclass correlations. Using a cluster definition of three points at <1% or two at 0.5% to define a scotoma on HVF, MRF detected 39/54 abnormal hemifields with a similar threshold-based criteria. Global indices were highly correlated between MRF and HVF: MD r2 = 0.80, PSD r2 = 0.77, VFI r2 = 0.85 (all P < 0.0001). For manifest glaucoma patients, correlations of regional mean thresholds ranged from r2 = 0.45-0.78, despite differing array of tested points between devices. ROC analysis of global indices showed reasonable sensitivity/specificity with (AUC) values of MD:0.89, (PSD:0.85) and (VFI:0.88). MRF retest variability was low with (ICC) values at 0.95 (MD and VFI), 0.94 (PSD). However, individual test point variability for mid-range thresholds was higher. MRF perimetry, despite using a completely different test paradigm, shows good performance characteristics compared to HVF for detection of defects, correlation of global indices and regional mean threshold values. Reproducibility for individual points may limit application for monitoring change over time, and fixation monitoring needs improvement.